WARP Project Forums - Wireless Open-Access Research Platform

You are not logged in.

#1 2014-Jul-21 10:58:43

pln32238
Member
Registered: 2014-Jul-21
Posts: 4

Delay problems between each ports

Hello,
I would like to ask you some problems about some measurements of the board “WARP v3”.
We have added a “FMC-RF-2X245” module in order to build a MIMO OFDM system. We call the RF ports (RF A & RF B) on the original “WARP v3 board” as Port_1 & Port_2 and  we call the RF ports (RF A & RF B) on the additional “FMC-RF-2X245” module as Port_3 & Port_4.
In our system, we want the exact same phase of the clock for all the DAC/ADC at each port. However, when we measure the signal delay between each port, we found that the delay is not same for each port. The process of the delay measurement is shown as follows:
We transmit the OFDM signal with pilot at all subcarriers. Then we use an attenuator to attenuate the signal. Then the signal is received by another port. We calculate the phase change of each subcarrier and estimate the delay of the transmission between each port (Measured results are averaged by 500 OFDM symbols).
The measured delays (samples) are shown as follows:
(1 ) Port_1 -> Port_2 = 38.454 samples;        (2 ) Port_2 -> Port_1 = 38.454 samples
(3 ) Port_1 -> Port_3 = 38.303 samples;        (4 ) Port_3 -> Port_1 = 37.512 samples
(5 ) Port_1 -> Port_4 = 38.307 samples;        (6 ) Port_4 -> Port_1 = 38.531 samples
(7) Port_3 -> Port_4 = 36.718 samples;        (8 ) Port_4 -> Port_3 = 37.777 samples
In the results, we can see that we can only get exact same delay for Port_1 and Port_2(For example results (1) and results (2)). It can be explained that Port_1 and Port_2 are on the same board (WARP v3). For other ports, they cannot keep the same delay for each measurement (Such as (3) and (4)). In addition, we found that for Port_3, it has some problems about the DAC. The measured results show that the DAC of the Port_3 always has shorter delay than other ports (For example comparing the results in (4) & (3) and the results in (7) & (8)).
We would like to ask following questions:
Why the DAC of the Port_3 has a significantly different delay than other ports? Can we solve this problem by hardware configurations?
Besides the Port_1 and Port_2, why other ports cannot keep the same delay? Can we solve this problem by hardware configurations?
Is it normal for this problem? Do we need something to take care when “FMC-RF-2X245” is connected with the “WARP v3”?
Thank you very much for your response!
Linning

Offline

 

#2 2014-Jul-21 16:59:05

chunter
Administrator
From: Mango Communications
Registered: 2006-Aug-24
Posts: 1212

Re: Delay problems between each ports

What you are measuring is end-to-end delay through two full RF paths. This delay is influenced by a variety of sources including manufacturing tolerances. The delay is influenced by the passive components in the RF path that are responsible for impedance matching, and those passive components are themselves influenced by temperature.

If your application is sensitive to offsets this small, you can consider building a sub-sample delay filter in front of outgoing/incoming I and Q. The coefficients used in this delay filter would need to be dynamically loaded based on your own calibrations. These calibration values will likely change across hardware and design revisions, so you might consider using the on-board EEPROM for non-volatile storage.

Offline

 

