Seeking advice on selecting a low-frequency, multi-channel, A/D converter to interface with TI DSP

Hi Walter,

Take a look at the ADS7951, ADS7952 or ADS7953 - these are 8/12/16 channel 12-bit devices that can be placed into an auto scan mode.  They also provide data back to you in 16-bit format where the 4MSBs contain channel address information.  The ADS1178 may interest you as well.  We'll look around to see what other parts might fit your application if you can provide a little more detail.  What is the nature of the signals you are trying to sample?  Do you have a preference in converter type (delta-sigma or SAR)?

Thanks Tom. 

That 16 channel part is looking interesting.  Is there perhaps a 14-bit or 16-bit version available? 

I need to measure DC inputs (down to zero Hertz) and up to perhaps 50 Samples per second per channel -- which is not fast.  I can sacrifice the speed in favor of resolution, accuracy, and 16-channels of data. I don't currently prefer one type of converter over another -- either delta-sigma or SAR would work for me, I suspect.  Though I'm not as familiar yet with the delta-sigma converters -- I presume they can make DC measurements. ???  Indeed, isn't the difference between these two types of little consequence to me, the user?  Or should I be aware of some peculiarity of the delta-sigma converters? 

Also, I see the ADS7953 is available in an EVM, which is helpful to me.  I need to use its SPI port to interface directly to a TI DSP TMS320C6748EVM board.  I presume that is straightforward to do.  Is that correct? 

The schematic for the A/D EVM board shows a DC connection from the inputs to the A/D converter.  I presume the EVM system as a whole can handle DC measurements.  ???

And is the ADS7953EVM low enough in power that it can operate from the 5Volts delivered from the DSP-EVM board?  So I don't need a separate power supply? 

I hope the DSP software will be low-overhead.  If this approach meets with your blessing, I'll pursue a deeper understanding the SPI interface to the DSP.

Thanks again for your help.

ADDED NOTE:  I'm looking at the TMS320 Cross-Platform Daughtercard Specification, Revision 1.0, spra711.pdf, which seems like an exciting way to easily interconnect various evaluation modules with a TMS320 DSP board.  That is very interesting to me, yet it appears this spec does not apply to my EVM board -- the LogicPD TMS320C6748EVM board.  Is that correct?  Is there some easy way to interconnect the A/D converter EVMS with my DSP-EVM? 

Hi Walter -

At the moment, there is no 14 or 16-bit version of the ADS795x parts.  Both the SAR and the delta-sigma can be used for DC measurements.  Here is a quick read on the differences between the two technologies Choosing SAR vs high-speed delta-sigma ADCs

Yes - there is an EVM for the ADS7953.  It has male/female connectors on the top/bottom side of the board for the analog input, the digital control lines and the power.  It should be relatively easy for you to connect into your TMS320C6748 via jumper wires (keep them short!).  Unfortunately your LogicPD kit is not going to be compatible with the SPRA711 document so it won't be a simple as just plugging the two boards together. 

The EVM can deal with DC inputs and the 5V requirement is on the order of a few millamps.  There are two supplies defined on the EVM - one 5V for the analog side of things, and 3.3 or 5V for the digital.  You could use the same 5V supply for both - just beware that the power from your LogicPD board may be a little noisy which can have an effect on the performance.

Tom,

You article on Choosing SAR vs high-speed delta-sigma ADCs was helpful, thank you. It sounds like a delta-sigma converter would work just fine for my application.  I don't see an EVM board listed for the ADS1174 or ADS1178 parts.  Is that correct?

Do I correctly understand that the ADS7953 needs its inputs to see source impedances of <= 50 ohms?  (ads7953.pdf, page 40, Figure 57) In my application that would require external buffers on each of the inputs.  Eight to sixteen buffers!  Ugh!  Or is there an acceptable way to put the buffer BETWEEN the multiplexer and the A/D converter on the chip?  Can that be done on the EVM board, using the on-board buffer? I suppose that solution would require extra settling time after the multiplexer switches inputs but before the A/D conversion begins. That could be a considerable change in timing -- is that configurable on the chip? 

