New Rigol 16-bit function generators DG800/900 series

[timber23]

I'm looking for compatible usb-lan adapter.
Could you share working models ?
Already tried Lenovo FRU03X6903 (RTL8152) and Dlink DUB-1312 (AX88179) with no success "adapter not found".

P.S. tv84 reported that r8188eu.ko module is included, does it means that usb-wifi is possible too?

I have tried this one. It is not working, just shows the LXI-symbol, but does not get IP address.

Code: [Select]

Bus 001 Device 002: ID 0bda:8152 Realtek Semiconductor Corp. RTL8152 Fast Ethernet AdapterIt is a noname device, looks white with "75.007.24/JP208" written on the side.

[toshas]

I'm sorry FRU03X6903 has RTL8153.

[toshas]

Also tried TP-Link UE200 (RTL8152B) with no luck.

Finally found this one : D-Link DUB-E100 rev. D1 (AX88772), vid:pid 2001:1a02 and it works!

[rtv]

DG811 pcb pictures: https://app.box.com/s/rjgizdvupdsnn4do5hnmw8jmyqdargsz

[thm_w]

For whatever reason I didn't have any luck with Ultra Sigma to send the SCPI command. *IDN? was able to receive a response so that part was working OK. Tried restarting PC and now Ultra Sigma won't even connect, it just crashes when I try to right click the DG811, or closes the window.

So I tried in pyvisa and it worked OK. If anyone else wondering, there is no response back from the instrument after the command is sent. Then the model will be changed right away in sys info, after this you can restart the instrument and limits/UI will then be updated.

rm = pyvisa.ResourceManager()   #'@py' will not find the DG800
print(rm.list_resources())

gen = rm.open_resource('USB0::0x1AB1::0x0643::DG8A2xxxxxxxx::INSTR')

#Request ID string
print(gen.query('*IDN?'))

#Change model
gen.write(':PROJ:MODE DG992') #works but no response: print(gen.query(':PROJ:MODE DG992'))

#Close the connection
gen.close()

[frozenfrogz]

Since I am getting quite some personal messages regarding how to perform the upgrade I thought it might be useful to elaborate on the method that was kindly put forward by tv84.

dd bs=512 seek=2000000 skips 2*10⁶*512 bytes of memory before writing the contents of the image file. Hence, the USB stick needing to have more memory than 1Gb.

In short: When the DG### sees the contents of the image file on a USB medium from byte 1024000000 onward, it will be listening to the "special" SCPI commands.

My first try was formatting with OS X Diskutility (GUI) and then write the image via dd on the command line. However, SCPI hack did not work afterwards.
Second try I deleted the Volume with diskutil from the command line and then wrote the .img via dd and my DG812 would start as a DG992 on next reboot.
There is no confirmation / ACK sent back by the device though (at least I did not get a response code via SCPI) and you will only see if it worked after restarting the device.

In case you do not know how to use diskutil, how to identify the correct mount point and how not to screw up your other hard drives / media connected to the computer you are working on, please get familiar with what dd and diskutil actually do. It is quite easy to shoot yourself in the foot with those tools.

Please let me know if you need more information.

Also, there is a dd port for Windows: http://www.chrysocome.net/dd
It is quite an old release and I did not try it myself, but maybe this is of use to some of you.

Kind regards
Frederik

[gossamer]

I had some issues with preparing the partition on the usb flash drive from OSX. I ended up using linux VM and followed this tutorial: https://www.redips.net/linux/create-fat32-usb-drive/

after which I dd the file and everything went smooth.

[maxwell3e10]

I got mine to upgrade. I used HxD to edit the USB drive and MATLAB tmtool to connect to the instrument. It went smoothly. You can tell that the upgrade worked when the color scheme changes and even the output LEDs switch color from green to red.

I tested briefly the DG811(->DG992) in comparison with Keysight 33500B series. There are obvious limitations of the DG811: overshoot and time jitter are worse, I will post some data later. So, it can't replace a higher-quality generator, but probably a good deal for 1/4 the price of a used Keysight.

[TurboTom]

Just noticed that using CH2 and the frequency counter are mutually exclusive. Apparently, the hardware resources (FPGA cells / grunt) won't permit contemporary operation. Considering the market that Rigol's new(ish) low end AWGs are intended for, the integration of a frequency counter may make some sense, especially since Rigol's implementation is quite nice, providing statistics and graphical display. But including it and providing an extra BNC input for the counter and then skimping on the hardware in that way is ridiculous! 

