Searching \ for '[OT] [EE] 24-bit A/D. Are We in the Twilite Zone H' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/io/atod.htm?key=a%2Fd
Search entire site for: '[EE] 24-bit A/D. Are We in the Twilite Zone H'.

Exact match. Not showing close matches.
PICList Thread
'[OT] [EE] 24-bit A/D. Are We in the Twilite Zone H'
2000\04\24@165527 by Dan Michaels

flavicon
face
Wagner wrote:
.....
>I use Max641, 586 or 587 at my devices (the first one has an internal
>catch diode). EMI? yes, but not enough to cause any problems at the 22.5
>bits ADC installed less than 20 mm from the coil.  Just for info, this
>ADC resolution is about 0.12uV (120 nano Volts), the interference should
>exist but it is not noticeable.
>

Yeah, 0.12 uV. 22.5 bits. Right.

Wagner, I noticed your mention of using the LTC2400 a couple of weeks
ago, and downloaded the datasheet. I'm still digesting it. I've built
amps for amplifying 2-5 uV before, and it's not all that easy to get
good results at those levels. Now, you and Linear Tech come along and
say, hey, we got 0.12 uV sensitivity here, and it works with a buck
regulator ckt, too. Oy, yoy, yoy.

I've got one board with opAmps and a 12-bit LTC1400 running up to
400 Ksps [with 20 Mhz PIC on-board, of course], and damn if I can
get the noise level below about 2 mV, no matter what I do. I've
read everything ever written on low-noise analog and high-speed
digital design, and tried 100 different things, and it's still
2 mV [that's **MILLI-volts**].

On a thread a few weeks ago, I asked a simple question of someone
about how they got their Vref to be noise-free at 12-bits, got
no response back from that guy, but did get some other clowns
telling me "... seems some people are re-learning things the hard
way ...." and ".... yes folks ohm's law is a reality ...", or
some such blather.

Forget the switcher. 0.12 uV. And with digital circuitry within
200 *yards*. Are you kidding?

I realize that the LTC2400 has differential input and 3 hz
bandwidth, etc/etc, whereas I have single-ended input, opamp ckt
with Rin = 1M and BW = 1 Mhz, plus A/D SPI running at 6.4 Mhz,
but still -- 2 mV noise. What's a poor fella to do? [beyond the
100 things he's already tried].

Am I at the tradeoff point in my design? Is there some magical
Noise-vs-??something??  formula here I forgot about [like trading
gain for bandwidth]? Also, they are putting 16-bit sound cards
into PCs nowadays. What the heck do those guys know, that they
can accomplish that?

Cheers,
- Dan Michaels
Oricom Technologies
===================

2000\04\24@175950 by dal wheeler

flavicon
face
Along these lines (minus the head butting), what are people actually *doing*
to compensate for the switcher noise when hooking up to analog circuitry.
Yes, I know it's better to avoid the issue and go with a linear supply, but
what have people done to "clean up" the analog portion when using a
switcher.

{Quote hidden}

2000\04\24@194816 by Wagner Lipnharski

flavicon
face
Well Dan, noise is a tricky thing.

Right now I am producing a pressure calibrator that uses the AD7713,
well, somehow extinct and replaced (supersede) by the AD7714, it is a
22.5 bits ADC, with 3 fully differential selectable inputs.  The
pressure transducer supply 100mV with the full scale pressure, but if I
supply the sensor bridge with 1/3 of its required constant current, it
gives me only around 32mV at full scale.  I am being reading 0.1PSI with
this 30PSI transducer, so it means around 100uV of analog signal, and
the 7713 is pretty steady and accurate.

The biggest problem I found years ago with low voltage is really noise
in voltage reference generators.  At this point I found out that in this
particular point "you have what you pay for".   The LM358 1.2 and 2.5V
reference has several versions, going from 91 cents up to $4.66
(30ppm/degC).  The LM4040 series are pretty good too, but you need to go
for a .1%, as the LM4040AIZ-2.5-ND (Digikey) for $4.10 per (TO-92)
unit.  The most important part of the tech specs of those vref's is the
"noise" part.  Some of them just obscure it, other just misinform, but
some state it. Some of them have milli-volts of noise (white noise,
pink, whatever), it is an intrinsic silicon noise, can't be removed,
can't be filtered, there is NO capacitor can avoid it, stabilize or
level it.  You just need to check the manufacturer document (specs
sheet) for the one you are using.  The above information is from
Digikey, you can get better products around.

Years ago I made a nice portable instrument, one of the functions were
to generate current and voltage, in micro-volts range.  I suffered like
hell trying to locate why the output voltage was so noisy. The output
was programmable from 0 to 25V in 10 micro-volts steps, so you could
specify 24.999,99 but the high precise HP unit I was using to measure it
always showed last two digits floating and jumping.  I changed
everything, from the solderless protoboard (first mistake), op-amps,
capacitors, resistors to 20ppm 0.1%, and so on.  Using a signal booster
I could see the vref noise. I was using a 1.25Vref, noise was in the
60uV range (according to the manufacturer), but the DAC was multiplying
by almost 20 times to get the 25V output, so the noise went up to almost
1.2mV!!! sometimes higher.

So I learned that there are two very, very important things in all
electronics... the bases... Clocks for digital and Vref's for analog.
They can cost up to 1/20 of the total equipment, now I recognize and pay
it with pleasure, they ARE what the product will offer as quality and
accuracy anyway.  Doesn't make sense to use a 91 cents Vref in a product
that cost $1500 to the customer, and that cheap Vref will define the
product accuracy and quality... you know what I mean.

I never thought to use a thermal oven for a crystal oscillator in life,
but I did once. It runs under 45¡C, needs 48 hours to stabilize before
calibration, and needs 4 calibrations, once each 6 hours after the
initial warm-up. Can't use a regular metallic screwdriver to touch the
cap inside the crystal chamber since it can change the internal temp...
and of course frequency, isn't that funny?  The plastic screwdriver
should be kept in another 45¡C chamber... scary. But when you go for
critical and high accurate things, there is no other way.

Calculations:   1.25Vref with 60uV of noise, means:

625mV  = 1 bits
312mV  = 2 bits
156mV  = 3 bits
78mV  = 4 bits
39mV  = 5 bits
19mV  = 6 bits
 9mV  = 7 bits
 4mV  = 8 bits
 2mV  = 9 bits
 1mV  = 10 bits
610uV  = 11 bits
305uV  = 12 bits
152uV  = 13 bits
76uV  = 14 bits
38uV  = 15 bits

So, the noise kills the unit at 14 bits, can't use it with a 16 bits ADC
or DAC.

Now, other thing. accuracy (percentage of error).
The LM4040 unit described above, has 0.1% or error. It means that in
1.25000V vref, it can be in real 1.249000 or 1.251000, now, if you look
at the above table, find the 1/1000 part, you will see the 1mV at 10
bits, so all above 10 bits will be not valid.

Now, look for ppm, a 30ppm/degC, means the unit will change its value in
30 parts per million per each celsius degree the ambient temperature
changes from the based 20¡C.  Then, suppose the room is at 25¡C, it will
be 150 parts per million, 150/1000000 or 0.00015, in other terms it is
0.015%, pretty high for only 5¡C in change, it is 0.003% per ¡C.  The
original 1.250000 V will change to 1.249250 V at zero ¡C, add a possible
.1% of negative error, will go to 1.248V, so it is already 2mV, drop
your accuracy to below 9 bits... pretty scary, huh?

We are not talking about a cheap 91 cents Vref, but about the 30ppm (one
of the best at the market) LM4040, with the best possible available at
Digikey with 0.1% of error, the most expensive one, around $5 per unit,
and that thing can't really goes up to 12 bits at 0¡C???? Yes!

What happens here is that you need to deal with all the possibilities.
First, you need to compensate and should trust in calibration. If the
unit is actually with 0.1% of error, the calibration will fix it, IF the
unit keep the error permanently.  Second, temperature offset and
deviations should also compensated by software, for that you need to use
a temperature sensor and calibrate the unit for different temperature
ranges, so the software would know where to look and how to do.

Going to your problem.
----------------------
2mV noise is pretty high.  How do you get it? How do you measure it?  at
the ADC output? or with an oscilloscope?. What is your signal source?
what impedance it offers?  To work with very low signals, it is *VERY*
important to match impedances, if not, bounced signals can kill the
original and clean one.  Sometimes a capacitor creates more problems
than solve. A capacitor is not a God's creation. It leaks, it farts, it
dirts all over the place in very low voltage.  
At 24 bits, nanoVolts range, a single electron leaking at the cap plates
can creates a bouncing signal at the ADC input.  
Here again, expensive and good polypropylene caps makes a hell of a
difference from cheap ones.

What happens if you short circuit the ADC input, still having the noise?

The trade-in here is to avoid at any cost high gain amplifiers, try to
deal with the signal as pure as it is, avoid intermediate circuits,
resistors, dividers, filters, whatsoever that can transform your life in
a live hell.  Signal treatment is a long and extensive learning at any
electronic university, and it is not any teacher of this subject that
can build a nice front-end amplifier to work with a 24 bits A/D.

Answering your subject question; Yes, at 24 bits A/D, we really ARE at
the twilight zone, doomed, alone, in total fear, no support, no
experience around, no books, no fancy "do it by yourself multicolored
animated gif playing at your screen"... just plain despair!

Keep contact.
Wagner.

PS: Just try and turn off the PC and that fluorescent lamp over your
bench, and see if the 2mV disappear... :)  


Dan Michaels wrote:
{Quote hidden}

2000\04\24@202143 by David VanHorn

flavicon
face
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


>Am I at the tradeoff point in my design? Is there some magical
>Noise-vs-??something??  formula here I forgot about [like trading
>gain for bandwidth]? Also, they are putting 16-bit sound cards
>into PCs nowadays. What the heck do those guys know, that they
>can accomplish that?

Begin at the beginning :)

Decouple your supplies into the op-amp.  A series R and local C to ground
should work well.

Check the op-amp PSRR AT the frequency of your main interference.

Use a spectrum analyzer to check everything, not just a scope.

Using a SMPS?, trigger the scope with the SMPS, and then go hunting noise
with the other input.

Make sure your local ground is quiet, and does not have someone else's
return current running through it. On my check reader board, I had two
700mA chopped stepper drivers less than an inch away, and got zero pickup
from that. Or the uP, or the SMPS. Un-measurable.

In the circuit itself, lower the impedances if you can, keep layout tight,
don't ground or decouple to any foreign points.

- --
Are you an ISP?  Tired of spam?
http://www.spamwhack.com  A pre-emptive strike against spam!

Where's Dave? http://www.findu.com/cgi-bin/find.cgi?kc6ete-9

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.2 for non-commercial use <http://www.pgp.com>

iQA/AwUBOQT++IFlGDz1l6VWEQIpYQCeLq58NHtbpb1ZJRA5wQ9ltNdR3AsAn1Pq
LkMoQ/GbkYuY6XVm610qkEyx
=Tu/C
-----END PGP SIGNATURE-----

2000\04\24@203607 by David VanHorn

flavicon
face
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

U want fun?

My current bugaboo:

I have a board that has to deliver a 30mV (full scale) signal to another
board.
That board emits 6 watt RF pulses at 916 MHz on an antenna 0.25" from my
board, and it takes 2A current pulses 50uS wide, with nice sharp edges.

The return current was the major problem, since the mic is a single ended
input. Transformer coupling solved that one, now I'm dealing with direct
rectification effects. I go tomorrow up to ames labs to evaluate the new
PCB.  We'll see what happens!


- --
Are you an ISP?  Tired of spam?
http://www.spamwhack.com  A pre-emptive strike against spam!

Where's Dave? http://www.findu.com/cgi-bin/find.cgi?kc6ete-9

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.2 for non-commercial use <http://www.pgp.com>

iQA/AwUBOQUCC4FlGDz1l6VWEQL47ACfTTNm97sHTd7Ip68HeBIb5ekAlFIAn3kF
eM3z4N82sHY9gxKwjMotaD+/
=qEjU
-----END PGP SIGNATURE-----

2000\04\24@214002 by Sean Breheny

face picon face
Hi Wagner,

At 07:48 PM 4/24/00 -0400, you wrote:
>some state it. Some of them have milli-volts of noise (white noise,
>pink, whatever), it is an intrinsic silicon noise, can't be removed,
>can't be filtered, there is NO capacitor can avoid it, stabilize or
>level it.  You just need to check the manufacturer document (specs

I don't see why you couldn't low-pass filter the output of the reference?
Prodiving that the thing that is using the ref isn't drawing much current
from it, you could just put a resistor in line and add a low-leakage cap
after it.

Even if the thing that needs the reference needs to draw a good deal of
current, you could buffer with a low-noise op-amp and still do the LPF.

