Searching \ for '[EE]: Filter Capacitors' in subject line. ()
Help us get a faster server
FAQ page: www.piclist.com/techref/logic/dsps.htm?key=filter
Search entire site for: 'Filter Capacitors'.

Exact match. Not showing close matches.
'[EE]: Filter Capacitors'
2000\06\06@090115 by

In many products, I see a large electrolytic capacitor in parallel with a
small disc or dipped capacitor.  Why is that done?  Obviously, the pair
does not act as a single large capacitor, or aonly a single one would be
used.  Just how does that arrangement work?

Mark P

hi Paul
The same way as multiple capacitors on a circuit board. Interference can be
from one chip to the next via the supply, so you put a capacitor across the
supply pins of each chip as close to the chip as possible.
The large capacitor can remove some ripple from the supply while the smaller
capacitor deals with spikes the value and where placed play apart in this.
As does the path of tracks around the board.
Regards Art

----- Original Message -----
From: Mark Peterson <markpCANNONTECH.COM>
To: <PICLISTMITVMA.MIT.EDU>
Sent: Tuesday, June 06, 2000 5:58 AM
Subject: [EE]: Filter Capacitors

> In many products, I see a large electrolytic capacitor in parallel with a
> small disc or dipped capacitor.  Why is that done?  Obviously, the pair
> does not act as a single large capacitor, or aonly a single one would be
> used.  Just how does that arrangement work?
>
> Mark P

Simply, Low frequency noise or fluctuations on the power supplies are
smoothed over as the LARGE capacitor sources extra current into the circuit
or sinks extra current keeping the supply to the ciruit more or less
constant.
It acts as a big reservoir of power.

High frequency noise is filtered out by adding a smaller capacitor which
dumps the excess noise onto the ground (0V)

There's probably some equation for it but I basically always put a 100uF cap
on for the low freq's and a 100nF for the high frequecies. It's worked on
all my projects so far.

Pete

> {Original Message removed}
The smaller cap generally has different characteristics than the large.  For
instance, you will rarely (if ever) see two ceramic caps in parallel, or two
electrolytics in parallel (unless the designer wanted a weird value, or space
considerations dictated such a design) for electrical reasons.  What you may see
is an electrolytic (higher internal resistance, can't react very fast to
small/short transients) in parallel with a ceramic (low capacitance, low
resistance, can't handle large/long transients).

Mark Peterson wrote:
>
> In many products, I see a large electrolytic capacitor in parallel with a
> small disc or dipped capacitor.  Why is that done?  Obviously, the pair
> does not act as a single large capacitor, or aonly a single one would be
> used.  Just how does that arrangement work?
>
> Mark P

Hi,

Put simply: The large electrolytic capacitor acts as a short for low frequencies.
and the small disc or dipped capacitor acts as a short for high frequencies.

In other words the Impedance/Frequency  characteristics of the two
types of caps are vastly different.

MikeW

>>> markpCANNONTECH.COM 06/06/00 02:58PM >>>
In many products, I see a large electrolytic capacitor in parallel with a
small disc or dipped capacitor.  Why is that done?  Obviously, the pair
does not act as a single large capacitor, or aonly a single one would be
used.  Just how does that arrangement work?

Mark P

At 07:58 AM 6/6/00 -0500, you wrote:
>In many products, I see a large electrolytic capacitor in parallel with a
>small disc or dipped capacitor.  Why is that done?  Obviously, the pair
>does not act as a single large capacitor, or aonly a single one would be
>used.  Just how does that arrangement work?
>
>Mark P

It does act as a single large capacitor. Have you ever (deliberately or
inadvertently) ;-)  taken an electrolytic capacitor apart? The plates are
coiled together. This construction has a significant amount of
inductance, which is like a series inductor (and there is a series
resistance as well) in series with an ideal capacitor.

So, at low frequencies the added impedance of the series inductance is
negligible, the e-cap acts pretty much like an ideal capacitor (
just some series resistance). At very high frequencies
the impedance of the e-cap may start to rise significantly because of the
inductance, and the ceramic capacitor shunting it may begin to dominate the
parallel impedance. Some e-caps are better than others, this is not
always necessary.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Spehro Pefhany --"it's the network..."            "The Journey is the reward"
speffinterlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com
Contributions invited->The AVR-gcc FAQ is at: http://www.bluecollarlinux.com
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

Large capacitors are to 'slow' to handle fast transients.
So often you see a small (10uF) next to a large one.
Just filtering spices and all other high freqentcy crap out of
the power supply that the large one can't handle.