#3 2014-Jul-22 09:18:38

pln32238
Member
Registered: 2014-Jul-21
Posts: 4

Re: Delay problems between each ports

Hello,
Thank you very much for your fast response. We agree with you that this delay sometimes is related to the passive components in the RF path. However, our problem is that the maximal difference of these delay measured at around 1 sample (25ns). We consider that this delay difference may not only due to the variations of the passive components in the RF path. In each measurement, we have warmed the device after 15 minute before the measurements. All the measurements have been averaged by 500 OFDM symbols and we found similar results in several measurements during different days.
I will explain our problem in details:
In the first step, we measure the transmission from Port_1 to Port_2. We use the Tx RF path + DAC in Port_1 and Rx RF path + ADC in Port_2. The measured delay is 38.454 samples (about 961ns). Then we measure the transmission from Port_2 to Port_1. We use the Tx RF path + DAC in Port_2 and Rx RF path + ADC in Port_1. The measured delay is also exactly 38.454 samples (about 961ns). As we got the exactly same delay in the measurement of Port_1 to Port_2 and Port_2 to Port_1, we can consider that the delay of the Tx RF Path at Port_1 + delay of the Rx RF Path at Port_2 equals to the delay of the Tx RF Path at Port_2 + delay of the Rx RF Path at Port_1. In this case, we can find that the clock of the DAC/ADC at Port_1 is exactly the same as the clock of the DAC/ADC at Port_2 in both frequency and phase. Therefore, we consider that this result is OK because these 2 ports are on the same board. The delay from Oscillator to DAC/ADC is the same for the 2 ports.
In the second step, we measure the transmission from Port_1 to Port_4. The measured delay is 38.307 samples (about 958ns). Then we measure the transmission from Port_4 to Port_1. The measured delay is 38.531 samples (about 963ns). As we considered that the delay of the Tx RF Path at Port_1 + delay of the Rx RF Path at Port_4 equals to the delay of the Tx RF Path at Port_4 + delay of the Rx RF Path at Port_1, the difference of this measured delay could be explained as the clock phase of the DAC/ADC at Port_1 is NOT the same as the clock phase of the DAC/ADC at Port_4. We consider that this result is also OK. Seems that the Port_1 is on the Mother board and Port_4 is on the Daughter board. The delay from the oscillator (which is on the Mother board) to the DAC/ADC at Port_1 is not the same as delay from oscillator to DAC/ADC at Port_4 (which are on the Daughter board).
However, in the third step, we measure the transmission from Port_3 to Port_4. The measured delay is 36.718 samples (about 918ns). Then we measure the transmission from Port_4 to Port_3. The measured delay is 37.777 samples (about 944ns). In this measurement, the difference of the delay is around 1 sample (26ns). As both Port_3 and Port_4 are on the same Daughter board, we could consider that this situation is similar as the first step (Port_1 and Port_2) in the aforementioned measurements. Therefore, we wonder that this large difference come from somewhere we don’t know.
In addition, we have got two “WARP v3” board + “FMC-RF-2X245” module. We got similar results in another board with all the Step 1 + Step 2 + Step 3. The measured results are shown as follows:
(1 ) Port_1 -> Port_2 = 37.993 samples;        (2 ) Port_2 -> Port_1 = 37.981 samples
(3 ) Port_1 -> Port_3 = 38.063 samples;        (4 ) Port_3 -> Port_1 = 37.256 samples
(5 ) Port_1 -> Port_4 = 37.757 samples;        (6 ) Port_4 -> Port_1 = 37.924 samples
(7) Port_3 -> Port_4 = 37.048 samples;        (8 ) Port_4 -> Port_3 = 38.034 samples
As the measured two boards have similar problem, we consider that this delay problem is not accidental. We would like to propose this problem and ask where is the source of the strange delay at Port_3?
Thank you very much for your response.

Offline

 

#4 2014-Jul-22 09:47:35

chunter
Administrator
From: Mango Communications
Registered: 2006-Aug-24
Posts: 1212

Re: Delay problems between each ports

How are you measuring these sub-sample propagation delays? If you introduce an artificial N sample delay at your transmission, do your measurements successfully see that change?

Offline

 

#5 2014-Jul-22 11:45:33

pln32238
Member
Registered: 2014-Jul-21
Posts: 4

Re: Delay problems between each ports

We estimate the end-to-end delay with the following process:
In the received frames, we roughly estimate the delay by detecting the first OFDM symbol sample. Then we got the fine delay estimation by calculating the OFDM phase variations of all the pilot subcarriers. If the OFDM system has a delay, the phase variation of each pilot subcarrier is fixed. We obtain this phase variation value and calculate the fine delay of the whole OFDM symbol.
We have tested the transmission with introducing an artificial N sample delay at the transmitter. We can successfully see this change in the measurements with our delay estimation.

Offline

 

#6 2014-Jul-23 23:28:11

welsh
Administrator
From: Mango Communications
Registered: 2013-May-15
Posts: 612

Re: Delay problems between each ports

Just to let you know.  We are actively looking at this and will try to get you a resolution as soon as possible.  However, that might not be for another day or two.  Thanks for your patience.

Offline

 

#7 2014-Jul-29 14:12:08

welsh
Administrator
From: Mango Communications
Registered: 2013-May-15
Posts: 612

Re: Delay problems between each ports

I apologize for taking so long to respond.  We have an updated WARPLab design that tweaks the configuration of a few clocks in the design.  Would you mind re-running your experiments with this design?  That well help us further diagnose what is going on.

You can find a WARPLab 7.4.0 4RF bitstream with the updates here.  Thanks for your help.