Sean

|
| Sean Breheny
| Amateur Radio Callsign: KA3YXM
| Electrical Engineering Student
\--------------=----------------
Save lives, please look at http://www.all.org
Personal page: http://www.people.cornell.edu/pages/shb7
spam_OUTshb7TakeThisOuTspamcornell.edu ICQ #: 3329174

2000\04\24@215650 by Sean Breheny

face picon face
Hi Dan,

I think you went past the BW consideration too quickly. There is a BIG
difference between getting  uV resolution in a 3HZ BW and in a 1 MHz BW.

I have been working on Cornell's Autonomous Helicopter project and we are
pushing MuRata rate gyros to their limits. They give 640uV per deg/sec. I
measured their output noise using the best low-noise amp I could make and I
got 0.05 deg/sec noise floor in about a 50Hz BW (amp noise + gyro noise)
which corresponds to 32uV of noise.

I then designed a circuit on a double-sided circuit board with the top
almost dedicated to ground plane. I kept the digital and analog GND planes
separate except at the connection point of the main GND wire. I placed the
recommended bypass caps for the AD974 (the 16-bit 200ksamp/sec ADC we are
using) and I bypassed the supplies to the op-amps in the circuit). I got a
final noise of 0.066 deg/sec for the whole system, which corresponds to 42
uV. I obtained this on the first try, by just following the manufacturers
recommendations and using careful layout and limiting BW when I could (I
think this is key). I have an (AVR ;-) ) micro running on the same board,
inches away.

Also, I think I was able to improve the noise performance by amplifying my
signal so that my minimum resolution corresponded to a couple of bits on
the ADC, so I wasn't pushing the ADC to its limits.

Sean

At 02:49 PM 4/24/00 -0600, you wrote:
{Quote hidden}

| Sean Breheny
| Amateur Radio Callsign: KA3YXM
| Electrical Engineering Student
\--------------=----------------
Save lives, please look at http://www.all.org
Personal page: http://www.people.cornell.edu/pages/shb7
.....shb7KILLspamspam@spam@cornell.edu ICQ #: 3329174

2000\04\24@225408 by Wagner Lipnharski

flavicon
face
Yes, I thought the same, did several different kind of passive filters.
Used from pico to microFarad capacitors, different capacitors, diferent
resistors. Looks like the microVolts noise increase with some
capacitors... leaking?  Used a filter like this, no change at all.

      1k       1k                   1k       470
+5V----R---o----R-----o-----o----o---R---o-----R----o-----o--->
          |          |     |    |       |          |     |
         ===        ===   ---  ---     ---/       ===   ---
         ---        ---   ---  ---      A   1.2V  ---   ---
          |          |     |    |       |          |     |
         Gnd        Gnd   Gnd  Gnd     Gnd        Gnd   Gnd

       100uF      220uF  0.1uF  82pF            330uF  0.1uF

When you read the documentation you will understand why they offer it in
different noise levels, still the same unit, cost 3 or 4 times more.
The noise can reach low frequency, in some rare situations I noticed it
as low as 0.5Hz.  National website pdf file about the LM385
http://www.national.com/ds/LM/LM185-1.2.pdf state at page #4, two graphs
about noise, the second show a single pole low pass filter where the
noise is reduced from 60uV to 55uV at 100kHz noise, with a sharp cutoff
filter it is reduced to around 35uV... so no game, noise still there.
The low voltage noise is difficult to remove.

Wagner

Sean Breheny wrote:
{Quote hidden}

2000\04\25@121404 by Dan Michaels

flavicon
face
Joe, Grif, Wagner, Dave, thank you all for the responses.
I think I have been around the track a couple of times with
pretty much everything you all mentioned.

Bypass caps everywhere, as much physical separation between
analog and digital sections as possible given a 3"x5" pcb,
"separate gnd busses" for analog and digital, analog gnd
single point connection returned directly to power entry point,
as much gnd planing as possible on a 2 layer pcb, wide power
traces, separate analog regulators, lowest feasible R values
on opamp ckts, 2 opamp inverting stages in series, not using
non-inverting stages, ferrite bead in power in lead and Rs-232
line.

Also, pcb is in an EMI-shielded ABS case, and with case
removed, the noise *is* always 2x to 3x to 5x higher. [I
wonder if this isn't "the" crucial clue?]. I assume this
means pickup from local digital sections of pcb is what is
being shielded. I don't think it's coming from outside.
=================

Some specifics:
- 2 mV noise is referenced to input, with input grounded
 directly and measured by reading A/D output, not with scope.
- for this measurement, input range = +/-1.25v, so 2 mV
 represents only about 9 bits - so am losing ~3 A/D LSBs.
- Power train:
 - wall wart -> 7812 --> 7805 (digital).
                 "   --> 78L06 (analog +).
                 "   --> 7662 (-12v) --> 79L06 (analog -).
 - all regs bypassed, filtered high side, low side, etc.
- Gnd planes present under noisy 7662 (766x derivative). Noise
 referred to above is broadband noise, not correlated to 7662
 switching.
- Analog circuitry: 2 inverting op amp stages (superior CMR to
 non-inverting), currently using TLE2072, 4 gains set via '4052.
 - Lowest range: +/-1.25v, Av1=0.4, Av2=4.
 - R values: 1st stage - 1M in, T-section feedback to lower
   effective R, with 82K-21K-68K. 2nd stage 12K in, 50K fb.
 - both stages switch gain via '4052 mentioned above.
 - output to A/D thru small R.
 - opamp/CMOS switch bypassed directly at pins.
- A/D: LTC1400, SOIC, 12-bit, 8-pin, internal Vref, SPI control
 at 6.4 Mhz. Vref bypassed directly at chip. Vdd, Vss bypassed
 close to chip.
- Digital control: PIC76 @ 20 Mhz, bypass caps in several places,
 gnd planes under chip, gnd island around xtal.
 - no hi-speed signals passing near analog circuitry.
 - 2 DC control signals from PIC to '4052, bypassed.
- RS-232 connection to pcb: lots of noise was originally measured
 on the TX line from the PC, and now has a ferrite bead inline.

Note - overall, this is a "system-level" problem. Mixed
 analog-digital, small space, noisy 7662, R values as small as
 I can make them, RS-232 to PC, wall wart power, etc. I have
 isolated and dealt with any number of noise sources already.

Working on next pcb layout:
- moving analog regs further away from PIC.
- removing 7812 from inline with 7805.
- considering moving single-point analog gnd connection from
 78L06/79L06<->power_entry_point to under A/D.

Note - I have several times fiddled with the opamp ckt design
 and R values, and they are about as low as I can go now, and
 still keep Rin = 1M.

I may need to go to a completely different front-end design, ie
FET follower amp - unfortunately this is much more complicated,
requires lots more space, it drifts, etc/etc. I was hoping to
keep the front-end design simple and straightforward, since
BW is only 1 Mhz, but I am beginning to wonder.
=============
=============

Re what Dave said:
>Decouple your supplies into the op-amp.  A series R and local C
to ground
>should work well.

Bypass caps are "everywhere", but I haven't tried series R in
the opamp power leads. However, they have their own regulators,
bypassed and separate from the digital regs, plus a separate
analog gnd, too. So I don't think a series R here will help much.
=================

>Check the op-amp PSRR AT the frequency of your main interference.
>
>Use a spectrum analyzer to check everything, not just a scope.

Measured noise is pretty wideband, not correlated to PIC xtal
or 7662 - [xtal is outside passband, 7662 has been dealt with].
===============

>In the circuit itself, lower the impedances if you can, keep
layout tight,
>don't ground or decouple to any foreign points.

I already have lowered the R values in the analog section as
far as I think I can go. The 2 mV noise was measured with input
BNC shorted.

"... gnd or decouple to a foreign point". What does "foreign"
point" mean?
====================
====================

Re what Joe mentioned:
>And yes, you may be missing something.  You can trade off bandwidth for
>noise! The narrower the bandwidth, the lower the noise figure. Every time
>you reduce bandwidth to 50% noise is decreased  to 35%.  Its one of those
>square law formulas I learned in Signals.

I am pretty familiar with signal theory, but this is a wideband ckt,
and the analog has to be made to tow the line. BW here is 1 Mhz, and
I wish I could make it 20 Mhz!!  [I could probably get Walter's 0.12
uV, if I LP filtered down to 3 hz. Yeah, right].

------->  My wondering was whether there isn't some magical formula
saying something like "for a 1 Mhz BW, no matter how damn hard you
try, you aren't ever gonna get the noise below ___ mV, unless you
do ........". [as in, "make the Rs all <= 1K", whatever].
===============

>There are other exotic methods for squeezing the last little bit out of a
>signal path, synchronous detectors, digital filtering, extremely long (over)
>sampling, etc.

My case is 1-shot waveform captures, and I want to make the ckt
work as well as possible *before* the DSP stuff comes in [which it
does later on].
===============
===============

RE what Wagner mentioned, I am still mulling over your comments.
Lots of good info there. The Vref on my LTC1400 A/D is inside
the chip, bypass caps outside. Noise measured via A/D binary
output, with front-end amp tied hard to gnd.

>Here again, expensive and good polypropylene caps makes a hell of a
>difference from cheap ones.

My only caps are ceramic bypass and electrolytic P/S filtering.
Where did you use polypropylene? For coupling?
==============

Whew, what a marathon. Given all this, if there is something
I have clearly overlooked, ....... ????????

Best regards,
- Dan Michaels

2000\04\25@121408 by Dan Michaels

flavicon
face
Appendix to marathon note:
........
>>Here again, expensive and good polypropylene caps makes a hell of a
>>difference from cheap ones.
>
>My only caps are ceramic bypass and electrolytic P/S filtering.
>Where did you use polypropylene? For coupling?
==============

Oops, forgot about the small caps [1-20 pF range] across the op
amp feedback Rs, used for overshoot & BW control. Ceramic.
Would polypro work better here?

regards,
- Dan Michaels
==============

2000\04\25@121434 by Tom Handley

picon face
  Dan, I'm currently working on a PIC-based DAQ system that uses the
LTC2400. Others have mentioned noise problems and techniques with
low-voltage measurement. I think the key here is that the LTC2400 family are
extremely accurate converters that can provide very accurate measurements at
lower resolutions. I'm using 16-18 Bits for display while averaging all the
Bits. The software provides for calibration.

  Wagner mentioned reference noise which is a good point that is sometimes
overlooked. I'm using a MAX6341 4.096V reference with a 1ppm/C Tempco, 0.02%
initial accuracy, and around 2.5uV of noise. It provides a pin for an
external cap which augments wide-band noise reduction. With 24 Bits the
resolution is ~244nV. As is pointed out in one of Analog Devices'
applications books; 600nV is the Johnson Noise in a 10KHz BW of a 2.2K
Resistor @ 25C. So, yes, this really is the "Twilight Zone" ;-)

  The LTC2400 is extremely easy to talk to but you must decide which mode
to use. In my case, I use the External SCK/Single Cycle/Reduced Data mode.
With /CS initially High and SCK initially Low, the basic flow is:

     Wait for End of Conversion:
     - Set /CS Low
     - Wait for SDI Low

     Shift-in first 4 Bits (EOC -> Dummy -> Sign -> Extended Range)

  At this point, you have Bits 31..28 in Data register Bits 3..0 where:

     Bit 31 (3) = End of Conversion Bit
     Bit 30 (2) = Dummy Bit
     Bit 29 (1) = Sign Bit
     Bit 28 (0) = Extended Range Bit

  The next Bit is the MSB of the conversion where Bits 27..4 are the 24-Bit
result and Bits 3..0 are sub LSB's that can be ignored or used in averaging.

  - Tom

At 02:49 PM 4/24/00 -0600, Dan Michaels wrote:
{Quote hidden}

------------------------------------------------------------------------
Tom Handley
New Age Communications
Since '75 before "New Age" and no one around here is waiting for UFOs ;-)

2000\04\25@122723 by Don Hyde

flavicon
face
Well, I did software on a project where my code turned the switcher off long
enough to make the actual measurement -- took a lot of C, and a linear
regulator to clean up the droop, but it worked.  We needed the switcher to
get 2.4V or whatever from an end-of-life 3v cell up to 5v, but the op-amps
and digitals could go on 4.5 or so.

> {Original Message removed}

2000\04\25@133222 by Craig Lee

flavicon
face
You could also put little RCs at the source of digital signals to
attenuate the high frequency components that are more likely to
radiate.  Requires that you know your maximum frequency to select
the RC. (has this been said already?) ie. LSB changes fastest...

