This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

2Mhz differential ADC input with hardware FFT options

Other Parts Discussed in Thread: ADS9224REVM-PDK, ADS9224R

Hi,

I was not sure where to post this inquiry or who to ask locally. My local FAE no longer supports TI.

The project is essentially a 0-1Mhz spectrum analyser with smart FFT analysis functions that can give statistics about peaks (shape, SNR, mean amplitude over time, switching frequency etc) on each FFT frame over ~200mS of 2Mhz sampled data from a differential ADC input (12bit or better). We will likely need to do FFT’s on the time series over 200mS of data i.e. FFT on specific bins or all bins to detect what frequency each bin is switching at.

Some of these features can be dropped or reduced to keep cost under control. Ideally the processor / DSP will be under USD$10, max USD$20.

Specific requirements:
- 1 x Differential 10bit (10 usable bits) ADC sampling at minimum 2Mhz (really needs to be usable 12bit+)
- Most likely requires Hardware FFT support as we have to do minimum 1024 (preference 2048 point) with windowing on the 2Mhz data (Hanning etc), may also require FFT’s on each of the 1024/2048 bins for the 200mS of sample data
- Nice to have DMA on the ADC so the 2Mhz samples ADC fills a ping pong type ring buffer so while one buffer fills up we process the other buffer and enough RAM to take in 200mS of 2Mhz samples data which is 2 bit data (800KB on board ram, plus the requirements for the operating system etc so I imagine around 1MB). Or any other technique that allows the results of the FFT to happen on 2Mhz sampled data at minimum 1024-point FFT.
- External RAM for storing results of calculations and data for analysis after the ~200mS sample.

Not sure if a TI FAE can contact us directly to discuss? This project is based on a patent and has investment. We have already completed a proof of concept over the last 12 months using software defined radio hardware to get raw data samples and a low powered processor to do 512 point FFT’s at 500Khz so we have a pretty clear idea on what we need to achieve to go to stage 2 prototypes. 

  • Carl,
    thanks for posting your questions here. I have contacted our apps team and they will post our reply here. Sorry for the delay, I was traveling last week.
  • Carl, 

    I received the inquiry today and just read the description. Could you confirm functions to be supported by the processor SOC:

    1. Require integrated ADC to the processor SOC?

    2. Require smart FFT functions

    3. Require graphical display of spectrum statistics?

    Thanks

    Jian

  • 1. Yes, we would prefer differential 12bit ADC that can sample at 2Mhz+, it could be external, however it would need to have a total BOM cost around USD$20 with ADC and processor. Our front end is differntial interface.

    2. Not 100% but it the unit will do mostly FFT's so it will be optimal to have hardware FFT. We may be doing 1024/2048point FFT on the 2Mhz sample data with 50% ooverlapd, we may also want to do 64 point FFT on the same data etc

    3. No graphics required.

    Happy to arrange call if easier. I think the details are all in the orginal post of the minimum specification but happy to supply as much information as you need so we can select a solution that meets the cost requirments and processing requirments.

  • Also, I think we will also be sampling IQ which will require two differential ADC inputs both sampes at 2Mhz.

  • None of our intergrated ADC can support differential input and required sample rate. So we will need an external ADC. Is there particular interface you have in mind to bring the samples in? Also what FFT result need to be send out?
    A call may be easier to understand what has been assessed so far. Which time zone are you in?
    Jian
  • GMT+10 (Sydney Australia).

    We will send out custom data, it will be details of the waveform for another device to analyse.
    We have a custom front end / antenna.
    we can convert our differential to single ended output using an opamp.

    please email meeting to carl@edsengineering.com.au if that works, happy to send you a zoom hosted meeting, skype, phone call...
  • Hi,

    I have been trying to work on some options and have found that it is not likely that we will be able to use the internal ADC on the processors from TI (or anyone).

     

    We are sampling IQ and require simultaneous sampled ADC. So we are trying to also find a single ended or diff ADC that will do 16 effective bits @ 250Khz or so.

    We would then likely use an SPI interface into the processor to get the sample data and perform hardware FFT’s in the processor.

    Can you please help us select at least the processor that would allow us to do FFT (IQ sampled) at a reasonable speed (2 channels sampled at 1Mhz).

