I've done more testing. Here are some plots with the spectrum analyzer in zero span and a real sawtooth wave being transmitted.
The first plot is with 3.11.1.0 and sawtooth amplitude = 1.0. /Perfectly 
linear./ This also shows that it's entirely possible to have baseband 
amplitudes approaching -1.0 to 1.0 without distortion (at least on the 
B2X0). BTW, the TX gain was set to 74 to limit the output to just under 
10 dBm.
The second plot is with 3.11.1.0 and sawtooth amplitude = 1.05. You can 
see the AD9361 clamp the output level.
The third plot is with 3.13.0.1 and sawtooth amplitude = 1.0. Something 
is horribly wrong.
The fourth plot is with 3.13.0.1 and sawtooth amplitude = 1.05. Again, 
you can see the AD9361 clamp the output level.
This is clearly a bug/regression.

ramp 3.11 1.0

ramp 3.11 1.05

ramp 3.13 1.0

ramp 3.13 1.05

Ron

On 09/10/2018 07:59 PM, Marcus D. Leech via USRP-users wrote:
On 09/10/2018 10:48 PM, Ron Economos via USRP-users wrote:
It seems like it was correct before and wrong now. It's always been said that baseband levels must be restricted to -1.0 to 1.0. Now it's something different?
I consider this issue to be the same as a patch to the Linux kernel 
that breaks user space. And you know how Linus feels about that.
Ron
Yes, I can see where you're coming from here, and I'll leave it to Michael West to comment further on the "fix".
But there has *NEVER* been a guarantee that in any randomly-chosen 
situation a baseband level near MAX can always be guaranteed not
  to drive analog components into non-linear operating states. It's 
just a convenient scaling ceiling.  Now, I'll allow as how moving from
  "close to 1.0" to 0.67 as the highest achievable baseband level is a 
significant jump, and perhaps more attention needs to be paid
  to whatever scaling fix was implemented.

Consider for example that mixers and RF amps are being asked to operate over a couple of decades of bandwidth.  So a mixer baseband level   that is "nicely maxed without unpleasant artifacts" at one frequency might be "ouch" at another.  And these edge conditions are notoriously   manufacturing-batch sensitive, and to a certain extent usage-sensitive as well.  Theoretically, one could iterate over the entire parameter space   in an exhaustive calibration exercise, and have calibration tables that "hide" all of that.  But it would be an expensive exercise, I'm willing to bet.

On 09/10/2018 07:26 PM, Marcus D. Leech via USRP-users wrote:
On 09/10/2018 10:21 PM, Ron Economos via USRP-users wrote:
I have to reduce the baseband amplitude from 0.96 to 0.67 to 
achieve the same IMD.
This is a very annoying issue for me, especially since I've set 
levels in the Gnu Radio gr-dtv example flow graphs based on 
previous behavior.
Ron

I think it was Michael West who worked on the scaling issues in B2xx.   The internal implementation had been wrong for a long time,   from what I understood.  This has the unfortunate side-effect that previous carefully-set baseband levels may no longer work properly.
Unfortunately, there's no "set the baseband magnitude and RF gain as 
loud as you can without producing excessive IMD" operator in UHD.

On 09/10/2018 05:37 PM, Marcus D. Leech via USRP-users wrote:
On 09/10/2018 08:34 PM, Ron Economos via USRP-users wrote:
I'll try that. It's going to take a while though since I have to 
rebuild.
Ron

Understood. Sorry for the PITA....


On 09/10/2018 05:30 PM, Marcus D. Leech via USRP-users wrote:
On 09/10/2018 08:27 PM, Ron Economos via USRP-users wrote:
No change with 3.13.0. Haven't tried 3.14, since it's not 
released yet.
Ron

Is this effect independent of baseband magnitude?


On 09/10/2018 05:21 PM, Marcus D. Leech via USRP-users wrote:
On 09/10/2018 08:17 PM, Ron Economos via USRP-users wrote:
I had previously noted that the B210 TX power had increased 
by about 5 dB going from UHD 3.11.1.0 to UHD 3.12.0. Today, I 
was doing some two tone IMD testing and noticed that the IMD 
performance has degraded *severely* from UHD 3.11.1.0 to UHD 
3.12.0.
Here are the spectrum analyzer plots at roughly the same 
power level. The 3rd order two tone IMD level changes from 
almost -50 dB in 3.11.0.1 to a hideous -20 dB with 3.12.0. 
Something is terribly wrong.


I know that there had been some "futzing" with scaling in the B2xx at around that time.  What about 3.13 or 3.14?

_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Reply via email to