Rigol RSA3000N/RSA5000N VNA software needs some work

[bayjelly]

So I've bought a Rigol RSA3015N. My choice fell on this device because I wanted a Realtime Spectrum Analyzer and a VNA, and with the current promotions (that include the EMI mode and the Preamplifier Option, which I want both) I felt that's what I got, for an attractive price.

I don't think I regret my purchase so far, but I do think that the VNA software really needs some more work. It is my impression that the VNA mode might be a later addition and therefore a newer part of the software, so I really hope future software updates will fix it. It's not unusable, but some things do really get in the way.

Here is a comprehensive list of problems that get me. Not all things are VNA mode specific, but apply to other modes as well. The VNA specific once are mostly about S11 measurements, since that's what I have done most so far. This is all with version 00.03.04, the current version at time of this writing.

• Fixed in 00.03.05: R+jX mode now shows capacitance/inductance as well.There is no way to directly read off capacitance or inductance from the Smith Chart mode traces (as e.g. "76Ω 80nH"), or I have not found any. Many other VNAs offer that, even the NanoVNA. There are Smith Chart modes for R+jX (complex impedance) and G+jB (complex admittance) however, so don't be fooled by the initial mode which shows reflection coefficient magnitude and phase, you do not have to calculate that by hand.
• Fixed in 00.03.05: There is now an option to export a trace as SNP. I have not tried importing that trace anywhere, but the file looks sane at first glance.I did not find a way to export a Smith Chart trace as Touchstone, CSV, or anything else (unlike in other modes where trace export is usually available). I worked around that by saving "Trace&State", copying the state from the VNA, and writing a parser for the idiosyncratic INI-like file to convert it to CSV (what I needed), which I may post here, but this seems like relatively basic functionality.
• Mostly in VNA S11 mode, but in some other modes as well, the maximum peaks in a trace are uninteresting, and you really care about the minimums. The device has a single "Minimum Search" command in the "Peak" menu in general, and VNA mode adds a "Cont Min" command (under "Cont Peak") that keeps the marker on the current minimum, but all the other Peak-related commands like "Next Peak", "Next Peak left/right" still work on Maximum Peaks, which makes them mostly useless for e.g. the Return Loss and SWR traces.
• There is also no Peak Table anymore in VNA mode. Since that presumably would be for the maxima as well, maybe Rigol realized that it does not make much sense for VNA Mode? It would be most useful to have a Minimums Table instead.
• Semi-Fixed in 00.03.05? The Smith Chart does seem to have proper acceleration of the rotary encoder, but the S11 Log Mag trace for example does not. Not having proper acceleration on the rotary encoder (globally, not only in VNA mode) makes the two aforementioned points about lacking multiple minimums search and table much more painful. It's also painful in the Smith Chart trace.
• Fixed in 00.03.05: The R+jX value now updates while turning the knob, without jumping to 50Ω/0.00mΩ until let off.Speaking of the Marker in the Smith Chart: When turning the knob, while the visual Marker itself moves swiftly to its new position, the reading of the marker needs to "settle". For a while, it will show a reading as if the marker was in the center of the chart (e.g. 0.00aU/0.00° or 50Ω/0.00mΩ depending on trace mode) before showing the actual value. If you turn the knob quickly, at some point all you see if "50Ω/0.00mΩ" until you let off.
• Improved in 00.03.05: You can now save "CalData&State", and after a quick test that seems to save the cal kit parameters as well. Still would be nice to save multiple cal kits, but this works for me.There is only one "Custom" calibration kit entry, and no way separately save it. You have to save the whole preset or state. As a slight workaround I saved my "favorite" Cal Kit as the Preset used at Power On, but since both Presets and States include all the other settings as well, this makes working with multiple Cal Kits rather cumbersome.
• I'm pretty sure I saw "Port Ext" in the summary box at the top right either saying that Port Extensions was off when it was on, or the opposite. I think it had something to do with restoring a State or Preset, and upon cycling the setting the display reflected reality again. A bit dangerous, since thinking there might be a port extension when there isn't (or opposite) might ruin things.
• If you change the frequency span after calibration, you get completely bogus results. I am not expecting the VNA to do some magic here: Recalibration is most likely necessary, at best the VNA could drop the number of points when narrowing the frequency span to keep only the calibrated points within the new span (but this would change the number of points to acquire behind your back, so I think I'm okay with having to always recalibrate). However, the VNA seems to internally just keep a list of calibration points with no frequency association, and when you make a new measurement with changed span, the result is a wildly bogus trace that makes no sense at all. This can be enormously confusing if you didn't catch the tiny "C ?" note at the top of the screen that indicates you "may" need new calibration. You definitely do here. Made worse by the fact that it seems that sometimes "C ?" does indeed mean that your calibration is still good, see the next points.
• If I enable a Port Extension, I get the "C ?" note that indicates I may need recalibration. Is that how it's commonly done? I thought a port extension would not invalidate the current calibration?
• The "C ?" seems to be "sticky" even if you revert to the settings that were calibrated, e.g. flipping the port extension (whether you'd need recalibration for that or now) on and then back off. This further diminishes the importance of the "C ?" note, despite it being so crucial if you changed the frequency span.
• Switching to another mode (e.g. GPSA) and back into VNA seems to keep every setting but not the calibration data. Again the only indication for having to recalibrate is the tiny "C ?" note, or a wildly inaccurate trace with your next measurement.
• In all modes, when entering a file name, the name is pre-filled and I found no way to just clear it all at once. No little "x" button as with R&S units, no long press of the backspace key, nothing. This is especially infuriating in VNA mode, where the UI seems to be slower and a backspace press takes a long time to react. Unless I am happy with the pre-filled name (I'm usually not), that means that every time I want to save something, I have to spend an awfully long time repeatedly pressing backspace. This seems like a very elemental UI inadequacy.
• Again in all modes, this gets me every time: If a setting where you enter a number is selected, you can start entering a number on the number pad, and the soft keys switch over to offering what unit to complete your entry in, e.g. you type in "10" and can then press the soft button for "-dBm" and have entered "-10dBm". Good! However, if the setting was already selected/highlighted, and you press the soft button again, the touchscreen number pad pops up, because the unit assumes you want to use the touchscreen (same as if you would have touched the item on the screen instead of using the soft key). Now, you can still use the non-touchscreen keypad, but the soft keys do not switch over to the units when you do now, forcing you to use the touchscreen buttons to complete your entry with a unit. This adds additional cognitive load, because I now need to be careful not to select an already selected soft key, or my muscle memory of using the soft keys won't work. This would for example be entirely solved by just also switching the soft keys to the unit keys, or alternatively only popping up the touchscreen entry if reselected using the touchscreen instead of with a soft key.
• Nitpick: I'm not a fan of the default presets in VNA mode. The two top traces are "Smith (Lin/Phase)" and "Polar (Lin/Phase)". Both show exactly the same trace, and both read reflection coefficient mag and phase. The only difference is the graticule in the background: Smith left and Polar right. I personally set that to Smith (R+jX) left and Smith (G+jB) right. The trace is still the same, but now I have different readings (impedance/admittance respectively) in addition to different graticules, and arguably a more useful one instead of Polar. Also, a 201 point measurement by default seems a bit low, maybe that was chosen because that's almost exactly 1 second with the default/lowest bandwidth of 1kHz? And finally the third trace, which shows reflection loss, has 0dB in the middle instead of near the top. But eh, all of that can be changed once and saved as a power on preset, so it's not a big deal.

