isolation for data aquisition

classic Classic list List threaded Threaded
14 messages Options
Reply | Threaded
Open this post in threaded view
|

isolation for data aquisition

Al-57
Well, I am finding out that it is no simple task to isolate analog voltages
to feed to a data aquistion card and maintain accuracy. I looked at linear
opto-couplers, V to F converters, and isolated op-amps. I could use a
separate A/D converter and opto-isolate its data output. The problem I see
is having to use relatively noisy DC/DC converters as power supplies. I
suppose with extra filtering they could work ok. If you are trying to read a
shunt down to fractions of a millivolt, the DC/DC spec sheets show 50 to 100
millivolts ripple on the output, not good. If you try to amplify the shunt
voltage, you run into another can of worms as far as noise and DC accuracy
are concerned. I guess I need to quantify just how accurate the various
readings need to be to have useful data. I'm pretty sure systems like the
Paktrakr are not terribly accurate. They don't even post specs on their
website. I'll have to check out Victors Evision.

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Peter VanDerWal
I've been working on the idea of breaking up the pack into smaller,
logical, packs.  Say around 30 volts each.
Use a small embedded device(PIC, etc.) to measure the voltage of the cells
in the sub pack and perhaps control a BMS.  It can draw it's power from
the batteries it's monitoring.
Optically isolate the databus (maybe I2C or CanBus) and send them back to
a central unit for display (and possibly charger control).  The central
unit can measure current, and perhaps the lower 30 volts of the pack.
With fewer units (instead of one per battery/cell) it makes it easier to
disconnect/turn off, if you need to store the EV for an extended period.
Should keep the costs down too.

Anyway, that's my idea.

> Well, I am finding out that it is no simple task to isolate analog
> voltages
> to feed to a data aquistion card and maintain accuracy. I looked at linear
> opto-couplers, V to F converters, and isolated op-amps. I could use a
> separate A/D converter and opto-isolate its data output. The problem I see
> is having to use relatively noisy DC/DC converters as power supplies. I
> suppose with extra filtering they could work ok. If you are trying to read
> a
> shunt down to fractions of a millivolt, the DC/DC spec sheets show 50 to
> 100
> millivolts ripple on the output, not good. If you try to amplify the shunt
> voltage, you run into another can of worms as far as noise and DC accuracy
> are concerned. I guess I need to quantify just how accurate the various
> readings need to be to have useful data. I'm pretty sure systems like the
> Paktrakr are not terribly accurate. They don't even post specs on their
> website. I'll have to check out Victors Evision.
>
> _______________________________________________
> General EVDL support: http://evdl.org/help/
> Usage guidelines: http://evdl.org/help/index.html#conv
> Archives: http://evdl.org/archive/
> Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev
>
>


_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Toby B
In reply to this post by Al-57
Yeah- thus, the small 12v motorcycle (or gel or similar) battery, with the
dc- dc as a charger.
Then some intelligent wiring, and smaller caps to filter the HF component
that the
battery can't catch.
Bonus is, if your dc-dc pops, you have a few amps to get you home...

That's my version...
Toby

On Wed, Jul 30, 2008 at 6:58 PM, Al <[hidden email]> wrote:

>  The problem I see
> is having to use relatively noisy DC/DC converters as power supplies. I
> suppose with extra filtering they could work ok.
_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Victor Tikhonov
In reply to this post by Al-57
The front end of EVision is BRUSA measurement board that does V->F and
I->F conversion, and that is accurate within +/-1% of the top scale.

Overall accuracy is defined by that board - EVision only measures this
frequencies as raw data in (thus noise is no longer an issue),
interprets and displays it and also does come math to derive other values.

Granted, measurement board is galvanically isolated from EVision and its
remote display (and that, in turn, is isolated from serial port
streaming out data).

Victor


Al wrote:

> Well, I am finding out that it is no simple task to isolate analog voltages
> to feed to a data aquistion card and maintain accuracy. I looked at linear
> opto-couplers, V to F converters, and isolated op-amps. I could use a
> separate A/D converter and opto-isolate its data output. The problem I see
> is having to use relatively noisy DC/DC converters as power supplies. I
> suppose with extra filtering they could work ok. If you are trying to read a
> shunt down to fractions of a millivolt, the DC/DC spec sheets show 50 to 100
> millivolts ripple on the output, not good. If you try to amplify the shunt
> voltage, you run into another can of worms as far as noise and DC accuracy
> are concerned. I guess I need to quantify just how accurate the various
> readings need to be to have useful data. I'm pretty sure systems like the
> Paktrakr are not terribly accurate. They don't even post specs on their
> website. I'll have to check out Victors Evision.
>
> _______________________________________________
> General EVDL support: http://evdl.org/help/
> Usage guidelines: http://evdl.org/help/index.html#conv
> Archives: http://evdl.org/archive/
> Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev
>
>
>

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Lee Hart
In reply to this post by Al-57
Al wrote:
> Well, I am finding out that it is no simple task to isolate analog
> voltages to feed to a data aquisition card and maintain accuracy.