Ries

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

At 07:58 AM 6/6/00 -0500, Mark Peterson wrote:
>In many products, I see a large electrolytic capacitor in parallel with a
>small disc or dipped capacitor.  Why is that done?  Obviously, the pair
>does not act as a single large capacitor, or aonly a single one would be
>used.  Just how does that arrangement work?

They are telling you that they expect to draw large current pulses
(probably driving something connected to the I/O pins) in addition to the
switching currents for the uP.
The large cap is good at providing a lot of current over a relatively long
time, but it's internal impedance is too high to provide a lot of current
over very short time intervals, so they add the small cap.

Warning: Bypassing is very often done wrong, so don't take what you see as
gospel.

- --
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/AwUBOT04AYFlGDz1l6VWEQK1lgCgg+ZksO9eWwz1Is6Sny26u1H/Q2cAnjaN
LA7FBC6MD7pOGedUP1n+Wh7a
=gxlK
-----END PGP SIGNATURE-----

The equivalent series resistance and inductance on the electrolytics are
fairly high, making them not do much for high frequencies. The ceramic
capacitor presents a low impedance for high frequencies while the
electrolytic presents a low impedance for low frequencies.

Harold

On Tue, 6 Jun 2000 07:58:32 -0500 Mark Peterson <markpCANNONTECH.COM>
writes:
> In many products, I see a large electrolytic capacitor in parallel
> with a
> small disc or dipped capacitor.  Why is that done?  Obviously, the
> pair
> does not act as a single large capacitor, or aonly a single one
> would be
> used.  Just how does that arrangement work?
>
> Mark P

FCC Rules Online at http://hallikainen.com/FccRules
Lighting control for theatre and television at http://www.dovesystems.com

________________________________________________________________
YOU'RE PAYING TOO MUCH FOR THE INTERNET!
Juno now offers FREE Internet Access!
Try it today - there's no risk!  For your FREE software, visit:
dl.http://www.juno.com/get/tagj.

Dave,

> Warning: Bypassing is very often done wrong, so don't take what you see as
> gospel.

Can you describe what you see very often that's wrong?

-- Mitch

What he is trying to say about a rule of thumb. Common practice is a few
extra capacitor does not hurt.
Most of the cases filtering is over done and try to compensate for bad
layout, grounding systems....
Nice example is a high-end audio preamplifier, takes about 50 ma current
from the power supply, which was filtered by 22000 uF capacitors....
Peter

-----Original Message-----
From: Mitchell D. Miller [mdmiller2HOME.COM]
Sent: Tuesday, June 06, 2000 10:30 AM
To: PICLISTMITVMA.MIT.EDU
Subject: Re: [EE]: Filter Capacitors

Dave,

> Warning: Bypassing is very often done wrong, so don't take what you see as
> gospel.

Can you describe what you see very often that's wrong?

-- Mitch

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

>Can you describe what you see very often that's wrong?

Well...

Bypasses should be located at the chip's ground pin.
Power should be routed to the cap, and then to the chip from the cap.
Anything else compromises the bypass.

Sockets with caps in them are marginally better than nothing at all (see
above), but it's even worse, because you've provided a path past the cap on
the ground side as well.

It's hard to show this without pictures, but a cap with tracks or leads out
to power and ground tracks, especially thick ones, is much less than
optimum.  You really want the current to be forced to walk across that cap.
as ground, and the other two treated as "in" and "out".

In many systems, you'll see a pair of tracks between rows of chips, with
caps across the tracks, and one track supplying VCC to row 2 while the
other supplies gnd to row 1.
While this system looks nice at first glance, it really is very noisy.
(look at where the current flows when a given chip needs a pulse of current)

The common practice of putting a 0.1uF on everything is probably better
than not bypassing anything, but it's far from ideal. The size of bypass
caps is related to what frequencies you want them to be low impedance for.
You need to tailor the bypassing for each "consumer".

Adding large value tantalums without a specific need is mostly a waste of
money.

