RoadTest: RoadTest the TEKTRONIX RSA306B
Evaluation Type: Independent Products
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: Keysight FieldFox
What were the biggest problems encountered?:
Before i begin, this is a early early review. This was just after about two hours of use, I know Tektronix was looking for a review so i wanted to get a initial review and keep updating this review quite periodically as i compare it to the Keysight Fieldfox.
To begin, i would really like to thank Element14 & Tektronix for giving me the opportunity to review the RSA306B. I have spent quite a bit of timing using the network analyzers from leading manufacturers. My most basic experience is with the Keysight Fieldfox N9917A and use equipment up to 4 THz. I work in a terahertz research and use this equipment everyday and i also have taught classes on using this type of equipment. For my road test I was going to compare it to various 67 GHz spectrum analyzers but came to the conclusion of why compare this to equipment that cost well over 10 times its cost, so I will be comparing the Tektronix RSA306B to the Fieldfox N9917A.
From a initial stand view the Fieldfox is a all inclusive Network / Spectrum analyzer that operates from 30 KHz to 18 GHz. It is a fully portable system that works Okay for its intended use. But for the price of it (~$20,000 USD), it's going to be a great comparison for the Tektronix RSA306B.
Whats included in the box? Well,The unit its self, a USB flash drive that contains all of the documents and software to install, a calibration certificate, and a USB 3.0 Cable. This all can be seen below. For me i really liked the fact that everything you needed as on a flash drive because written manuals are almost useless now adays when you can easily use the "Find" feature in a PDF reader to get to the exact point you need.
The software I installed was the Tektronix SingalVU-PC i installed this on a few PCs to see how it runs.
The first PC I installed it on was a Laptop (4 gig ram, 1.2 GHz dual core) and it ran really well, it installed in a couple minutes and had zero issues running it on. This PC was just to see how it would run on a 5 year old laptop and yet again it ran with no problem, The first issue i ran into here was that the Tektronix uses USB 3.0 and my laptop only had USB 2.0 and the Textronix would absolutely not run in USB 2.0 mode. I will talk about this later on.
The second PC i installed this software on was my Personal desktop that i built a few months ago and use to run HFSS on. The PC has 64 Gig Ram and 6 Cores that run at 4.2 GHz and the software installed in less then 45 seconds and ran with no issues as expected. This PC has USB 3.0 ports and as soon as i connected the Tektronix it appeared on the SingalVu software almost immediately and worked really well and this was the PC i used to initially test it out. I did run into a issue of it randomly disconnecting and reconnecting right away, again I will talk about this issue a little later on.
The third PC i installed it on was the lab PC (12 Cores, 128 gig ram) and it installed no issue and ran perfectly. The PC case has usb 3.0 ports but i could not get the RSA306B to connect to the PC, I did not build the PC so i don't know entirely the internals of the PC or if it was security that was keeping it from connecting but i could not get the Tektronix RSA306B to connect. I will investigate this more, I intended to use the PC in the Lab because all of the equipment is down there but will just bring the Fieldfox and Agilent arbitrary waveform generator to my office to further test this more.
The software for this product was really nice, I can't say much about it because I've used this type of equipment and it only took me about 20 minutes to get familiar with the user interface and sorta learn my grounds with it. It'll be hard for me to really review this interface because across industry it is really similar, From moving from one product to this product it was really easy and anyone experienced in this field will think the same. They did do a great job with the layout and making it easy to navigate. I really like how easy it was to just go back to preset because it has a main button on the screen. My main issue with this software was the way it exported data. Since i was comparing the RSA306B with other equipment I am going to plot the data from both units on one plot. The SingnalVU exported the CSV file with about 130 lines of information and then only 1 column of the data. The CSV file did not contain the frequency points so for me to be able to plot it i had to go and manually add this to all the CSV files i exported. No matter what equipment I've used previously it always exports the data with both the frequency and gain. I will further look into this to see if there is a option for this. I have not checked the other file formats, but this was disappointing to see because it added an extra step into this when all other equipment and even simulation software exports both.
Since i had that issue with the LAB PC for this inital test i tested the RSA306B and Fieldfox with the a military spec voltage controlled oscillator (VCO). To begin with the RSA306B has a N-type connector so i had to use a N-type to SMA connector. This was the same with the Fieldfox as both can be seen below. The N-type to SMA connector and SMA cable are both rated for 18 GHz. So both of these are being tested under the same condition. The VCO was powered by a Agilent B2900A precision source.
Warm up time.