Absolutely! There aren't many situations noisier than an EV.

> I looked at linear opto-couplers, V to F converters, and isolated
> op-amps. I could use a separate A/D converter and opto-isolate its
> data output. The problem I see is having to use relatively noisy
> DC/DC converters as power supplies.

You don't *need* to use switchers. Other options: Use a linear supply.
Derive power directly from the battery being measured. Use non-contact
devices for measuring current (like hall effect parts). Analog meters
(which don't require power supplies) are available with switch or
optical outputs.

> If you try to amplify the shunt voltage, you run into another can of
> worms as far as noise and DC accuracy are concerned.

It takes special opamps, with very low input offset voltage specs
(microvolts instead of millivolts). Some are hand-trimmed; others are
chopper-stabilized. They aren't that much more expensive than regular
opamps.

Be aware that special circuit design techniques are needed as well.
Thermoelectric voltages can cause all sorts of problems if you are
unaware of them.

> I guess I need to quantify just how accurate the various readings
> need to be to have useful data.

Yes! It's important to know what you are going to do with the data. Does
it matter if a battery's voltage is 12v, 12.3v, 12.34v, 12.345v, etc?
Two digits is only good for good/bad indications. Three digits only
provides a very rough indication of state of charge. Four digits is
about the minimum for serious measurements (accurate to a few percent).
Five digits is for more advanced R&D work, like measuring subtle
temperature effects etc.
--
Ring the bells that still can ring
Forget the perfect offering
There is a crack in everything
That's how the light gets in    --    Leonard Cohen
--
Lee A. Hart, 814 8th Ave N, Sartell MN 56377, leeahart_at_earthlink.net

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Morgan LaMoore
> Be aware that special circuit design techniques are needed as well.
> Thermoelectric voltages can cause all sorts of problems if you are
> unaware of them.

Could you recommend a resource for learning about these effects?

Also, could you recommend a resource about learning to design more
noise-immune circuits? I get the basics like good filtering and short
traces from bypass capacitors to component supply/ground pins, but I'd
like to learn more.

Thanks,
Morgan LaMoore

On Thu, Jul 31, 2008 at 10:09 AM, Lee Hart <[hidden email]> wrote:

>
> Al wrote:
> > Well, I am finding out that it is no simple task to isolate analog
> > voltages to feed to a data aquisition card and maintain accuracy.
>
> Absolutely! There aren't many situations noisier than an EV.
>
> > I looked at linear opto-couplers, V to F converters, and isolated
> > op-amps. I could use a separate A/D converter and opto-isolate its
> > data output. The problem I see is having to use relatively noisy
> > DC/DC converters as power supplies.
>
> You don't *need* to use switchers. Other options: Use a linear supply.
> Derive power directly from the battery being measured. Use non-contact
> devices for measuring current (like hall effect parts). Analog meters
> (which don't require power supplies) are available with switch or
> optical outputs.
>
> > If you try to amplify the shunt voltage, you run into another can of
> > worms as far as noise and DC accuracy are concerned.
>
> It takes special opamps, with very low input offset voltage specs
> (microvolts instead of millivolts). Some are hand-trimmed; others are
> chopper-stabilized. They aren't that much more expensive than regular
> opamps.
>
> Be aware that special circuit design techniques are needed as well.
> Thermoelectric voltages can cause all sorts of problems if you are
> unaware of them.
>
> > I guess I need to quantify just how accurate the various readings
> > need to be to have useful data.
>
> Yes! It's important to know what you are going to do with the data. Does
> it matter if a battery's voltage is 12v, 12.3v, 12.34v, 12.345v, etc?
> Two digits is only good for good/bad indications. Three digits only
> provides a very rough indication of state of charge. Four digits is
> about the minimum for serious measurements (accurate to a few percent).
> Five digits is for more advanced R&D work, like measuring subtle
> temperature effects etc.
> --
> Ring the bells that still can ring
> Forget the perfect offering
> There is a crack in everything
> That's how the light gets in    --    Leonard Cohen
> --
> Lee A. Hart, 814 8th Ave N, Sartell MN 56377, leeahart_at_earthlink.net
>
> _______________________________________________
> General EVDL support: http://evdl.org/help/
> Usage guidelines: http://evdl.org/help/index.html#conv
> Archives: http://evdl.org/archive/
> Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev
>

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Lee Hart
Morgan LaMoore wrote:
>> Be aware that special circuit design techniques are needed as well.
>> Thermoelectric voltages can cause all sorts of problems if you are
>> unaware of them.
>
> Could you recommend a resource for learning about these effects?

