06-26-2019 11:38 AM
I want to use two ADC channels in Tile 0 for coherent sampling at 4 GSPS ( real data) and I find that the non-MTS design does not yield coherent sampling of the two channels in the same tile. I would think that for a single tile operation, I need not use the MTS design as it is for synchronization between different tiles and not within the same tile? Am I right in this assumption or am I missing something here and have to use MTS design? I would want to use the non-MTS design because I can use the DDR mode to capture a larger volume of data. Can anyone help me with this?
07-12-2019 05:20 AM
Can you tell us what you mean by coherent sampling of 2 channels in the tile.
To me coherent sampling refers to the the FFT being coherent.
I think you mean that the samplese should be aligned in the time domain. You are correct. Channels in a single tile should always be aligned.
I wonder if what you are seeing is down when the capture in the fabric happens?
07-12-2019 08:42 AM - edited 07-12-2019 08:43 AM
Thanks for the response @klumsde !
Yes. What I mean is coherency in time domain. For my application, I have to correlate signals from two different paths and from the correlation coefficient I obtain it seems the samples are not time domain coherent. My device is working fine since I get the expected results using another high-speed ADC data capture solution.
Could you please explain what you mean by "I wonder if what you are seeing is down when the capture in the fabric happens? "
07-15-2019 01:37 AM
My two cents here.
ADC channel in the same tile should align in time domain even without MTS. MTS is for multi-tile synchronization.
I checked non-mts project at my side, I can see the same issue. The only thing I guess is because the data comes out from BRAM which is used to store ADC result is not tirggered at the same time for non-MTS project.
07-15-2019 02:03 AM
I did a bit search on this issue. Here you might find the reason.
07-15-2019 10:08 AM - edited 07-16-2019 10:54 AM
Thanks for the @zhendon
I am using the 2019.1 version. Do these statements still hold for that design?
There seem to be a couple of conflicting statements. "In the subsequent version data is simultaneously captured for all the 8-ADC channels and provided to GUI sequentially". This would mean the sampling is coherent and I would not have the problem I am facing.
But this statement "This design has individual channel start/trigger signals and the common channel start/trigger signal required for MTS is removed" essentially says the sampling is not coherent since the start signal is for each individual channel and not common.
So, is there a way to overcome this issue to make the two channels sample coherently?
Also going to the Appendix A Performance table ADC Performance Matrix - Does it mean that if I am sampling all 8 channels with Real Data the max sampling rate I can get on any channel is 0nly 0.512GSPS equalling a max data volume of 128MB? Also, Does it also indirectly say that I cannot have all 8 channels to sample at 4GSPS even for a very short time?
07-16-2019 06:53 PM
It does look like the subsequent version after 2018.2 should have this function added from the statement. Let me check this and get back to you later.
If this function is still under development, I am sorry that you may have to use the MTS design or the RF analyer for the moment. (with BRAM only).
07-18-2019 08:16 PM
Sorry for the delayed reply.
2019.1 trd design still doesn't have simultaneous capture function provided. But the next version of trd will plan to add this function.
For the performance matrix, your understanding is correct. The max sample lenght is total 128Mb so the data duration will be 2ms if all 8 ADC sample at 4.096 GSPS.
07-18-2019 08:37 PM
Thanks for the reply @zhendon
I just want to clarify - the mode of operation wherein I sample all 8 channels simultaneously at 4.096GSPS is possible with the data length of 128MB and 2.048ms with the DDR mode? From the appendix, I got the feeling that the max sample rate that will be supported when operating all 8 channels is only 0.512GSPS at DDR mode? That is why I asked.
So, even if I use the MTS design in the 2019.1 TRD, I will not be able to get simultaneous data capture from two channels? Is that the case?
07-19-2019 01:58 AM
I can almost assure you that all 8 channel can run at Max sample rate 4.096GSPS in TRD design. But I haven't try sample the max data length. I suspect that the 128MB and 2 ms would be fine because I don't see any other bottle neck for archive this.
MTS of 2019.1 can work fine, you can get simultaneous data captured from two channels.
The simultaneous capture function mentioned in my previous reply is just for non-MTS case. Sorry for the confusion.
07-19-2019 01:25 PM
So to summarize -
if I use the non MTS image in DDR mode I can get all 8 channels at 4.096GSPS ( 128MB and 2.048ms) but not coherent data. This is the case even though the appendix says otherwise, i.e., 8 channels, 128MB or 2.048ms, max sample rate 0.512GSPS). I guess, my confusion still remains on what is the meaning of the Max sampling rate in the ADC Performance matrix table ( for DDR mode)
If I use the MTS image in BRAM mode, I can get both channels at 4.096GSPS although only 32K samples per channel ( as per the performance table) and if I want all 8 channels can also be coherently sampled in this mode (32K samples per channel only). There is no DDR mode with the MTS design and hence no way of getting longer data blocks in this mode.