SDS1104X-E Bode Plot with non-Siglent signal generators

e.g. the HP 8904A Multifunction synthesizer

or the Agilent E4421B ESG-A RF signal generator

See this page on how to set up GPIB on a Raspberry Pi! If that's already done, or if you're controlling your RF signal generator over serial/RS232, USB/USBTMC or Ethernet/VXI-11, keep reading! (Or if you don't want to read further, just download sds_pyvisa_bode.py.)

I finally decided to pay up and get myself a modern oscilloscope, the Siglent SDS1104X-E. Not only does it have exceptional value for money, it is easily modified into the 200 MHz version (or so I hear, I have not done that myself—at least not yet). But the clincher for me was the Bode plot function. Ostensibly, it requires a Siglent arbitrary waveform generator to work (which also have great value for money), but I began thinking... Could I use one of my existing signal generators to emulate a supported Siglent AWG, using a microcontroller or a Raspberry Pi to translate SCPI commands between the two?

Well, it turned out that Dmitry 4X1MD had already had that idea, and had done all the heavy work for me! There's a long EEVblog discussion on the subject, and his Python code does exactly what I wanted: it identifies itself to the oscilloscope as a Siglent AWG, and then translates commands sent by the oscilloscope (over Ethernet) into commands that several other non-Siglent arb generators (connected over a serial/RS232 interface) can understand. All I would have to do was to write a simple "driver" module to interface with my Agilent E4421B ESG-A series RF signal generator, which can produce signals from 100 kHz upwards, way past the scope's maximum. It is not an arbitrary waveform generator, but the scope only ever requests sine waves from them anyway. The only real limitations are that this generator will not work down into the audio range, and it only has one output channel, so a splitter will be needed to provide a signal both into the DUT, as well as into the scope's reference input. (Pretty soon after that was working, I added support for my HP 8904A multifunction synthesizer as well. That works from 10 Hz up to 600 kHz.)

Dmitry's code includes a "dummy" driver which only responds to the oscilloscope's commands and reports its activity on the terminal, without actually interfacing to any signal generator. I used that driver as a first test, connected the oscilloscope onto my network, and saw that it would indeed happily generate a Bode plot (albeit one showing noise only, as there are no signals to measure), apparently convinced that it's communicating with some model of Siglent AWG. On a whim, I connected a probe to CH2, hooked it up to the probe compensation terminal, and made a Bode plot around its 1 kHz frequency. The phase plot (red) is garbage, of course, as the reference channel (CH1) was unconnected, but the amplitude plot (purple) shows a clear peak at 1 kHz! The Bode plot function apparently uses frequency discrimination (probably the same FFT routine that's already built into the math functions) instead of just measuring a simple and stupid peak-to-peak amplitude of the raw input signal.

The SCPI commands needed to control the RF generator are extremely basic: *RST and SYST:PRES to reset the generator to its default state OUTP ON to turn on the output FREQ 12400000 HZ to set the frequency to 12.4 MHz, for example POWER 8.1 DBM to set the output amplitude to 8.1 dBm, for example SYST:ERR? to check for an error These commands are so basic, in fact, that I'd be surprised if this driver wouldn't work with pretty much any RF signal generator that understands SCPI!

All seemed easy enough, but when I began to examine Dmitry's code, I couldn't make head or tail out of it.  :)  Granted, I'm a total newbie with Python, I don't know anything about object-oriented programming, and I've never been involved with programming projects of any significant size. No doubt more experienced programmers would have no problem with this. But also, I was intending to use the PyVISA library for communicating with the generator over GPIB, and that required Python 3.6+. Alas, Dmitry's code did not run on Python3. Eventually I bit the bullet and rewrote the entire code in my own style (i.e. in simple linear fashion) using Python3 from the start. I learned a lot about sockets and servers, and about Python in general, in the process.

I first wrote the program to work with my Agilent E4421B RF signal generator, and later added support for my HP 8904A multifunction synthesizer. I tried to keep everything relating to the RF generator as model agnostic as possible, in the hopes of having the code work as-is with any RF generator that talks kosher SCPI. The older 8904A, however, uses its own command language, so it's a special case. The program will detect the 8904A based on its ID string and use its specific commands. With any other generator, the program will use the generic SCPI commands shown above.

