sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
<GitHub130> [artiq] sbourdeauducq commented on issue #737: @cjbe Do you still have the problem with 3.5.3? Or will it not install for some reason? https://github.com/m-labs/artiq/issues/737#issuecomment-305381564
rohitksingh_work has joined #m-labs
gric has quit [Ping timeout: 258 seconds]
ohsix has joined #m-labs
ohsix has quit [Read error: Connection reset by peer]
ohsix has joined #m-labs
gric has joined #m-labs
sb0 has quit [Quit: Leaving]
rohitksingh_wor1 has joined #m-labs
rohitksingh_work has quit [Ping timeout: 240 seconds]
ohsix has quit [Ping timeout: 240 seconds]
ohsix has joined #m-labs
hartytp has joined #m-labs
<hartytp> rjo: your scheme basically works.
<hartytp> Using the numbers you suggest gives a residual 30mHz error because the DRG takes 513 cycles, not 512 cycles
<hartytp> small phase steps are visible on a scope
<hartytp> -55dBc at 7.6kHz
<hartytp> how do you plan to do phase offsets in this mode as, IIRC, the DRG overrides the POW so one can't use both?
<rjo> hartytp: ok. thanks for testing. weird that it takes 513 cycles. i can't see why. on a scope? or on a spectrum analyzer? yes. the -55dBc is consistent with the phase step in that test. you'd have to use the timing of the DRG accu clear and DDS accu clear for controling the phase offset independently.
<hartytp> mixer + scope/frequency counter
<rjo> hartytp: i see.
sb0 has joined #m-labs
<hartytp> DDS at 124.99998533MHz, synth at 125MHz, DRG as discussed on GitHub
<hartytp> Beat note frequency is approx 0.204Hz
<rjo> hartytp: that's the residual beat with the DRG on?
<rjo> hartytp: so 204 mHz and not 30 mHz? positive or negative?
<sb0> what is "contact size 16"?
<sb0> tenths of mm?
<sb0> there seems to be a standard unit but i don't see it documented
<sb0> really?
<hartytp> AWG50=1mil, doubles every 6 reduction in AWG
<hartytp> 124.99998487MHz DDS gives 0.262Hz beat
sb0 has quit [Quit: Leaving]
<rjo> hartytp: that's 0.204 Hz with 0x1fffffc1 and 0.262 Hz with 0x1fffffbf? the register values are more useful. otherwise i have to guess the rounding of that app.
<rjo> hartytp: and in both cases that's the beat with the DRG doing the 0x0080(0000) step, 0x8000 interval ramp?
<rjo> hartytp: ok. then for 0x1fffffc0 it's -29mHz. fits the model.
<rjo> hartytp: one more thing: is the DRG operation disturbed if you update random other parameters (assuming DRG (auto) accu clear is off, DDS auto accu clear off, LRR@IO_UPDATE is off, etc)? or does it continue nicely?
rohitksingh_work has joined #m-labs
rohitksingh_wor1 has quit [Ping timeout: 260 seconds]
rohitksingh_wor1 has joined #m-labs
rohitksingh_work has quit [Ping timeout: 240 seconds]
<hartytp> 124.99998487 = 1FFFFFBF
<hartytp> 124.99998533 = 1FFFFFC1
<hartytp> (i.e. +- 1LSB from 1FFFFFC0)
<hartytp> I tried updating the frequency of profile 1 (switch from 0 to FFFFFFFF and use software to IO_UPDATE) but don't see anything on my mixer
<hartytp> profile 0 is active
<hartytp> (don't see anyting means my beat note continues without visible glitches)
<hartytp> any other tests you want me to do?
rohitksingh_wor1 has quit [Read error: Connection reset by peer]
<hartytp> changing the active profile's ASF by a LSB also doesn't do anything particularly visible...
<rjo> hartytp: all good. thanks! if you have a bit of time, i'd like to make a test that determines the origin of the 512/513 interval discrepancy: use a large DRG interval (0xffff) and a very large DRG step (0x4000 (0000) up and 0xc000 (0000) down) so that it "wraps" every 4 DRG intervals. you should be able to resolve everything on the scope.