If I am just missing something that is there, or misunderstood something, please let me know.

[bayjelly]

So very recently, a new firmware version came out, with the following release notes:


00.03.05
   - Fix some bugs
   - Add screenshot feature in  Web Control
   - Add SCPI error query command
   - Add capacitance or inductance value display in VNA's R+jX and G+jB mods
   - Add continuous minimum value point search in GPSA mode


I have not had the time to try it out yet, but it looks like this addresses my first complaint regarding a missing feature: The ability to read capacitance and inductance directly off the Smith chart.

There is no telling whether any other of the issues have been addressed without trying the new version out (I really wish Rigol would put out more detailed notes than "Fix some bugs"), which I will do when I find the time.

It's interesting that GPSA mode got continuous minimum search now. There is a vague hope on my part that this may also affect the several issues I have with minimum search in VNA mode being "incomplete", but we will see.

[bayjelly]

Alright, I updated to 00.03.05 and modified my original post to write down what was fixed and improved. I was actually happy to see that some of the points above were addressed, this makes the device much more useful and I'm glad to see some forward progress.

Note that I could not do extensive testing. Some of the things I marked down as "fixed" appeared to be fixed to me after a quick test and may not be, but also some of the open points might be fixed and I just was not able to test it.

[ormandj]

How are things going with this device now for you? I've been considering picking one up for much the same reason as you, but definitely am interested in the VNA aspect for my HF amateur radio uses.

[bayjelly]

I should first add that for many reasons, I cannot give an assessment either way on whether I recommend the device for professional use. I purchased my device for hobby purposes, and that's the lens I'm viewing it through. But since you mentioned amateur radio uses, it seems that that's okay.

So, I'm glad some of the more jarring issues from my first post were fixed, but I am still anxiously awaiting more software updates with fixes for the remaining jarring issues. Overall however, I still don't regret my purchase and mostly like using the device. I wish it was as polished as my R&S oscilloscope for example, but it's got a lot of "bang for the buck" making up for it, considering that I use it for hobby use only.

However, and this may be important, you explicitly mention "HF amateur radio". If you literally mean all of the HF band between 3 and 30MHz, and you plan to use the VNA feature in that band, you may be a bit disappointed. In the lower frequencies, the device measurements appear very... "noisy". This is better explained with the attached picture, where I sweeped a (completely mismatched) antenna from 3 to 30MHz. You can see that the lower it goes, the more the individual points jump around erratically in the Smith chart. This is with averaging enabled: The data points seem to be somewhat stable between measurements?