There are some design choices the user needs to be aware of:

• Except for the HP 8904A, the program does not check the generator's make and model against any database of supported or unsupported generators. It either will work, or it won't.
• No frequency or amplitude limits are hard coded into the driver! The oscilloscope will happily draw grabage output if trying to work outside the generator's supported ranges.
• Any errors reported by the generator are only relayed to the terminal where the program is running, so it may be good to watch. The output messages fall under various "log levels". Unless there's strange problems occurring, it's best to use the INFO level, which will only report real problems such as out-of-range errors from the generator. The HP 8904A will not report any errors to the terminal, but will beep instead.
• The signal amplitude assumes a load of 50Ω! Since the program cannot know what kind of of splitter is used at the generator's output, and what terminations (if any) are installed at the oscilloscope's inputs, the user is responsible for compensating the signal amplitude correspondingly.
• The HP 8904A driver tries to drive both outputs of the generator with the same settings. If the generator only has one output, a splitter of some kind will be needed.
• For some reason, the oscilloscope never commands the generator to turn its output OFF, but the program does that when execution is ended with Ctrl+C (but not if execution ends due to some error, probably).

• A minimum frequency of 100 kHz (the scope's Bode Plot II minimum frequency is 10 Hz).
• A maximum amplitude of +20 dBm into 50Ω—just over 6.3 Vpp before the power splitter. That means about 3.2 Vpp after, whereas the Siglent generators can provide up to 20 Vpp on both two channels.

• A maximum frequency of 600 kHz.
• A maximum aplitude of 10 Vpp into 50Ω—more if the DUT is high impedance and no terminations are used.
• Some HP 8904As have two output channels (OPTION 002), but mine is missing that option, so a splitter is needed.

I wanted to control both signal generators through the GPIB bus, because that's the way they're connected. I have several older instruments connected that way, in fact. And that's one reason why I chose to use the PyVISA library. The E4421B does also have a serial interface, so I could have used that exactly like the other drivers do in Dmitry's software, but PyVISA has the advantage of transparently supporting all GPIB, serial/RS232, USB/USBTMC and Ethernet/VXI-11 interfaces! By the way, the E4421B's serial port (labeled "AUXILIARY INTERFACE") is a 9-pin female D-connector, which one might plug straight into a PC's serial port with a straight cable, just like one would connect a modem. Or one might plug a USB-serial adapter directly into that port. But no, that would be too logical! In the manual, Agilent specifies you need to use a straight cable plus a null modem adapter and a female-female adapter to make the connection (and they specify Agilent part numbers so you can order them at exorbitant prices). In all, the pinout is the same as on a PC host, but the gender is opposite! Talk about non-standard... Was this an actual design choice, or a simple engineering booboo?

I dedicated an old Raspberry Pi 3B (soon replaced with a Pi 2B I found in my cupboard, to leave the more powerful 3B for more demanding use) to serve as the interpreter between the scope and the generator, and used an Agilent/Keysight 82357B USB-GPIB host adapter to connect the RF generator and a bunch of other instruments. See this page on setting up GPIB, and using PyVISA in general, on a Raspberry Pi. (I also segregated my scope, and possible future instruments with an Ethernet connection, to its own private network. Using a tagged 802.1Q VLAN interface, the Pi works as a single point of entry between the instrument network and my main Linux network. The Pi does not route any outgoing connections, so—call me paranoid—the oscilloscope won't be "phoning home" without my permission!)

