ADC @ 16 bits

[Jean-François L]

Hello,

I actually have a ServoPod and a MiniPod. For a specific application, I really need to aquire 8 ADC channels at 16 bits and to do some calculations at 4096 Hz.

I know that the DSP56F807 has 16 ADC channels at 12 bits, but I would like to find a hardware that would fill my requirements.

Do you have any products or do you know of any products which I could use?

If not, what do you think about this solution:
- Use a ServoPod for the main calculations
- Use an independant 8 channels at 16 bits ADC module
- Communicate with that module by using for example the CAN or SPI protocol.

Thank you,

[RMDumse]

Yes, we have such a module, which has 8 16-bit A/D's on it, that communicate via the SPI module. The A/D's can convert at rates up to 250 kHz. And over all, we are very pleased with these A/D's performance. We've now used them on several projects and they have performed well in each application. I don't think we've added this module to our price list. Perhaps nmitech could put it on the list and provide a link.

[nmitech]

8-Channel,
http://www.newmicros.com/cgi-bin/store/order.cgi?form=prod_detail&part=NMIN-1864-8

4-channel,
http://www.newmicros.com/cgi-bin/store/order.cgi?form=prod_detail&part=NMIN-1864-4

1-channel,
http://www.newmicros.com/cgi-bin/store/order.cgi?form=prod_detail&part=NMIN-1864-1

[ngkdc]

I'm interested ... very interested. Right now, I'm using a ServoPod (because I need 16 A/D channels); this might be helpful. Major question though, how is the sample-to-sample accuracy? Assuming a 0.5 vollt input across all eight channels, what is the output result delta (from the ideal)? Right now, with the 12-bit A/D, I'm seeing anything from +/- 4 counts to +/- 40 counts (out of 4096) with a lab supply and no other processes running. I could really use the additional accuracy, as long as it isn't buried in conversion noise.

Another question, can these boards be daisy chained on the SPI bus? If I really wanted 16 channels of 16-bit A/D, could I easily do this? I'm assuming I'd lose the use of PE2 for the CANBus; that doesn't really bother me in this application. I really want to avoid adding any additional interface boards if I can help it ... I guess I could multiplex as many A/D boards with an external address decoder ... but I'm trying to cut down my interface board requirements!

Lastly ... any plans for a +/- 5 volt bipolar input for this card? You give me 16 channels of +/- 5 volts and/or 0-5 volts, with less than 1 mV conversion noise and I'll be a very happy camper indeed! As it appears right now, 0-5 volts into 16-bit A/D should result in 76.3 uV per step. I'll handle the isolation amplifiers as needed (though Linear Technologies LT1990A would fit the board nicely).

*****

I know you all are busy making tomorrow happen today in your lab ... but how, other than shooting blind questions your way, can I find out what is available and/or under development? Eventually I see things on the web site, but it's far from a complete catalog.


Thanks much for all your hard efforts.


Rick

[nmitech]

Hi Rick,

The most number of error count differences that i have seen on
my NMIN-1864-8, 16-Bit A/D board is 20. By taking 100 readings, then I took a largest reading value subtracts to the smallest reading value and get the average error range of +/- 10 counts. Lower the input voltages produce lower the error counts. About +/- 4 counts for 1V input. The error counts are appeared to be consistent on all 8 channels.

The NMIN-1864-x can not be used for daisy chained on the SPI bus. Originally, this A/D board was designed for a specific custom application. If more than 8 channels are needed, it will require a new layout. We'll consider the daisy chained for future designs. Port PE2 is not used on the NMIN-1864 card, so it won't affect your can bus control signal.

No plans for a +/- 5 volt bipolar input for this card at this time, but we will add this to our wish list for future consideration.

[ngkdc]

Thanks for the information.

Regards,

Rick

[RMDumse]

Don't overlook the idea of having us make exactly what you want. About a third of our business comes from custom products. And perhaps another third from semicustoms. For custom products we make devices available only to the requesting customer. Of course, they pay full development cost, and production prices. Semi-custom is where we work in cooperation with a customer, who picks up some of the development cost, but gets to guide specs, and then we sell the to anyone at production prices.

I don't think it would take too much to make you a 16-ch x 16-bit A/D card. Bipolar would be a little trickier, requiring an analog front end ahead of this converter, but it is still quite doable.

The best way to know what is happening in our design center, is to work with us on "our" next design.

Talk to us if the idea is appealing.