How to record two signals simultaneous(by using ADC) ?

[samaaa]

Hi:

I need to record a two analog signals, by using an ADC(of 20 bits resolution or higher), the two signals are very weak signals of signal to noise ratio about -30 db, so the two signals are very effective by the noise.

the problem is : how to record these two signals at the same time(simultaneous )?
If i will use multiple channel ADC, then the two signals will be sampled not at the same time exactly, because the ADC of multiple channels use the multiplexing to sampling multiple channels.

so, how can i record(by using ADC) the two signals at the same time exactly??

 

[UltrafastPED]

You need "simultaneous sample and hold" circuits; and since you need 20 bit resolution they will be pricey.

Here is a 16 bit simultaneous sample and hold with 3 differential channels:
http://www.ti.com/lit/ds/sbas362c/sbas362c.pdf

Contact TI or Analog Devices directly for advice on a 20 bit system.

 

[samaaa]

You need "simultaneous sample and hold" circuits; and since you need 20 bit resolution they will be pricey.

Here is a 16 bit simultaneous sample and hold with 3 differential channels:
http://www.ti.com/lit/ds/sbas362c/sbas362c.pdf

Contact TI or Analog Devices directly for advice on a 20 bit system.

OK, if ADC "simultaneous sample and hold" of 20 bits is unavailable:

can i use two ADC chips, to record the two signals?
can i ensure that the two signals will be sampled in the same time exactly?

 

[UltrafastPED]

Did you read the TI spec sheet? Is it clear to you?

 

[meBigGuy]

You need to define "same time exactly" since true "exactly" is impossible.

Two discrete A/D converters will vary based on process. Two A/Ds on a single chip will be better matched. What frequency are you sampling at, and what timing jitter and offset can you tolerate? Note that even single A/Ds have inherent timing jitter and aperature uncertainty. For example the TI data sheet referenced has 150ps of delay and jitter matching.

Also, how are you affected by gain error and offset matching? What about matching in your anti-alias filters?

 

[samaaa]

Did you read the TI spec sheet? Is it clear to you?

yes i read it, i understand that this chip use one ADC for each channel

You need to define "same time exactly" since true "exactly" is impossible.

i am not sure, but may be i need about 1 micro second difference between sampling the two signals

What frequency are you sampling at

not less than 200 Hz.

what timing jitter and offset can you tolerate? Note that even single A/Ds have inherent timing jitter and aperature uncertainty. For example the TI data sheet referenced has 150ps of delay and jitter matching.

Also, how are you affected by gain error and offset matching?

i will search for these questions

 

[meBigGuy]

1 micro second is a long time. But, 20 bits is a lot of bits. Are you sure you need 20 bits. Can you explain your rationale for choosing 20 bits.

Here is a 20 bit converter. http://www.ti.com/lit/ds/symlink/ddc101.pdf I'm not sure how well matched 2 in parallel would be since I didn't fully read the acqusition description yet. Maybe acquisition can be matched. Look at it and try to fully understand how it acquires and produces a result. Also look at some other 20 bit converters. Sorry I can't give you more detail just yet.

 

[berkeman]

Hi:

I need to record a two analog signals, by using an ADC(of 20 bits resolution or higher), the two signals are very weak signals of signal to noise ratio about -30 db, so the two signals are very effective by the noise.

the problem is : how to record these two signals at the same time(simultaneous )?
If i will use multiple channel ADC, then the two signals will be sampled not at the same time exactly, because the ADC of multiple channels use the multiplexing to sampling multiple channels.

so, how can i record(by using ADC) the two signals at the same time exactly??

Why are the signals weak going into your ADC system? Aren't you amplifying and filtering them? With a sample rate of 200Hz, you can put a pretty strong LPF on the front end. What breakpoint are you using for your anti-alias filter? What polynomial are you using, and what order filter? You should probably be picking your polynomial to minimize delay skew between the two channels, and you may need to hand-screen the components (especially caps) to match the delay better...

 

[samaaa]

1 micro second is a long time. But, 20 bits is a lot of bits. Are you sure you need 20 bits. Can you explain your rationale for choosing 20 bits.

Why are the signals weak going into your ADC system? Aren't you amplifying and filtering them? With a sample rate of 200Hz, you can put a pretty strong LPF on the front end. What breakpoint are you using for your anti-alias filter? What polynomial are you using, and what order filter? You should probably be picking your polynomial to minimize delay skew between the two channels, and you may need to hand-screen the components (especially caps) to match the delay better...

ok, let me explain:

ECG signal is obtained by using preamplifier as figure below:

i want to record ECG signal but without using preamplifier, first i will record the two signals come from right and left arms, then inside the computer(by using matlab) i will obtain the ECG signal mathematically "ECG=Left arm-right arm", as figure below:

so i need ADC of a high resolution because the ECG signal is very weak (about 1 mV),
also i need to record the two signals (come from right and left arms ) at the same time to make the subtraction operation inside the Matlab.

i can't use amplifier before the ADC because the noise is very large (about 0.6 to 1 V P.P)

 

[AlephZero]

There was a previous thread about this. https://www.physicsforums.com/showthread.php?t=707046

But you seem intent on doing it the hard way, instead of filtering out the narrow band noise before you digitize the signals.

