> At 08:43 PM 5/15/2005 +0200, you wrote:
>
>> I need some advice comments on this circuit
>> http://www.vscp.org/images/barometer001.gif  You probably laugh your
>> guts out but but when you finished laughing please tell me what I
>> have done wrong.
>
>
> First off, you've got the + and - inputs swapped. When
> that's fixed, the gain will not be 10, but approximately
> 100/11.8 = 8.47. The PIC output buffer will enter into it a bit.
> If you don't like that, put a unity-gain op-amp buffer after the RC.
> (with a dual op-amp, it's doesn't add any components).
> Better yet make it a two or three pole filter with a time constant
> in the tenths of a second. A single pole filter can only do so much.
> Eg. if you use 100K and 1uF and PWM at 16kHz, then in one half-cycle
> (31usec) the voltage will change by about t* (2.5/100K)*1/C = 0.7mV
> at midscale, representing 7mV at the output or about 1.5 counts on a
> 10-bit ADc.
> The ripple with your setup will be outside the range of my linear
> approximation, many many counts.
>
> There's nothing inherently wrong with your concept, however just using a
> 12-bit ADC would be much better for most purposes.
>
>
>> Is this a solution that is worth trying?  I can take away any static
>> error during calibration but is there any sources for temperature or
>> other dynamic errors that I need to take care of.
>
>
> If the pressure transducer is ratiometric from the supply, then the PWM
> is not a bad way of subtracting from that. You'll have to consider
> the sensor accuracy itself. If you use a decent op-amp with low offset
> voltage and decent resistors, they should not contribute substantially to
> the drift.
>
> Best regards,
>
> Spehro Pefhany --"it's the network..."            "The Journey is the
> reward"
> .....speffKILLspam.....interlog.com             Info for manufacturers:
> http://www.trexon.com
> Embedded software/hardware/analog  Info for designers:  
> http://www.speff.com
>
>
>
>

>Seems complicated to me! I'd be concerned about the stability of the
>differential amplifier (if you're trying to build the diff amp yourself),
>stability of the PWM output, ripple of the PWM output, etc. IF the
>transducer indeed has the resolution and accuracy you're seeking, it seems
>a simpler approach is to use an external A/D that has more resolution. I
>did a precision thermometer using a Data Acquisition System chip from
>Maxim. It, as I recall, has a 16 bit delta sigma A/D, a 10 bit D/A, and a
>real time clock. Simple to interface to the PIC with SPI.
>
>Harold
>  
>

>Ake Hedman <EraseMEakhespam_OUTTakeThisOuTeurosource.se> wrote:
>  
>
>>I need some advice comments on this circuit
>>http://www.vscp.org/images/barometer001.gif  You probably laugh your
>>guts out but but when you finished laughing please tell me what I have
>>done wrong.
>>    
>>
>
>You have the inputs to the opamp swapped. Another way of saying this
>is that R4 needs to be the one connected to the opamp output and R3
>needs to be grounded. And also R2 needs to be 10K, not 100K.
>
>-- Dave Tweed
>  
>

> Interesting idea.
>
> The MPX sensor output has a large amount of noise.  You
> will need to filter it extensively before applying it to your circuitry.
> I seem to recall the app notes specifying a 750 Hz or so low
> pass filter.  You will probably want an even lower cutoff frequency
> than suggested.  A single pole RC is probably not sufficient for what
> you are after - a multiple pole op-amp filter may be better.
> Similarly, the PWM will probably need more filtering as well.
> Extremely clean power to the circuit is a must.  Also, choose your
> sampling frequency carefully to avoid harmonics.  Watch out for
> non-linear regions of the op-amp near ground and 5V rails.
> Finally, once you've treated it as best as is reasonable in
> hardware, remember that software filtering is your friend.
>
>> Is this a solution that is worth trying?  I can take away any static
>> error during calibration but is there any sources for temperature or
>> other dynamic errors that I need to take care of.
>
>
> I'd try it, with additional filtering of course.  Temperature will
> effect the results, you will need to determine the temperature
> range and the accuracy required over that range to see if you
> exceed the expected results (a rather complicated calculation,
> considering all the "analog" parts and their tolerances).
>
> klumiaspam_OUTadelphia.net
>

> At 11:10 PM 5/15/2005 +0200, you wrote:
>
>> Spehro,
>>
>> First of all thanks for noticing  the +/. swap. Had it on paper and
>> the component in eagle was the other way.  Probably should have taken
>> me several hours to find on a board... ;-(
>>
>> When you say adding a 2nd pole filter do you mean to replace the PWM
>> LPF?  My thought was to also remove ripple in software. In this case
>> it has known properties so this should be possible.
>
>
> If it's large you can run into nonlinearities when you don't expect to.
>
>> Where does the 11.8 come from in 100/11.8?
>
>
> In the DC path, the 1.8K is in series with the 10K (think of the
> capacitor
> as being open circuit for DC). If you put a buffer in there,
> as I suggested, it will reduce that to zero and eliminate the effect
> of the
> PIC port output impedance.
>
> That's the gain looking from a fictional DC PWM source. The gain from the
> sensor will be half of (that number plus 1) because of the  voltage
> divider
> on the non-inverting input.
>
>
> Best regards,
>
> Spehro Pefhany --"it's the network..."            "The Journey is the
> reward"
> KILLspamspeffKILLspaminterlog.com             Info for manufacturers:
> http://www.trexon.com
> Embedded software/hardware/analog  Info for designers:  
> http://www.speff.com
> >>>Inexpensive test equipment & parts
> http://search.ebay.com/_W0QQsassZspeff
>
>