Edit: Some more playing around seems to indicate that the situation is still way more severe if the counter function is enabled: If statistics is active, I won't be able to enable sweeps and other more complex functions on channel 1! Moreover, the generator always reverts to sine waveform. When changing to square wave, duty cycle appears to be limited. I guess a major firmware fix is required from Rigol to set this straight! Shame, I thought Rigol this time got a half-way bug-free instument out but this seems to prove me wrong.
Very strange: After power cycling the instrument and reconfiguring the aformentioned setup, I wasn't able to reproduce my findings and the AWG appears to work as expected. 

[maxwell3e10]

I played a bit more with mine DG811(DG922) and seem to have found that it has terrible voltage noise level. If  confirmed by others, it would indicate that 16 bit resolution is nothing but a marketing gimmick!

Here is the simple setup: 10Vpp (Hi-Z) sine wave at 1 kHz with 5 Vdc offset. So the bottom of the sine wave is at 0 V. Now zoom onto the bottom of the wave with an oscilloscope set to 50mV/div or 100 mV/div. Note that not all oscilloscopes can do it, it needs to have a good overload recovery. Among the ones I tried, Owon and Tek could zoom in, while Keysight and Micsig could not.
Here are a couple of pictures comparing the noise level  of DG811 (channel 1, top trace) with Agilent 33500 and with Rigol DG4162, both are much better than DG811.
I have mistakenly tested the noise level while adding extra noise using the waveform combine feature. See message below.

[timber23]

I played a bit more with mine DG811(DG922) and seem to have found that it has terrible voltage noise level. If  confirmed by others, it would indicate that 16 bit resolution is nothing but a marketing gimmick!

I repeated your measurement using Siglent SDG2042x as a reference. The DG800's trace is displayed in yellow and the SDG2042x's trace is displayed in blue. I do not see any terrible noise level. Everything seems fine.

[TurboTom]

@maxwell3e10 -

I replicated your tests with an ultra-low frequency square wave (since then the overdrive tolerance of the scope input isn't as relevant to the test) and found more or less similar noise figures for the DG811 and the DG4102, both averaging at round about 850µVRMS. Interestingly (but not unexpectedly), the output of the Siglent SDG6000X is much more noisy with figures round about 3.1mVRMS (bandwidth of at least 500MHz). The spectrum of the DG811 appears to be just white noise without the sampling filter drop at ~120MHz which leads to the assumption that the noise is really generated in the analog output section. I think adding screenshots doesn't make much sense since they look rather boring, the only important figure is the RMS measurement...

Since my Rigol scope isn't as tolerant to input overdrive than the model(s) you used, I cannot replicate your test with the sine wave output directly. And I'm too lazy to pull out the TEK2465 from the darkest corner of the basement for the test...

[maxwell3e10]

Duh! Yesterday I was playing with the feature of combining two waveforms on one channel and added noise to it, then forgot to turn it off! Thanks for checking it and setting me straight, TurboTom.

So now that I turned off extra noise, the noise level of DG811 is just a bit worse than Keysight 33500 (see below). It is very similar to the level of noise from DG4162.

[maxwell3e10]

Here is a more detailed comparison of the residual noise spectra for Keysight 33500B and DG811 generators. It is taken again near the bottom of the sine wave. I assume the noise rise around 200 kHz is due to switching noise of the ADC. So it appears that DG811 actually has less ADC noise while it has more analog broadband noise, but that can potentially be filtered out.

[maxwell3e10]

With low pass filtering at 30 kHz using Micsig scope one can indeed see that DG811 (recorded on Ch 1) is quieter at low frequency than Keysight 33500B (Ch 2). This is recorded for a 0.1 Hz 10Vpp sine wave. I was hoping to see individual ADC steps, which should be equal to 0.15 mV for 16 bit ADC with 10V full range. But the noise is just a bit too high.

[timber23]

With low pass filtering at 30 kHz using Micsig scope one can indeed see that DG811 (recorded on Ch 1) is quieter at low frequency than Keysight 33500B (Ch 2). This is recorded for a 0.1 Hz 10Vpp sine wave. I was hoping to see individual ADC steps, which should be equal to 0.15 mV for 16 bit ADC with 10V full range. But the noise is just a bit too high.

I have repeated your measurement using Rigol MSO5000 to display the sine-wave. You have to change vector-view to dot view. Then the individual ADC steps are visible.