Offline

 

#8 2014-Sep-01 11:51:48

yvan
Member
Registered: 2012-Apr-23
Posts: 12

Re: Delay problems between each ports

Hello,
Thank you for your help, I worked with “pln32238”, but he no longer works with us now. So I took over the result of his work.

As you requested, i configure the WARP v3 board with the WARPLab 7.4.0 4RF bitstream.
I started measurement delay between port.
It worked at first but I could not finish because the port A of the FMC board stops responding.
Now Port A of FMC board does not work a all.

This is what i tried to isolate the problem :

When the board is configured with the WARPLab 7.4.0 4RF bitstream :
   - D1 and D2 (FMC Board) blinks
   - D5, D6, D7 (FMC Board) are ON
   - D3 and D4 (FMC board) are OFF
   Ping to v3 Board works

When the board is configured with the WARPLab 6.5 4RF bitstream
   - D1 and D2 (FMC Board) blinks faster than the WARPLab 7.4.0
   - D5, D6, D7 (FMC Board) are ON
   - D3 and D4 (FMC board) are OFF
   Ping to v3 Board does not works

When i removed the FMC Board and configured the v3 Board with the WARPLab 6.5 2RF bitstream, ping to v3 Board works.

Have you any idea where the problem could come?

Thank you very much for your response!

Offline

 

#9 2014-Sep-02 09:00:07

welsh
Administrator
From: Mango Communications
Registered: 2013-May-15
Posts: 612

Re: Delay problems between each ports

Can you run the WARPLab 7.4.0 Reference 4RF bitstream and post the UART messages that the board prints during boot?  Thanks.

Offline

 

#10 2014-Sep-02 12:30:47

yvan
Member
Registered: 2012-Apr-23
Posts: 12

Re: Delay problems between each ports

Hello,
Thank you very much for your fast response.
This is what i get when i configure the board with WARPLab 7.4.0 4RF bitstream :

[0C]WARPLab v7.4.0 (compiled Feb 25 2014 00:00:11)
Configured for 4 RF Interfaces - FMC-RF-2X245 FMC module must be installed

Clock config 3:
  RF: On-board
  Samp: On-board
  Off-board Outputs: Disabled

Radios didn't lock! RC stat reg: 0x18101818
Configuring transport...
  Set MAC Address ETH A via EEPROM: 40:D8:55:04:21:65
  ETH A IP Address: 10.0.0.1
  Configuring ETH A for AXI DMA mode with 9014 byte buffers (2 receive, 1 send)
  Listening on UDP ports 9000 (unicast) and 10000 (broadcast)

Waiting for Ethernet link...

Initialization Successful - Waiting for Commands from MATLAB
[08][00]

Thank you in advance

Offline

 

#11 2014-Sep-02 12:58:50

welsh
Administrator
From: Mango Communications
Registered: 2013-May-15
Posts: 612

Re: Delay problems between each ports

Since the radios did not lock, please contact Mango Communications about the hardware problem.  Thanks.

Offline

 

#12 2014-Sep-17 02:45:47

yvan
Member
Registered: 2012-Apr-23
Posts: 12

Re: Delay problems between each ports

Hello,

OK, It’s done for the RF Board issue, we’ll return it to Mango. Thank you very much.

I re-run the delay between port measurements with another FMC Board (which works fine), here are the results :
Delays are given in number of samples

For a WARP Board v3 with WARPLab 6.5 :
(1)    Port 1 -> Port 2 : Delay = 37.191;     Port 2 -> Port 1 : Delay = 37.209; delta delay = -0.018
(2)    Port 1 -> Port 3 : Delay = 37.346;    Port 3 -> Port 1 : Delay = 36.156; delta delay = 1.19
(3)    Port 1 -> Port 4 : Delay = 37.356;    Port 4 -> Port 1 : Delay = 37.161 ; delta delay = 0.195
(4)    Port 2 -> Port 3 : Delay = 37.408;    Port 3 -> Port 2 : Delay = 36.199 ; delta delay = 1.209
(5)    Port 2 -> Port 4 : Delay = 37.414;    Port 4 -> Port 2 : Delay = 37.202 ; delta delay = 0.212
(6)    Port 3 -> Port 4 : Delay = 36.385;     Port 4 -> Port 3 : Delay = 37.377 ; delta delay = -0.992