All this RF equipment is thermally sensitive so i gave it about a 45 minute warm up time since i couldn't really find any information on it in the documents. While measuring the VCO with the RSA306b i noticed that it was still showing that it was warming up but i continued with measurements about a 90 minutes since start up it finally went away and it finally had warmed up so i started back from the beginning with all the measurements which i didn't mind because i just got the chance to get more familiar with it. From a practical stand point it felt like quite a bit of time for reference the Fieldfox only needs about 15 minutes.
To just start i began by testing the noise floor of the RSA306B and compared it to the Fieldfox. The lower the noise floor the better the spectrum analyzer is. On high end equipment you can get ~ -180 dB noise floor and a noise floor of - 120 dB is considered good enough. Looking below you can see the noise floor for both the RSA306B and Fieldfox. The noise floor of the RSA306B was about -95 dB and had rather large spikes across the entire bandwidth that was measured. The Fieldfox had a rather smooth noise floor but was only around - 97 dB across the entire bandwidth (Note: the Fieldfox can be pushed to ~ -110 dB but I kept the resolution bandwidth high to measure, it becomes way to slow when you push the noise floor down to -110 dB). Both had slightly poor noise floor and would not be practical for any application that needed meet significant specification but I'm sure this was not the intent of the RSA306B. When i compare this to the Fieldfox mind you a device that costs 5 times as much as this, I think the RSA306B is a great competitor and when the RSA306B is pushed to its lowest resolution bandwidth it still updated really fast when compared to the Fieldfox.
The VOC have a center frequency of 2.256 GHz with a tolerance of .01%. The VOC's control pin was shorted to ground to just test at the center frequency. The Fieldfox measured a peak at 2.2557 GHz and the RSA306B showed 2.259 GHz. Again we see that RSA306B has great comparison to the Fieldfox. There was only a 0.15% difference in accuracy in measurement. But the one thing the RSA306B still maintains the spikes outside of the center frequency (same ones that appear in the noise floor) and this is one thing that
Initial Thoughts- (Like i said its a quick review! I will keep updating this. Please don't think this is the final review)
Until i update this is what I have gathered.
1) The SingalVu software works really nicely and works great with the RSA306B. For the time i used it i did get this weird occurrence of it disconnecting and reconnecting. I did have the USB3.0 screwed into the RSA306B and my PC never showed it disconnecting from the USB port just the software showed that.
2) The SingalVu should really output the frequency with the data so people can plot it easier.
3) The filtering in this system should really improved having a noise floor of only -95 dB is kind of low.
The main question I'm hoping to answer with further testing is, What is this good for?
I know this seems redundant but i don't entirely know what to think of this and what Tektronixs hope for this is. With a noise floor of -95 dB its not good enough for practical applications. Are they hoping to get into hobby level, but that raises the question of a price tag of $3800 to much for hobby level. This showed great comparison against equipment that is significantly more expensive but the Fieldfox is a full fledged 2 Port Network and Spectrum analyzer and is all self contained and not needed to be tethered to a PC. I will be testing the Fieldfox and RSA306B with the Agilent arbitrary waveform generator and hope to find the what the RSA306B is useful for or where it really shows its advantages. As of right now i find this as a great toy for a hobbyist but don't see the practical use behind it.
Update #1 I tested accuracy and retested Noise floor.
With the first update since I posted the initial review, I sorta began to like this a lot. One thing is its response rate to a low-resolution bandwidth was super fast compared to any of my other spectrum analyzer. It was responding as fast as our Agilent Network/Spectrum analyzer (67GHz ~$500,000 USD). But here are the few things I tested.
Noise Floor (With 50Ohm Load)
To make sure that the Tektronix was not picking any noise from the lights or anything I retested the Noise floor with a precision 50 ohm load. Yet again, it yielded no difference in results.
Note:I was able to push down the noise floor to ~- 110 dB but that required a resolution bandwidth of 1Hz and the system responded extremely fast to it. But this rose a question to me, Is this actually sampling at 1Hz or is it using some digital signal processing like interpolating this make this go much faster. The field fox took about 1 minute at 100 Hz resolution bandwidth to go from 2 GHz to 4 GHz, the Tektronix took about 5 seconds to do at 1 Hz from 2 GHz to 4 GHz. So I want to further look into what it’s doing so I can make sure it’s a accurate noise floor. I will talk about this more in a future update.
Now one of the most important things I wanted to test was accuracy. So I used our Agilent Arbitrary waveform generator to create a sinusoidal signal at 3 GHz (There was no modulation) just so I can verify accuracy. The system was tested from +20 dB (maximum) to -90 dB (lowest). The results excluding -90 dB was very good. The waveform generator was calibrated and is extremely accurate so this is a great source for this test. I tested at 10 dB increments and the Tektronix was extremely accurate at every level, See pictures below of a few of the tests., But when I tested at – 90 dB it was impossible to distinguish the signal from the noise. But besides that this was really nice for this equipment, I was really surprised and made me like it a lot more!