[maxwell3e10]

Yes, there are steps, but they are due to time discretization not ADC bits. Notice the step size is about 1 mV. Apparently Rigol used only 2^15 time steps in calculating the sine wave. Here are the measurements using DMM7510 with very slow frequency of 0.1 mHz. Near the top and bottom of the wave the signal is smooth. But on the slope one can see the time steps and even a glitch when the sine wave crosses zero.
Perhaps they can update the firmware, for a 16 bit generator one needs about 2^18 time steps. Otherwise, one can generate a arbitrary waveform with more points.

[geezer458]

Two things about this unit I do not like:

They advertise 250 MHz sampling for ARB or at least strongly imply that. In fact, ARBs that you load up as signed 16 bit integers are played out at a maximum rate of 60 MHz. If there is any trick to do this better I'd like to know it.

If you use external modulation it is filtered and then (apparently) sampled at 1 MHz and then digitally multiplied without interpolation, so unless you like ugly waveforms it's not really useful for anything above audio frequencies. I admit that the second item is a bit of an optimistic stretch on my part but it is (as usual) very misleading to state that the analog modulation has a BW of 1 MHz because that makes you think you could use it at 100 kHz and it would be OK. Not!

I don't think they understand what an arbitrary waveform is. They provide a bewildering array of strange waveforms that they made up from math functions and other stuff and you can select an arbitrary waveform from that group. And everything from that group has their 250 MSPS mumbo jumbo going on, but an actual arbitrary waveform is left out in the cold, so to speak.

[maxwell3e10]

Welcome to the forum.  In the spec for DG800/900 it says the external modulation bandwidth is 50 kHz. For comparison 33500B has 100 kHz analog modulation bandwidth. I haven't tested that feature.

[thm_w]

Yes, there are steps, but they are due to time discretization not ADC bits. Notice the step size is about 1 mV. Apparently Rigol used only 2^15 time steps in calculating the sine wave. Here are the measurements using DMM7510 with very slow frequency of 0.1 mHz. Near the top and bottom of the wave the signal is smooth. But on the slope one can see the time steps and even a glitch when the sine wave crosses zero.
Perhaps they can update the firmware, for a 16 bit generator one needs about 2^18 time steps. Otherwise, one can generate a arbitrary waveform with more points.

The step size is 200uV in his screenshot. I'm seeing 70uV as the lowest discrete step, which is a limitation of the oscilloscope (1mV * 10 / 255 = 40uV, close enough). To use a DMM makes more sense as you've done I think.

I can't see any noticeable difference between the MSO5000 internal generator (AD9744 14-bit), and the DG811 for sine wave (1kHz, 0.1Hz), in terms of noise, etc.

If I look at the horizontal time step, DG800 has a discrete step of 32us (15-bit as you've measured). Internal generator has a step of ~76us (14-bit) for 1Hz sine.
So might report to Rigol, see what they say.

edit: 8Mpts memory should be capable of 125ns division with a 1Hz signal. 1Mpts = 1us. So the limitation shouldn't be that memory, even if you half it for 2 channels. Could it be it takes long to load the memory?
edit2: sent info to rigol.

[thm_w]

They advertise 250 MHz sampling for ARB or at least strongly imply that. In fact, ARBs that you load up as signed 16 bit integers are played out at a maximum rate of 60 MHz. If there is any trick to do this better I'd like to know it.

But it doesn't say 250MHz in the spec anywhere, its very clear:
https://beyondmeasure.rigoltech.com/acton/attachment/1579/f-08a0/1/-/-/-/-/DG800%20Datasheet.pdf
https://beyondmeasure.rigoltech.com/acton/attachment/1579/f-08a1/0/-/-/-/-/DG900%20Datasheet.pdf

240MHz frequency counter, 35MHz for sine, 20MHz for arbitrary waveform.

You will find the same advertising on Keysight and other generators, in terms of Msps figures.

[geezer458]

Thanks    I will own the fact that I didn't read every line of the DG900 datasheet I quote below in a jpeg attachment
(not sure how to make it inline). I think you could see why after viewing the imaged snippet that I was satisfied that I
could use the modulation feature at 100 kHz. There were duplicate sets of lines after that referencing Internal/External
and a 1 MHz BW over and over, and I had already read something about the input impedance in another place. I had
seen what I was looking for already and didn't find my way down to the "catch" where they say the BW is 50 kHz. I'm
not sure that I would like the results of modulation at 50 kHz either but at the end of the day I would agree that the
fault is mine. Caveat Emptor.