Walter,

Glad you found the article helpful!  Take a look at the ADS7953 data sheet page 41.  If you have 16 DC inputs you want to sample, you could add a buffer between the MUX output and the ADC input - one of the nice features of this chip.  If you have similar input impedances on all channels, you can share a single buffer.  The MUX switches to the next channel in the conversion sequence while the current conversion is taking place so that you have ample settling time.  The EVM did not put the buffer between the MXO and AINP pins, so you would not be able to easily see that functionality on the board.  Maybe we'll change that in a future hardware revision...

There is an EVM for the ADS1178 - you can find details on that board here: http://focus.ti.com/docs/toolsw/folders/print/ads1178evm.html

Tom Hendrick said:

Take a look at the ADS7953 data sheet page 41.  If you have 16 DC inputs you want to sample, you could add a buffer between the MUX output and the ADC input - one of the nice features of this chip.  If you have similar input impedances on all channels, you can share a single buffer.  The MUX switches to the next channel in the conversion sequence while the current conversion is taking place so that you have ample settling time.  The EVM did not put the buffer between the MXO and AINP pins, so you would not be able to easily see that functionality on the board.  Maybe we'll change that in a future hardware revision...

I looked at the schematic for the EVM (slau262a.pdf, page 27).  It seems straightforward to use the on-board buffer in the manner you described above.  It would require:

• Break the on-board connection between MXO and AINP (by cutting the copper trace).  [NOTE: There is a typo on the schematic.  It says, "MXD", when it should say, "MXO".]  And solder a short wire to each side of the break.  This is the hard part! Can you comment on the do-ability of this?
• Remove the jumper from W4. (Easy!)  Then connect the previously mentioned two short wires to appropriate places on header W4.  (This shouldn't be too hard, since the header is made to accept a jumper.)

Other than the obvious CYA, "Never do that!  Because it voids your TI warranty.  Etcetera" -- could you comment on the wisdom of the above?  And, is the existing on-board buffer (and surrounding components) suitable for this purpose?  (Answer:  It seems to be.) 

Hi again Walter,

The ‘cut trace’ thing is not going to be so easy.  Pins 7 and 8 of the ADS79xxEVM essentially share a common track and one test point via.  You would have to lift one pin or the other and then wire them into W4.  The pitch on the chip is pretty fine, but it is doable - which ever pin you don't lift can use the via, so you would only have to solder a wire to one ADS7953 pin.  CH0 from P1.16 would always go through the AMP so you'd have to cut the input trace to (or lift and wire directly to) U1.3 as well.

Gerber files of the actual board are attached for your reference.  The added wires may degrade performance a bit, and there is always the 'situation normal' aspect to worry about regarding warranties and such.  Definately do-able though if you want to give it a try. 

I could use a bit more of your advice, if I may. (My eyeballs are getting weary from reading all the data sheet details, searching for the right converter.)

I found a converter recently -- the ADS8332 -- that has many desirable properties for me. (8 channels, 16-bit resolution, SPI interface, low-frequency [DC to 50 Samples per second per channel].) Then I find that its EVM board "requires" plus and minus ten volt power supplies (!) -- which leaves me out because I want to power it from the DSP-C6748-EVM board that has a 5-volt supply. 

Please clarify or suggest.  (And thanks again for your help.)

Hi Walter,

The ADS8332 is a 5V converter. The reason that the EVM requires +/-10V supplies is to power the op amp that we decided to use to drive the ADC input. On the EVM board we used the OPA211 with +/-10V supplies. The reason we decided on bipolar supplies and a bipolar op amp was so that the amplifier driving the ADC had the ability to reach the full range of the converter. We wanted to make sure the input could be true ground forcing us to use dual supplies. Built onto the EVM is a +5V regulator which uses the +10V supply to create +5V to power the ADC.

There are a fair amount of converters in our portfolio. I would say the first decision you would need to make is deciding what type of converter fits your application. Either using a delta sigma or maybe using a SAR converter. From there we can narrow down some choices for you. The converter to use will depend on your application. Maybe if your using the converter to sample low frequency or even DC signals, a delta-sigma may fit your application. Maybe look into using the ADS1256 or the ADS1158/1258 as options.

Regards,

Tony Calabria