Using fat power tracks can be bad.  Narrow tracks are higher impedance to
high frequencies, but are the same DC resistance (close anyway).  I use a
wide track from the chip to it's bypass cap(s) and a narrow track from
there back to system power. This is because in the high frequency domain, I
want the current to come from the bypass cap, not from the system buss.

The ground return is just as important as the power. You need to make sure
that the chip not only has a low Z ground, that ground needs to get
directly back to where VCC came from.  If for example, you have a chip
bypassed perfectly, but it has a long path to get back to the sourcing cap
at the regulator, you will have a noisy system.  The noise current will
flow around the system, and find it's way back to the reg cap eventually.
If you have a system that is noisy, and it gets worse with cables plugged
in, or with long leads attached to "ground" pads, this is usually why.  If
you have a chip driven by fast signals from another chip, then there needs
to be a low Z ground between them as well. It becomes clear when you think
of cmos inputs as capacitors (which they are) driven by switches on the
other end of the line. That charging current coming out of the destination
chip's ground has got to return to the source chip somehow!  Wouldn't you
rather pick the path?

Microprocessor crystal caps should be connected by an isolated track to the
uP's ground pin. Not to anywhere else.  The current in that oscillator can
make a nice transmitter, especially if it's coupled into a nice low
impedance ground track an inch or so away.  I've seen systems fail part 15
for that alone, when the ground track happened to be a resonant length on
some harmonic of the xtal.

I may have missed a couple, but that's a pretty good "rouge's gallery" of
bypassing mistakes.

- --
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/AwUBOT1Wp4FlGDz1l6VWEQLyKACfY6Mf425OVWnmSL/6Eqb2yvyC35AAniIk
bBoENVQjXHVvu4uJsC+0+any
=jx5j
-----END PGP SIGNATURE-----

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

At 10:53 AM 6/6/00 -0700, Peter Schultz wrote:
>What he is trying to say about a rule of thumb. Common practice is a few
>extra capacitor does not hurt.
>Most of the cases filtering is over done and try to compensate for bad
>layout, grounding systems....

Extra caps will not save a bad layout. It's like bigger engines on the
titanic.
Occasionally, accidentally, you may hit something that makes the problem
look better by shifting it somewhere else.

>Nice example is a high-end audio preamplifier, takes about 50 ma current
>from the power supply, which was filtered by 22000 uF capacitors....

If these large caps were perfect, then this would be a good idea.
It would be very interesting to sweep them, and see where they are really
acting as low impedance capacitors. I suspect they are very effective at
60-120 hz, (of course good regulation would do far more to eliminate PS hum
than large caps)  but up at 10-40kHz, they may not be any more effective
than maybe 100uF (not sure here, I haven't made the measurement on such
large caps, and it would be pretty specific to type)

- --
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/AwUBOT1YWYFlGDz1l6VWEQIH3gCgu0ZwTg75FuZOOj/yCEwf9hKXAL8AoNMp
rriXFBdbPclqfUtu0xSTuKfb
=lxGs
-----END PGP SIGNATURE-----

Dave,
You absolutely right, it will not save a bad layout, but that is what people
think.
We have oscillation on this board, let put more filtering, isn't that sounds
familiar ?
Peter

{Original Message removed}
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

At 11:38 AM 6/6/00 -0700, Peter Schultz wrote:
>Dave,
>You absolutely right, it will not save a bad layout, but that is what people
>think.
>We have oscillation on this board, let put more filtering, isn't that sounds
>familiar ?
>Peter

Oh yeah!!

Worst is where I see them added!  Like putting a cap to ground (some
arbitrary point in the ground system) from a signal line, to "smooth it
out"... Grrr. Now the driver needs much more current on each transition,
and the problem is worse.

Resistance in series solves this problem, but it's not so convenient to
implement usually. (Tip: high speed clock lines should always have a
resistor at the source. Even if later you choose the value to be zero,
though likely 100 ohms will be better)

- --
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>

p5nPF+ukK56CFlpWgyO8bjuh
=wKD7
-----END PGP SIGNATURE-----

I seem to recall reading of the development of the first disk drives at
IBM inSan Jose. Rather than look at control system theory and stable
feedback loops, someone figured that a vibrating head positioner could be
fixed by putting a capacitor across the actuator. After the "fix", it
shoved the head through the side of the drive.

