[USRP-users] Phase drift issue of N310

Marcus D. Leech patchvonbraun at gmail.com
Thu Mar 21 08:43:20 EDT 2019


On 03/21/2019 07:08 AM, Damon wrote:
> Hi Marcus,
>
> Thank you for your reply.
>
> I compared the USRP configuration of your grc file with that of mine. 
> I found that the key to solve this issue is set_time_next_pps or 
> set_time_unknow_pps should be called.
>
> I don't know why   those two functions have an effect on phase error. 
> I haven't found any special description of these two functions in the 
> N310 manual.
>
> Best regards,
>
> Damon
They'll have an effect on initial phase error, because each half of the 
N310 has its own timekeeper.  If those aren't brought into alignment, then
   the sample streams will be unaligned.  But that should have no effect 
on ongoing phase drift--ONLY on initial phase error.


>
> On 2019/3/21 上午1:07, Marcus D. Leech wrote:
>> On 03/20/2019 12:55 PM, Damon wrote:
>>> Hi Marcus,
>>>
>>> Thanks for your reply.
>>>
>>> Yes, it happen with any other frequency changes, not only 460.01M 
>>> and 469.03Mhz.
>>>
>>> I'm building a 4-channel coherent receiving system with n310. It 
>>> needs to compensate the phase errors of multiple receiving channels. 
>>> So I need to measure the phase errors first and then compensate 
>>> them. But in the range of 100MHz bandwidth (from center_freq-50MHz 
>>> to  center_freq+50MHz), the phase difference of two receiving 
>>> channels of different dboards varies too much with frequency, so 
>>> it's very difficult to compensate the phase errors. As a contrast, 
>>> the phase errors of the two receiving channels of X310 with ubx vary 
>>> very samll with frequency, so it is easy to compensate the phase 
>>> difference.
>>>
>>> Looking forward to your test results
>>>
>>> Best regards,
>>>
>>> Damon
>> I used the attached script.
>>
>> Now, I'm only looking at +/- 2MHz, rather than your 100Mhz bandwidth.
>>
>> I found that the phase noise and offset did not change noticably 
>> tuning across the entire 4Mhz band.  I don't have a machine fast 
>> enough here to
>>   sweep my TX across 100Mhz, but with the N310 RX at a fixed tuning, 
>> and sweeping the TX (in this case, a Marconi transceiver test set), I 
>> did not see
>>   any significant phase shift or change in mutual phase noise as the 
>> TX swept across the 4MHz band.
>>
>>
>>>
>>> On 2019/3/20 上午10:40, Marcus D. Leech wrote:
>>>> On 03/19/2019 09:12 PM, Damon wrote:
>>>>> Hi Marcus,
>>>>>
>>>>> The phase  responses of two channels of different dboards 
>>>>> (ubx-160) in  a X310  are very consistent. When the frequency of 
>>>>> transmitting signal changes from 460.01MHz to 460.03Mhz, the phase 
>>>>> difference between two RX channels of different dboards in X310 
>>>>> remains unchanged. But for the phase difference between two RX 
>>>>> channels of different dboards of N310,  there are dozens or 
>>>>> hundreds degree of changes.
>>>>>
>>>>> Can you try to reproduce the test and help me to figure out how to 
>>>>> solve it?
>>>>> I'm not sure if it's a hardware bug, a driver bug or USRP setup 
>>>>> problem.
>>>> I cannot imagine a hardware or driver bug that could produce this 
>>>> behavior.  It would mean that the receiver was somehow changing LO
>>>>   frequency when the TX frequency changed.
>>>>
>>>> Does this happen with *other* frequency changes, or just 
>>>> 460.01<--->460.03.  I wonder if you have an interfering signal that 
>>>> is being
>>>>   "uncovered" by TX frequency change, and you're simply measuring 
>>>> that interfering signal?
>>>>
>>>> The only other thing that occurs to me is filtering that is 
>>>> extremely non-linear phase.  But that would create such a mess that 
>>>> most applications
>>>>   would likely not work--clearly they do.
>>>>
>>>> I can try to reproduce in my lab tomorrow, but, like the last tests 
>>>> I did, I very much expect to not be able to reproduce.
>>>>
>>>>
>>>>>
>>>>> Best regards,
>>>>> Damon
>>>>> On 2019/3/19 上午6:10, Marcus D. Leech wrote:
>>>>>> On 03/18/2019 01:26 PM, Damon wrote:
>>>>>>> Hi Marcus,
>>>>>>>
>>>>>>> Sorry, I can't reproduce the first observation in this 
>>>>>>> discussion thread. A new problem about phase response has arisen.
>>>>>>> I am testing the phase coherence performance of four receiving 
>>>>>>> channels of N310. A B200 is transmitting single tone continuous 
>>>>>>> wave to a one to four splitter. The 4 outputs of the splitter 
>>>>>>> are connected to 4 RX channels of N310. Attached please find the 
>>>>>>> GRC file of this test.
>>>>>>>
>>>>>>> The RX frequency of 4 channels of  N310 is set to 460MHz, and 
>>>>>>> keep running in the test.
>>>>>>>
>>>>>>> The TX frequency of B200 is set to 460.02MHz first, and then to 
>>>>>>> 460.03MHz. I thought the phase difference between different 
>>>>>>> dboards would change very little when the signal frequency 
>>>>>>> difference is very small, similar to the performance of X310. 
>>>>>>> However, the fact is that the phase difference between the two 
>>>>>>> dboards of N310 varies considerably with the signal frequency 
>>>>>>> transmission. For example, in the attached picture, when the 
>>>>>>> signal frequency is 460.01MHz, the phase difference between 
>>>>>>> channel 2 and channel 0 is -118 degrees, the phase difference 
>>>>>>> between channel 1 and channel 0 is 0 degrees; when the 
>>>>>>> transmission frequency is adjusted to 460.03MHz, the phase 
>>>>>>> difference between channel 2 and channel 0 is 117 degrees, and 
>>>>>>> the phase difference between channel 1 and channel 0 is 0 
>>>>>>> degrees. It is very difficult to understand that the phase 
>>>>>>> difference of two receiving channels of two different dboards 
>>>>>>> has changed by 235 degrees with the signal frequency change of 
>>>>>>> 20 KHz. The phase difference of two receiving channels of the 
>>>>>>> same dboard is basically unchanged.
>>>>>>>
>>>>>>> Best regards,
>>>>>>>
>>>>>>> Damon
>>>>>>>
>>>>>> Since the LO on the daughterboards has no idea that you've 
>>>>>> changed input frequency, this is clearly a measurement thing, and 
>>>>>> it's up to you
>>>>>>   to understand what you're measuring, and why.
>>>>>>
>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On 2019/3/16 上午7:47, Marcus D. Leech wrote:
>>>>>>>> On 03/14/2019 04:37 PM, Damon wrote:
>>>>>>>>> Hi Marcus,
>>>>>>>>>
>>>>>>>>> The UHD Version is v3.14.0.0-rc1.
>>>>>>>>>
>>>>>>>>> Best regards,
>>>>>>>>>
>>>>>>>>> Damon
>>>>>>>>>
>>>>>>>> I don't see this issue at all, using v3.14.0.0-rc3
>>>>>>>>
>>>>>>>> How are you measuring phase, what does you flow-graph look 
>>>>>>>> like? Have you increased the gain enough to assure that the 
>>>>>>>> inherent system
>>>>>>>>   noise is not dominating your phase measurements?
>>>>>>>>
>>>>>>>>
>>>>>>>>>>> Hi Ali,
>>>>>>>>>>>
>>>>>>>>>>> The daughterboards have their own clock generators, but they 
>>>>>>>>>>> are not
>>>>>>>>>>> exactly 'independent'. At least they don't have to be, as 
>>>>>>>>>>> they share the
>>>>>>>>>>> same reference clock. Look at the block diagram:
>>>>>>>>>>>
>>>>>>>>>>> https://kb.ettus.com/images/9/9d/USRP_N310_N300_DB_Schematic.pdf 
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> and "Ref Clock" block. I don't have N310 and I know that 
>>>>>>>>>>> reality can be
>>>>>>>>>>> a bit far from expectations (i.e. look at my "What makes 
>>>>>>>>>>> sense and what
>>>>>>>>>>> doesn't in the way carrier frequency is set for TwinRX 
>>>>>>>>>>> currently?" post).
>>>>>>>>>>>
>>>>>>>>>>> But maybe the daughterboards can be configured to use that 
>>>>>>>>>>> reference clock.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Best Regards,
>>>>>>>>>>> Piotr Krysik
>>>>>>>>>> The LMK clock generator uses the reference clock from the 
>>>>>>>>>> mainboard, so
>>>>>>>>>> there should not be any mutual phase-jitter/drift issues. I 
>>>>>>>>>> can test this
>>>>>>>>>>     on my N310 in the coming day or two.
>>>>>>>>>>
>>>>>>>>>> What version of UHD is in use?
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>
>>>>>>
>>>>
>>>>
>>




More information about the USRP-users mailing list