	As there is only one output on the generators, I made this splitter to provide input signals for both the device under test, as well as for the oscilloscope's reference input. The rudimentary resistive splitter, made from three 15.9Ω resistors in star configuration, has 6 dB loss (which means an amplitude ratio of 0.5), and likewise 6 dB isolation between outputs (so if a significant amount of signal is reflected from the DUT due to impedance mismatch, part of it may end up in the reference channel). Yes, the construction is awful, but should be ok for this low frequency use. With both outputs properly terminated, the SWR was below 1.5:1 up to 120 MHz—quite bad, but good enough. But feel free to do this with SMD components on a proper etched PCB, and to add attenuators to improve matching and isolation, and LNAs to compensate for loss. And by all means, email me with your design and results!  :)
	Especially at RF frequencies, you should use 50Ω feedthrough terminations at the oscilloscope end of the signal cables. I did not have any in shelf, so I substituted BNC tees and ordinary terminators (both of which I had accumulated in masses when the Ethernet networks at my old university were upgraded from 10base2 coax to 10baseT twisted pair—yes, I'm really that old). I measured the terminators with my RigExpert AA-1000 antenna analyzer, and found that they work just fine—the oscilloscope itself causes a much worse mismatch to 50Ω than the tee and terminator do! So no need to go overboard with expensive inline terminations! If you need a better impedance match, and can afford an attenuation of 20 dB (which corresponds to the 1:10 amplitude ratio found in most oscilloscope probes), then a 20 dB BNC attenuator should work just fine as an inline terminator even without any final terminator in place. With the other end left open, it will present a 51Ω impedance to the source, corresponding to just 1.02:1 SWR! (The oscilloscope's capacitive input will make the SWR a tad worse, I think, but still better than the tee and terminator.)
	Once I had my program working, I was eager to see how my Agilent E4421B RF generator plays along with the Siglent oscilloscope. So even before making the splitter, I hooked up a 10 MHz crystal between the signal generator and the scope's DUT channel, taking the reference channel directly from the generator. I connected to the crystal using alligator clips and oscilloscope probes at 1× setting, with no 50Ω terminations anywhere... Not a proper measurement setup! But this is the Bode plot I got—the system definitely works!
	And this is measured using the HP 8904A multifunction synthesizer, feeding into a series LC circuit comprising a 33 mH inductor and a 1 μF capacitor (also without any proper terminations, impedance matched splitter, etc...). And it also works! (These screenshots, by the way, were obtained over the network with another simple script using PyVISA, which I made as a short exercise in Python. I find that much more convenient than having to plug a USB thumb drive first into the oscilloscope and then into my PC, just to obtain a screenshot and transfer it to where I need it.)

The program

Since the program binds to privileged ports, it must be run as root. To test the program, simply run ./sds_pyvisa_bode.py, which will run it in "dummy mode"—communicating with the oscilloscope but not with the signal generator. Configure the scope's Bode plot function source (Utility → Bode Plot II → Config → Source) to use the LAN interface and the Raspberry Pi's (or whatever computer you run the program on) IP address. The Bode plot function should now run happily, but draw garbage. After stopping the Bode plot function, hit Ctrl+C to exit the program. (I have once seen the oscilloscope hang if I killed the program while the Bode plot was in progress. Pressing and holding the power button for some 5 seconds shut down the hung oscilloscope cleanly enough, but regardless, I think it's better to stop the Bode plot before ending the program.)

To actually use the program, you must specify the PyVISA resource ID of the signal generator. The resource ID is a string like "GPIB0::7::INSTR" (instrument number 7 on the first GPIB bus, i.e. GPIB0), "ASRL/dev/ttyUSB0::INSTR" (a serial port instrument on /dev/ttyUSB0), "TCPIP::10.42.47.99::INSTR" (an Ethernet connected instrument at address 10.42.47.99) or something like "USB0::0x9876::0xABCD::SOMETHING1234::INSTR" (a USBTMC instrument). You can scan for resource IDs (other than Ethernet) by running this directly in Python3 with root privileges (ignore all the "Invalid descriptor" errors):

   >>> import pyvisa
   >>> rm = pyvisa.ResourceManager()
   >>> rm.list_resources() 

   ./sds_pyvisa_bode.py GPIB0::19::INSTR

NOTE: This information is provided as-is without any guarantees that it will work with your oscilloscope or signal generator. I'm definitely not an expert on SCPI or Python. But this kludge did work with my new SDS1104X-E oscilloscope (software version 6.1.37R8, hardware version 09-06) and my Agilent E4421B RF generator (middle-aged, obviously, since it's branded Agilent, not the brand-spanking-new Keysight, nor the venerable Hewlett-Packard we all used to love) and my HP 8904A (really old, obviously). By all means, do let me know if you get it working with your hardware, especially if you're using some other brand or model RF generator! Also, sorry if my Python looks like it's written with my left foot. I'm a complete newbie in this.  :)