Harold

On Tue, 6 Jun 2000 11:38:08 -0700 Peter Schultz <PeterSMINIMED.COM>
writes:
{Quote hidden}

FCC Rules Online at http://hallikainen.com/FccRules
Lighting control for theatre and television at http://www.dovesystems.com

________________________________________________________________
YOU'RE PAYING TOO MUCH FOR THE INTERNET!
Juno now offers FREE Internet Access!
Try it today - there's no risk!  For your FREE software, visit:
dl.http://www.juno.com/get/tagj.

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

At 12:40 PM 6/6/00 -0700, Harold M Hallikainen wrote:
>        I seem to recall reading of the development of the first disk
drives at
>IBM inSan Jose. Rather than look at control system theory and stable
>feedback loops, someone figured that a vibrating head positioner could be
>fixed by putting a capacitor across the actuator. After the "fix", it
>shoved the head through the side of the drive.

ROTFLMAO!!

- --
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/AwUBOT18o4FlGDz1l6VWEQL7JgCguYlVmk7m89eTCSBj8M2NIbiE820An3Z/
7FqeeFta25t/2ojoMmWKtbf7
=PVu+
-----END PGP SIGNATURE-----

>         I seem to recall reading of the development of the
> first disk drives at
> IBM inSan Jose. Rather than look at control system theory and stable
> feedback loops, someone figured that a vibrating head
> positioner could be
> fixed by putting a capacitor across the actuator. After the "fix", it
> shoved the head through the side of the drive.

Am I correct in understanding, then, that this fixed the vibration problem?

<smirk>

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

>Am I correct in understanding, then, that this fixed the vibration problem?
>
><smirk>

This could indeed be seen as a specification problem, rather than a design
problem.
I'm sure that after the application of the "fix" vibration was no longer a
problem.

- --
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/AwUBOT2A3YFlGDz1l6VWEQJGcgCgvSCx6OnNxl+DpNyGUXQQXQjFHDkAoIv0
UxTlw9ja9e1Od394ibDiNhHh
=uSQr
-----END PGP SIGNATURE-----

FYI This has to do with the ESR of the caps.  For an ideal capacitor model,
a large cap would work even better for high freq. than a small cap.  In
practice there is an internal resistance that makes large electrolytics
good for damping low freq. (near DC) noise, and small ceramic caps better
for getting rid of high frequency noise.

At 02:16 PM 6/6/00 -0700, you wrote:
>hi Paul
>The same way as multiple capacitors on a circuit board. Interference can be
>from one chip to the next via the supply, so you put a capacitor across the
>supply pins of each chip as close to the chip as possible.
>The large capacitor can remove some ripple from the supply while the smaller
>capacitor deals with spikes the value and where placed play apart in this.
>As does the path of tracks around the board.
>Regards Art
>
>{Original Message removed}
David VanHorn wrote:
....
>Sockets with caps in them are marginally better than nothing at all (see
>above), but it's even worse, because you've provided a path past the cap on
>the ground side as well.
>

I've read that the long tiny leads on the caps in this case
represent such a large inductance that this kind of socket is
quite useless at higher frequencies.
==============

>It's hard to show this without pictures, but a cap with tracks or leads out
>to power and ground tracks, especially thick ones, is much less than
>optimum.
......
>Using fat power tracks can be bad.  Narrow tracks are higher impedance to
>high frequencies, but are the same DC resistance (close anyway).  I use a
>wide track from the chip to it's bypass cap(s) and a narrow track from
>there back to system power. This is because in the high frequency domain, I
>want the current to come from the bypass cap, not from the system buss.
>

Interesting observations here. But this makes it sound like going to
"ground planes" would be the worst thing imaginable. I've read in the books
that narrow buss traces might cause a problem regarding "return currents"
for the "signals" routed from one chip to the next. Narrow traces -->
larger inductive loops --> more EMI/noise transmission.

Can you comment on this ??? [also see next]
================

{Quote hidden}

This kinda sounds opposite to what you said in the previous paragraph.

Solution --> narrow Vcc, but wide gnd leads ????? Can you illuminate??