Craig

{Original Message removed}

2000\04\25@140314 by Wagner Lipnharski

flavicon
face
Dan Michaels wrote:
[snip]
> My only caps are ceramic bypass and electrolytic P/S filtering.
> Where did you use polypropylene? For coupling?
> ==============

As I said before, working with Volts, even mV, it doesn't care so much
what type of capacitor you use, but goint to microVolts or even
NanoVolts, things change drastically. Very low leakage filter capacitors
are very important.  At 500nV, how many electrons do you think would be
necessary to leak from one capacitor's plate to another to bouce the
voltage to 510nV? not a bunch.  How many bits into the A/D it means? a
24 bits ADC would suffer with that change at the vref. Some
Polypropylene Film ECQP(U) (Panasonic) offers a minimum isolation of 45
GigaOhms, probably your circuit board will have less isolation between
tracks.

Just a question: THe 2mV readed at the A/D output is a jumping noise or
a steady noise, that can be confused with a/c or d/c A/D input offset?

Another point is that I never use a 16 bits ADC to read 16 bits
signals... I just ignore last 3 or 4 bits, so I pay more for a steady
conversion... using an expensive 22 bits AD7713 to really read 18
bits... going now to the LTC2408 (24 bits that in real are 28 bits) to
use just 20 bits... that's the deal, come on, it is a 6 digits counter
with 20 bits!!! I can have a resolution of 10nV at 10mV range...
(9.999,99)

> The Vref on my LTC1400 A/D is inside the chip,
> bypass caps outside. Noise measured via A/D binary
> output, with front-end amp tied hard to gnd.

What about short to ground the LTC1400 inputs? not just the front end
amp inputs...  You need to isolate and make sure where is the origin of
the noise.

Wagner.

2000\04\25@140728 by Wagner Lipnharski

flavicon
face
Even a good polypropylene or polyester cap can generate noise if you are
out of luck.
A good way to check those guys is increase the op-amp gain to the max
rail limits and check the noise generation, so you could play with
components until you find out the ones that would benefit you with low
noise, then reduce op-amp gain to the correct gain.

I already made few op-amps high impedance gain, that the input coupling
cap + resistor could not touch the circuit board... yup, just in the
air, the best isolator I know.

Wagner.

Dan Michaels wrote:
{Quote hidden}

2000\04\25@141554 by Wagner Lipnharski

flavicon
face
Tom Handley wrote:
>
>    Dan, I'm currently working on a PIC-based DAQ system that uses the
> LTC2400. Others have mentioned noise problems and techniques with
> low-voltage measurement. I think the key here is that the LTC2400 family are
> extremely accurate converters that can provide very accurate measurements at
> lower resolutions. I'm using 16-18 Bits for display while averaging all the
> Bits. The software provides for calibration.
>
>    Wagner mentioned reference noise which is a good point that is sometimes
> overlooked. I'm using a MAX6341 4.096V reference with a 1ppm/C Tempco, 0.02%
> initial accuracy, and around 2.5uV of noise. It provides a pin for an
> external cap which augments wide-band noise reduction. With 24 Bits the
> resolution is ~244nV. As is pointed out in one of Analog Devices'
> applications books; 600nV is the Johnson Noise in a 10KHz BW of a 2.2K
> Resistor @ 25C. So, yes, this really is the "Twilight Zone" ;-)

Yes, I forgot to say, using a high voltage Vref, like 4.096, the silicon
noise is the same from the smaller ones, so the relation Vref/noise is
increased.  I did it once, even using 0.1% 10ppm matched pair resistors
(If I remember well) to reduce the Vref it worths, the final noise is
also reduced.

The Max62xx/63xx units are very steady ones, of course they need to
justify the cost above $10 :)

2000\04\25@150803 by andy howard

flavicon
face
From: "Dan Michaels" <.....oricomKILLspamspam.....LYNX.SNI.NET>

> Joe, Grif, Wagner, Dave, thank you all for the responses.
> I think I have been around the track a couple of times with
> pretty much everything you all mentioned.
> Bypass caps everywhere, as much physical separation between
> analog and digital sections as possible given a 3"x5" pcb,
> "separate gnd busses" for analog and digital, analog gnd
> single point connection returned directly to power entry point,
> as much gnd planing as possible on a 2 layer pcb, wide power
> traces, separate analog regulators, lowest feasible R values
> on opamp ckts, 2 opamp inverting stages in series, not using
> non-inverting stages, ferrite bead in power in lead and Rs-232
> line.
> Also, pcb is in an EMI-shielded ABS case, and with case
> removed, the noise *is* always 2x to 3x to 5x higher. [I
> wonder if this isn't "the" crucial clue?]. I assume this
> means pickup from local digital sections of pcb is what is
> being shielded. I don't think it's coming from outside.



Hi Dan.

I don't know if you've seen
http://www.channel1.com/users/analog/noise.html.  It's a useful
discussion of noise reduction, mostly focussed on PCB layout issues.

In fact the whole site is very interesting with loads of useful hints
and tutorials about analogue electronics with a very practical bias.
Well worth a look IMO.


Cheers

Andy.

2000\04\25@162657 by Dan Michaels

flavicon
face
Thanks to Sean, Don, Craig who also responded here. Some
additional thoughts regarding my noise problem:

Grif wrote:
.........
Have you tried moving the first op amp into a separated, shielded box, with
.....
 Feed thru's on the
power into the box,,, and signal in and out of the box. and a separate
regulator, make sure you don't let any left over noise on the input side to
the regulator couple to the ground or output of the regulator.
.........

Thanks, for the many suggestions. I think I have already incorporated
most of them. However, thermal control - not on this thing!! Also, the
front-end amp isn't separately shielded, but is as far from the digital
cktry as possible, has gnd planing under all the chips/Rs/etc, has the
EMI-shielded ABS box overhead, consists of 2 inverting amps in series,
has separate V regs and busses, low-Z 2nd amp, etc.

The one thing I might try on the new layout is returning the bypass
caps on the analog V-reg hi-sides straight back to the power entry
point. They return to the analog gnd buss now. Might help.
=====================
=====================

Wagner wrote:
.........
>The biggest problem I found years ago with low voltage is really noise
>in voltage reference generators.
.........
>the twilight zone, doomed, alone, in total fear, no support, no
>experience around, no books, no fancy "do it by yourself multicolored
>animated gif playing at your screen"... just plain despair!
..............
>PS: Just try and turn off the PC and that fluorescent lamp over your
>bench, and see if the 2mV disappear... :)
>

This was good info you provided on your experiences with V-references.
[And I'm gonna ignore all that "total fear" & "plain despair" talk.
Sun's out today. I need to get < 1 mv noise, before worrying about
10 uV. :-)].

Considering I am using the LTC1400 with builtin Vref, there isn't much
I can do in this arena, except external bypassing, heavy traces, etc,
which I do now. I would think the 1400's Vref would be commensurate
with 12-bit ops.

Is it your experience, Wagner, that onboard Vref is worse than
external? After all, the A/D's 6.4 Mhz internal SPI cktry is only
microns away.

I did discover the PC sends piles of hash out the RS-232 line, and
of course, the monitor is a great broadcaster. I did some controls for
this, added a ferrite and shielded cable to the RS-232 line, and started
using EMI-shielded ABS cases [huge improvement, as noted before].

I probably should go thru another round of testing, to try to localize
the noise source(s) better. It is pretty broadband, but thinking about
it, I should do more scope/triggered/source_correlation measurements,
since end-to-end measurements suffer from aliasing in the A/D.
================
================

Anyone have any ideas regards the great noise reduction I see
when the EMI case is put on?

I do think this is shielding LOCALLY-generated noise, rather than
external. I presume "radiated" E-noise, because conducted and
H-noise would not be affected as much by the shield. Correct?

I do know that the shielding greatly cuts radiated E-noise from
the 7662 and local VCO chip [those sources were already dealt
with, BTW, via gnd planing and separating gnds, and easy to
measure since can be correlatd]. I do not believe I have any
serious inductive loops in the "digital" section - the power
and gnd busses and signals all run pretty much in close
parallel formation there.

Please don't anyone suggest "multi-layer" pcb for this project.
I'll have to accept 2 mV noise over that option.

Cheers,
- Dan Michaels
Oricom Technologies
===================

2000\04\26@002623 by Dan Michaels

flavicon
face
Craig Lee wrote:
>You could also put little RCs at the source of digital signals to
>attenuate the high frequency components that are more likely to
>radiate.  Requires that you know your maximum frequency to select
>the RC. (has this been said already?) ie. LSB changes fastest...

Yeah, good point. Thanks. I was actually thinking about this
after I mentioned yesterday that I have 2 DC control signals
going from the PIC over to the '4052 that select the 4 gains
in the analog section. The lines are bypassed at the '4052,
but a series R would probably help too, to bandlimit possible
hash. [what the heck - adds only a few more components to the
200 already on the board].
================
================

Sean Breheny wrote:
....
>I think you went past the BW consideration too quickly. There is a BIG
>difference between getting  uV resolution in a 3HZ BW and in a 1 MHz BW.
....
>got 0.05 deg/sec noise floor in about a 50Hz BW (amp noise + gyro noise)
>which corresponds to 32uV of noise.
....
>final noise of 0.066 deg/sec for the whole system, which corresponds to 42
>uV. I obtained this on the first try, by just following the manufacturers
>recommendations and using careful layout and limiting BW when I could (I
>think this is key). I have an (AVR ;-) ) micro running on the same board,
>inches away.
....

16-bit, 42 uV, AVR inches away. "first try". Shoot, that's just
too easy!! How can you learn anything when it's so easy? :-).

Yeah, I realize BW = 50 hz vs 1 Mhz is a major factor. Noise voltage
is related as what, sqrt[BW1/BW2]. Yes ?????

So, 2mV * sqrt[50/1,000,000] = 2mV * 0.007 = 14 uV. Sounds good, but
something is still off. My 100 Mhz Hitachi scope has < 1 mV of
noise with the input dead shorted.
===============
===============

Tom Handley wrote:
..........
>   Wagner mentioned reference noise which is a good point that is sometimes
>overlooked. I'm using a MAX6341 4.096V reference with a 1ppm/C Tempco, 0.02%
>initial accuracy, and around 2.5uV of noise. It provides a pin for an
>external cap which augments wide-band noise reduction. With 24 Bits the
>resolution is ~244nV. As is pointed out in one of Analog Devices'
>applications books; 600nV is the Johnson Noise in a 10KHz BW of a 2.2K
>Resistor @ 25C. So, yes, this really is the "Twilight Zone" ;-)
>

The largest resistor in my analog ckt is the 1M input on the 1st
stage, so Johnson noise for it at 1 Mhz = 0.12 mV. Well below my
2 mV measured noise. So, that's not a problem here.

Tom, from your experience, do you think the internal Vref ckts
found in [some of] the A/D converters are as noise-free/stable/etc
as the external Vref ckts you've been using???
==================
==================

Well, all in all, it's sure good to find out I'm the only guy
around here who can't get his stuff to work down into the uV
region. !!!! Hmmm, guess it's just a matter of finding that magic
configuration, afterwards which all is golden. Back to the lab.

Cheers and thanks for all the input,
- Dan Michaels
Oricom Technologies
===================

2000\04\26@010738 by Plunkett, Dennis

flavicon
face
26/4/2000


No this is not the twilite zone at all, infact relality!
You cna keep the 24bit resoultion and still be OK.
Note that we are talking of while noise whos bandwidth is infinite in both
directions and the effective sum of the power is (Zero) think before you
respond.
Ok so what can we do about its, as often we will have this noise flip and
change the state of a signal, very much unwanted!
OK remember that the effective power is Z-E-R-O? but at any one time it may
be at the maximum or at the minimum correct also? So what would happen if
your where to mix in another while noise source of equal maximum and minimum
amplitude swings over a period of time? Would there not be a high chance
that the resultant mixing of these two white noise signals be also or near
zero for most of the time?
Now here is the tricky question, over the same infinte period is the
resultant signal power also zero?
You will find that this is often called bit jittering!


Dennis





> {Original Message removed}

2000\04\26@042044 by Russell McMahon

picon face
Dan,


1.    It's vital that your op-amps have good gain at the common mode
frequency.
ie - even if you are amplifying a dc or low frequency signal, if you have
digital (eg clock, uP data, SPI signals etc) appearing as common mode
signals then your op-amps MUST be able to reject these actively . ie you
should be using amps with a 10MHz plus gain bandwidth and really the more
the better! A low gbw opamp will pass the common mode signals as noise in an
uncontrolled way.

2.    Consider using parts like the AD7730 or the newer AD7731 with internal
gain switching. I haven't used these personally but they are very well
spoken of by friends of mine with much experience in this area.

3.    If I was doing this (and I'm glad I'm not just at present :-)) I would
consider building a minimum component configuration just to see how good I
could make it with the IC used.
eg

- No gain switching 4052
- No potentially noisy adjacent 766x adjacent.
- Power rails sourced from some rather external stable source with copious
local filtering.
- SPI control chip farish away with decoupled well shielded control lines.
- House in ferrous box.(need not be pretty or fancy.*
- etc

ie get rid of everything possible

Once you are sure of how well you can make the simplest possible circuit
perform you can then extend it and see where it starts to degrade.


