Rigol DP832 remote command update rate????..............

[Smokey]

My setup is I'm using the LXI LAN connection on the Rigol DP832 power supply to a python script setting the voltage and reading the current.

I'm trying to essentially do a voltage sweep incrementing the set voltage on one channel (gate) and reading the current back on the other channel (drain) to make a curve tracer of sorts.  The problem is I can't reliably do a set_voltage+read_current cycle in less than about 0.7seconds, which would make the test take WAY too long.  Is this just the fastest I can set and read from the instrument or is there something else I can do?  I'm open to other communications data capture methods if that's faster. 

[LaurentR]

Can't find anything in the manuals, but I have a similar experience.
A couple of years ago I ran sweeps across the voltage ranges to characterize the setting and measurement error (drive with the DP832, read with the DP832 and a 3446x and compare) and settled on a 1.5-2 sec cycle time as the response  from the DP832 was slow.

Note that in all the native sequencing available from the front panel, all times are in seconds (no fractions of a second), so it doesn't natively do anything fast. Not sure where the delay is.

[RoGeorge]

I'm trying to essentially do a voltage sweep incrementing the set voltage on one channel (gate) and reading the current back on the other channel (drain) to make a curve tracer of sorts.

I guess the best approach to do that will be to use the "Timer and Delay" internal feature of the DP832, then use LXI to read all channels measurements at once while drawing the charts on the PC.
"Timer and Delayer" is described in the user manual, it has the possibility to output automatically a custom defined "waveform", or predefined waveform like sinus, pulse or ramp.

LATER EDIT:
Sorry, the minimum time for each output value is 1 second, so I guess the timer feature doesn't help.

Another option
You could try to use a kind of hack, based on the assumption that the LXI reading of the measurements is faster (I didn't tested how fast the measurements can be read):
- set a very low current limit, like 5mA, for the channel destined to feed the gate with voltage
- then, turn the channel from 0 to something like 10V - this will create a voltage ramp at output, because the filter capacitor from the output of the channel is big, so it will take some time to charge up to 10V with only those 5mA.
- while the voltage is raising at the channel that feeds the gate, start reading as fast as possible the measured values for the gate voltage channel and for the current drain channel.
Maybe this will work fast enough to make a curve tracer, but I didn't tested it.

[RoGeorge]

It happens that I am working right now on something similar with what you are doing.

I just found that the measured values are updated only twice per second. I can request by LXI the measurements much faster, even 100 times per second, and it will respond, but it will respond with the last measured values, so I can not have more than 2 distinct measured values per second.

It seems that a curve tracer can not be done faster than 1-2 steps per second.
 

[Smokey]

Damn.  That all matches with what I've been seeing.  That sucks.  I wonder why they didn't allow faster remote updates.  Time for plan B....

[CustomEngineerer]

My guess would be because its a DC power supply, not a waveform generator. I thought I remembered either the user guide or the programming guide listing how fast the voltage could be changed through programming. I also seem to remember that its essentially limited to how quickly the power supply can actually change the voltage anyways, and since the rigol has a decent amount of output capacitance, its fairly slow to change the output. I could of course be wrong, its not something I've played around with because its not something I've needed to do.

[Smokey]

True.  I'm not saying the thing is useless because it can't do this.  I know I'm pushing it.  It's still a good power supply.

I'm not talking about ns voltage changes here though.  If I could do 50ms changes in the 10mA step range I would probably be happy. 
And with that said there is no reason why you can't get the measurements out quicker.

[rs20]

Indeed. But, from 30,000ft at least, you're asking for the features of an SMU, not a basic DC power supply.

[lundmar]

Yeah, it's a real shame. I don't think I have seen any DC power supplies with quick update rate and very fast LXI response.

I wish I could find an open source hardware DC power supply with a proper internal data logger with a big memory buffer supporting up to 1kHz precision sampling rate of voltage/current that can last for at least 30 minutes and then be able to read it out via LXI. Oh, and it needs to be compact and fanless so it fits perfectly on the desk or lab of every engineer and hacker out there. If only such a project existed..

[RoGeorge]

The main problem to make the curve tracer is not the settling time, but the slow measurements of Iout and Vout, which are measured only twice per second.

A curve tracer can still be done if an LXI oscilloscope (or some other ADC capable to measure faster) is added to the DP832 setup .
- Generate Vgs using the trick of linear charging the output filter capacitors from DP832 by a mA range current set.
- Use constant Vds and a small series resistor to measure Ids
- Use an oscilloscope to simultaneously display Vgs, Vds and Ids
- Force the oscilloscope to trigger, then turn on the power supply channel corresponding to Vgs
- Read the oscilloscope traces, then plot nice charts

A full curve can be traced once at each 2 seconds, which is not that bad, like in one point per second.

