bbn-q / qlab Goto Github PK

The problem is that if you have AWGs enabled that aren't used when calibrating a particular qubit, then you will either get an error in PulseCalibration.loadSequence() or you will have bad/dummy data coming out of an AWG. We should disable those AWGs not used in calibrating a particular qubit.

See requests in #43 discussion.

It would be great if this were automatically rather than having to manually copy these values into the appropriate field.

Can probably do this by disabling the DAC clocks.

• move from PatternGen to Pulse class
• keep as separate files
• rename from xPulse to Pulse.x

These parameters need to be inserted into the settings structure by default:

prevSettings.ExpParams = struct('digitalHomodyne', [], 'filter', [], 'softAvgs', 1);
prevSettings.deviceName = '';
prevSettings.exptName = '';
prevSettings.data_path = '';

Also, the experiment quickpicker has the same issue.

You get a permissions error when loading the library.

An experiment class should be composed of

• an instrument manager
• a data file handler
• a looping manager
• add a measure method

Generally it should be much easier to create new experiments.

The StateFilter code is still using the "childFilter" architecture that is not valid anymore. There is corresponding fix needed in the experiment GUI: BBN-Q/PyQLab#70

Should explicitly write the loop points to file. If a loop is such that it can write just a start and step, it should do so.

Construct the PatternGen object from a settings structure.

Need to update driver to look at "channelDataFor" and only load the appropriate channels.

The analog channels are significantly delayed compared to the marker channels (I've seen a ~52 ns delay = 62 samples). To fix this we could put marker data into a FIFO buffer before it is output. Ideally, this buffer would have a size which is configurable with a register.

If a file named something like 0.h5 is present in a data folder, DataNamer.m will fail at parsing it with regexp.

I may submit a PR fixing this later today.

Find out if the bottleneck is the handshaking or the throughput of USB 1.0.

As reported by IBM and Colm, occasionally the PLL chip on the APS seems to lose lock with the reference. Symptoms include SSB frequencies and pulse widths being incorrect.

Next time this happens, we should check the status bits on the PLL chip. Does it think it is still locked up? If not, we can expose these bits to the driver and check them in init().

SingleShot measurement filter returns empty array when X6 is put in digitizer mode. averager mode is fine, though. Expected behavior??

• Move more sequence file boilerplate to parameter files
• No hardcoding of marker channels
• Move to logical rather than IQkey keys
• Consolidate fixedPts and cycleLengths

Add a measAffineFile field to the MeasFilter settings. Then load the file in the constructor.

APS driver should "remember" what is currently in waveform memory.

Before version 1 we should max out the block RAM on the FPGA and push SSB off to a later version.

Not sure if the API allows this but we've wasted many hours with strange behaviour that we eventually track down to an unlocked LabBrick.

There are a couple ways to do this:

1. Previously I've run a updating estimate based on how long things are taking.
2. If we control the experiment rep rate then it is mostly calculable. The trouble with this is that the AWG load/start/stop times are harder to include.

inverse is used in QST_LSQ.m here. Should this be inv or \? Maybe I'm missing a library

The BNC driver has few loose ends that need tidying up:

• setting reference source
• setting external reference frequency
• setting reference out frequency
• setting reference out on/off
• checking reference lock state
• cleanup workaround for no output after toggling pulse mode issue after BNC fixes issue

When loading, there is no link between data and the respective measurement channels

Convert pulse functions into pulse objects. Overload the disp operator to show humanized labels.

There is no current method for specifying the outputs on the marker channels of the APS. We would like to support two modes:

• Buffer pulse - trigger at edges of buffer pulse
• Single trigger with start and width

Changing the number of devices connected causes crash on enumerate.

Let's not re-invent the wheel. There is a robust format out there, so use it.

I've tried running calibratepulses and after it compiles the first 8- sequences and takes that data if fails with this error message:

Attempted to access data(0); index must be a positive integer or logical.

Error in PulseCalibration.analyzeRabiAmp (line 20)
if data(end/2) > offset

Error in PulseCalibration/PulseCalibrationDo (line 46)
[piAmpGuesses(1), offsetPhases(1)] = obj.analyzeRabiAmp(data(1:end/2));

Error in PulseCalibration/Do (line 207)
obj.PulseCalibrationDo();

Error in calibratePulses (line 32)
pulseCal.Do();

So it looks like it just isn't taking any data. I changed the ExpParams.measurements from M1 to the new M1_KernalIntegrator that we are using. Not sure if this is correct though.

Any Ideas?