* For "proof of principle" in suspect environments I have used light "steel"
element covers intended to make range tops look prettier when elements are
not in use. These make a nice flat steel sheet with a turned up edge and 2
can be nested (bottom one upside down) to make an easy open enclosure. Cut
holes, use feed throughs etc as required You do have a complete annular open
ring when these are nested but it is vanishingly small - especially if you
put a weight on top. Aint pretty but it helped me.

     *********************
*    *                              *    *
*    *                              *    *
*    *                              *    *
*****************************





     Russell McMahon
_____________________________

>From other worlds - http://www.easttimor.com
                               http://www.sudan.com

What can one man* do?
Help the hungry at no cost to yourself!
at  http://www.thehungersite.com/

(* - or woman, child or internet enabled intelligent entity :-))


{Original Message removed}

2000\04\26@124824 by Dan Michaels

flavicon
face
Dave Van Horn wrote:
.....
>Check the op-amp PSRR AT the frequency of your main interference.
.....

Dave, it hadn't escaped my attention that you mentioned this the
other day. This may be much of the problem. I am currently using
the TLE072 dual op amp, with gain-BW product = 10 Mhz. It's about
the best thing I could find that looked comparable to the ckt reqs.
I upgraded from the slower LF353 when I started pushing the BW up.

I just re-checked the specs. The CMRR of the TLE072 drops to 50
dB at 1 Mhz, the PSRR+ to 50 dB, and the PSRR- to only 20 dB.

Note, I earlier canned the LF353 because its PSRR- drops way down
to 2 dB at 1 Mhz! Also, I originally went with the inverting
opamp config because of the CMRR dropoff.

This may be the real problem with my board. So what is the
solution?

1. Anyone know of a good dropin DIP-8, dual op amp replacement
  with better PSRR?

2. Would it help to work on the analog -V regulator [79L06] a
  little harder? I already have bypass and filter caps there,
  and on the opamp, plus 50 mil power traces. OTOH, it does
  get its power from the noisy ole 7662, and on the present
  pcb, its fairly close to the PIC. [on the new pcb I had
  already moved it away].

3. Maybe something other than 79L06 would be better.

best regards,
- Dan Michaels
Oricom Technologies
===================

2000\04\26@141013 by Grif\ w. keith griffith

flavicon
face
<x-flowed>I guess the next step is start breadboarding the thing one part at a
time.  Do the front end,,, no digital stuff at all, battery powered,
running clean.  Then start adding stuff.

Big snips,,,


>Grif wrote:
>.........
>Have you tried moving


'Grif'   N7IVS

</x-flowed>

2000\04\26@142424 by Dan Michaels

flavicon
face
Dennis wrote:
>No this is not the twilite zone at all, infact relality!
>You cna keep the 24bit resoultion and still be OK.
>Note that we are talking of while noise whos bandwidth is infinite in both
>directions and the effective sum of the power is (Zero) think before you
>respond.
>Ok so what can we do about its, as often we will have this noise flip and
>change the state of a signal, very much unwanted!
>OK remember that the effective power is Z-E-R-O? but at any one time it may
>be at the maximum or at the minimum correct also? So what would happen if
>your where to mix in another while noise source of equal maximum and minimum
>amplitude swings over a period of time? Would there not be a high chance
>that the resultant mixing of these two white noise signals be also or near
>zero for most of the time?
>Now here is the tricky question, over the same infinte period is the
>resultant signal power also zero?
>You will find that this is often called bit jittering!
>

I think I have the general idea of what you are talking about here.

>From information theory, encryption works because you can take a
non-random bit stream and XOR it with a random stream, and the result
comes out random. Likewise, I have read that you can add a small
amount of random noise to an image that has non-random hash on it,
like streaking, banding, washedout areas, pixellation, etc, and
greatly improve the overall image quality.

So it stands to figure that you can possibly add some random noise
to a waveform that has some non-random interference on it, and
effectively "mask" that interference. I have actually thought
about trying this in the past. Might actually work for small
"nibby" noise, and leave the longer-term waveforms being captured
more or less intact.

Cheers and thanks for the input,
- Dan Michaels
Oricom Technologies
===================

2000\04\26@144723 by Andrew Kunz

flavicon
face
This works quite well.  We use it to get better resolution on communication
signal levels over RF.

Andy










Dan Michaels <EraseMEoricomspam_OUTspamTakeThisOuTLYNX.SNI.NET> on 04/26/2000 02:21:47 PM

Please respond to pic microcontroller discussion list <PICLISTspamspam_OUTMITVMA.MIT.EDU>








To:      @spam@PICLISTKILLspamspamMITVMA.MIT.EDU

cc:      (bcc: Andrew Kunz/TDI_NOTES)



Subject: Re: [OT] [EE] 24-bit A/D. Are We in the Twilite
         Zone Here?








Dennis wrote:
{Quote hidden}

I think I have the general idea of what you are talking about here.

>From information theory, encryption works because you can take a
non-random bit stream and XOR it with a random stream, and the result
comes out random. Likewise, I have read that you can add a small
amount of random noise to an image that has non-random hash on it,
like streaking, banding, washedout areas, pixellation, etc, and
greatly improve the overall image quality.

So it stands to figure that you can possibly add some random noise
to a waveform that has some non-random interference on it, and
effectively "mask" that interference. I have actually thought
about trying this in the past. Might actually work for small
"nibby" noise, and leave the longer-term waveforms being captured
more or less intact.

Cheers and thanks for the input,
- Dan Michaels
Oricom Technologies
===================

2000\04\27@133631 by Dan Michaels

flavicon
face
Grif wrote:
>I guess the next step is start breadboarding the thing one part at a
>time.  Do the front end,,, no digital stuff at all, battery powered,
>running clean.  Then start adding stuff.

Yeah, I do this with my s.w., when I have something I can't fix any
other way. With h.w., it's so much harder. I have done a lot of
testing in the past, mostly on the finished pcbs, removing other
chips, adding caps here and there, etc.

I think this is more a "systems-level" problem and has to be fixed
in place, so I've been going the route via first principles.

That is, I read everything I can find on low-noise analog and
high-speed digital design, and then incorporate those ideas into
the pcb design, etc, and then test and work with that. I think
sometimes the sum is greater than the parts [gee, sounds like
something *you* would say - yes?].
==============

Wagner wrote:
>
>Just a question: THe 2mV readed at the A/D output is a jumping noise or
>a steady noise, that can be confused with a/c or d/c A/D input offset?
>
>Another point is that I never use a 16 bits ADC to read 16 bits
>signals... I just ignore last 3 or 4 bits, so I pay more for a steady
..........

Not an offset issue, I think, rather fuzz that overwhelms the lowest
2 LSBs. And for a 12-bit A/D I was hoping to get better than 9.5 bits.
============

>> The Vref on my LTC1400 A/D is inside the chip,
>> bypass caps outside. Noise measured via A/D binary
>> output, with front-end amp tied hard to gnd.
>
>What about short to ground the LTC1400 inputs? not just the front end
>amp inputs...  You need to isolate and make sure where is the origin of
>the noise.
>

Yeah, I need to go back and do some more testing. Try to isolate
and correlate noise with possible sources - exactly. All the older
pcbs are gone now, and I am finishing a new layout, and so will
do more tests once I get those. Also, see above answer.

Thanks for the suggestions,
- Dan Michaels
================

2000\04\27@133827 by Dan Michaels
flavicon
face
Andy Howard wrote:
.....
>I don't know if you've seen
>http://www.channel1.com/users/analog/noise.html.  It's a useful
>discussion of noise reduction, mostly focussed on PCB layout issues.
....

Thanks for the link. I looked at it, and it's a very good overview
of the issues. I think I've dealt with most of them, based upon
other readings. Other useful links in this area are:

www.national.com/apnotes/apnotes_all_1.html
AN643 - EMI/RFI Board Design

www.fairchildsemi.com/apnotes/ap_notes_all.html
AN-389: Follow PCB Design Guidelines For Lowest CMOS EMI Radiation

2000\04\27@133829 by Dan Michaels

flavicon
face
Andy wrote:
>This works quite well.  We use it to get better resolution on communication
>signal levels over RF.
>

Good, do you add the noise in before or after the A/D [if there
is one]?

Not wanting to spend the next 6 weeks looking at theory, I was
conjecturing - let's say you have some high-frequency non-random
interference getting into your system, and being aliased into
the A/D. Even though aliased, it will still have its same spectrum
folded around a few times, but still "coherent" [???????].

So, if you add the "masking" noise to the A/D output [ie, via s.w.
postprocessing] rather than prior to the A/D input, it should
work as well. Yes/no/maybe????

Best regards,
- Dan Michaels
Oricom Technologies
===================

>From ORIGINAL:
>So it stands to figure that you can possibly add some random noise
>to a waveform that has some non-random interference on it, and
>effectively "mask" that interference. I have actually thought
>about trying this in the past. Might actually work for small
>"nibby" noise, and leave the longer-term waveforms being captured
>more or less intact.
>

2000\04\27@150308 by Dan Michaels

flavicon
face
Russell McMahon wrote:

>1.    It's vital that your op-amps have good gain at the common mode
>frequency.
>ie - even if you are amplifying a dc or low frequency signal, if you have
>digital (eg clock, uP data, SPI signals etc) appearing as common mode
>signals then your op-amps MUST be able to reject these actively . ie you
>should be using amps with a 10MHz plus gain bandwidth and really the more
>the better! A low gbw opamp will pass the common mode signals as noise in an
>uncontrolled way.
>
>2.    Consider using parts like the AD7730 or the newer AD7731 with internal
.....
>

My present opamp, TLE072, has 10 Mhz GBW but only 20 dB PSRR at 1 Mhz,
and I've been looking for a better drop-in replacement. Natl has the
LM61x2 series, going to 75 Mhz GBW, but amazingly, the PSRR on every
one of these goes into the mud (ie, <= 20 dB) right near 1 Mhz <-- this
seems to be a magical number.

I checked the AD7730, $28/ea from Arrow. Ugh.

I'm beginning to think you just can't do it all with a simple opamp
ckt lie I am using. To do better than 2 Mv noise, would need to go
with FET follower front-end feeding onto expensive low-impedance,
differential-mode opamp ckts, like they do it inside the scopes.
================

>3.    If I was doing this (and I'm glad I'm not just at present :-)) I would
>consider building a minimum component configuration just to see how good I
>could make it with the IC used.
....
>ie get rid of everything possible
>
>Once you are sure of how well you can make the simplest possible circuit
>perform you can then extend it and see where it starts to degrade.
>

Yeah, Grif and Wagner suggested this too. I'm going to try building
up one of my new pcbs [when I get one] one circuit at a time, and
test, test. This way it will more or less be "in system", maybe I
can pinpoint the key problem aspect this way.

Thanks and best regards,
- Dan Michaels

2000\04\27@162020 by Scott Dattalo

face
flavicon
face
On Thu, 27 Apr 2000, Dan Michaels wrote:

{Quote hidden}

PSRR - is the Power Supply Rejection Ratio. Right? If you have NO noise on your
power supply then you can live with such a small value. But it's one of those
theory vs. practice things. So the question I have is, can you reduce the noise
on your power supply and (this is equally important) reduce the power supply
source impedance? The former is probably obvious, however the latter is
important because the output stage of your opamp will cause current to flow
through the power supply. If the power supply has a high impedance, this current
will be converted to a voltage and cause the supply to ripple synchronously to
the opamp's output. If there's a lot of gain in the circuit, then you'll get an
oscillator! (Don't ask how I know :).

It's fairly straight forward to calculate the noise voltage that would be
induced on your power supply leads due to the opamp switching. The 'filter
capacitance' along with equivalent series resistance (esr) can be used to
approximate the filter's impedance. If you look at the max current swing in your
output and say that it's a sinusoidal current source driving (or sinking) the
filter then the voltage swing will be approximately:

 V ~ I * sqrt(R^2 + (1/(2*pi*C))^2)

Which says you want zero R and infinite C...

--------

But, even if you did this filtering perfectly, you need to also look at the
other side of the filter. Ground is not always ground. If the ground to which
this filter is referenced is bouncing around like a trampoline, then it will
couple right into the opamp's supplies. Bummer.

Scott

2000\04\28@051905 by Russell McMahon

picon face
Dan,

I've attached some hopefully helpful comments from my colleagues.  who are
expert in this area. As you will read, they are achieving results that would
be very acceptable to you.


     Russell McMahon
_____________________________

>From other worlds - http://www.easttimor.com
                               http://www.sudan.com

What can one man* do?
Help the hungry at no cost to yourself!
at  http://www.thehungersite.com/

(* - or woman, child or internet enabled intelligent entity :-))


____________________________________________________________


Russell,

Further to our recent discussion re high-resolution ADC's, we have for some
time been manufacturing a universal digitiser based around an AD7730
sigma-delta ADC which achieves much better noise performance than your
PIClist colleage appears be observing.