[JDubU]

The DP832 spec sheet lists the "Command Processing Time" as:   <118ms (The maximum time required for the output to change accordingly after receiving the APPLy and SOURce commands).

http://beyondmeasure.rigoltech.com/acton/attachment/1579/f-01c1/1/-/-/-/-/DP800%20Datasheet.pdf

[Smokey]

Yeah, it's a real shame. I don't think I have seen any DC power supplies with quick update rate and very fast LXI response.

I wish I could find an open source hardware DC power supply with a proper internal data logger with a big memory buffer supporting up to 1kHz precision sampling rate of voltage/current that can last for at least 30 minutes and then be able to read it out via LXI. Oh, and it needs to be compact and fanless so it fits perfectly on the desk or lab of every engineer and hacker out there. If only such a project existed..

That's what I'm talking about!
Ya I realize I'm trying to approximate an SMU.  I don't need nearly the precision of a real SMU though. 
I call shenanigans on the DP832 not being able to measure fast enough.  The internal protection does not take 0.5 sec to detect a limit condition.  It's measuring fast, but not giving me access to the measurements faster than 2 per sec.

I thought about the scope data capture, but that's only 8bit resolution for the data which unfortunately is lower than I need.  The measurement on the DP832 is much better than that.

[alm]

I find it ironic that people expect a DC power supply to sample quickly. If it is truly a DC signal, you only need to sample it once .

There are some high-speed power supplies like Keithley 230x and HP/Agilent/Keysight 663xx that can sample at least the current up to a few kS/s or so. Carefully examine their datasheets to learn specifics and limitations (some modes need to be triggered, for example). I imagine the Rigol designers did not consider sampling a signal with a few mF of output capacitance across it at a high sampling rate very useful. High-speed power supplies (and SMUs) will have a very low output capacitance with the associated higher noise, decreased stability and increased complexity.

If you want to sample the signal faster, connect a DMM across the power supply. A decent bench DMM should be sample up to 1 kS/s or so with decent resolution. That is what people used to do before SMUs and power supplies with advanced measurement functions were available. Or you could try the resolution from your scope by oversampling and averaging multiple sweeps.

If you want to make the sourcing part faster without buying a $$$ SMU, you could look into combined power supply / amplifiers like KEPCO BOP, Lambda BOSS and HP 682xA and drive them from a function generator. They will not give you any high-speed measurements, so you would still need one or two external DMMs to sample the output. SMUs were designed to combine those functions into one instrument, and speed up the source-measure cycle through integration.

You could also look into those 'toy SMUs' like the Analog Devices ADALM1000.

[Smokey]

Resurrecting this old thread.  Got an ad from Rigol for the new DP2000 series.  Check out the specs on programmability and response times (pg 7)....

https://beyondmeasure.rigoltech.com/acton/attachment/1579/f-531346d9-45a9-40b2-9b3e-e9120b59e01d/1/-/-/-/-/DP2000_DataSheet_en.pdf



This both confirms all the stuff we were speculating about the DP800 series, and more interestingly has specs that make all the desired functionality possible on new DP2000 series (and some of it with the existing DP900 series too). 

10ms Command processing time and 1ms dwell opens a lot of doors for interesting scripted experiments here. 

[alm]

The power supply does not have an active down-programmer, so the actual voltage programming times, especially going from 100% down to 0 V with no load are a lot less impressive:

[Smokey]

True.  Important to point out.  We need to remember it's still a bench supply after all. 
I don't think that would be much of an issue for my original application of a FET classifier.

[alm]

There are systems power supplies with an active down-programmer that can slew much faster, because in a systems application that might translate to increased throughout.

But for a curve tracer you're not likely to need to slew very fast, indeed, as long as you're not doing pulsed measurements. How about the readback sample rate? Is it also faster?

[RoGeorge]

A power supply can only source current, it can not sink current.  It has electrolytic capacitors at output, and thus the slower discharging time when the output is left without a load.  Otherwise I found it fast enough for a power supply.

When remote controlled, the '*OPC?' command is supposed to tell when a command is completed (in this case when the voltage settled to the requested value), though I never tested that in my DP832.

For fast settling there are 4 quadrant power supplies that can both source or sink current, often called SMU, but that's a different type of instrument.

[alm]

A limited sinking capability is common in more high end power supplies. The old HPAK 6632A/B for example. As I said before, they do this because it's important in automation. This only provides two quadrants: no negative voltages without changing connections. GW-Instek also has a power supply that can sink current. I would expect it to also have a faster down programming speed. But clearly not in these supplies.