A>D conversion of a DC signal

[OddHarmonic]

Here at the National Center for Voice and Speech, we make lots of acoustic measurements. SLMs, and microphones are common, and I'm quite well versed in how that stuff works throughout the signal chain. However, we also occasionally make some instrumentation measurements, namely pressure and flow using a mask system. Currently, our computer interface hardware is made by a company called Kay Elemetrics. My lab is using the CSL 4500 box, which consists of a Lynx-2 card in the computer, and a breakout box developed by Kay to offer more control over the signals. These controls include variable gain, a few special calibrated, voice-specific gain settings, and the option to DC-couple the last two input channels, allowing one to get signals with DC content into the computer.

For a variety of reasons, we are working on putting together a new data collection system that will be USB/Firewire/PCMIA. We hope to improve on a few features, but we also need portability.

My question is, how can we get DC signals into the computer using a system that also can handle audio. The problem I have found, is most companies are on one side of the fence. They deal entirely with instrumentation, and have these great little boxes that have 8-16 A>Ds (often 12/14 bit) for DC signals, and a sample-rate system that allows you to spread the sample rate over several channels. (1 chan. @ 200K, 2 chan. @ 100K, etc...) These boxes never seem to have good audio implementation. THEN, on the other side of the fence there are lots of companies that do great audio, but don't seem to ever touch the DC world. KAY seems to be the only exception, but like I said above, we are drifting away from them.

I need to be able to record both microphone signals and pressure/flow at the same time. Lynx has offered to direct-mod a Lynx card for me, but again, PCI limits us to our desktops. They said that the removal of a few DC-coupling caps and removing the high-pass from the signal chain is all that is needed to allow a A>D converter to capture DC.

Anyone know of a company that already deals with this? Anyone know of a small-enough (i.e. personable) converter company that would be willing to mod a unit in-house, instead of me hacking away at SMD components and ruining something before I can make one measurement? Anyone know of a reliable way to perform the aforementioned operation without too much danger?

Thanks for the help!
Andrew

 

[bcarso]

One of the problems with having a company modify a piece of equipment is that they are married to it indefinitely thereafter. In fact, even if they merely advise what capacitors, etc. to remove---assuming the resulting circuit is practical in terms of freedom from excessive drift and offset etc.---they have opened a line of communication that the customer will probably use indefinitely. Such open-ended consultation is a direct and opportunity cost that few companies can afford, much as some might like to help.

Unless you can either find an existing system that fits your needs, or persuade a manufacturer that there is a significant market for one, I suspect that you are out of luck.

However: you could encode your d.c. signals as a.c. signals in various ways and use existing hardware. Voltage-to-frequency converters used to be quite popular, among other things, for allowing simple isolation using transformers or optoisolators. Most of them provided pulse outputs which might not be optimal for delta-sigma A/D's, but some conditioning could be applied.

Then, on the back end, use either a hardware freq-to-voltage converter, or extract the information in software.

One would need to know the range of voltages to be encoded and the requirements for frequency response, noise, and drift, to get started. You would encode with an offset so that for the smallest input the frequency would still be comfortably greater than zero.

 

[Crusty2]

One of my first jobs was modifying S*ony PCM-702's at UCLA to do this back in the '80's. A little scary voiding the warranty on a brand new $5,000 box, but data aquisition systems capable of recording D.C. started at around $30k.

It involved D.C. coupling the A/D convertor and tapping the data/word clocks out to a new wire wrapped card to interface to the computer (also carrying dual D/A convertors) and writing some software. Pretty simple really.

The researcher who came up with it never had help from So*ny; he used a service manual to figure it out. The app. engineers at B*urr Brow*n pitched in though.

A small company took up the guantlet and started doing the mod and reselling them to other researchers.

I wonder if you could find a A/D chip manufacturers "evaluation board" that you could easily modify? It would already have the support stuff sorted for you.

 

[gyraf]

Observe, that many modern-day audio-aimed A/D converters has built-in digital high-pass filters, cancelling out DC components..

e.g. the AL1101 ADAT A/D, the built-in digital HPF is at 2.5Hz:

http://www.profusionplc.com/static/images/data%20sheets/al1101g.pdf

Jakob E.

 

[Samuel Groner]

You might try www.weiss.ch. I believe that the ADC2 is DC-coupled already, but I might be wrong.

Samuel

 

[burdij]

If your frequency response requirements for the analog values is not too high, you could convert the analog signal to FM or PCM before running it into the standard audio card.

 

[OddHarmonic]

Bcarso,

I've experienced that with companies, and I do understand their unwillingness to dive deep into the muck we are looking at. A few years back when we were working on a customized Pocket PC to do field data collection, we tried to get Compaq involved. They gave us a pointer or two, and dissappeared...

I'm looking at the best solution at this point being some sort of modulation circuit to get the data into an AC signal as you said. However, I don't have the time at work right now to be DIYing such a box, and troubleshooting it. I've begun searching for something that already exists, and I can't imagine that in a world this large, someone doesn't make one, but I haven't found anything yet. Any pointers?