I don't know how common this is across VNAs, and whether I could have gotten a better result with some settings. At about 15MHz (where I put the marker) is where it gets stable enough that I consider it really usable for my purposes. Below that it's not necessarily completely unusable I guess, but very limited. I added the second screenshot where I sweep the 50 ohm load of my "cal kit" (just a bunch of Rosenberger components) for illustration. This should be flat at 1:1 SWR, and yet in the lower right pane, you can see that while the SWR is still staying below 1:1.50, it's getting dangerously close to that line. Unless I did something wrong (feel free to point that out), that's about the precision at which you can expect to tune an antenna for those lower frequencies. (Pay no attention to the seemingly erratic phase however, that's expected to jump around with a matched load, because the phase loses meaning the closer the magnitude goes to 0, as it should with a perfectly matched load.)

For comparison, I also added a screenshot of a sweep from 50MHz to 1.5GHz (this time of an "antenna" haphazardly made of some cable lying around to get an interesting pattern). You can see that there is no issue there, and while my "piece of string" antenna does not show it, I have reached SWRs of 1:1.02 and lower when sweeping antennas for the 2m band and others.

[bayjelly]

Ah. It seems that Rigol even corrected the data sheet for that. In this data sheet from 2020, page 13 claims the range for the VNA is from 100kHz to the maximum range, albeit with a footnote that says "In S11 measurement, the performance becomes worse when the carrier frequency is smaller than 10 MHz.".

The current data sheet however just says that the range starts at 10MHz, and the footnote is gone. You can still sweep below 10MHz, though.

Both data sheet revisions show a graph for the return loss starting at 1MHz. You can see it starting out at only -15dB for 1MHz before getting better, so that should directly explain what SWR you can expect there.

[Andrey_irk]

I don't know how common this is across VNAs...

Most professional VNAs have frequency ranges starting from 10MHz. The limit comes from the directional coupler architecture. There are a few exceptions from that, KS extended the E5071C's frequency limit down to 300k but the accuracy is far worse in that region.
There are resistive couplers that have very low starting frequency limit (Anritsu has that in their VNAs which go down to 70kHz), but they aren't popular because you lose quite a bit of power on them (might be other limitatins also).

[switchabl]

That used to be the case with microwave VNAs but most current <20 GHz models from R&S, Keysight and also others like Copper Mountain go down into the kHz range with a somewhat degraded dynamic range. Beyond that, it's still mostly 10 MHz unless there is a separate measurement path for low frequencies (like on the PNA-X with 900 Hz LFE option).

[G0HZU]

If it helps, I've been using HP/Agilent/Keysight VNAs at work all my career including the 8753/8714/E507x/PNA models.
The 8753 models can operate down to 30kHz and the 8714 and E507x models typically operate down to 300kHz although some E5071C models can work down to 9kHz. These are all models that cover up to either 3GHz, 6GHz or 8.5GHz.

There is some degradation in performance below 3MHz with these VNAs but it is fairly minor once a full cal/correction is done. By comparison, the sub 10MHz plots of the Rigol VNA look to be really awful in terms of noise. The performance looks to be really poor here.

It's generally the >=14GHz lab VNAs that are spec'd above 10MHz. Some are spec'd for use above 50MHz.

If you want to keep the Rigol RSA/VNA, I would recommend buying something like a Hugen nanovnaH for VNA use below 50MHz. These little VNAs have excellent performance across LF through to about 50MHz. If the nanovnaH doesn't have enough capability then maybe buy an old HP VNA for low frequency stuff.

[EE-digger]

In keeping with the original title, I can vouch for the fact that in 2024, VNA software still needs work.

I'm in the process of returning a new RSA3015N after 2 intensive days of evaluation.  I won't go down the long list here but the final straw that broke the camel's back is that in S11, there are NO marker functions.  You can't mark a valley and  perform a bandwidth measurement, it just ain't there.

This is ridiculous that something like bandwidth of a filter (band or notch) or response of an antenna can't be measured.

This coupled with high noise below 300MHz or so in S11 are just a few of the reasons it's going back.  Rigol's support also shows that they are new to VNAs.  I will be kind and leave it at that.  Two lengthy sessions were a complete waste of time with lack of fundamentals.

Kinda sad because it seems to be a solid product, built "tighter" than Siglent, but Siglent has it all over them for speed, performance and software (with its own minor "niggles").  But taken further, if I had paid close to $10k for an RSA5065N, I would be totally pi$$ed that the product is in this shape.

added - for lower frequency S11, the Siglent SVA1032X (created from a younger sibling  ) cals to under -50dB or less from 100kHz to 3.2GHz.  If you're working in the low end, then calibrate in the low end.  This will allocated more points to the cal.  The same holds true for a log sweep.  Cal and test with 1601 points or more for enough data points at the low end of the log sweep.

disclaimer:  I have not received any compensation for any glowing comments on Siglent, but would sure like to    For the Rigol side, I would be glad to help with constructive input for development but at the prices being charged, the products should already be complete (more or less).