Nothing specific jumps to mind; but Analog Devices, Burr-Brown, and
Linear Technologies all have quite a number of publications that discuss
them. Also look specifically for articles authored by Bob Pease, Bob
Widlar, and Jim Williams; they are masters of analog design.

> Also, could you recommend a resource about learning to design more
> noise-immune circuits? I get the basics like good filtering and short
> traces from bypass capacitors to component supply/ground pins, but I'd
> like to learn more.

This is a *huge* field. I don't know of any specific "short cut"
references. What I know has come from many sources, over many years. By
nature, noise is random and unpredictable, and the measures you use to
deal with it are often case-specific or of a statistical nature (they
reduce the likelihood or severity of it, but don't entirely eliminate
the problem.
--
Ring the bells that still can ring
Forget the perfect offering
There is a crack in everything
That's how the light gets in    --    Leonard Cohen
--
Lee A. Hart, 814 8th Ave N, Sartell MN 56377, leeahart_at_earthlink.net

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Josh Wyatt-2
Lee Hart wrote:

> Morgan LaMoore wrote:
>>> Be aware that special circuit design techniques are needed as well.
>>> Thermoelectric voltages can cause all sorts of problems if you are
>>> unaware of them.
>> Could you recommend a resource for learning about these effects?
>
> Nothing specific jumps to mind; but Analog Devices, Burr-Brown, and
> Linear Technologies all have quite a number of publications that discuss
> them. Also look specifically for articles authored by Bob Pease, Bob
> Widlar, and Jim Williams; they are masters of analog design.

Wow.  Bob Pease. Does Bob still publish "Pease Porridge" ?

Thanks,
Josh

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Lee Hart
>> Also look specifically for articles authored by Bob Pease...

Josh Wyatt wrote:
> Wow.  Bob Pease. Does Bob still publish "Pease Porridge"?

Yep; He's alive and well, and still writing his monthly column in
Electronic Design magazine.

--
Ring the bells that still can ring
Forget the perfect offering
There is a crack in everything
That's how the light gets in    --    Leonard Cohen
--
Lee A. Hart, 814 8th Ave N, Sartell MN 56377, leeahart_at_earthlink.net

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Tom Parker-12
In reply to this post by Morgan LaMoore
On Thu, 2008-07-31 at 10:31 -0500, Morgan LaMoore wrote:
> Also, could you recommend a resource about learning to design more
> noise-immune circuits? I get the basics like good filtering and short
> traces from bypass capacitors to component supply/ground pins, but I'd
> like to learn more.

I'll give you an example. The Brusa front end for the EVision. I believe
this circuit is "just" a voltage to frequency converter.

http://carrott.org/twiki/pub/ElectricMini/Instrumentation/BrusaShuntCircuit.jpg

The two screws attach it to each side of the shunt. The back side is
mostly ground plane. I don't know how many layers it has.

Does anyone have any recommendations for protecting it from the
automotive environment. I'm not sure if the current EVision ships with a
bare board on the shunt. I think of putting it in one of those sealed
electrical boxes will suffice. I'll have to check whether a plastic box
will get to hot.

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Lee Hart
Tom Parker wrote:
> I'll give you an example. The Brusa front end for the EVision. I believe
> this circuit is "just" a voltage to frequency converter.

Yes, but they use a lot of parts to get it.

> The two screws attach it to each side of the shunt. The back side is
> mostly ground plane... Does anyone have any recommendations for
> protecting it from the automotive environment.

I have one. It's just a "naked" PC board with no environmental
protection at all. Given the very small spacings between the surface
mount parts, it *has* to be mounted so as to keep it totally dry and clean.

At minimum, I would conformally coat the board with one of the thick
silicone rubber based materials, like Dow Corning 2577-1. This is a
clear thick coating that provides moisture and physical protection, yet
can be dug off if repairs are needed.

Or you can put it in a sealed box for better physical protection.
Nothing on this board gets hot, but the *shunt* does, so don't put it in
the same box.

--
Ring the bells that still can ring
Forget the perfect offering
There is a crack in everything
That's how the light gets in    --    Leonard Cohen
--
Lee A. Hart, 814 8th Ave N, Sartell MN 56377, leeahart_at_earthlink.net

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Victor Tikhonov
In reply to this post by Tom Parker-12
Tom,

I conformal coat these now (as I do all other EVision PCBs).