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

Harold wrote:
>        I seem to recall reading of the development of the first disk drives at
>IBM inSan Jose. Rather than look at control system theory and stable
>feedback loops, someone figured that a vibrating head positioner could be
>fixed by putting a capacitor across the actuator. After the "fix", it
>shoved the head through the side of the drive.
>

Without seeing the ckt, this sounds like they re-discovered the
integrator. You;d think even back in 1953 or whatever, IBM would know
you need a "zero-switch" to discharge the cap on an integrator.
Easily done here. Mount the switch on the head. When the head bangs
into the side of the drive, the switch engages and resets the
integrators as a sophomore. <:-)).

Cheers,
- Dan Michaels
==============

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

>I've read that the long tiny leads on the caps in this case
>represent such a large inductance that this kind of socket is
>quite useless at higher frequencies.

Yes.

>Interesting observations here. But this makes it sound like going to
>"ground planes" would be the worst thing imaginable. I've read in the books
>that narrow buss traces might cause a problem regarding "return currents"
>for the "signals" routed from one chip to the next. Narrow traces -->
>larger inductive loops --> more EMI/noise transmission.

You're forgetting the bypass. You want the ground fat, and from the chip to
the cap fat, but track from the cap to the rest of the system thinner. The
loop in this case is very small.

{Quote hidden}

Narrow VCC, up to the bypass. Fat from bypass to chip, fat, and
interconnected ground.

Watch interconnection, it can bite too.  I had a thermal printer that took
pulses of 19A 300uS wide. I did NOT let it's ground return run through my
processor logic in a plane.
Can you guess why? :)   I ran separate co-planar power and ground tracks,
200 mils wide to the head. The ground track was isolated from the plane
until the point where it hit the supply caps for the printhead power.

The PH designer had a clue too, they provided power ground and logic ground
on separate pins.

It's a science, but there's some art in it too :)

- --
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/AwUBOT3OkIFlGDz1l6VWEQJlxgCeKnsaRow3SIE3YG8ZBHwbH/YI5mkAoMt3
YcJC2eP915QciVisKZuj5KTz
=KsfA
-----END PGP SIGNATURE-----

On Tue, 6 Jun 2000, Dan Michaels wrote:

> Without seeing the ckt, this sounds like they re-discovered the
> integrator. You;d think even back in 1953 or whatever, IBM would know

They didn't even discover the voice-coil actuator until the late 70's, if
I recall correectly...

Dale
---
The most exciting phrase to hear in science, the one that heralds new
discoveries, is not "Eureka!" (I found it!) but "That's funny ..."
-- Isaac Asimov

Another approach I've used is the micro-Q capacitors from Circuit
Components, Inc. (see http://www.cci-msc.com/mq1000/mq1000.htm). These
capacitors fit under the chip or socket. They use the same through holes
as the power pins to the chip. The result is VERY SHORT leads (and low
inductance).
On most designs now, I've got a ground plane on both sides of the board.
A ceramic bypass capacitor is right at the power pin on the chip (typical
trace length is 0.1 inch). The other side goes to the ground planes.

Harold

On Tue, 6 Jun 2000 20:27:04 -0600 Dan Michaels <oricomLYNX.SNI.NET>
writes:
{Quote hidden}

________________________________________________________________
YOU'RE PAYING TOO MUCH FOR THE INTERNET!
Juno now offers FREE Internet Access!
Try it today - there's no risk!  For your FREE software, visit:
dl.http://www.juno.com/get/tagj.

David VanHorn  wrote:

>
>>Interesting observations here. But this makes it sound like going to
>>"ground planes" would be the worst thing imaginable. I've read in the books
>>that narrow buss traces might cause a problem regarding "return currents"
>>for the "signals" routed from one chip to the next. Narrow traces -->
>>larger inductive loops --> more EMI/noise transmission.
>
>You're forgetting the bypass. You want the ground fat, and from the chip to
>the cap fat, but track from the cap to the rest of the system thinner. The
>loop in this case is very small.
>

Ok, this is what you said initially, but I am still confused as this
would seem to imply that use of a "power" plane [I mis-spoke last time]
would be the worst approach, since complete opposite of small Vcc traces.
????????
=============

{Quote hidden}

Ok, I see this re Vcc, but can you address the issue more fully of the
"grounds". Fat everywhere, or thin? Gnd planes = best????
===============

For a 2-layer bd, I envision  --> a fat gnd running under the chips,
bypass caps near Vss [gnd] pins, thick trace cap to Vcc pins, cap
tied to thin Vcc trace, distributed all over the board. Comments?