Our digitiser operates as an addressable node on a four-wire (comms + power)
RS-485 bus running at 0.75Mbit/s and has digital I/O capable of switching up
to 400mA.  A single LM2940 5V analog regulator takes the bus power in and
provides +5V out for digital and analog power and transducer excitation.
The on-board 8051-derived microcontroller runs at 24MHz.  In addition to
accommodating mV-level bridge transducers, an on-board single-ended to
differential converter allows bipolar single-ended inputs up to 10V full
scale.  The unit is constucted on a 2-layer PTH PCB (120x60mm) with a
ground-plane occupying most of the component side of the board.

For reasons I won't go into the ADC is run quite fast with a (pk-pk
noise-defined) resolution of about 15 bits and then additional filtering is
done by the microcontroller which improves this to about 18 bits.  The
filtered 24-bit result is updated internally at about 400Hz but is normally
only read over the bus at about 20Hz.  Input bandwidth is around 10Hz under
small-signal conditions, but large-signal transitions are tracked much more
rapidly using the FASTStep feature of the AD7730 and supporting features in
our own digital filtering.

With a good quality 350-ohm loadcell connected in a typical industrial
environment (most of these are used in the dairy industry) and using the
most sensitive range (10mV full scale) we typically achieve an
input-referred resolution (defined by pk-pk noise) of well under 0.1uV.
Note that this resolution is achieved even in the presence of substantial
mechanical vibration (a primary design requirement). Using a 120kg-rated
loadcell we get better than 1 gram usable resolution in the presence of
floor vibration which generates acceleration forces on the loadcell (when
weighing a 25kg bag of milk powder) equivalent to about 2.5% of the loadcell
capacity (about 3kg pk-pk).

The digitised result is impervious to significant local electrical noise, is
unaffected by the use of a cellphone within 1 metre, and shows no
perturbations when the digital outputs switch inductive loads (pneumatics
solenoids).  The entire system (typically comprising 50 nodes) is powered
from a off-line switching supply.

Interestingly, during development, while we did not have any real problem
with noise limiting usable resolution, we did have difficulty approaching
the claimed drift figures for the AD7730 device.  The problem turned out to
be due to high-frequency noise injection into the ADC front-end being
aliased to very low frequencies which were indistinguishable from offset
drift.

The point I am making is that it can be done (though in our case development
took about 4x longer to get it right that we originally budgeted for).  I
don't know why your colleague might be having problems with noise (I haven't
used the LT sigma-delta converters myself), but has he taken the obvious
step of trying a second device to see if it behaves the same.  ESD damage
will often manifest itself as poor noise perfomance in analog devices.  Also
what happens if he shorts the input of the ADC (directly at the pins) ?  Is
the reference noisy and does the input-referred noise level vary if the
reference voltage is varied ?  Is the noise really analog (could the digital
comms with the ADC instead have a high bit-error rate).  What is he using as
a signal source  - could it be noisy ?

Hope my comments help.

Regards,

Ken Mardle

2000\04\28@232829 by David VanHorn

flavicon
face
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Try a different op-amp.

I just got back from a field trip where I discovered that my major noise
source was the op-amp itself rectifying RF. No tracks needed, the chip was
acting as an AM demod, and giving a very nice replication of the
transmitted pulses.

LMC6464 got rid of the problem, rectified pulses now unmeasurably low.
:)

- --
Are you an ISP?  Tired of spam?
http://www.spamwhack.com  A pre-emptive strike against spam!

Where's Dave? http://www.findu.com/cgi-bin/find.cgi?kc6ete-9

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.2 for non-commercial use <http://www.pgp.com>

iQA/AwUBOQpwqIFlGDz1l6VWEQJLOwCfQKUW6E0gaCYxjf0asWAv6KBc9Q4An25h
kCqWyf7Gg8dS0JRkerYne+Rq
=7dyc
-----END PGP SIGNATURE-----

2000\04\28@233711 by David VanHorn

flavicon
face
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


>2. Would it help to work on the analog -V regulator [79L06] a
>   little harder? I already have bypass and filter caps there,
>   and on the opamp, plus 50 mil power traces. OTOH, it does
>   get its power from the noisy ole 7662, and on the present
>   pcb, its fairly close to the PIC. [on the new pcb I had
>   already moved it away].

Couldn't hurt, (see my other note on op-amps as AM demodulators!)
It's probably layout driven, can you probe around with a spectrum analyzer
and see what surfaces? It may be that you can get 3-6dB just by re-routing
some tracks.

>3. Maybe something other than 79L06 would be better.

Hmm.. 79L10 followed by 79L06, with intermediate filtering?
Cheap, brute force, probably effective. Got that much headroom?


- --
Are you an ISP?  Tired of spam?
http://www.spamwhack.com  A pre-emptive strike against spam!

Where's Dave? http://www.findu.com/cgi-bin/find.cgi?kc6ete-9

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.2 for non-commercial use <http://www.pgp.com>

iQA/AwUBOQpyWoFlGDz1l6VWEQK0fgCfXcfOs8apgxW3p2vtSvhV0oyv3oEAoMo8
L8ln6J6MYf/lN4Trh3jhw8c3
=Lu5V
-----END PGP SIGNATURE-----

2000\04\29@133256 by Dan Michaels

flavicon
face
Dave Van Horn wrote:
>
>Try a different op-amp.
>
>I just got back from a field trip where I discovered that my major noise
>source was the op-amp itself rectifying RF. No tracks needed, the chip was
>acting as an AM demod, and giving a very nice replication of the
>transmitted pulses.
>
>LMC6464 got rid of the problem, rectified pulses now unmeasurably low.
.......

Sounds like you could use some front-end RF filtering too. Maybe
ferrite beads in the input signal lines, or inductors in the Vcc
lines, etc.

And I've looked at dozens of op amp specsheets in the past couple
of days. Turns out all of the JFET types have very bad PSRR- at 1
Mhz, only 0 to 20 dB. Even true for Natl LM6152 which is 75 Mhz GBW.

Maxim has nothing usable. Linear Tech has a lot of interesting
new op amps. The LT135x series have enormous slew rates [400-1000
V/us] at nominal GBW [12-70 Mhz], plus nice PSRR [50 dB @ 1 Mhz],
and fairly low Icc ~1-2 mA. Really nice amps.

Unfortunately, these things are very costly [$7, qty 1, Digikey],
and the I[input_bias] is up near 1 uA, so now I need a fancy way
to control DC offset jumps when changing between the 4 amp gains.
What fun!!

Well, in a few days, I'll get an LT1355 and plug it in.

I'm just starting to look at the Burr-Brown amps. No current
catalog here, so have to download datasheets and search thru for
the PSRR, etc, values.

Anybody know a nice opamp with GBW >= 10Mhz, CMRR and PSRR
>= 50 dB at 1 Mhz, low I(input_bias), and under $7? [Ha, see
comment below re Holy Grail].
==================

{Quote hidden}

On my new pcb layout, I've already moved the analog v.regs further
from the digital cktry, improved their hi-side bypassing, and
increased the area of gnd planing around them. I'll see if this
helps when I get in the new pcbs.

And there isn't really too much headroom on the 79L06.
7812 -> 7662 (with -10 -> -11v out under load), and 79L06 requiring
about -9 for regulation. If I go to a 7815 and higher Vin, then
I have even more heat problems dropping Vin down to 5v for the
7805 to digital.

I can't see anything left to do, viz-a-viz layout/etc, short of
going to a totally different P/S design.
============

Ha, this whole thing is beginning to look like the Quest for the
Holy Grail :-). Just call me Percival. I'm trying to improve the
analog system overall ENOB from 9.5 bits, and *at best* will
probably get only 10.5-11 bits. Spend 3 months, countless $$$$,
just to get pittance in improvement. Ahhh, wonderful engineering!!

Best regards, and thanks for the help,
- "Percy"
===================

2000\04\29@140208 by andy howard

flavicon
face
> From: "Dan Michaels" <KILLspamoricomKILLspamspamLYNX.SNI.NET>

> >>3. Maybe something other than 79L06 would be better.
> >Hmm.. 79L10 followed by 79L06, with intermediate filtering?
> >Cheap, brute force, probably effective. Got that much headroom?

> On my new pcb layout, I've already moved the analog v.regs further
> from the digital cktry, improved their hi-side bypassing, and
> increased the area of gnd planing around them. I'll see if this
> helps when I get in the new pcbs.

Ah, speaking of linear regs, I've often found they can exhibit very high
RF frequency instability. All my designs now have decoupling (10nF and a
few tens of pF) *directly* at the pins of the reg. I found that having
any significant length of track between pins and the decoupling didn't
always clear it up.
















.

2000\04\29@145255 by Dan Michaels

flavicon
face
Andy Howard wrote:
>> From: "Dan Michaels" <RemoveMEoricomTakeThisOuTspamLYNX.SNI.NET>
>
>> >>3. Maybe something other than 79L06 would be better.
>> >Hmm.. 79L10 followed by 79L06, with intermediate filtering?
>> >Cheap, brute force, probably effective. Got that much headroom?
>
>> On my new pcb layout, I've already moved the analog v.regs further
>> from the digital cktry, improved their hi-side bypassing, and
>> increased the area of gnd planing around them. I'll see if this
>> helps when I get in the new pcbs.
>
>Ah, speaking of linear regs, I've often found they can exhibit very high
>RF frequency instability. All my designs now have decoupling (10nF and a
>few tens of pF) *directly* at the pins of the reg. I found that having
>any significant length of track between pins and the decoupling didn't
>always clear it up.

Thanks, this is exactly what is on my new pcb layout. Hopefully, this
coupled with finding an opamp with better PSRR will help. [one can
only hope].

Too bad the turnaround on h.w. is so much more trouble than fixing s.w.
Boy, those s.w. guys got it so much easier than they can possibly know.
[oops, did I say that?].

2000\04\29@192206 by David VanHorn

flavicon
face
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


>Sounds like you could use some front-end RF filtering too. Maybe
>ferrite beads in the input signal lines, or inductors in the Vcc
>lines, etc.

BTDT. None of that made a noticable difference without changing the op-amp.
I'm going in again now, and re-evaluating all that in case it makes a
difference, but it was too small to observe before.

>And I've looked at dozens of op amp specsheets in the past couple
>of days. Turns out all of the JFET types have very bad PSRR- at 1
>Mhz, only 0 to 20 dB. Even true for Natl LM6152 which is 75 Mhz GBW.

Use the best you can afford, and then fix as much as possible on the layout.
RC filters specifc to the op-amp will help a lot, they will knock 1 MHz
right down.
Those Murata/Panasonic EMI filters are great down there, and cheap too $0.25.


- --
Are you an ISP?  Tired of spam?
http://www.spamwhack.com  A pre-emptive strike against spam!

Where's Dave? http://www.findu.com/cgi-bin/find.cgi?kc6ete-9

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.2 for non-commercial use <http://www.pgp.com>

iQA/AwUBOQuIQYFlGDz1l6VWEQIkLgCePHS7ak/LuoZn5z0JeHQvd/gmZIAAn07u
P2oVWoMTCIRz0VOK/Ddle/0o
=gmQX
-----END PGP SIGNATURE-----

2000\04\30@084011 by Russell McMahon

picon face
Dan,

Here is the latest response from my friend who has extensive experience
delaing with signals in very real world situations at levels well below what
you are dealing with. I can vouch for his experience and success
commercially. His comment on direct conversion without amplification using a
commercial Sigma Delta chip should be given some consideration.



regards



               Russell McMahon

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


Russell.

I don't mean to be rude but maybe this guy is flogging a dead horse.

Why doesn't he jut get rid of the op-amp front-end and use a sigma-delta
converter chip with the low-level signal conditioning built in.  At well
under US$15 in moderate quantities the AD7730 is very cost-efective when you
consider that it needs nothing in the way of analog support chips and (with
care) is capable of very high performance.

Analog Devices and Crystal Semiconductor (and probably others) have a whole
range of sigma-delta ADC devices with varying cost/performance tradeoffs,
including several which will give usable input-referred resolutions well
under 0.1uV.  I have used the AD7730, AD7714, and AD7715 and recommend them
highly for sensor applications where resolution is more important than high
speed conversion.

I have also used the AD7750 for power metering  - it is by far the easiest
and cheapest way to build an accurate AC power (or more correctly energy)
metering device.  I believe there is now an improved version (AD7751 ?)
which can better handle variable-frequency inputs (the AD7750 only works of
a narrow frequency range).

I'm not trying to sell AD parts here  - merely to point out that they (along
with other manufacturers) have some very useful devices which though not
cheap, may represent cost-effective solutions both in terms of the
performance they provide and in the substantial reduction in design time
they offer (especially important for low-volume niche products).