Crusty,

That is an interesting project. DAQ systems have come down considerably in price since then! (Cheap USB WinDaq systems are less than 100 bucks) I just wish the worlds of DAQ and audio weren't so splintered. As I alluded to above, although a mod to existing hardware OR building something from scratch is possible, I have little experience with digital circuits, and not a lot of free time at work to devote to such a massive project. Frustrating, it is.

Interesting note, we have an old Sony PC-108M, which is an eight channel DAT recorder built for this exact purpose. It has a digital output, but the data rides on a 25-pin D-SUB cable, and Sony never developed any specific protocal for it. THe manual gives the pin outs, but I would have to develop a box that took that data and streamed it into something like AES/EBU to pair up with the audio data coming in. Again, too much time...

Jakob,

Thanks for the note, I didn't know that chips were building that in for audio use, and the datasheet doesn't allude to any workarounds to skip the filter, so I'm guessing that many audio converter companies will from now on be held to such limits.

Samuel,

I checked out Weiss, no mention of DC coupling that I can see, but I'll be digging more.

RF,

We have used a lot of WinDAQ stuff in the past, but I've been wandering NI stuff as well now, thanks.

Everyone, thanks for the responses, I wish I could devote the DIY time to develop a solution, but other duties would sink if I attempted that. I feel that I HAVE to find a solution that is available on the market, or stick with KAY Elemetrics (groan...)

Andrew

 

[OddHarmonic]

[quote author="burdij"]If your frequency response requirements for the analog values is not too high, you could convert the analog signal to FM or PCM before running it into the standard audio card.[/quote]

Our Freq. requirements aren't that high. Any idea who might make such a box that will do the FM/PCM conversion? (standalone preferrable)

Thanks,
Andrew

 

[Samuel Groner]

I checked out Weiss, no mention of DC coupling that I can see, but I'll be digging more.

Just e-mail Daniel Weiss and say hello from me... I believe that they would be able to make a custom one if the standard unit is not DC-coupled already. At least if you don't need 24 bit DC precision!

Samuel

 

[gyraf]

Any idea who might make such a box that will do the FM/PCM conversion? (standalone preferrable)

A simple Voltage/frequency and frequency/voltage converter system could be used for this - Bruel&Kjaer often did this for sub-sonic measurement storage on tape recorders.

There are V/F-converters on-a-chip out there from several suppliers (I don't remember parts numbers right now) - and used with a Phase-Locked-Loop, you can have a frequency-to-voltage converter. There are some very good (National?) application notes on the 4046 C-mos PLL.

All depends on what DC precision and bandwidth you really want and need - but with a system like this you could use any kind of audio in/out system..

Jakob E.

 

[bcarso]

[quote author="OddHarmonic"]Bcarso,

I've experienced that with companies, and I do understand their unwillingness to dive deep into the muck we are looking at. A few years back when we were working on a customized Pocket PC to do field data collection, we tried to get Compaq involved. They gave us a pointer or two, and dissappeared...

I'm looking at the best solution at this point being some sort of modulation circuit to get the data into an AC signal as you said. However, I don't have the time at work right now to be DIYing such a box, and troubleshooting it. I've begun searching for something that already exists, and I can't imagine that in a world this large, someone doesn't make one, but I haven't found anything yet. Any pointers? ...snip....

Andrew[/quote]

Comp*q! Hah! You were lucky you got to talk to a real person or to get a return phone call! But don't get me started... Better trying to talk to them than Dhell.

As far as a commercial modulator system, I'm not immediately aware of such, but the process control industry may well have what you need. I agree that it would be folly to roll your own just yet at least. How about stating your requirements----bandwidth, noise/resolution, dynamic range, polarity (bipolar or unipolar), input impedance and type (differential or single-ended), etc.?

What is wrong with K*y Elemetrics by the way? Performance, cost, bad breath?? :razz:

 

[burdij]

National makes the LM331 which might work and has a low parts count (i.e. protoboard level). Their example shows a 50KHz circuit. You might want to use a lower high end frequency, top end of maybe 10KHz just to reduce the processor loading of the computer and to stay in the audio range.

Here is a data sheet:

http://cache.national.com/ds/LM/LM231.pdf

 

[bcarso]

[quote author="gyraf"]

Any idea who might make such a box that will do the FM/PCM conversion? (standalone preferrable)

A simple Voltage/frequency and frequency/voltage converter system could be used for this - Bruel&Kjaer often did this for sub-sonic measurement storage on tape recorders.

There are V/F-converters on-a-chip out there from several suppliers (I don't remember parts numbers right now) - and used with a Phase-Locked-Loop, you can have a frequency-to-voltage converter. There are some very good (National?) application notes on the 4046 C-mos PLL.

All depends on what DC precision and bandwidth you really want and need - but with a system like this you could use any kind of audio in/out system..

Jakob E.[/quote]

That's of course what I recommended at the outset---but he wants something truly off-the-shelf. Pease has written about V-F's extensively, and Nation*l apparently still makes the LM231/331 V-F chip. But it will still require an extensive design effort to come up with a working system.