Accuracy at Highest & Lowest Frequency
Now the Tektronix was rated from 9 KHz to 6.2 GHz so I wanted to test how well it performs at these frequencies. I found out that it still maintained quite good accuracy at both of the highest and lowest frequency. What I found interesting though was it actually worked to 8.5KHz as I have shown in the picture below. I tested at 9 KHz and 8.5 KHz and both had the same accuracy. It was about 4% off of the source rating and at 3 GHz I saw that it was less than 1% off. In the higher frequency, it was limited to 6.2 GHz and would not go above that rating. Both of these tests can be seen below. High frequency limit showed a 4% error as well.
Next update- In the next update i will test power to noise ratio between the Tektronix and FieldFox.
As some ideas, it would also be fun for the next installments to show some example modulation capture (it could even be a car door-lock remote) to explore how the radio signal changes over time (e.g. frequency…
It is not unusual to see such a high noise floor, because according to the graph you have a 2GHz span (1-3GHz). It would take forever on a normal swept analyzer to capture at a lower noise level…
Inderpreet Singh wrote:
Looking at the front end bandwidth and response characteristics, I don't think the Tek stands a chance again the KeySight.
To be fair to the RSA306B, the Keysight N9917A is priced…
So the response rate was extremely well for a 1Hz Resolution bandwidth... i mean it was really fast. So i raised the question is it actually doing 1Hz or is it using some digitial signal processing (interpolating) or whatever to achieve this. In Update 3 i will test power-noise ratio again the Field fox.
As some ideas, it would also be fun for the next installments to show some example modulation capture (it could even be a car door-lock remote) to explore how the radio signal changes over time (e.g. frequency drift, preamble, datastream and duty cycle. These are things that are much harder with older analyzers but could be possible with the RSA306b. It would make for a super-cool blog post. Also to what depth Bluetooth Smart can be explored since this could be a more popular thing, and also any high bandwidth radio sources that are available.
Also, screenshots of the software would be very interesting, and video if possible, because one of the benefits is the real-time aspect and that is harder to see with CSV files into Excel.
It is not unusual to see such a high noise floor, because according to the graph you have a 2GHz span (1-3GHz). It would take forever on a normal swept analyzer to capture at a lower noise level because it needs a small resolution bandwidth. With the RSA306B, as I understand, it doesn't behave like a traditional swept analyzer because it can capture massive chunks of bandwidth, so there are speed gains (perfect for looking at modulation in immense detail because it can be in real-time) but it still would be impossible to have much improvement on noise floor with such a huge (2GHz) range if you wanted the result in a practical time. Even my (relatively) low-cost analyzer can reach -150dBm (although this is using an amplifier) but completely not possible for a 2GHz span.
It is not possible (for both the Keysight and the Tektronix that you mention) to state they have poor noise floor without knowing what settings you used. All we have to go on is the 2GHz span trace, which of course would have a high noise level for the reason mentioned above.
You mention you didn't want to reduce the resolution bandwidth because it slowed the output rate, and again this is expected but it would be good to see (perhaps in a follow-up posting) what the speeds are, since they should be faster than a traditional analyzer, and the additional benefits of the massive capture bandwidth. These should be some of the key differentiators of the RSA306b (as I understand, I could be wrong, I've not looked at it in detail, and I don't really understand the architecture differences between it and the older analyzers, but this information should be out there).
Do check out The Signal Path blog as mentioned by Kas of you haven't already done so as he's done some great reviews of this and other spectrum analyzers in the past.
Inderpreet Singh wrote:
Looking at the front end bandwidth and response characteristics, I don't think the Tek stands a chance again the KeySight.
To be fair to the RSA306B, the Keysight N9917A is priced an order of magnitude higher as far as I can see so not surprising it's not going to match its performance. Even the much more expensive RSA507A wouldn't compete with it in outright performance but price/performance wise these Tektronix USB spectrum analyzers look really good to me.
Nice review, they do talk about what they are targeting I just don't 100% remember were. I think it's more field and educational applications then lab applications. So things like interference hunting (directional antenna and GPS attached to PC, basic troubleshooting, and basic standard testing, spurious events, out of band interference, etc. ). BTW if you look I think under licenses, I think, (not sure which drop down menu) there are legacy 30 day licences that allow you to do a whole bunch of other testing.
As for the "random" spikes, the signal path (youtube channel) goes into detail about that on the original RSA306.
I look forward to reading more of your review.