On the other subject of engineers doing silly (and often dangerous)
hings  - there seems to be a lot of it about and much of it is absolutely
hilarious.  Maybe there is a commercial opportunity here (modern versions of
Victorian morality tales ?)  - how about acting as a publishing editor for a
book of short stories (promoted via snippets posted on the web).

Sounds as though you could come close to filling it by yourself but I'm sure
there would be plenty of willing contributors.  Like the time I nearly
electrocuted myself under Mums' house while investigating a water leak, or
"the home made arc lamp incidents" (yes two separate ones  - melting the
glass in my bedroom window and burning a hole right through by bedding and
foam mattress to the wire-wove supporting it), or exploding insects on the
back lawn using a 1500uF capacitor bank charged to 2kV and a set of old
multimeter test leads, or the Tesla coil incident (or maybe I should let
Chris Paice tell his own story), or the pirate FM transmitter mounted on a
ramp on top of the U of A chemistry building with a remote release as an aid
to preventing it falling into the hands of the radio inspectors (involving
persons who are now respectively head of an F&P division and a senior
engineer at Talon).

Gavin Higgie could tell the story of the EPROM programmer and the 230VAC
supply (and an engineer at DSE whose name I can't recall could tell a
similar story involving literally hundreds of very expensive (and very dead)
NMOS chips in a top-secret digital signal processing system  - being a
military establishment the poor individual was required to wear the results
of his labours on an arm-band for several days).

You could even cast the net wider and allow non-electrical themes.  That
would admit a large number of "bomb" stories (including of course "RDM and
the Headmaster"), allow Ross to relate his story of the dangers of making
smoke bombs by melting ingredients on a gas stove, not to mention the
dangerous combination of boredom and compressed air (as epitomised while I
was working overtime assembling Eveready torches at Fountain Electronics
when I was 15), the
total destruction of the school chem lab fume cupboard (a small 2H + 0
explosion), and how the front 20 rows at the school drama production were
showered with burning newspaper.

Then there is just plain dumb stuff like having to explain to the NZ Forest
Products Ltd. Accomodation Officer a disassembled (and very oily) motorcycle
engine in the lounge of the company house provided for student vacation
employees, or the dye capsules which mysteriously found their way into the
public swimming pool, or the time I helped roll a spare wheel down one of
Wellington's steeper streets (no real harm done but the consequences could
have been horrendous), or the attempt to shoot a large rat with a spear gun
(stupid at the best of times but doubly so considering we were both in a
small tin dinghy at the time), or the time I broke the fish tank and flooded
the doctors waiting room.

There is also the "penguin" incident  - person on holiday at East Cape
sleeping peacefully in back of small van is woken up in pitch dark by cold
wet flapping fishy-smelly alien monster thing (introduced by so-called
"friends")  - penguin survives ordeal relatively intact but person sustains
deep gash on head which bleeds profusely and by light of feeble torches
makes entire scene look like horror movie.

Now there just has to be a market for book of that kind of stuff.

Regards

       Ken Mardle


'[OT] [EE] 24-bit A/D. Are We in the Twilite Zone H'
2000\05\03@003137 by Dan Michaels
flavicon
face
Scott Dattalo wrote:
>
>PSRR - is the Power Supply Rejection Ratio. Right? If you have NO noise on your
>power supply then you can live with such a small value. But it's one of those
>theory vs. practice things. So the question I have is, can you reduce the noise
>on your power supply and (this is equally important) reduce the power supply
>source impedance? The former is probably obvious, however the latter is
>important because the output stage of your opamp will cause current to flow
>through the power supply. If the power supply has a high impedance, this
current
>will be converted to a voltage and cause the supply to ripple synchronously to
>the opamp's output. If there's a lot of gain in the circuit, then you'll get an
>oscillator! (Don't ask how I know :).
>
>It's fairly straight forward to calculate the noise voltage that would be
>induced on your power supply leads due to the opamp switching.
.....

Hi Scott, thanks for your input. Yes, to all of your questions - re
PSRR and P/S source impedance. However, I wouldn't think the opamps
themselves [at least in my case] are responsible for creating much
P/S buss ripple, since they move a relatively small amount of current
- a few milliamps.

In general, putting .1 uF and 10-20 uF caps on v.reg high and low
sides is the usual way to lower P/S impedance. However, you are also
right about the "theory vs practice" thing. Caps are not ideal
devices, they have series R and L. And all leads/traces have
inductance. This is why they always say put the bypass caps as
close as possible to the Vcc/Vss pins.

But no matter how much bypassing you seem to have, and how small
1/(2*PI*F*C) may seem to be, you can always measure tons of hash
on digital busses - at least in my experience. And if the analog
busses pick up some of this noise, because they are on the same
pcb with the digital, run for inches and are inductive, and the
bypass caps aren't ideal, then you need good opAmp PSRR to help
deal with whatever pickup there is. PSRR = 20 dB [as in most std
opAmps at 1 Mhz] is only 10:1 rejection, so a mere 20 mV of buss
noise would translate into 2 mV of opAmp noise.

This is how I see the problem.
=============

>It's fairly straight forward to calculate the noise voltage that would be
>induced on your power supply leads due to the opamp switching. The 'filter
>capacitance' along with equivalent series resistance (esr) can be used to
>approximate the filter's impedance. If you look at the max current swing in
your
{Quote hidden}

As mentioned above, filtering is never perfect, since real caps aren't
perfect, plus all leads have some inductance. And correct, gnd is not
always gnd. So I guess the main solution here is to use wide traces,
physically separate analog and digital subsystems, don't track digital
lines over analog traces and gnds, and return the analog and digital
gnds separately to the power source with only a 1-point connection
between the two.

This last is interesting because some people say the 1-point
connection should be made at the power entry point, while others
say it should be directly under the A/D converter - ie, the place
where analog and digital meet - and that the analog currents should
be conducted thru the digital gnd plane, but via a physically
separate route than the high-frequency/high-current digital signals.
These things seem to be a little difficult to get a handle on.

best regards,
- Dan Michaels
Oricom Technologies
===================

2000\05\03@003146 by Dan Michaels

flavicon
face
Russell McMahon wrote:
.....
>I've attached some hopefully helpful comments from my colleagues.  who are
>expert in this area. As you will read, they are achieving results that would
>be very acceptable to you.
.......
>For reasons I won't go into the ADC is run quite fast with a (pk-pk
>noise-defined) resolution of about 15 bits and then additional filtering is
>done by the microcontroller which improves this to about 18 bits.  The
>filtered 24-bit result is updated internally at about 400Hz but is normally
>only read over the bus at about 20Hz.  Input bandwidth is around 10Hz under
>small-signal conditions, but large-signal transitions are tracked much more
>rapidly using the FASTStep feature of the AD7730 and supporting features in
>our own digital filtering.
>

Russell,

Thanks for going the extra distance and talking to your associate about
my problem. This entire thread has offered some really good advice for
successful application of 24-bit A/Ds. Unfortunately, the AD7730 won't
help me very much, since my system requires 1 Mhz BW, and the BW of
the 24-bit converters is a little too low - 10-50 hz. If I ever do get
into the 24-bit business, I'm going to come back and find this thread.
================
......
>Interestingly, during development, while we did not have any real problem
>with noise limiting usable resolution, we did have difficulty approaching
>the claimed drift figures for the AD7730 device.  The problem turned out to
>be due to high-frequency noise injection into the ADC front-end being
>aliased to very low frequencies which were indistinguishable from offset
>drift.
......

I find this observation pretty interesting. The front-ends of the
successive-approx A/Ds, like I use, tend to be quite wideband - the
LT1400 has a 400 Khz max samplg rate but 4 Mhz BW. So I think
aliasing of hi-freq pickup may be a problem - and I'm trying to
figure out how to deal with this other than just low-passing the
signal in. I don't have a drift problem, like your associate,
but hi-freq noise can still wrap around into the pass-band.

I found an article dealing with some of these issues, and am
perusing it:

"Attack the Noise Gremlins That Plague High-Speed ADCs"
http://www.elecdesign.com/magazine/1999/dec1799/analog/1217ao3.shtml

Best regards and thanks,
- Dan Michaels

2000\05\03@003155 by Dan Michaels

flavicon
face
Dave Van Horn wrote:
.....
>>Sounds like you could use some front-end RF filtering too. Maybe
>>ferrite beads in the input signal lines, or inductors in the Vcc
>>lines, etc.
>
>BTDT. None of that made a noticable difference without changing the op-amp.
>I'm going in again now, and re-evaluating all that in case it makes a
>difference, but it was too small to observe before.

I bet what fixed the problem is that the LMC6464 only has 50 Khz BW.
===============

>>And I've looked at dozens of op amp specsheets in the past couple
>>of days. Turns out all of the JFET types have very bad PSRR- at 1
>>Mhz, only 0 to 20 dB. Even true for Natl LM6152 which is 75 Mhz GBW.
>
>Use the best you can afford, and then fix as much as possible on the layout.
>RC filters specifc to the op-amp will help a lot, they will knock 1 MHz
>right down.
>Those Murata/Panasonic EMI filters are great down there, and cheap too $0.25.
>

Good suggestion, I've got just enough room to add these EMI
filters in the layout. So far, I've heard about 50 suggestions on
this thread, and about all of them are on the pcb now. I'm gonna
poop badly if the noise doesn't get any better after all this
wheel spinning :-).

cheers,
- Dan Michaels

2000\05\03@114746 by Dan Michaels

flavicon
face
At 03:24 AM 4/29/00 -0700,  Tom Handley wrote:
........
>PSBS: I had more to say on the LTC2400 and Dennis' comments about
>`Think before you comment' and "Bit Jittering" which I assume meant
>"Dithering" which I don't think is practical at the sub uV level but
>I've been busy and I'll try to follow up...
>

Tom,

I've been itching to hear your comments on signal "jittering/dithering".

Andy Kunz mentioned in passing that he uses this technique, but did
not elaborate on how it was done or what kind of improvement he saw.

best regards,
- Dan Michaels
==============

2000\05\03@120052 by Andrew Kunz

flavicon
face
I haven't been following this thread, just opened this one on a lark.

Give me your fax and I'll send you a sketch of it later.  Maybe someone will
scan it and post it somewhere then.

Andy









Dan Michaels <spamBeGoneoricomspamBeGonespamLYNX.SNI.NET> on 05/03/2000 11:41:51 AM

Please respond to pic microcontroller discussion list <TakeThisOuTPICLISTEraseMEspamspam_OUTMITVMA.MIT.EDU>








To:      RemoveMEPICLISTspamTakeThisOuTMITVMA.MIT.EDU

cc:      (bcc: Andrew Kunz/TDI_NOTES)



Subject: Re: [OT] [EE] 24-bit A/D. Are We in the Twilite
         Zone Here?








At 03:24 AM 4/29/00 -0700,  Tom Handley wrote:
........
>PSBS: I had more to say on the LTC2400 and Dennis' comments about
>`Think before you comment' and "Bit Jittering" which I assume meant
>"Dithering" which I don't think is practical at the sub uV level but
>I've been busy and I'll try to follow up...
>

Tom,

I've been itching to hear your comments on signal "jittering/dithering".

Andy Kunz mentioned in passing that he uses this technique, but did
not elaborate on how it was done or what kind of improvement he saw.

best regards,
- Dan Michaels
==============

2000\05\03@120241 by Andrew Kunz

flavicon
face
Oh, and it easily adds bits.  I do 256 samples and get 2-3 bits more GOOD data,
usually more (depends on what it's measuring) up to 4 bits.

Andy

2000\05\03@122316 by Alice Campbell

flavicon
face
Dan:
Think of it as the opposite of averaging data. to average,
shift right and get twice the number of samples.  To jitter,
shift left one space and add 1 bit randomly + or - to each
one.  Suddenly duplicate numbers slide off the top of one
another.  Particularly useful for plotting data or making
sounds,the blurry data is actually more usable than the
original.

alice


{Quote hidden}

2000\05\03@123942 by jamesnewton

face picon face
I don't understand "duplicate numbers slide off the top of one another".

I also can't see how this jittering could have any positive effect. I could
understand shifting it one bit left and then average with previous
readings... The random extra bit makes the reading #.5 rather than #.0. Is
that what we are talking about?

---
James Newton (PICList Admin #3)
jamesnewtonEraseMEspam.....piclist.com 1-619-652-0593
PIC/PICList FAQ: http://www.piclist.com or .org

{Original Message removed}

2000\05\03@134114 by David VanHorn

flavicon
face
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


>I bet what fixed the problem is that the LMC6464 only has 50 Khz Bw

Almost sorta.
I had limited the other amp to well below that with external RC, and was
unable to get rid of the problem. Actually, I was unable to affect it
significantly.  The 6464 dosen't exhibit the rectification problem at all.
This is definitely a situation where the opamp is rectifying ambient RF
into noise on my signal.