Cheers,
- Dan Michaels

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

At 07:29 AM 6/7/00 -0700, Harold M Hallikainen wrote:
>        Another approach I've used is the micro-Q capacitors from Circuit
>Components, Inc. (see http://www.cci-msc.com/mq1000/mq1000.htm). These
>capacitors fit under the chip or socket. They use the same through holes
>as the power pins to the chip. The result is VERY SHORT leads (and low
>inductance).
>        On most designs now, I've got a ground plane on both sides of the
board.
>A ceramic bypass capacitor is right at the power pin on the chip (typical
>trace length is 0.1 inch). The other side goes to the ground planes.

These are the ones that are a large slab with leads?
Definitely better than the "axial 0.1uF in a socket" and possibly as good
as an SMD part mounted at the ground pin. Ideally, you'd feed the VCC side
of this with a relatively narrow track (I use 8 mils for logic) which tends
to keep the noise isolated to that node.

As long as the chip ground is at the same point as the cap ground, then I'd
say these are a good (if somewhat pricey) answer. I think that an SMD cap
on the solder side with proper routing would be within a dB or so and less
expensive.

Haven't measured these yet, but they do have nice low Z looking construction.

- --
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/AwUBOT6YsYFlGDz1l6VWEQK1AwCeOLSZ11GteobITdjn962WaMGkLioAoL43
O7h48JzeoG7o7Zgz+yHj6DIk
=xdlH
-----END PGP SIGNATURE-----

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

>Ok, this is what you said initially, but I am still confused as this
>would seem to imply that use of a "power" plane [I mis-spoke last time]
>would be the worst approach, since complete opposite of small Vcc traces.
>????????

If you look closely enough, ground becomes "just another signal" :)
In the case of an isolated chip, say a microcontroller with it's xtal and
xtal caps, you could feed the whole system with 6 feet of wire-wrap wire on
both power and ground pins, provided you have proper bypassing at the chip.
In fact, at the other end of these 6' leads, you would see less noise than
you would at the chip power pins because of the series resistance and
inductance.

Unfortunately, we usually need to have a few other things in the system.

When prototyping, I use a sheet of copper clad board as my ground, install
bypasses as appropriate at the devices, and run power starred from the 5V
reg output cap on wire-wrap wire. An ideal PCB layout would mimic this.
The idea is that the impedance between devices is absolutely minimized in
the ground plane (because it is the return for all signals) and maximized
in the power system, keeping each chip's noise away from the others.

There are special cases like my printer, where putting large (absolute
magnitude or dI/dT) currents through the plane would be bad, so in those
cases I run a separate return twisted with the source lead in proto, and in
PCB a separate source track and coplanar return track.

>Ok, I see this re Vcc, but can you address the issue more fully of the
>"grounds". Fat everywhere, or thin? Gnd planes = best????

Almost always, a plane is best, but you can get really close with good
routing and making sure all the chips have good low Z (fat) connectivity to
each other.
I have never had to use multilayer, and I stay away from it because it
doubles the PCB cost.

>For a 2-layer bd, I envision  --> a fat gnd running under the chips,
>bypass caps near Vss [gnd] pins, thick trace cap to Vcc pins, cap
>tied to thin Vcc trace, distributed all over the board. Comments?

That's how I route them. Bypass caps should be AT the ground pins,  then I
flood all unused areas on both layers with ground, and pop in a few vias
between layers where I know I'm going to have high freq currents in the
ground.  Note on the VCC though, not just strung all over, but as many as
possible returning to the +5 reg output cap on discrete tracks. I usually
can't get every chip it's own track, so I group them. You might have a low
noise amplifier that gets it's own track, and a logic section with several
chips that has a shared track back to the reg.

Another approach would be to take a thicker track as a distribution bus,
and tap off it with small tracks to local sections. You could add bypasses
at the points where the feeders join the buss. This should be pretty quiet,
but watch where those caps ground to. They should connect to a point that
has very low Z back to the source, and dosen't have a lot of current
running through it.