For the same WARP Board v3 with WARPLab 7.4.0 :
(7)    Port 1 -> Port 2 : Delay = 37.800 ;    Port 2 -> Port 1 : Delay = 37.836 ; delta delay = -0.036   
(8)    Port 1 -> Port 3 : Delay = 38.468 ;    Port 3 -> Port 1 : Delay = 39.258 ; delta delay = -0.79
(9)    Port 1 -> Port 4 : Delay = 38.475 ;    Port 4 -> Port 1 : Delay = 39.262 ; delta delay = -0.787
(10)    Port 2 -> Port 3 : Delay = 38.551 ;    Port 3 -> Port 2 : Delay = 39.269 ; delta delay = -0.718
(11)    Port 2 -> Port 4 : Delay = 38.564 ;    Port 4 -> Port 2 : Delay = 39.307 ; delta delay = -0.743
(12)    Port 3 -> Port 4 : Delay = 37.984 ;     Port 4 -> Port 3 : Delay = 37.979 ; delta delay = 0.005

We can see that, indeed WARPLab 7 solved the delay issue between ports 3 and 4. (delay from port 3 to 4 is the same than delay from port 4 to 3, but not with WARPLab 6.5)
Can you tell me what have been made between WARPLab 7.4 and WARPLab 6.5 concerning the clock sharing ?

With WARPLab 7.4, we can also see that when we do measurement between a port on the mother board (port 1 or 2) and a port on the FMC Board (port 3 or 4) delays are not the same.
So is it possible to have the same delay, for example by adding a delay on the sampling clock in the “AD9513” clock buffer of the FMC Board ?

Thank you in advance for your time.

Offline

 

#13 2014-Sep-17 17:05:09

welsh
Administrator
From: Mango Communications
Registered: 2013-May-15
Posts: 612

Re: Delay problems between each ports

Just one clarification:  was this data gathered from the standard WARPLab 7.4.0 design or from the modified WARPLab 7.4.0 design I linked to in this post?

Offline

 

#14 2014-Sep-19 11:08:16

yvan
Member
Registered: 2012-Apr-23
Posts: 12

Re: Delay problems between each ports

Hello,

Oh sorry, you’re right, results given above were measured with the standard WARPLab 7.4.0 design.
I also realized that there was an error in my code for the fine delay estimation.
So here are the results with the good estimation for the standard WARPLab 7.4.0 and for the WARPLab 7.4.0 update_2014_07_29 :

For a WARP Board v3 with standard WARPLab 7.4.0 :
(1)    Port 1 -> Port 2 : Delay = 38.181;     Port 2 -> Port 1 : Delay = 38.155; delta delay = 0.026
(2)    Port 1 -> Port 3 : Delay = 37.531;    Port 3 -> Port 1 : Delay = 38.739; delta delay = -1.208
(3)    Port 1 -> Port 4 : Delay = 37.524;    Port 4 -> Port 1 : Delay = 38.736 ; delta delay = -1.212
(4)    Port 2 -> Port 3 : Delay = 37.448;    Port 3 -> Port 2 : Delay = 38.701 ; delta delay = -1.253
(5)    Port 2 -> Port 4 : Delay = 37.439;    Port 4 -> Port 2 : Delay = 38.696 ; delta delay = -1.257
(6)    Port 3 -> Port 4 : Delay = 38.019;     Port 4 -> Port 3 : Delay = 38.025 ; delta delay = -0.006

For the same WARP Board v3 with WARPLab 7.4.0 update_2014_07_29 :
(7)    Port 1 -> Port 2 : Delay = 38.184 ;    Port 2 -> Port 1 : Delay = 38.161 ; delta delay = 0.023   
(8)    Port 1 -> Port 3 : Delay = 38.041 ;    Port 3 -> Port 1 : Delay = 38.235 ; delta delay = -0.194
(9)    Port 1 -> Port 4 : Delay = 38.032 ;    Port 4 -> Port 1 : Delay = 38.230 ; delta delay = -0.198
(10)    Port 2 -> Port 3 : Delay = 37.965 ;    Port 3 -> Port 2 : Delay = 38.197 ; delta delay = -0.232
(11)    Port 2 -> Port 4 : Delay = 37.950 ;    Port 4 -> Port 2 : Delay = 38.190 ; delta delay = -0.240
(12)    Port 3 -> Port 4 : Delay = 38.019 ;     Port 4 -> Port 3 : Delay = 38.024 ; delta delay = -0.005

Thank you in advance

Offline

 

#15 2014-Sep-19 12:42:46

welsh
Administrator
From: Mango Communications
Registered: 2013-May-15
Posts: 612

Re: Delay problems between each ports

Great. We'll roll in those changes to the next WARPLab release.  For now, you can use the WARPLab 7.4.0 update_2014_07_29.

Let us know if you need anything else.

Offline

 

Board footer