- --
Are you an ISP?  Tired of spam?
http://www.spamwhack.com  A pre-emptive strike against spam!

Where's Dave? http://www.findu.com/cgi-bin/find.cgi?kc6ete-9

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.2 for non-commercial use <http://www.pgp.com>

iQA/AwUBORB+SIFlGDz1l6VWEQIJfQCguR3NpPb2xEc2EVLPqEae4Wxs0zMAnA1t
JakZzq3aTpmZTSwVTNFyE+aD
=S9Nf
-----END PGP SIGNATURE-----

2000\05\03@155513 by jamesnewton

face picon face
<BLOCKQUOTE AUTHOR="Alice">This refers to GRAPHING the numbers</BLOCKQUOTE>

<BLOCKQUOTE AUTHOR="Andy">Xunknown + Noise = Xreal (more closely,
anyway)</BLOCKQUOTE>

Ok, kids. Which is it? <GRIN>

Specifically, I understand Alice's point but is Andy talking about averaging
data or what? How else can an Xunknown be made more Xreal with noise?

Also, doesn't adding a "third dimension to the graph" have the same effect
as averaging during the time that the Xunknown is at that point? in that it
makes it clear that X was REALLY at this point for a while.

---
James Newton (PICList Admin #3)
EraseMEjamesnewtonspampiclist.com 1-619-652-0593
PIC/PICList FAQ: http://www.piclist.com or .org

{Original Message removed}

2000\05\03@162042 by Scott Dattalo

face
flavicon
face
On Wed, 3 May 2000, James Newton wrote:

> <BLOCKQUOTE AUTHOR="Alice">This refers to GRAPHING the numbers</BLOCKQUOTE>
>
> <BLOCKQUOTE AUTHOR="Andy">Xunknown + Noise = Xreal (more closely,
> anyway)</BLOCKQUOTE>
>
> Ok, kids. Which is it? <GRIN>
>
> Specifically, I understand Alice's point but is Andy talking about averaging
> data or what? How else can an Xunknown be made more Xreal with noise?
>
> Also, doesn't adding a "third dimension to the graph" have the same effect
> as averaging during the time that the Xunknown is at that point? in that it
> makes it clear that X was REALLY at this point for a while.
>

I believe that what they're saying is that at long as your A2D converter has
fairly stable transitions that you can add 'noise' to the signal, then you can
filter the noisy signal obtain a more accurate representation of the 'real'
signal. You can imagine a similar effect without noise. Suppose you have at your
disposal a saw-tooth wave form that varies in amplitude by an amount equal to
say one count of your A/D converter. Now further suppose that your able to
precisely add this to your unknown signal. If you poll the A/D converter while
you apply the triangle wave and measure how long it takes for the A/D converter
to increment, you can calculate how far between the quantized voltages that your
signal is.

This is wonderful in theory, but in practice it ain't so easy. For one, the
signal under test varies. You could add an additional sample and hold circuit in
front of the A/D but why add the cost (and [uncontrolled] noise)? However, if
you add a controlled noise source to the signal you can achieve the same
results as you can with the theoretically perfect saw tooth waveform. The idea
is that the random noise will cause the signal to bounce around and cross
several of the quantized voltage levels. The only thing you need to do is low
pass filter (i.e. average) the consecutive samples to achieve the more accurate
reading.

I can imagine a simple example where the controlled noise varies by 2 A/D
counts. If you measured your signal without the noise you'll get a count of
X. If you add the noise, your signal will vary between X-1, X, and X+1. If more
readings were obtained in the X+1 bin than the X-1 bin, then you can say that
the original 'X' value is between X+0.5 and X+1. This is essentially one more
bit of information. And certainly a longer average over many samples will
improve the estimate. (But you'll still need to take into account your varying
signals...)

Scott

2000\05\03@175025 by Plunkett, Dennis

flavicon
face
In this case your attempting to average out the LSB. Note that this is only
an advantage as the noise is nominaly equivalent to the LSB. While on this
matter we must also consider the noise source. In an ADC the noise generated
is ((1/root 12) * the LSB) (Noise typically has peeks 6 to 8 times the
average!) so this is the base noise floor of the ADC (General term). From
this we can see if the noise on the signal is greater than or less than
(Remember that noise is adative not subtractive)
Back to the original LSB stuff.
What we are attempting to do is to find the LSB value. Note that the ADC
qunatizes the sample so that it is to the next lowest LSB value. Thus if tou
are .1LSB lower then the result will be 0.1LSB less, however normal rounding
would consider this as 1LSB more!



Dennis




> {Original Message removed}

2000\05\03@201940 by Brandon, Tom

flavicon
picon face
>From what I've read dithering is most appropriate on slow moving signals.
Imagine you have a 4bit ADC with a slow moving analog input of 1-16units.
The inputs at say 8.4, you sample it, and get 8. Sample again, 8 again. You
sample 50 times, nothing but 8's, all you know is it's between 8 and 9. Now,
imagine there's noise (say 3LSB p-p (I believe this is somewhere around
1\2LSB noise power)) while you take your 50 measurements. Over those 50
measurements, with the possible 3 LSB's noise you now get values ranging
from say 5 to say 11. Average it out and you should (if the noise is
"random") get about 8.4. Et voila, increased accuracy. If you actually add
known noise you can get even better rsults by then subtracting the added
noise. But, why add noise when you couldn't get rid of it if you tried.

Tom.

{Original Message removed}

2000\05\04@015521 by Tom Handley

picon face
  Dan, I see Scott and Tom have explained dithering in detail. My point
was, using the LTC2400 at full resolution (24-Bits), I don't see where
you can get a good controlled noise source. I think the confusion here
is that I'm refering to a specific device which also has 4 sub-LSBs
that you can use for averaging. Sorry for the confusion. I'm guilty
of moving your original question to LTC2400-specific info...

  - Tom

At 09:41 AM 5/3/00 -0600, Dan Michaelsyou wrote:
{Quote hidden}

------------------------------------------------------------------------
Tom Handley
New Age Communications
Since '75 before "New Age" and no one around here is waiting for UFOs ;-)

2000\05\04@061621 by Kbek Tony

flavicon
face
Hi,
I'm intrigued by this discussion, and I'm probably
in over my head here :-)
Anyway regarding bit jittering, increasing accuracy etc.
If I understand correctly if one adds a random bit to the
reading then one possibly ( likely ? ) could increase the accuracy ?
Is this vaild for, let's say one has 6 bit's of noice ?
Would one then add several random bit's ?
Ore are this only valid for the LSB ?

Further how does one accomplish this 'random bit' ?
to be truly random I guess it would be very hard,
I've read the last month disscussion regarding
an random byte generation, but for a single bit
there must be an easier way.

Anyway very interesting thread, nice to read.

BTW Just 'playing' with the 24 bit AD7730 ;-) DS


/Tony



Tony KŸbek, Flintab AB            
ÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓ
E-mail: RemoveMEtony.kubekEraseMEspamEraseMEflintab.com
ÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓ

2000\05\04@124420 by jamesnewton

face picon face
Correct me if I'm wrong, but the noise for the system that Scott is talking
about has to be added BEFORE the signal is "quantized" from analog to
digital. After, the extra fractional information is already lost. So
shifting the A2D result one bit left and adding a random bit would not
improve your resolution.

Again, the only reason I can see for shifting one bit left and adding a
random bit would be A) in combination with averaging and B) to make the *2
into a *2 + 0.5. And I really don't see the need for B. As an example: lets
say you have an XUnknown (XU) of 8.5 and are adding random noise (RN)
between 0 to 1. Your A2D only measures units of 1 (IN). This gets shifted
left one (*2) and added to the average (+A) which is then shifted left one
(/2) and, if you had floating point divide, you would be able to read this
as having an extra bit of precision past the actual input value. (OUT)

 ANALOG  |         DIGITAL
XU      RN      IN      *2      "+A     /2      OUT
                                       0
8.5     0.5     9       18      18      9.00    4.5
8.5     1       9       18      27      13.00   6.5
8.5     0.5     9       18      31      15.00   7.5
8.5     0       8       16      31      15.00   7.5
8.5     0.5     9       18      33      16.00   8
8.5     1       9       18      34      17.00   8.5
8.5     0.5     9       18      35      17.00   8.5
8.5     0       8       16      33      16.00   8
8.5     0.5     9       18      34      17.00   8.5
8.5     1       9       18      35      17.00   8.5
8.5     0.5     9       18      35      17.00   8.5
9       0       9       18      35      17.00   8.5
9       0.5     9       18      35      17.00   8.5
9       1       10      20      37      18.00   9
9       0.5     9       18      36      18.00   9
9       0       9       18      36      18.00   9
9       0.5     9       18      36      18.00   9
9       1       10      20      38      19.00   9.5
9       0.5     9       18      37      18.00   9
9       0       9       18      36      18.00   9
9       0.5     9       18      36      18.00   9
9       1       10      20      38      19.00   9.5
4.7     0.5     5       10      29      14.00   7
4.7     0       4       8       22      11.00   5.5
4.7     0.5     5       10      21      10.00   5
4.7     1       5       10      20      10.00   5
4.7     0.5     5       10      20      10.00   5
4.7     0       4       8       18      9.00    4.5
4.7     0.5     5       10      19      9.00    4.5
4.7     1       5       10      19      9.00    4.5
4.7     0.5     5       10      19      9.00    4.5
4.7     0       4       8       17      8.00    4
4.7     0.5     5       10      18      9.00    4.5


Let me know if anyone wants the Excel spreadsheet used to make that.

Note that the averaging is absolutely necessary and has the effect of
reducing the maximum frequency that can accurately be sampled by a factor of
4 or so. You can get around that (at the cost of momentary loss of some
accuracy) by tossing the average (over writing it) when the input changes by
more than 1 from its previous value. Better yet, compute the delta between
the last reading and this, and divide the average by it, multiply the new
value by it before averaging so that small changes have no effect but large
changes quickly overcome the average.

Now, if the random noise was between -0.5 and +0.5 and you wanted an exact
result rather than just a more accurate proportional result, you would need
to add .5 in somewhere either by adding a 1 after the *2 every other time or
by just adding .5 to the OUT.

But there just isn't any reason to add digital noise to an a2d result (other
than graphing).

Right?

---
James Newton (PICList Admin #3)
RemoveMEjamesnewtonspam_OUTspamKILLspampiclist.com 1-619-652-0593
PIC/PICList FAQ: http://www.piclist.com or .org

{Original Message removed}

2000\05\04@131045 by Andrew Kunz

flavicon
face
James,

What we would do is measure an analog signal using the time-honored cap/res
delay time with a PIC.  Our time was typically 10-50, where our signal was
acquired (sample-and-hold setup).

Then we would add a random number to this in the range 0-15.  The random number
could come from any of a number of places; ours was the CRC of the last message
we received (we were continuously receiving messages).

These results were then continuously averaged (Avg = ((sum of samples) - Avg) +
newsample).

The result was surprisingly accurate.

Andy

PS.  To whomever I was going to fax yesterday, never mind.  The above is what I
really wanted to show you.

2000\05\04@133326 by Scott Dattalo

face
flavicon
face
On Thu, 4 May 2000, Andrew Kunz wrote:

> James,
>
> What we would do is measure an analog signal using the time-honored cap/res
> delay time with a PIC.  Our time was typically 10-50, where our signal was
> acquired (sample-and-hold setup).
>
> Then we would add a random number to this in the range 0-15.  The random number
> could come from any of a number of places; ours was the CRC of the last message
> we received (we were continuously receiving messages).
>
> These results were then continuously averaged (Avg = ((sum of samples) - Avg) +
> newsample).
>
> The result was surprisingly accurate.

James, now I know why you were confused. I have to admit that after Andy's
explanation I'm confused as well. First a couple of questions which Andy may
wish to answer. When you say 'time was typically 10-50', what does that
mean? 10-50 milliseconds, 10-50 samples per second? Are you sure that the
formula is correct? (There's a scaling factor missing.)

I suspect if you go back and repeat your experiments with the "digital random
number" stuff removed that you'll get the same results - perhaps even improved
results. From the information you've provided, I can confidently assert that
adding this digital noise is of no beneficial value whatsoever. Instead, the
averaging algorithm removes the random variations in the REAL analog signal
along with the digital variations that were artificially added. Theory aside,
let's look at it practically. If your A/D converter is absolutely solid and
your signal contains no noise, then you'll get the same value everytime you read
the A/D (assuming a DC input of course). The information from the A/D is
constant. There's nothing you can do to increase the information content. If you
add a random value that has a mean of zero, this is not going to affect the A/D
result.


Scott

2000\05\04@135750 by Andrew Kunz

flavicon
face
We were measuring an RF signal with a high noise component (0-15 was reasonable,
ie, 160% noise on 100% signal!)