Definitely, think of any "track" as an inductor, and resistor.
Supply currents cause droop in the voltage at the end of a track.
Return currents will induce a voltage along a ground track too, with the
sourcing end rising above ground, depending on the current magnitude,
dI/Dt, and the track impedance.

- --
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/AwUBOT6dDYFlGDz1l6VWEQLaxwCg+4icEqbm4r2n/x6D/wmnHpLs4aAAoPdD
=z5Q1
-----END PGP SIGNATURE-----

David VanHorn wrote:
{Quote hidden}

Yeah, try to include that R. After reading a lot of appnotes/etc,
I am convinced that using "source series resistors" is a great
advantage in hi-speed systems, esp those using CMOS chips like the
'AC series. 100 ohms or so right on the output pin greatly reduces
ringing/other effects due to Cin of downstream chips as well as stray
capacitance on the lines being driven. This problem becomes very real
above 10 or 20 Mhz or so.

ALso, in certain cases, like where digital pins are controlling
analog devices [eg, CMOS switches], I am also adding caps downstream
of the series R to make little RC filters. Helps keep digital noise
off the analog chips. Here, you can also control the BW. In previous
situations, where I had just used a cap, I had a littany of problems
[inc the notorious one involving PIC read-modify-write instructions,
where the cap holds the voltage fixed while a series of bsf/etc cmds
are given].

Cheers,
- Dan Michaels
==============

>Resistance in series solves this problem, but it's not so convenient to
>implement usually. (Tip: high speed clock lines should always have a
>resistor at the source. Even if later you choose the value to be zero,
>though likely 100 ohms will be better)
>

Yeah, try to include that R. After reading a lot of appnotes/etc,
I am convinced that using "source series resistors" is a great
advantage in hi-speed systems, esp those using CMOS chips like the
'AC series.

Nowdays, you can get chips aimed at driving buses, that already include
a series resistor in the output stages.  (TI's 74ABT series...)

BillW

Daivd VanHorn wrote:
...........
>
>>For a 2-layer bd, I envision  --> a fat gnd running under the chips,
>>bypass caps near Vss [gnd] pins, thick trace cap to Vcc pins, cap
>>tied to thin Vcc trace, distributed all over the board. Comments?
>
>That's how I route them. Bypass caps should be AT the ground pins,  then I
>flood all unused areas on both layers with ground, and pop in a few vias
>between layers where I know I'm going to have high freq currents in the
>ground.  Note on the VCC though, not just strung all over, but as many as
>possible returning to the +5 reg output cap on discrete tracks.
..........

Dave,

All in all, very interesting. I don't believe I have ever seen this
particular scheme [ie, thin Vcc busses + thick bypass leads/etc] described
in any appnotes. But I am sure you learned this empirically by trial and
error.

Certain A/D converter appnotes [roughly 50%] say analog and digital
gnds should be connected together at 1 point only, and only under
the A/D using a "thin" trace, since this trace will pass dc well but help
prevent hi-freq digital noise from getting into the analog area. Sounds
like you have pretty much discovered this same idea viz-a-viz Vcc traces.

Thanks for all the tips,
- Dan Michaels
==============

At 12:02 PM 06/07/2000 PDT, you wrote:
>    >Resistance in series solves this problem, but it's not so convenient to
>    >implement usually. (Tip: high speed clock lines should always have a
>    >resistor at the source. Even if later you choose the value to be zero,
>    >though likely 100 ohms will be better)
>    >
>
>    Yeah, try to include that R. After reading a lot of appnotes/etc,
>    I am convinced that using "source series resistors" is a great
>    advantage in hi-speed systems, esp those using CMOS chips like the
>    'AC series.
>
>Nowdays, you can get chips aimed at driving buses, that already include
>a series resistor in the output stages.  (TI's 74ABT series...)
>
>BillW
>

Yes, you can. For busses, chips with source series Rs are a great help.

But I was specifically referring to my general designs here, not necessarily
driving busses. With very hi-speed logic like 74ACxx chips, designed to
run 100-200 Mhz, those gate caps start to present a serious problem,
and a simple R upstream can help, even with glue logic.

Also, in my PIC designs, I now use Rs and RCs on certain lines [like
clock lines, as Dave mentioned] to help prevent ringing and to filter
noise - such as in control lines going to the analog section.