At the beginning (up to serial #15) I didn't touch shunt boards
and included them as is - BRUSA use to sell them like that as part of their
Ah counter for years and apparently there were no major
issues with that (else they'd change design, coat, put in
the box etc). I talked with engineer in charge and they told me
coating should not be an issue.

You can get a coating compound in a spray can and do it,
just protect connectors. That of course will prevent
moisture absorption by the PCB material - any leakages
on the board with uV-level offset op amps is not good.

OP-Amps may be very low offset, chopper-compensated and so on,
but it is not a substitute for a clean PCB.

If you install the measurement board in the enclosed box,
I suppose there is no need to coat it, but coating
won't hurt in any case.

Victor


Tom Parker wrote:

> On Thu, 2008-07-31 at 10:31 -0500, Morgan LaMoore wrote:
>> Also, could you recommend a resource about learning to design more
>> noise-immune circuits? I get the basics like good filtering and short
>> traces from bypass capacitors to component supply/ground pins, but I'd
>> like to learn more.
>
> I'll give you an example. The Brusa front end for the EVision. I believe
> this circuit is "just" a voltage to frequency converter.
>
> http://carrott.org/twiki/pub/ElectricMini/Instrumentation/BrusaShuntCircuit.jpg
>
> The two screws attach it to each side of the shunt. The back side is
> mostly ground plane. I don't know how many layers it has.
>
> Does anyone have any recommendations for protecting it from the
> automotive environment. I'm not sure if the current EVision ships with a
> bare board on the shunt. I think of putting it in one of those sealed
> electrical boxes will suffice. I'll have to check whether a plastic box
> will get to hot.
>
> _______________________________________________
> General EVDL support: http://evdl.org/help/
> Usage guidelines: http://evdl.org/help/index.html#conv
> Archives: http://evdl.org/archive/
> Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev
>
>
>

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Victor Tikhonov
In reply to this post by Lee Hart
Lee Hart wrote:

> At minimum, I would conformally coat the board with one of the thick
> silicone rubber based materials, like Dow Corning 2577-1. This is a
> clear thick coating that provides moisture and physical protection, yet
> can be dug off if repairs are needed.

As I mentioned to Tom, I got authorizing OK from BRUSA to coat
their boards now. You can do it to yours.

> Or you can put it in a sealed box for better physical protection.
> Nothing on this board gets hot, but the *shunt* does, so don't put it in
> the same box.

NO! The measurement board must stay on the shunt the way it is.

Any extension of wiring will inevitably introduce the error
not taken care of during calibration and instability in measured value.
Yes, I realize according to the text books no current flows from shunt
toward high input impedance OP-Amps, so no appreciable
voltage drop develops, but in practice the noise induced on any
extension wiring, quality of solder joints, temp influence on the
extension wiring will negate all efforts to achieve high accuracy,
esp. in presence of very noisy battery current conductors near by.

If you mount measurement board separately from the shunt,
don't complain you get worse than 1% accurate measurements.

Victor

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev

Reply | Threaded
Open this post in threaded view
|

Re: isolation for data aquisition

Neon John
In reply to this post by Victor Tikhonov
On Fri, 01 Aug 2008 12:31:59 -0700, Metric Mind <[hidden email]> wrote:


>You can get a coating compound in a spray can and do it,
>just protect connectors. That of course will prevent
>moisture absorption by the PCB material - any leakages
>on the board with uV-level offset op amps is not good.
>
>OP-Amps may be very low offset, chopper-compensated and so on,
>but it is not a substitute for a clean PCB.

True.  Nuclear radiation detectors, particularly ion chambers, output signals
in the picoamp range.  We discovered long ago that plain old Krylon Krystal
Clear spray is one of the best low-leakage, non-hygroscopic conformal coatings
available in a spray can.  It can be removed with a Q-tip dipped in toluene or
simply soldered-through.

The normal procedure after a repair on an ion chamber input board was to bake
it under low vacuum at about 150 deg for several hours in an environmental
chamber and then apply the first coat of Krylon while the board was still
warm.  That precludes any moisture absorption by the board.   The vacuum has
to be limited to prevent damage to electrolytic capacitors.

John
--
John De Armond
See my website for my current email address
http://www.neon-john.com
http://www.johndearmond.com <-- best little blog on the net!
Tellico Plains, Occupied TN
Give Blood.  8 Billion Mosquitoes can't be wrong.

_______________________________________________
General EVDL support: http://evdl.org/help/
Usage guidelines: http://evdl.org/help/index.html#conv
Archives: http://evdl.org/archive/
Subscription options: http://lists.sjsu.edu/mailman/listinfo/ev