The 10-50 was counts; I don't have the formula which converted counts to dBmV.

It worked much better than w/o the noise correction.

Andy









Scott Dattalo <RemoveMEscottTakeThisOuTspamspamDATTALO.COM> on 05/04/2000 01:32:11 PM

Please respond to pic microcontroller discussion list <EraseMEPICLISTspamspamspamBeGoneMITVMA.MIT.EDU>








To:      RemoveMEPICLISTKILLspamspamMITVMA.MIT.EDU

cc:      (bcc: Andrew Kunz/TDI_NOTES)



Subject: Re: [OT] [EE] 24-bit A/D. Are We in the Twilite
         Zone Here?








On Thu, 4 May 2000, Andrew Kunz wrote:

> James,
>
> What we would do is measure an analog signal using the time-honored cap/res
> delay time with a PIC.  Our time was typically 10-50, where our signal was
> acquired (sample-and-hold setup).
>
> Then we would add a random number to this in the range 0-15.  The random
number
> could come from any of a number of places; ours was the CRC of the last
message
> we received (we were continuously receiving messages).
>
> These results were then continuously averaged (Avg = ((sum of samples) - Avg)
+
> newsample).
>
> The result was surprisingly accurate.

James, now I know why you were confused. I have to admit that after Andy's
explanation I'm confused as well. First a couple of questions which Andy may
wish to answer. When you say 'time was typically 10-50', what does that
mean? 10-50 milliseconds, 10-50 samples per second? Are you sure that the
formula is correct? (There's a scaling factor missing.)

I suspect if you go back and repeat your experiments with the "digital random
number" stuff removed that you'll get the same results - perhaps even improved
results. From the information you've provided, I can confidently assert that
adding this digital noise is of no beneficial value whatsoever. Instead, the
averaging algorithm removes the random variations in the REAL analog signal
along with the digital variations that were artificially added. Theory aside,
let's look at it practically. If your A/D converter is absolutely solid and
your signal contains no noise, then you'll get the same value everytime you read
the A/D (assuming a DC input of course). The information from the A/D is
constant. There's nothing you can do to increase the information content. If you
add a random value that has a mean of zero, this is not going to affect the A/D
result.


Scott

2000\05\04@141852 by Andrew Warren

face
flavicon
face
Andrew Kunz <PICLISTSTOPspamspamspam_OUTMITVMA.MIT.EDU> wrote:

> What we would do is measure an analog signal using the
> time-honored cap/res delay time with a PIC. .... Then we would add a
> random number to this in the range 0-15.
>
> These results were then continuously averaged (Avg = ((sum of
> samples) - Avg) + newsample).
>
> The result was surprisingly accurate.

Andy:

I haven't read any of the previous messages in this thread, so I may
be WAY off-base here... But adding random noise AFTER the A/D
conversion (i.e., adding DIGITAL noise) doesn't give you any more
information than you'd have if you just left the signal alone.  If
you're trying to increase the resolution of your conversion by
dithering, you need to do that BEFORE the conversion, by adding
ANALOG noise.

As I said, I'm not sure what the exact point of the discussion is, so
please ignore this if you were talking about something other than
dithering the signal to achieve higher resolution.

-Andy


=== Andrew Warren - spamBeGonefastfwdSTOPspamspamEraseMEix.netcom.com
=== Fast Forward Engineering - San Diego, California
=== http://www.geocities.com/SiliconValley/2499

2000\05\04@143327 by Andrew Kunz

flavicon
face
I had nothing to do with the algorithm other than coding it.

There may have been something about the designer's concept that isn't here,
mainly because I didn't need to understand the concept.

Andy

2000\05\04@150723 by Andrew Warren

face
flavicon
face
Andrew Kunz <KILLspamPICLISTspamBeGonespamMITVMA.MIT.EDU> wrote:

> I had nothing to do with the algorithm other than coding it.
>
> There may have been something about the designer's concept that isn't
> here, mainly because I didn't need to understand the concept.

Hmm... Maybe adding the digital noise made that pseudo-average
algorithm converge faster?  I'd have to think about it for a bit;
maybe Scott Dattalo can answer more quickly.

-Andy


=== Andrew Warren - EraseMEfastfwdspamEraseMEix.netcom.com
=== Fast Forward Engineering - San Diego, California
=== http://www.geocities.com/SiliconValley/2499

2000\05\04@203718 by Brandon, Tom

flavicon
picon face
In regards to adding more noise, remember, you're not actually adding any
accuracy (repeatability), in fact you're decreasing the accuracy of the
individual readings as succesive readings of the same value (without noise)
will have more variation. What you are doing is adding precision. The
averaged result is now closer to the 'true' value.

However, I think this would only apply with analog noise. You need
fractional noise or it falls apart. Adding 1LSB of digital noise would just
give you 1\3rd correct values, 1\3rd up 1 LSB and 1\3rd down 1LSB (assuming
true Gaussian noise). Average that out and it just returns the correct
value.

Yes, averaging the values would reduce the noise. If it's natural noise then
it's Gaussian, thus it should average out to 0 over multiple samples. But
the digital noise is irrelevant. Assuming the digital noise is also Gaussian
it too will be averaged out, but it will not increase accuracy or precision
(but as suggested it may speed the averaging).

In terms of how you 'add' the random noise, you don't. Ever designed a
(mixed signal) circuit with 0 noise? You don't add analog noise, you just
use the existing noise (assuming it's Gaussian, which almost all naturally
occuring noise is). You could digitally produce the noise and then subtract
it out for even better accuracy but this is rarely used as it isn't
justified. After spending 6months trying to eliminate the #$%$ing analog
noise, it's nice to be able to put it to good use.

In terms of digitally producing Gaussian noise, you just apply the Central
Limits Theroem. It states that:
A sum of random numbers becomes normally distributed as more and more of the
random numbers are added together. The Central Limit Theorem does not
require the individual random numbers be from any particular distribution,
or even that the random numbers be from the same distribution.

This is why Gaussian noise is so common. Whenever multiple random processes
(any distribution including pseudorandom computer generated numbers)
interact, the overall output will show Gaussian distribution.

Tom.

-----Original Messages-----

>From Andy:
<SNIP>
It worked much better than w/o the noise correction.

Andy

>From Scott:
<SNIP>
I suspect if you go back and repeat your experiments with the "digital
random
number" stuff removed that you'll get the same results - perhaps even
improved
results.
<SNIP>

Scott

>From KŸbek Tony:

Hi,
I'm intrigued by this discussion, and I'm probably
in over my head here :-)
Anyway regarding bit jittering, increasing accuracy etc.
If I understand correctly if one adds a random bit to the
reading then one possibly ( likely ? ) could increase the accuracy ?
Is this vaild for, let's say one has 6 bit's of noice ?
Would one then add several random bit's ?
Ore are this only valid for the LSB ?

Further how does one accomplish this 'random bit' ?
to be truly random I guess it would be very hard,
I've read the last month disscussion regarding
an random byte generation, but for a single bit
there must be an easier way.

Anyway very interesting thread, nice to read.

BTW Just 'playing' with the 24 bit AD7730 ;-) DS


/Tony



Tony KŸbek, Flintab AB            
ÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓ
E-mail: @spam@tony.kubek@spam@spamspam_OUTflintab.com
ÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓÓ

2000\05\04@211346 by Scott Dattalo

face
flavicon
face
On Thu, 4 May 2000, Andrew Warren wrote:

> Andrew Kunz <spamBeGonePICLISTspamKILLspamMITVMA.MIT.EDU> wrote:
>
> > I had nothing to do with the algorithm other than coding it.
> >
> > There may have been something about the designer's concept that isn't
> > here, mainly because I didn't need to understand the concept.
>
> Hmm... Maybe adding the digital noise made that pseudo-average
> algorithm converge faster?  I'd have to think about it for a bit;
> maybe Scott Dattalo can answer more quickly.

I can't see how it'd possibly converge more quickly. If you view the average
operation as a low pass filter and assume that the noise is 'white' (that is, it
has energy at all frequencies) then there will be noise present right up to the
filter's cutoff. If you had a sine wave with a frequency close to the filter's
cutoff, you'd notice that the filter would let some of it through. I think the
same would be true here as well. In other words, I think it would take LONGER to
settle. Now the noise is not truely white since its DC component is absent
(presumably - otherwise you're just adding an error to the signal). The reason
this scheme works in the analog domain so well is that 1) the noise is (or
should be) absent of a any DC component 2) the low pass filter bandwidth is so
low that the amount of energy in the noise over that frequency band is
negligible.

>From this observation it's clear to conclude that adding random noise only works
well if the averaging filter or low pass filter has a very low frequency
cutoff. If you were to use this technique to acquire higher bandwidth data like
acoustical data, this technique would reduce the digitization accuracy. However,
if noise is added in a frequency band beyond the frequency at which your signal
resides, you may again apply these concepts. In this case it will become
necessary to over-sample the data, low-pass filter it, and then decimate it to
the sample rate you would have sampled without using this technique.

Perhaps a simpler approach would be to add a known error signal to your analog
signal. This signal could be sine wave or triangle wave with frequency just
outside the frequency of the signal of interest. Satisfy Nyquist for this known
analog signal (by sampling at 2 or 3 times the frequency of the sine or triangle
wave), then digitally subtract it out, and low pass filter the result. The idea
is that +1 -1 = 0, only the addition is analog and the subtraction is digital.
Now, the digitized sine wave subtracted from the sampled signal needs to have
more resolution than just one bit of your A/D converter other wise you wouldn't
benefit from the dithering.

Caveats - of course I've never tried this, but in theory...

Scott

2000\05\05@071308 by Peter L. Peres

picon face
Hi,

wrt. Andy Kunz's adding digital noise: I think that I know what is on.

He said that he has a very poor S/N at the input. I infer that he tried to
obtain a steady readout of the average RF level measured while receiving
modulated (AM ? - he does not say) rf signals. For this, he needed to
smooth the input somehow, to obtain something resembling RMS. By adding
'fast' noise I think that he faked oversampling, and then averaged the
result. This would be a median filter, done otherwise, imho. He also said,
that 'the result was surprisingly accurate' which means that he was
measuring the same signal also using other means, which means for RF are
either a specan or a rf milivolt/miliwattmeter. All three devices read out
RMS integrated over some time unless set up otherwise (i.e. peak or
instant power etc).

am I way off ?

Peter

2000\05\05@073422 by Andrew Kunz

flavicon
face
WOW, now that makes sense.

The signal was FM/FSK on a TV cable, with our carrier too close to Channel 2.

The "standard" was a SAM-II meter which measures signal level in dBmV.  Standard
equipment for cable TV.

The goal was averaged signal level, not instantaneous.

Glad somebody knew what I was talking about - I sure don't!

Andy









"Peter L. Peres" <.....plpspam_OUTspamACTCOM.CO.IL> on 05/04/2000 06:46:38 PM

Please respond to pic microcontroller discussion list <TakeThisOuTPICLIST.....spamTakeThisOuTMITVMA.MIT.EDU>








To:      TakeThisOuTPICLISTKILLspamspamspamMITVMA.MIT.EDU

cc:      (bcc: Andrew Kunz/TDI_NOTES)



Subject: Re: [OT] [EE] 24-bit A/D. Are We in the Twilite
         Zone Here?








Hi,

wrt. Andy Kunz's adding digital noise: I think that I know what is on.

He said that he has a very poor S/N at the input. I infer that he tried to
obtain a steady readout of the average RF level measured while receiving
modulated (AM ? - he does not say) rf signals. For this, he needed to
smooth the input somehow, to obtain something resembling RMS. By adding
'fast' noise I think that he faked oversampling, and then averaged the
result. This would be a median filter, done otherwise, imho. He also said,
that 'the result was surprisingly accurate' which means that he was
measuring the same signal also using other means, which means for RF are
either a specan or a rf milivolt/miliwattmeter. All three devices read out
RMS integrated over some time unless set up otherwise (i.e. peak or
instant power etc).

am I way off ?

Peter

2000\05\05@103327 by Robert A. LaBudde

flavicon
face
<x-flowed>I may be naive, but:

1. How do you generate a 24-bit voltage reference for a 24-bit ADC?

2. How do you know whether or not you have obtained 24-bits of accuracy?

================================================================
Robert A. LaBudde, PhD, PAS, Dpl. ACAFS  e-mail: .....ralspamRemoveMElcfltd.com
Least Cost Formulations, Ltd.                   URL: http://lcfltd.com/
824 Timberlake Drive                            Tel: 757-467-0954
Virginia Beach, VA 23464-3239                   Fax: 757-467-2947

"Vere scire est per causas scire"
================================================================

</x-flowed>

More... (looser matching)
- Last day of these posts
- In 2000 , 2001 only
- Today
- New search...