    If you could help finding an ADC that can do >= 16 effective bits simaltaneously sampled that would be great (must be SAR ADC with SPI or other high speed serial to get the data into the processor).

  • Carl,

    Please give me a day to look into external ADC options. Just to clarify potential alternatives based on your previous emails:
    Option 1: differential ADC, 10-12bit, 2MHz sampling rate;
    Option 2: dual ADC with simultaneous sampling, 10-12bit, 2MHz sampling rate.
    Additionally, need a single-ended 16-bit ADC @250KHz.

    If using external ADC, then data rate for both channels will be 48-56 Mbit/s. where I assumed 14bit for the 56Mbit/s.
    The 250KHz ADC can be on a separate SPI port.

    Please confirm above understandings.

    On the processor device, have you get a chance looking into the core benchmarks for FFT, at:
    www.ti.com/.../core-benchmarks.html
    we may also need to estimate required internal memory for your FFT design.

    Some of our Keystone devices have FFT coprocessor, but I am not sure if they will meet your cost target. you may take a look at:
    www.ti.com/.../66ak2l06.pdf
    as a starting point. The user guide for the FFTC is at:
    www.ti.com/.../sprugs2c.pdf

    let me know your thoughts on DSP vs. HWA while I look into ADCs.

    regards
    jian
  • Hi,

    For the ADC, we could do 1M sampling. we really need to get closer to 16 effective bits and it must be able to simaultanous sampling (IQ sampling)

  • You do release the part you spec'd is USD$2XX per chip...? Obviously we can just put the most expensive part and know we will meet requirments, but that is way above the budget we have of USD$20
  • I was thinking osmething more like this processor
    www.ti.com/.../description

    It can do 1Kfft's in about 25uS and cost USD$11

    I cant see much detail on the eval kit, the website wont let me order or see technical documentation, I have tried two computers on two networks and that page appears broken.
  • Carl,

    The link for 66AK2L was only to point you to the FFTC coprocessor. For the processor options, I am currently aftering:

    1. lower-end singlec-core DSP in the class of C6652 (there is no FFTC on this device so need to estimate if the FFT can sustain the speed):
    www.ti.com/.../
    2. A lower end processor with an FFTC
    3. An ARM core processor with integrated ADC - this device is beyond cost target and ADC may have a few issues.

    For the ADC, can you browse a bit on:
    www.ti.com/.../products.html
    one example may be:
    www.ti.com/.../ads9224r.pdf

    Assuming C6652 can do the realtime FFT, along with the above ADC, we are still around $33 based on pricing from ti.com. But let me know if these two devices are on the right direction.

    regards
    Jian
  • Thanks for your help, we also narrowed it down to the ADS9224R which has a perfect development board the ADS9224REVM-PDK has a reference design for everything we need behind our matching / front end.

    How do I find more information about the Precision Host Interface (PHI) on the ADS9224REVM-PDK? I have been searching for a couple of hours with no luck.

    For a prototype we can probably make out own dev board and use the Precision Host Interface (PHI) if we can find some more information on it. Like an API so we can talk directly to it, some details on the EEPROM on the host board (can we just put a blank on one our own board and program it with the Precision Host Interface (PHI) board / software? I have not been able to find much in the way of details of the board, how it works, and what i can do with it. It would be perfect for first cut of our prototype as it appears to have all the functions we need for a "quick test". If we could get those PHI to output some other signals it would be absolutely ideal i.e. if we can get a couple of outputs to clock at preset rates and talk to another SPI device, even without that, if we can just use it to read our own ADC board that would be great.

  • ADS9224REVM just arrived, we plugged it in and used the jumpers to short the inputs, and, left them floating and the best noise we can get is -10dbm! what could we be doing wrong, it shows -140dbm in the data sheet / user guides.

    I feel like i must have missed something obvious, i have looked at the data sheets, and looked at it with two other engineers and we cannot work out anything. I tried a laptop with no power supply connected, the best we can get with inputs grounded is around -10db

    I have been able to put a loop antenna in and see signals from my signal source, and i have been able to drop the noise floor by putting a large loop antenna to around -30db. But with both inputs grounded we should see close to the noise floor?

  • Carl,
    The ADS9224 is from a different product group. Can you open a new e2e post so it can be routed to the right team for bring-up support? Let me know if you have issues finding the right expert, once you have a new e2e post.
    jian