Zurich instruments HDAWG driver

[terrorfisch]

PROBLEMS:

• Long programs take forever (approx.) to compile
• Short waveforms lead to tiing issues if sequenced by a complex program with many instructions

TODO:

• Offset handling
• Utilize playWaveformIndexed with for loops to shorten seqc programs
• Increase minimal Waveform length (to 192 samples?). The exact minimal length depends on the choosen translation pattern.
• Basic tests for AWG class

Implements #431

[coveralls]

Pull Request Test Coverage Report for Build 1902

• 510 of 890 (57.3%) changed or added relevant lines in 8 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-4.6%) to 85.291%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
qupulse/utils/types.py	18	19	94.74%
qupulse/_program/_loop.py	2	5	40.0%
qupulse/hardware/awgs/base.py	5	33	15.15%
qupulse/_program/seqc.py	309	480	64.38%
qupulse/hardware/awgs/zihdawg.py	171	348	49.14%

Totals
Change from base Build 1900:	-4.6%
Covered Lines:	28605
Relevant Lines:	33060

---

💛 - Coveralls

[peendebak]

@ekammerloher @terrorfisch The following example (modified from zihdawg.py) fails for me:

hdawg.reset()

entry_list_zero = [(0, 0), (100e-9, 0, 'hold')]
entry_list_step = [(0, 0), (50e-9, .5, 'hold'), (100e-9, 0, 'hold')]
marker_start = TablePT({'P1': entry_list_zero,  'marker': entry_list_step})
tpt1 = TablePT({'P1': entry_list_zero, 'marker': entry_list_zero})
spt2 = SequencePT(marker_start, tpt1)
    
p = spt2.create_program()

ch = ('P1', None)
mk = ('marker', None)
voltage_transform = (lambda x: x,)*len(ch)
hdawg.channel_pair_AB.upload('table_pulse_test5', p, ch, mk, voltage_transform) 

The error message is:

...
  File "d:\users\eendebakpt\projects\qupulse\qupulse\hardware\awgs\zihdawg.py", line 539, in to_file
    raise HDAWGIOError('{} already exists'.format(file_path))

HDAWGIOError: C:\Users\controller\Documents\Zurich Instruments\LabOne\WebServer\awg\waves\DEV8049_AB_7318615092299051717_m.csv already exists

Otherwise the PR itself works nice. I will continue testing.

Will there be besides the channel_pair_AB also a channel_pair_ABCDEFGH for when the channel grouping of the HDAWG8 is set to 8x1?

[terrorfisch]

Will there be besides the channel_pair_AB also a channel_pair_ABCDEFGH for when the channel grouping of the HDAWG8 is set to 8x1?

As far as I remember from talking to a ZI engineer, the sequencer program will be duplicated to the 4 sequencers if one uses the 8x1 mode. I think the only downside of pushing the program to the four channel pairs seperatly is that we need to compile it four times (unless we optimize this and hash the program structure). Or is there anything that I miss?

[ekammerloher]

The last 5 commits were not tested on the HDAWG yet, so the failing example is not a big surprise to me. I don't have access to the HDAWG at the moment, but will try to fix it soon.

…

On Tue, 14 May 2019 at 22:07, Pieter Eendebak ***@***.***> wrote: @ekammerloher <https://github.com/ekammerloher> @terrorfisch <https://github.com/terrorfisch> The following example (modified from zihdawg.py) fails for me: hdawg.reset() entry_list_zero = [(0, 0), (100e-9, 0, 'hold')] entry_list_step = [(0, 0), (50e-9, .5, 'hold'), (100e-9, 0, 'hold')] marker_start = TablePT({'P1': entry_list_zero, 'marker': entry_list_step}) tpt1 = TablePT({'P1': entry_list_zero, 'marker': entry_list_zero}) spt2 = SequencePT(marker_start, tpt1) p = spt2.create_program() ch = ('P1', None) mk = ('marker', None) voltage_transform = (lambda x: x,)*len(ch) hdawg.channel_pair_AB.upload('table_pulse_test5', p, ch, mk, voltage_transform) The PR itself works nice. I will continue testing. Will there be besides the channel_pair_AB also a channel_pair_ABCDEFGH for when the channel grouping of the HDAWG8 is set to 8x1? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#454?email_source=notifications&email_token=AATX4SICHAKYMQBATDGNYFDPVMLYRA5CNFSM4HJ4T4FKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODVMUMJI#issuecomment-492389925>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATX4SLU4XRYO2RA57DFMTDPVMLYRANCNFSM4HJ4T4FA> .

[peendebak]

Will there be besides the channel_pair_AB also a channel_pair_ABCDEFGH for when the channel grouping of the HDAWG8 is set to 8x1?

As far as I remember from talking to a ZI engineer, the sequencer program will be duplicated to the 4 sequencers if one uses the 8x1 mode. I think the only downside of pushing the program to the four channel pairs seperatly is that we need to compile it four times (unless we optimize this and hash the program structure). Or is there anything that I miss?

@ekammerloher terrorfish For multi-device synchronization it might be essential to generalize the 2x4 concept. It also takes a little bit more effort from the user to use the 4 channel pairs, otherwise I am not yet aware of other downsides.

[peendebak]

@terrorfisch @ekammerloher Is seems this driver accepts Pulses that are defined with seconds as the unit for the time domain. In general qupulse works with ns as the unit. Should we change this to ns? if not, iis there a method to change the unit of the time domian from ns to s (and back)?

And: is there already a slack channel for qupulse?

[terrorfisch]

#483 tackles parts of this