 

[analogdesign]

i can't use amplifier before the ADC because the noise is very large (about 0.6 to 1 V P.P)

Can you use a very narrow band-pass filter to reject most of this noise (kind of like a preselector in an RF front-end?). Then, after you've knocked the noise down, you CAN use a preamp and make your life MUCH easier.

 

[meBigGuy]

I'll just accept the fact that you want to record raw data (for now). A/D converters require anti-alias filters. Do you understand the aliasing issues. Frequencies higher than 1/2 your sample rate fold back into your signal bandwidth. You need to remove the out-of-band noise.

regarding a 20 bit converter:
Assuming 1V p-p 20 bits is 1uV. Since your signal is 1mV you can use fewer bits. Especially if you process it to remove the noise. In theory you can recover signals that are less than the step size of the converter if the input has adequate noise (dither). I'm guessing 20 bits is way overkill. 16 bits would give you 15uV steps.

If, in the process of anti-alias filtering, you reduce the noise and add gain, you will be able to use much fewer bits.

 

[samaaa]

But you seem intent on doing it the hard way, instead of filtering out the narrow band noise before you digitize the signals.

i know that i am using the hard way, i do that because my supervisor want me to do that.

Can you use a very narrow band-pass filter to reject most of this noise (kind of like a preselector in an RF front-end?). Then, after you've knocked the noise down, you CAN use a preamp and make your life MUCH easier.

the main noise is come from electrical 50 Hz, even when i removed it, the noise still large value(about 0.1 to 0.2 V)

I'll just accept the fact that you want to record raw data (for now).

yes i want to record a row data.

A/D converters require anti-alias filters. Do you understand the aliasing issues. Frequencies higher than 1/2 your sample rate fold back into your signal bandwidth. You need to remove the out-of-band noise.

yes i understand the aliasing issues , i will use anti-alias filter before the two ADC.

Since your signal is 1mV you can use fewer bits. Especially if you process it to remove the noise. In theory you can recover signals that are less than the step size of the converter if the input has adequate noise (dither). I'm guessing 20 bits is way overkill. 16 bits would give you 15uV steps.

the main ECG signal is about 1 mV, but the ECG signal has important features(components) of about 0.05 to 0.1 mV
so , can i record this signal by using 16 bits ADC, and then recover it efficiently ?

 

[marcusl]

One reason you are swamped by noise is that you have no ground reference. You need a ground reference electrode and a high-impedance instrumentation amplifier. More sophisticated systems that drive the body reference electrode are also available, but start simple. Suggest you read one of the many good books on biomedical instrumentation to understand how to do it before building anything.

Since you are dealing with human subjects, safety is a paramount issue. It is standard practice in medical electronics to absolutely isolate the circuits that contact a person from any and all equipment that connects to the power mains. You can do this by running your front end preamps/filters/samplers from batteries and transmitting the digital signals to your computer via opto-isolators. There are also high isolation power supplies available that are human-safety-rated. This must be taken seriously. Again there is much information available on how to do it.

 

[berkeman]

i know that i am using the hard way, i do that because my supervisor want me to do that.

One reason you are swamped by noise is that you have no ground reference. You need a ground reference electrode and a high-impedance instrumentation amplifier. More sophisticated systems that drive the body reference electrode are also available, but start simple. Suggest you read one of the many good books on biomedical instrumentation to understand how to do it before building anything.

Since you are dealing with human subjects, safety is a paramount issue. It is standard practice in medical electronics to absolutely isolate the circuits that contact a person from any and all equipment that connects to the power mains. You can do this by running your front end preamps/filters/samplers from batteries and transmitting the digital signals to your computer via opto-isolators. There are also high isolation power supplies available that are human-safety-rated. This must be taken seriously. Again there is much information available on how to do it.

@samaa -- marcusl's post is excellent. I suggest that you and your supervisor talk it over in detail. If your supervisor does not understand anything that marcusl has said, that is very worrisome...

 

[meBigGuy]

I concur with the safety points that have been raised. That is actually the major design issue. The amplifiers, digitizers and post processing issues are actually pretty simple. The safety issues are difficult, and very important.

 

[samaaa]

thank you guys :)