Forums

Using a stereo audio ADC for baseband sampling.

Started by Eric Jacobsen September 23, 2015
This may ultimately be a dumb question, but I need to ask it.

I'm looking at using a stereo audio ADC as an I/Q baseband sampler in
a low-rate communication system.   One like this:

http://www.ti.com/product/tlv320adc3001

I seem to recall some horror stories of using audio ADCs for general
use, but I can't really seem to see any reason to not go this way.

Am I missing something?   This does not appear to do any curve
compression or anything like that, and I don't really need near as
many bits as it puts out.   Thought I'd query the brain trust to see
if I'm missing anything.




Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com
eric.jacobsen@ieee.org (Eric Jacobsen) writes:

> This may ultimately be a dumb question, but I need to ask it. > > I'm looking at using a stereo audio ADC as an I/Q baseband sampler in > a low-rate communication system. One like this: > > http://www.ti.com/product/tlv320adc3001 > > I seem to recall some horror stories of using audio ADCs for general > use, but I can't really seem to see any reason to not go this way. > > Am I missing something? This does not appear to do any curve > compression or anything like that, and I don't really need near as > many bits as it puts out. Thought I'd query the brain trust to see > if I'm missing anything.
Does it respond down to DC? -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
On 9/23/2015 10:43 AM, Eric Jacobsen wrote:
> This may ultimately be a dumb question, but I need to ask it. > > I'm looking at using a stereo audio ADC as an I/Q baseband sampler in > a low-rate communication system. One like this: > > http://www.ti.com/product/tlv320adc3001 > > I seem to recall some horror stories of using audio ADCs for general > use, but I can't really seem to see any reason to not go this way. > > Am I missing something? This does not appear to do any curve > compression or anything like that, and I don't really need near as > many bits as it puts out. Thought I'd query the brain trust to see > if I'm missing anything.
I'm not sure what you mean by "general" use. ADCs are ADCs. There are a number of types. "Audio" is not really an ADC type, rather it implies a frequency range. Often they are SD type converters... is that what you mean? They can also be successive approximation. What sorts of horror stories have you heard? BTW, the part you selected is not just an ADC, it is a DSP with a stereo ADC. This one sounds a lot like the line of devices for hearing aids and similar apps where power and size are major concerns. They used to only be programmable using a graphic package rather than coding, but it looks like they fixed that and this device can be assembly coded. I assume you want the DSP as much as the ADC? -- Rick
On Wed, 23 Sep 2015 11:11:53 -0400, Randy Yates wrote:

> eric.jacobsen@ieee.org (Eric Jacobsen) writes: > >> This may ultimately be a dumb question, but I need to ask it. >> >> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in a >> low-rate communication system. One like this: >> >> http://www.ti.com/product/tlv320adc3001 >> >> I seem to recall some horror stories of using audio ADCs for general >> use, but I can't really seem to see any reason to not go this way. >> >> Am I missing something? This does not appear to do any curve >> compression or anything like that, and I don't really need near as many >> bits as it puts out. Thought I'd query the brain trust to see if I'm >> missing anything. > > Does it respond down to DC?
From the datasheet: "Independent digital highpass filters ... can be disabled entirely." Other issues to consider: - 1/f noise - DC accuracy - DC drift with temperature - DC offset may change with PGA setting All of those would normally be taken care of by the HPF, which (thanks to Murphy) is going to be disabled for this application. It shouldn't be a problem if there is some way of compensating for the DC offset / low frequency drift downstream in the demodulator. Regards, Allan
On 23.09.15 16.43, Eric Jacobsen wrote:
> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in > a low-rate communication system. One like this:
There is nothing wrong when using audio ADC for other purposes, as long as you stay in the desired frequency range, usually 20..20k.
> I seem to recall some horror stories of using audio ADCs for general > use, but I can't really seem to see any reason to not go this way.
I made high precision measurements with conventional 48kHz PC sound devices. (eg. 10ns delay or capacity measurements with 0,1%) It is simply a question of the symmetries in your setup. You cannot expect the response to be absolutely flat or even have high absolute precision. Furthermore you should not rely on the low frequency or DC response. As long as you stay within this limits, you get some pretty accurate results. Maybe you need some calibration to compensate for differences between the two channels. Marcel
On Wed, 23 Sep 2015 15:44:51 +0000, Allan Herriman wrote:

> On Wed, 23 Sep 2015 11:11:53 -0400, Randy Yates wrote: > >> eric.jacobsen@ieee.org (Eric Jacobsen) writes: >> >>> This may ultimately be a dumb question, but I need to ask it. >>> >>> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in >>> a low-rate communication system. One like this: >>> >>> http://www.ti.com/product/tlv320adc3001 >>> >>> I seem to recall some horror stories of using audio ADCs for general >>> use, but I can't really seem to see any reason to not go this way. >>> >>> Am I missing something? This does not appear to do any curve >>> compression or anything like that, and I don't really need near as >>> many bits as it puts out. Thought I'd query the brain trust to see >>> if I'm missing anything. >> >> Does it respond down to DC? > > From the datasheet: > > "Independent digital highpass filters ... can be disabled entirely." > > Other issues to consider: > - 1/f noise - DC accuracy - DC drift with temperature - DC offset may > change with PGA setting > > All of those would normally be taken care of by the HPF, which (thanks > to Murphy) is going to be disabled for this application. > > It shouldn't be a problem if there is some way of compensating for the > DC offset / low frequency drift downstream in the demodulator.
- The ADC will be designed to work well over a limited range of sample rates. Don't go outside the recommended range; instead decimate the output if you only need a lower sample rate. - I read the datasheet for that part. It has registers for everything, but I couldn't find one to turn off the HPF. Possibly this is done in one of the miniDSP blocks, and you just need to program it correctly. - It does have a register bit to disable the AGC. I guess you will need to have the AGC (designed for voice) disabled for your application. - It also has a register to provide a DC offset independently to the L and R inputs. 15mV steps, +/- 105mV range. That may or may not be handy for your application, perhaps to calibrate out the DC offset on the quadrature downconverter. - Depending on the source and how good you are at PCB layout, you might get better results with an ADC with differential inputs. This device can't do that. Regards, Allan
On Wed, 23 Sep 2015 16:31:51 +0000, Allan Herriman wrote:

> On Wed, 23 Sep 2015 15:44:51 +0000, Allan Herriman wrote: > >> On Wed, 23 Sep 2015 11:11:53 -0400, Randy Yates wrote: >> >>> eric.jacobsen@ieee.org (Eric Jacobsen) writes: >>> >>>> This may ultimately be a dumb question, but I need to ask it. >>>> >>>> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in >>>> a low-rate communication system. One like this: >>>> >>>> http://www.ti.com/product/tlv320adc3001 >>>> >>>> I seem to recall some horror stories of using audio ADCs for general >>>> use, but I can't really seem to see any reason to not go this way. >>>> >>>> Am I missing something? This does not appear to do any curve >>>> compression or anything like that, and I don't really need near as >>>> many bits as it puts out. Thought I'd query the brain trust to see >>>> if I'm missing anything. >>> >>> Does it respond down to DC? >> >> From the datasheet: >> >> "Independent digital highpass filters ... can be disabled entirely." >> >> Other issues to consider: >> - 1/f noise - DC accuracy - DC drift with temperature - DC offset may >> change with PGA setting >> >> All of those would normally be taken care of by the HPF, which (thanks >> to Murphy) is going to be disabled for this application. >> >> It shouldn't be a problem if there is some way of compensating for the >> DC offset / low frequency drift downstream in the demodulator. > > > - The ADC will be designed to work well over a limited range of sample > rates. Don't go outside the recommended range; instead decimate the > output if you only need a lower sample rate. > > - I read the datasheet for that part. It has registers for everything, > but I couldn't find one to turn off the HPF. Possibly this is done in > one of the miniDSP blocks, and you just need to program it correctly. > > - It does have a register bit to disable the AGC. I guess you will need > to have the AGC (designed for voice) disabled for your application. > > - It also has a register to provide a DC offset independently to the L > and R inputs. 15mV steps, +/- 105mV range. That may or may not be > handy for your application, perhaps to calibrate out the DC offset on > the quadrature downconverter. > > - Depending on the source and how good you are at PCB layout, you might > get better results with an ADC with differential inputs. This device > can't do that.
Some more: - The input voltages need to be biased to +1.35V above ground to read zero. - The actual value of 1.35V isn't guaranteed. The datasheet does not give max and min values, only a typical voltage. I guess it's set from an untrimmed bandgap reference, so it's unlikely to be worse than +/-5%. Bad luck if you're doing something medical or aerospace or a field which requires worst case design. - The +1.35V voltage does not appear on a pin. You'll need to generate your own bias voltage for the inputs. The uncertainty in this voltage will also add to the DC errors. BTW, It might be possible to program the IN2L pin to output this voltage, but it will have a high impedance and need buffering. These are issues with this particular ADC, but not audio ADCs in general. As an example, I recently had a reason to use an audio ADC in a DC coupled noise measurement application, and chose one with differential inputs *and* and a common-mode voltage output pin. I think the part number was AK5385. It's a lot bigger than your TI part though. Regards, Allan
On 9/23/2015 1:16 PM, Allan Herriman wrote:
> On Wed, 23 Sep 2015 16:31:51 +0000, Allan Herriman wrote: > >> On Wed, 23 Sep 2015 15:44:51 +0000, Allan Herriman wrote: >> >>> On Wed, 23 Sep 2015 11:11:53 -0400, Randy Yates wrote: >>> >>>> eric.jacobsen@ieee.org (Eric Jacobsen) writes: >>>> >>>>> This may ultimately be a dumb question, but I need to ask it. >>>>> >>>>> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in >>>>> a low-rate communication system. One like this: >>>>> >>>>> http://www.ti.com/product/tlv320adc3001 >>>>> >>>>> I seem to recall some horror stories of using audio ADCs for general >>>>> use, but I can't really seem to see any reason to not go this way. >>>>> >>>>> Am I missing something? This does not appear to do any curve >>>>> compression or anything like that, and I don't really need near as >>>>> many bits as it puts out. Thought I'd query the brain trust to see >>>>> if I'm missing anything. >>>> >>>> Does it respond down to DC? >>> >>> From the datasheet: >>> >>> "Independent digital highpass filters ... can be disabled entirely." >>> >>> Other issues to consider: >>> - 1/f noise - DC accuracy - DC drift with temperature - DC offset may >>> change with PGA setting >>> >>> All of those would normally be taken care of by the HPF, which (thanks >>> to Murphy) is going to be disabled for this application. >>> >>> It shouldn't be a problem if there is some way of compensating for the >>> DC offset / low frequency drift downstream in the demodulator. >> >> >> - The ADC will be designed to work well over a limited range of sample >> rates. Don't go outside the recommended range; instead decimate the >> output if you only need a lower sample rate. >> >> - I read the datasheet for that part. It has registers for everything, >> but I couldn't find one to turn off the HPF. Possibly this is done in >> one of the miniDSP blocks, and you just need to program it correctly. >> >> - It does have a register bit to disable the AGC. I guess you will need >> to have the AGC (designed for voice) disabled for your application. >> >> - It also has a register to provide a DC offset independently to the L >> and R inputs. 15mV steps, +/- 105mV range. That may or may not be >> handy for your application, perhaps to calibrate out the DC offset on >> the quadrature downconverter. >> >> - Depending on the source and how good you are at PCB layout, you might >> get better results with an ADC with differential inputs. This device >> can't do that. > > > Some more: > > - The input voltages need to be biased to +1.35V above ground to read > zero. > > - The actual value of 1.35V isn't guaranteed. The datasheet does not > give max and min values, only a typical voltage. I guess it's set from > an untrimmed bandgap reference, so it's unlikely to be worse than +/-5%. > Bad luck if you're doing something medical or aerospace or a field which > requires worst case design. > > - The +1.35V voltage does not appear on a pin. You'll need to generate > your own bias voltage for the inputs. The uncertainty in this voltage > will also add to the DC errors. > BTW, It might be possible to program the IN2L pin to output this voltage, > but it will have a high impedance and need buffering. > > > These are issues with this particular ADC, but not audio ADCs in > general. As an example, I recently had a reason to use an audio ADC in a > DC coupled noise measurement application, and chose one with differential > inputs *and* and a common-mode voltage output pin. I think the part > number was AK5385. It's a lot bigger than your TI part though.
I use an AKM part in a product, AK4556. I find they work well and the one I picked is *very* simple to use, but you might not figure that out from reading the data sheet. It is hard to read. I don't remember if the HPF can be turned off or not, but I don't have a need for that. I add decoupling caps on the input and output so DC is not part of the I/O. -- Rick
Allan Herriman <allanherriman@hotmail.com> writes:

> On Wed, 23 Sep 2015 11:11:53 -0400, Randy Yates wrote: > >> eric.jacobsen@ieee.org (Eric Jacobsen) writes: >> >>> This may ultimately be a dumb question, but I need to ask it. >>> >>> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in a >>> low-rate communication system. One like this: >>> >>> http://www.ti.com/product/tlv320adc3001 >>> >>> I seem to recall some horror stories of using audio ADCs for general >>> use, but I can't really seem to see any reason to not go this way. >>> >>> Am I missing something? This does not appear to do any curve >>> compression or anything like that, and I don't really need near as many >>> bits as it puts out. Thought I'd query the brain trust to see if I'm >>> missing anything. >> >> Does it respond down to DC? > > From the datasheet: > > "Independent digital highpass filters ... can be disabled entirely." > > Other issues to consider: > - 1/f noise > - DC accuracy > - DC drift with temperature > - DC offset may change with PGA setting >
Yup! All good points to consider. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
On 23 Sep 2015 17:16:20 GMT, Allan Herriman
<allanherriman@hotmail.com> wrote:

>On Wed, 23 Sep 2015 16:31:51 +0000, Allan Herriman wrote: > >> On Wed, 23 Sep 2015 15:44:51 +0000, Allan Herriman wrote: >> >>> On Wed, 23 Sep 2015 11:11:53 -0400, Randy Yates wrote: >>> >>>> eric.jacobsen@ieee.org (Eric Jacobsen) writes: >>>> >>>>> This may ultimately be a dumb question, but I need to ask it. >>>>> >>>>> I'm looking at using a stereo audio ADC as an I/Q baseband sampler in >>>>> a low-rate communication system. One like this: >>>>> >>>>> http://www.ti.com/product/tlv320adc3001 >>>>> >>>>> I seem to recall some horror stories of using audio ADCs for general >>>>> use, but I can't really seem to see any reason to not go this way. >>>>> >>>>> Am I missing something? This does not appear to do any curve >>>>> compression or anything like that, and I don't really need near as >>>>> many bits as it puts out. Thought I'd query the brain trust to see >>>>> if I'm missing anything. >>>> >>>> Does it respond down to DC? >>> >>> From the datasheet: >>> >>> "Independent digital highpass filters ... can be disabled entirely." >>> >>> Other issues to consider: >>> - 1/f noise - DC accuracy - DC drift with temperature - DC offset may >>> change with PGA setting >>> >>> All of those would normally be taken care of by the HPF, which (thanks >>> to Murphy) is going to be disabled for this application. >>> >>> It shouldn't be a problem if there is some way of compensating for the >>> DC offset / low frequency drift downstream in the demodulator. >> >> >> - The ADC will be designed to work well over a limited range of sample >> rates. Don't go outside the recommended range; instead decimate the >> output if you only need a lower sample rate. >> >> - I read the datasheet for that part. It has registers for everything, >> but I couldn't find one to turn off the HPF. Possibly this is done in >> one of the miniDSP blocks, and you just need to program it correctly. >> >> - It does have a register bit to disable the AGC. I guess you will need >> to have the AGC (designed for voice) disabled for your application. >> >> - It also has a register to provide a DC offset independently to the L >> and R inputs. 15mV steps, +/- 105mV range. That may or may not be >> handy for your application, perhaps to calibrate out the DC offset on >> the quadrature downconverter. >> >> - Depending on the source and how good you are at PCB layout, you might >> get better results with an ADC with differential inputs. This device >> can't do that. > > >Some more: > >- The input voltages need to be biased to +1.35V above ground to read >zero. > >- The actual value of 1.35V isn't guaranteed. The datasheet does not >give max and min values, only a typical voltage. I guess it's set from >an untrimmed bandgap reference, so it's unlikely to be worse than +/-5%. >Bad luck if you're doing something medical or aerospace or a field which >requires worst case design. > >- The +1.35V voltage does not appear on a pin. You'll need to generate >your own bias voltage for the inputs. The uncertainty in this voltage >will also add to the DC errors. >BTW, It might be possible to program the IN2L pin to output this voltage, >but it will have a high impedance and need buffering. > > >These are issues with this particular ADC, but not audio ADCs in >general. As an example, I recently had a reason to use an audio ADC in a >DC coupled noise measurement application, and chose one with differential >inputs *and* and a common-mode voltage output pin. I think the part >number was AK5385. It's a lot bigger than your TI part though. > >Regards, >Allan
Thanks to everybody for the input, but Allan really hammered the important points here. It's these sorts of details that make the difference. The application is cost sensitive in both development and recurring cost, so this is a tough one. I owe Allan a beer. Or two. ;) Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com