> Ake Hedman wrote:
>
>> First of all thanks for noticing  the +/. swap. Had it on paper and the
>> component in eagle was the other way.
>
>
> If you are using Eagle you have another problem.  The lack of a dot at
> the
> opamp output shows that it is not really connected to the wires.
>
>
> *****************************************************************
> Embed Inc, embedded system specialists in Littleton Massachusetts
> (978) 742-9014, http://www.embedinc.com

> Ake Hedman wrote:
>
>> I need some advice comments on this circuit
>> http://www.vscp.org/images/barometer001.gif
>
>
> Hmm, where to begin.  I admit I didn't read all of your post (a bit
> long) so
> maybe some of this is explained or addresed there.  However, several
> things
> look seriously wrong if this is supposed to do what I think you intended.
>
> 1  -  This circuit performs thresholding with hystersis, which is
> contrary
> to feeding the output into an A/D input.  You probably have the opamp
> inputs
> backwards.
>
> 2  -  If this is supposed to be a linear amplifier and the opamp
> inputs are
> flipped, then why attenuate the input?  R2 and R4 divide the input
> voltage
> by 2, which doesn't make sense if you ultimately want to amplify it.
>
> 3  -  The R5 and C1 PWM filter has a rolloff frequency of about
> 900Hz.  The
> PWM frequency needs to be well above that to filter out the individual
> PWM
> pulses.  You need the residual PWM noise to be about 1/2 LSB on the
> A/D or
> less, and I doubt this filter will do that for reasonable PWM
> frequencies.
> You probably want a two pole filter at least.
>
> 4  -  If this is supposed to be a diff amp with the PWM setting the "0"
> level, then R1 is definitely wrong.  Since the PWM filter output
> impedence
> will vary with frequency and will be impossible to adjust for, it
> should be
> buffered.
>
>
> *****************************************************************
> Embed Inc, embedded system specialists in Littleton Massachusetts
> (978) 742-9014, http://www.embedinc.com

> Ake Hedman wrote:
>
>> I was trying to keep away from 12-bit A/D's due to cost reason.
>
>
> There is no need for this.  The 30F3013 is nicely priced and comes
> with 12
> bit A/Ds.  Designing the circuit so that all errors amount to less than 1
> part in 8000 is the tricky part.  At that level you have to consider
> offset
> voltages caused by currents thru ground paths and a bunch of other stuff.
> This is certainly doable, but not something a newbie is likely to get
> right
> the first time.
>
>
> *****************************************************************
> Embed Inc, embedded system specialists in Littleton Massachusetts
> (978) 742-9014, http://www.embedinc.com

>The Maxim part (16 bit data acquisition system) was not too bad, as I
>recall. In the thermometer circuit (which uses a precision thermistor), I
>use a 0.1% pull-up on the thermistor, then go directly into the A/D. The
>top of the pull-up is the on-chip reference, so reference errors cancel.
>This one part eliminated the need for differential amplifiers and a bunch
>of precision parts. You connect it - it works. I often say "The ideal
>design has zero parts." If you can get one part to do the work of 10, I
>use the one. It may appear cheaper to use the 10 cheap parts, but the
>hassles in getting it to work reliably, deal with adjustments, etc. is
>just not worth it to me. I ended up with a circuit with no adjustments. It
>just works...
>
>Harold
>"The ideal design has zero parts."
>
>
>  
>
>>Harold,
>>
>>I was trying to keep away from 12-bit A/D's due to cost reason. I have
>>the uP and it have a 10-bit A/D. If there are any 12-bit A/D's available
>>at a good price oh well its another situation. Suggestions?
>>
>>/Ake
>>    
>>
>
>
>
>  
>

> Interesting idea.
>
> The MPX sensor output has a large amount of noise.  You
> will need to filter it extensively before applying it to your circuitry.
> I seem to recall the app notes specifying a 750 Hz or so low
> pass filter.  You will probably want an even lower cutoff frequency
> than suggested.  A single pole RC is probably not sufficient for what
> you are after - a multiple pole op-amp filter may be better.
> Similarly, the PWM will probably need more filtering as well.
> Extremely clean power to the circuit is a must.  Also, choose your
> sampling frequency carefully to avoid harmonics.  Watch out for
> non-linear regions of the op-amp near ground and 5V rails.
> Finally, once you've treated it as best as is reasonable in
> hardware, remember that software filtering is your friend.
>
> > Is this a solution that is worth trying?  I can take away any static error
> > during calibration but is there any sources for temperature or other
> > dynamic errors that I need to take care of.
>
> I'd try it, with additional filtering of course.  Temperature will
> effect the results, you will need to determine the temperature
> range and the accuracy required over that range to see if you
> exceed the expected results (a rather complicated calculation,
> considering all the "analog" parts and their tolerances).
>
> TakeThisOuTklumiaEraseMEspam_OUTadelphia.net
>