DanM

I wrote:
.....
>Also, in my PIC designs, I now use Rs and RCs on certain lines [like
>clock lines, as Dave mentioned] to help prevent ringing and to filter
>noise - such as in control lines going to the analog section.

Oh yeah, the I2c/SPI clock is another signal that can benefit
from a series R or RC. A 5 Mhz clock will have harmonics at 15,
25, ... Mhz. If the signal trace is longer than a couple of inches,
this can be a serious problem - if driving an A/D converter/etc.

On the subjects of power system design and signal integrity a good starting
point is:

its archives plus other highly informative sites as Howard Johnson's
High-Speed Digital Design, and Printed Circuit Design Magazine.

Power Distribution System Design Methodology and Capacitor Selection for
Modern
CMOS Technology - IEEE Transactions on Advanced Packaging - August 1999 at:
ftp://ftp.qsl.net/pub/wb6tpu/si_documents/decap_whitepaper.pdf
contains a particularly detailed analysis of the subject.

Gary Crowell
Micron Technology

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

>ALso, in certain cases, like where digital pins are controlling
>analog devices [eg, CMOS switches], I am also adding caps downstream
>of the series R to make little RC filters. Helps keep digital noise
>off the analog chips. Here, you can also control the BW. In previous
>situations, where I had just used a cap, I had a littany of problems
>[inc the notorious one involving PIC read-modify-write instructions,
>where the cap holds the voltage fixed while a series of bsf/etc cmds
>are given].

Huge difference here though.  You added plenty of series R, so the high
speed current through the driving chip is limited by that R. The C behind
it only limits the rise and fall times now, and isn't pulling any fast
current, or dumping any into "ground".

- --
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/AwUBOT7aSIFlGDz1l6VWEQJrGACgyItpGqMK6fqEXyk31tu5SCmUCqsAoItm
sSxiQyhspxr+3M1G45NkJ2I4
=viCM
-----END PGP SIGNATURE-----

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

>Dave,
>
>All in all, very interesting. I don't believe I have ever seen this
>particular scheme [ie, thin Vcc busses + thick bypass leads/etc] described
>in any appnotes. But I am sure you learned this empirically by trial and
>error.

More by trial and measurement, but yes.

>Certain A/D converter appnotes [roughly 50%] say analog and digital
>gnds should be connected together at 1 point only, and only under
>the A/D using a "thin" trace, since this trace will pass dc well but help
>prevent hi-freq digital noise from getting into the analog area. Sounds
>like you have pretty much discovered this same idea viz-a-viz Vcc traces.

Same theory.  When I do something like that, I use something like AGND in
the schematic for all affected nodes, then a zero ohm resistor to tie AGND
to GND, and place those pads at the appropriate point. This way the floods
don't connect, and I can use a ferrite, or solder bridge, or resistor at
that point, whatever works best.

If you get the chance to pick up a spectrum analyzer (like the 7L14) cheap,
then do it, it answers SO many questions!

It's a nice feeling when the part 15 test guys ask "Are you sure it's on?" :)

- --
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/AwUBOT7bMoFlGDz1l6VWEQKOlwCeNr6wWSt09ltd5bJ1CV4dy0cyXWwAn2g5
XY0OYmfM6PRJYwx55aAn8/B5
=47ul
-----END PGP SIGNATURE-----

David VanHorn wrote:
.....
>If you get the chance to pick up a spectrum analyzer (like the 7L14) cheap,
>then do it, it answers SO many questions!
>
>It's a nice feeling when the part 15 test guys ask "Are you sure it's on?" :)
>

Yeah, I had pretty much figured you have a scope probe in one
hand and spectrum analyzer probe in the other. Keep up the good
work. Keep the feds happy!!

- Dan Michaels

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

>Yeah, I had pretty much figured you have a scope probe in one
>hand and spectrum analyzer probe in the other. Keep up the good
>work. Keep the feds happy!!

And us hams!  I don't want to pollute my own spectrum!

- --
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/AwUBOT8F5IFlGDz1l6VWEQLgYACcCEUnKvrUGbuf1Ag5TmD1s0rzpnQAn3GD
cbvDZzmuw7bZxukIM+DaeEGu
=WIEQ
-----END PGP SIGNATURE-----

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