About the sampling, here is the content of the current advertisement on Amazon:

100MHz Function Generator, 2 Channel, 250MSa/sec, 16Bit Resolution, 16M Memory
100 MHz function / arbitrary waveform generator
16 Mpts memory depth per channel for arbitrary waveforms

and here is what is at the top of the data sheet mentioned by thm_w:
----------------------------------------------------------------------------------------
Unique SiFi II (Signal Fidelity II) technology: generate the arbitrary
waveforms point by point; recover the signal without distortion; sample
rate accurate and adjustable; jitter of all the output waveforms (including
Sine, Pulse, etc.) as low as 200 ps
16 Mpts memory depth per channel for arbitrary waveforms
Standard dual-channel with the same performance, equivalent to two
independent signal sources
...
Sample rate up to 250 MSa/s, vertical resolution 16 bits
Arbitrary waveform sequence editing function available; arbitrary
waveforms also can be generated through the PC software
--------------------------------------------------------------------------------------
I don't remember, but I must have read the material below that and perhaps
wondered what they meant by:

Basic Waveforms Sine, Square, Ramp, Pulse, Noise, DC, Dual-tone
Advanced Waveforms PRBS, RS232, Sequence
Built-in Arbitrary Waveforms 160 types of waveforms, including Sinc, Exponential Rise, Exponential Fall, ECG, Gauss, HaverSine,
Lorentz, etc.
.....
Sequence 2 k to 60 MSa/s
Noise (-3 dB) 100 MHz bandwidth
Arbitrary Waveform 1 μHz to 15 MHz 1 μHz to 20 MHz 1 μHz to 20 MHz

It looked like what they meant by ARB waveform was their collection of functions so I chose, perhaps correctly,
to connect the 20 MHz limit with those things, believing that they referred to a limit for playing out predetermined
lists of points. And that may not be wrong.

What I should have understood is that, by their terminology what I was doing was an "Advanced" "Sequence"
to which the limit 2 k to 60 MSa/s applies.

Well, all right.

In the end, I made a generous interpretation based on the headlines and what I wanted it to be. I didn't call them
because it was likely to be a waste of time. Very few companies anymore let you talk to people who aren't reading
canned scripts. I have bought exactly two pieces of Rigol gear and was very favorably impressed with their capabilities
so this time when the item provided less value than I expected rather than more it made me unhappy.

Unfortunately I didn't know that there was a big thread about this unit before I bought it and I think reading that might
have stayed my hand. I am not a regular reader of EEVBLOG, shame on me!

[maxwell3e10]

Edit: Added the screen shot of my sample rate test. I used an arbitrary waveform with as many steep edges as possible for that. The reason is that the "controlled edge technology" that Rigol calls "SiFi2" is not operational in Arb mode. This means, slopes are defined by just a single sample interval, also recognisable by the slightly faster rise/fall times and (even) more ringing. Now I selected a frequency at which the signal cannot be reproduced by full integers of the sampling frequency, resulting in edge jitter. I measured the time between the two traces that are shown when the scope "averages" over several scans and calculate the frequency which should resemble actual sample rate. And voila, 250MHz!

So it seems that it can output an arbitrary waveform at 250MHz, but I am not sure if its one of the pre-programmed ones, or a custom one.

[maxwell3e10]

Here is another screenshot of the steps in sine wave. This is for a 1 Hz 20Vpp wave. DG811 is yellow, 33500B is blue. If I didn't know any better, I would say at first glance DG811 is not really a 16-bit sine wave, more like 13 bit. But it is due to 2^15 time steps. The Keysight has 4 times as many, 2^17 steps.



This affects only low-frequency sine signals. For frequencies above 7.6kHz one cannot output more than 2^15 steps anyway at 250MS/sec.  Still, generating high-fidelity signals in the audio band is something it should be capable of doing. So hopefully Rigol can address this with a firmware upgrade. Do you know who on this forum represents Rigol?

[thm_w]

This affects only low-frequency sine signals. For frequencies above 7.6kHz one cannot output more than 2^15 steps anyway at 250MS/sec.  Still, generating high-fidelity signals in the audio band is something it should be capable of doing. So hopefully Rigol can address this with a firmware upgrade. Do you know who on this forum represents Rigol?

Oh good point I didn't consider the frequency aspect of that. No one on this forum represents Rigol, but I have updated my case that it would only effect lower frequencies (250M/32768 = 7.6kHz).

They must have something in the software already to decimate the 15-bit calculations when outputting, say a 30MHz sine right? So hopefully its not too hard to add.