>Anthony Van Herrewege wrote:
>  
>
>>Hm, that's not true. I do everything in Eagle and it doesn't place
>>dots unless you're connecting 2 wires together (which isn't the same
>>as connecting a wire to a pin).
>>    
>>
>
>Yes, but as I remember it, the OP had a junction of two lines right at the
>opamp output.  There should be a dot where two wires and a pin connect.  I
>also remember seeing a slight offset.  The lines are probably not connected
>because he wasn't using the right grid when drawing the schematic.
>
>
>*****************************************************************
>Embed Inc, embedded system specialists in Littleton Massachusetts
>(978) 742-9014, http://www.embedinc.com
>  
>

>Ake Hedman wrote:
>  
>
>>It's just a sketch so I didn't put to much efforts into such things.
>>    
>>
>
>Argh!  Neatness and attention to detail count.  If you don't care, why
>should anyone else?  Next time you ask 2000 people for advice on a sloppily
>drawn sketch, please say so.  That way I can avoid wasting any time on it up
>front.
>
>
>*****************************************************************
>Embed Inc, embedded system specialists in Littleton Massachusetts
>(978) 742-9014, http://www.embedinc.com
>  
>

> Ake Hedman wrote:
> > It's just a sketch so I didn't put to much efforts
> into such things.
>
> Argh!  Neatness and attention to detail count.  If
> you don't care, why
> should anyone else?  Next time you ask 2000 people
> for advice on a sloppily
> drawn sketch, please say so.  That way I can avoid
> wasting any time on it up
> front.
>
>
>

> -----Original Message-----
> From: RemoveMEpiclist-bouncesEraseMEEraseMEmit.edu [RemoveMEpiclist-bouncesspam_OUTKILLspammit.edu] On Behalf
> Of Wouter van Ooijen
> Sent: Tuesday, May 17, 2005 12:00 AM
> To: 'Microcontroller discussion list - Public.'
> Subject: RE: [EE] Comments on circuit
>
> > Completely uncalled for.  The diagram is neat enough
> > for people to understand.  Even you did.  Enough
> > people responded that it is obvious to me that it is
> > quite clear to others as well.  I find it hard to see
> > that the omission of a single dot is deserving of this
> > diatribe.
>
> It is clearly an Eagle circuit drawing. How about the possibility that
> the OP would use this drawing (or a modified version) as starting point
> for creating a PCB?
>
> Wouter van Ooijen
>
> -- -------------------------------------------
> Van Ooijen Technische Informatica: http://www.voti.nl
> consultancy, development, PICmicro products
> docent Hogeschool van Utrecht: http://www.voti.nl/hvu
>
>

> -----Original Message-----
> From: RemoveMEpiclist-bouncesTakeThisOuTspammit.edu [EraseMEpiclist-bouncesspamspamBeGonemit.edu] On Behalf
> Of Michael Rigby-Jones
> Sent: Tuesday, May 17, 2005 1:48 AM
> To: Microcontroller discussion list - Public.
> Subject: RE: [EE] Comments on circuit
>
>
>
> >-----Original Message-----
> >From: RemoveMEpiclist-bouncesKILLspammit.edu [piclist-bouncesSTOPspamspam_OUTmit.edu]
> >Sent: 17 May 2005 04:36
> >To: 'Microcontroller discussion list - Public.'
> >Subject: RE: [EE] Comments on circuit
> >
> >
> >We'll never have the opportunity to know how many excellent
> >minds he has driven away.
> >
> >I control an engineering department and we make some pretty
> >interesting things, but I long ago decided to simply read and
> >avoid the nastiness of posting. When he's right he's right,
> >but when he's wrong he's extremely wrong. All that mind
> >negated by personality.
>
> Please don't turn this into yet another slanging match. If you have a
> problem with Olin then please sort it out off-list.
>
> Mike
>
> =======================================================================
> This e-mail is intended for the person it is addressed to only. The
> information contained in it may be confidential and/or protected by
> law. If you are not the intended recipient of this message, you must
> not make any use of this information, or copy or show it to any
> person. Please contact us immediately to tell us that you have
> received this e-mail, and return the original to us. Any use,
> forwarding, printing or copying of this message is strictly prohibited.
> No part of this message can be considered a request for goods or
> services.
> =======================================================================
>

> At 08:32 AM 5/19/2005, Olin Lathrop wrote:
> >Gerhard Fiedler wrote:
> >>>But it also produces all kinds of noise, like
> 'vss connected to gnd'
> >>>and 'only one pin on net'. As it is I do not find
> it usefull.
> >>
> >>Can't you configure that?
> >
> >Not that I know of.
>
> Good old dos orcad.. Totally configurable, and
> realtime DRC that
> actually works.
> 12 years obsolete, and still going strong.
>