Topic: FMC107 reading problem  (Read 8234 times)

pieryves February 14, 2013, 07:18 PM

  • Member
  • *
  • Posts: 11
Hello,
I recently ordered a FMC107 board. I’ve been trying for 2 days to make a simple data acquisition. I refered to the « Getting Started Guide » section 8 to do so.
I injected signal  in the Adc0 (Labeled « A ») with these properties :
  • Type : Sine wave
  • Frequency : 100 KHz
  • Voltage : 0 to 1V (1Vpp centered on 0.5V)
  • Sampling rate : 65 Mhz

The data acquired seems to be corrupted, which can be downloaded here
Here are some informations I have noticed :
  • Data seems to be left shifted by 4 bits
  • Data is only made of these 4 characters:  0x5,  0x6,  0x9, 0xA
  • There is data in ADC1.xxx as well. (Only injected data in ADC0 channel)
  • By making a graph with the data data (using excel and .txt file), we can observe that the period seems good. (i.e. 650 sample per period)
Before going any further with the card, I’d like this application the give me good results. I’ve been trying to manipulate the data to understand what was happening, but I couldn’t figure what this behaviour is. It look like the data between the two channel of the ADC are mixed in some sort of way ; can’t figure how.
I would appreciate help from you. Maybe am I using the application the wrong way.

arnaudNL February 15, 2013, 05:52 AM (#1)

  • 4DSP Staff (EU)
  • Administrator
  • Member
  • *****
  • Posts: 7110
Dear Pierre-Yves,
Actually you should use a clock above 3MHz as there are transformers and you are likely to be filtered out, or in non linearity zone at 100kHz. You would need a DC coupled card to go under 3MHz
The data is indeed left aligned, on your previous post you were masking out the 4 MSB bits which is obviously not correct, you should shift the samples to the right and not remove 4 bits out of the 12 bits, looking at the graph on your old post, it looks to me as if you have missing bits.
I think you should go back to the initial software, don't tamper with samples and the number of bits. Actually the samples are left aligned because then you have a sign bit which is convenient, of course the samples have a bigger value but otherwise one would need to implement sign extension. Also leave matlab a side so far, it is too easy to wrongly represent the samples. I would initially recommend you to use a fantastic tool made by Analog Devices. This tool is called Visual Analog and is free of charge. I am attaching a Visual Analog project file which is actually able to use buffer generated by our FMC10xAPP. But this assume you have the initial software where the samples are left aligned without any bits masked out.
Place a 10MHz signal, 2Vpp centered on 0V, run FMC10xApp, launch Visual Analog, open the project, browse to your buffer and press the run button. You will see the time domain but also frequency domain and FFT post analysis.
Best Regards,
Arnaud
 
http://www.analog.com/en/converters-tools/adc-tools/topic.html <== Visual Analog can be downloaded from there.

pieryves February 15, 2013, 12:47 PM (#2)

  • Member
  • *
  • Posts: 11
Hello Arnaud,
Thanks for the answer.


You are right about the AC and DC coupling. What I need is a DC coupled card, I'll take care of that issue.


Then, I removed my older post because I realized that I was on the wrong path with the bit masking. I still have your original application, I took care of not overriding it.


I tried visual analog with a 10 MHz and 5MHz signal centered on 0V with 2Vpp. Indeed, it is a great tool ! I can see my signal, but still it seems that bits are missing. I attached to this message the data I gathered and a screenshot of the graph. Note that the red line is ADC0 and blue line is ADC1. I only injected a signal in ADC0. It looks like both signal are complementary. Could it be that data from both channel are interpreted as one value ? Can you confirm that what I see is not a normal behavior ?


Regards,
Pier-Yves Lessard

arnaudNL February 15, 2013, 01:33 PM (#3)

  • 4DSP Staff (EU)
  • Administrator
  • Member
  • *****
  • Posts: 7110
Dear Pier-Yves,
Can you verify the DIP switch settings on your board? The user manual describes the various configuration possible (The FMC10x, is several cards in one). Maybe the board is configured as a FMC108 and not FMC107?
Best Regards,
Arnaud
 

pieryves February 15, 2013, 03:44 PM (#4)

  • Member
  • *
  • Posts: 11
Hello Arnaud,


Indeed, the Dip Switch were configured as a FMC108. The value I get now does make more sense.


Still, I can't achieve to get a clean sine wave. I attached 2 graph showing data at 5MHz and 10MHz. Sample rate at 65MSPS. I validated the quality of the signal injected with a scopemeter.


Thank you
Regards

arnaudNL February 15, 2013, 04:42 PM (#5)

  • 4DSP Staff (EU)
  • Administrator
  • Member
  • *****
  • Posts: 7110
Bonsoir Pier-Yves,
Great news on the DIP switch configuration, it looks like the last problem is only a representation problem. I looked at the refence software and it configures the ADS62P49 in two's complement output.
If you look at the settings in the "Input Formater" block in Visual Analog, it is configured as "Offset binary" which is I think is not correct, try to configure the input formatters as "Two's Complement".
The .vac also have a wrong sampling frequency, it should be 65 and not 1250..
Best Regards,
Arnaud
 
 
 
 

pieryves February 15, 2013, 05:05 PM (#6)

  • Member
  • *
  • Posts: 11
Thank you Arnaud,

You were right, that was the problem. I am not very used to your product and I expected to read a signal from 0 to 4095 not -2048 to 2047.

I does make sense now. Will it be the same case for a DC coupled card ?
The user manual specify an input range of 2Vpp, but not more. Will a DC coupled card expect -1V to 1V or 0 to 2V ?

Thank you for the time you spent with me.

Regards
  • « Last Edit: February 15, 2013, 06:01 PM by pieryves »

arnaudNL February 18, 2013, 09:47 AM (#7)

  • 4DSP Staff (EU)
  • Administrator
  • Member
  • *****
  • Posts: 7110
Hello Pier-Yves,
The two's complement / offset binary settings is an ADC chip, the output format selection. in twos complement mode, the bit sign is at mid range and in offset binary there is not bit sign but only a value ranging from 0 to something. It is only about how the ADC chip represents his values.
I am not able to tell whether the DC coupled card have the center set to 0 or max range divided by 2, I would expect the DC coupled card input to be ranging between 0 and 2V but I am not sure, it might be an ordering option. Our sales department should be able to confirm that for you.
Best Regards,
Arnaud