<rjo> DRG limits still railed (0xffff and 0x0000 in the 32-bit truncation of that app).
rohitksingh has joined #m-labs
<rjo> if you do that test, you should see a five (instead of the four expected) DRG interval periodicity in your demodulated phase. maybe the only change required to fix the behavior is to lower the "down" DRG interval from 0xffff to 0x0001 (or maybe 0x0000?).
<hartytp> It spends two cycles at the top
<hartytp> what site do you use to dump images?
<hartytp> synth and DDS at 125MHz
<hartytp> DRG as you described
<hartytp> falling interval to 0 gives a flat line
<hartytp> 0x0001 interval:
<rjo> hartytp: imagebin.ca seems to be down now...
<rjo> hartytp: let's verify that the ramp down takes one DRG cycle more than expected, could you give these two configurations a test? FTW=0x1ffff0000, DRG step up/down=0x0200/0xfe00, DRG interval=0x0080/0x0080 and then the same with DRG interval=0x0080/0x0001
<rjo> the first should have a 119 Hz beat and the second at most 0.9 Hz but the exact value is of interest.
<rjo> hartytp: that last screeshot of yours is interesting. that's only three DRG cycles, right?
<rjo> in that case for that second test, do DRG interval=0x0080/0x0040...
<rjo> hartytp: in any case, i am sufficiently convinced that one could use the DRG to extend the FTW. the wrinkles are the slightly complicted way of determining optimal DRG parameters and the lack of/difficulty with phase offset control.
<hartytp> hmm the links work okay for me
<rjo> hartytp: works now for me.
<hartytp> I think it does something like dwell for an extra cycle at the turning point
<rjo> yes. but then only at one of the two turning points. and it dwells one "down" interval.
<hartytp> yes
<hartytp> is this the config you want?
<rjo> hartytp: 0x0C: FE 00 00 00 02 00 00 00, 0x0D: 00 80 00 80, FTW0: 1F FF 00 00, DRG step: 357.1875 deg down, 2.8125 deg up, 0.512 µs interval.
<rjo> that should be 119 Hz beat. and then basically adjust the down interval to zero out the beat frequency. from what we have seen 0.256 µs down interval (0x0D: 00 40 00 80) should be zero beat.
<hartytp> ack. There was a typo in the previous FTW you posted.
<rjo> ack. one F too many.
<hartytp> also, pasting 0x into the eval software crashes it
<rjo> ha... from a couple of older eval boards i remember them having bundled python software.
<hartytp> okay, that config gives 118.3Hz
<hartytp> hmm. decreasing the falling step interval decreases the beat note frequency
<hartytp> 0x0003 takes it to 5.6Hz
<hartytp> then 0x0002 gives 116.3Hz
<hartytp> 0040 (0.2560) gives 59.4Hz
<hartytp> (ish, could use more averaging)
<rjo> i can give you more extreme paramters that make this easier...
<hartytp> sure
<rjo> FTW0: 1F F0 00 00 (124.755859375 MHz), step 0x0C: FC 00 04 00, interval 0x0D: 00 10 00 10 and then tweak for min beat.
<rjo> the step is 354.375, 5.625 deg
<rjo> and the intervals are 64 ns.
<hartytp> tweaking falling step interval to give the slowest beat
<hartytp> 0x0004 gives 949.9Hz
<hartytp> (949.96Hz)
<rjo> hartytp: thanks. i think i have all the data i want.
<hartytp> okay
<rjo> oh.
<hartytp> I'll leave it set up for the next few days if you think of anything else
<hartytp> ?
<rjo> hartytp: ack.
hartytp has quit [Quit: Page closed]
sb0 has joined #m-labs
mumptai has joined #m-labs
rohitksingh has quit [Quit: Leaving.]
sb0 has quit [Quit: Leaving]
mumptai has quit [Remote host closed the connection]