Also the output of the Nation*l part is a pulse train---it would be nice if it were more monochromatic.

EDIT: Thanks burdij---you beat me to it on the P/N.

 

[OddHarmonic]

[quote author="bcarso"]
Comp*q! Hah! You were lucky you got to talk to a real person or to get a return phone call! But don't get me started... Better trying to talk to them than Dhell.

As far as a commercial modulator system, I'm not immediately aware of such, but the process control industry may well have what you need. I agree that it would be folly to roll your own just yet at least. How about stating your requirements----bandwidth, noise/resolution, dynamic range, polarity (bipolar or unipolar), input impedance and type (differential or single-ended), etc.?

What is wrong with K*y Elemetrics by the way? Performance, cost, bad breath?? :razz:[/quote]

I found it a miracle that we got an engineer's ear, but he sure didn't care what we had to say. They do have a development program where you can get access to schematics, pinouts, etc... but you are usually on your own after that. Unless they see some profit in it, I guess.

So, in my whirlwind of a day, I didn't consider one important fact about the data collection we are using. The bandwidth I need for this unit is from DC-3.4 KHz (Let's call it 4K) I feel that derails the F-V converter, as anything above DC would be lost in the shuffle.

Just for the record, here are the stats on the signal conditioner I'm using with the two transducers:

Glottel Enterprises MS-100A2
DC-3.4 KHz
SNR: 40-60 dB (depending on transducer being used)
Output Impedance: 100 R
Single-Ended, on BNCs

I found an interesting article about using a VCO to encode the low freq data into a midband sine. Would this be effective?

http://www.elecdesign.com/Articles/Index.cfm?AD=1&AD=1&ArticleID=2641

Thanks,
Andrew

 

[gyraf]

I feel that derails the F-V converter, as anything above DC would be lost in the shuffle.

Nope, you should be able to easily frequency-modulate a DC-to-4KHz onto a standard digi-audio track.. Think of it as FM-transmitting in audible range..

Jakob E.

 

[burdij]

3.5KHz modulation bandwidth shouldn't be a problem going into a regular audio card. As I recall, the energy of the modulation will be concentrated in an infinite series of harmonics of the modulation frequency + or - the carrier frequency for FM. As long as you pass a few of these through to the audio card, you should get a fairly accurate representation of the modulation.

The XR2206 is a chip that is or was used fairly extensively in analog electronic synthesizers as a voltage controlled oscillator. It produces a fairly clean sine output which might simplify the subsequent conversion to digital.

 

[bcarso]

Agree with Jacob and burdij. Put the carrier up in the 15kHz range and you should be able to handle 3.4kHz signal bandwidth easily. Any much higher and you may run afoul of phasey things in the soundcard antialiasing filter. Also, you will want to know what the filter characteristics of the signal conditioner are, since out-of-band aliasing-type effects will occur in the modulator if you get significant energy above the passband.

The use of a triangle-wave VCO which is shaped to sine, as described in the ED article, is not a bad approach. I've often tempted to describe the freebie designs in those magazines as worth about what you pay for them, but this appears to be an exception. However, you are still getting into the construction business, which was not what you are after IIRC.

Also, you have only one end of the chain in a sense, as you need to address decoding your data. This is made easier by the relatively modest accuracy requirements and known zero signal frequency. You may find some packaged lab math functions with a bit of additional code to be quite adequate (DFT with suitable windowing, track the peak, etc.). Or look at zero-crossings and run a software counter, etc., and invert for the frequency. You could also calibrate out your modulator errors with some work, to the extent that they are not drifting.

 

[mhelin]

Instead of going the V/F converter route you could also build an industrial (non-audio) A/D converter and interface it with a S/PDIF transimitter. That's a lot more work though, but you wouldn't need to decode the signal after conversion.

 

[OddHarmonic]

Just a few updates and more questions, thanks again for the many and varied responses! After googling far deeper then I had ever gone (40+ pages into a search!), finding myself mysteriously on a bunch of equipment manuf. mailing lists that I have no interest in, and even getting a couple cold-calls from said companies, I still have not found an on-market solution to this problem. I refuse to believe that someone out there doesn't sell a solution in a box....

Since that is so far unsuccessful, I have thrown together a PCB layout for the VCO schematic I posted earlier. I have never needed to design my own PCB, so I would appreciate any comments about how it looks. Be gentle, this is my first go at it. A few notes, I used ExpressPCB software, and although this is a 2-layer board, I neglected to use two layers. Is it worth pulling the Ground/Power lines onto their own layer? If so, which layer, top or bottom?

Also, a few questions about the circuit:

Will standard film caps work for everything non-polarized, and standard Electrolitic for the rest?

Thanks everyone,
Andrew

 

[bcarso]

It's interesting that you were able to route entirely on one layer---you may be a natural at this!

Typically ground and possibly power would be on the component side and most of the routing on the bottom. In addition area fills wherever with ground plane can sometimes help noise issues.