EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: jpb on December 01, 2018, 12:39:39 pm
-
I wasn't sure whether to put this under "Beginners" or "Test Equipment" but plumped for test equipment.
I'm looking to get a low-end spectrum analyser but one that doesn't take up my entire workbench so a cheap modern one probably. This is NOT yet another thread on looking for buying advice as there are a lot (which I've read) already on the forum.
I just have a few quick questions on the consequences in general of the choice I make, namely:
1.) Can the bad phase noise of a low end, say Rigol, SA be improved with a low phase noise external reference or are you stuck with it?
2.) Looking at the Signal Path review of the Rigol DSA1030A, the sweep became incredibly slow when the RBW was reduced to 10Hz so the span had to be made very narrow. I know from comments from owners that DSA1030 is slow but I'm wondering how I translate sweep rates in data sheets (typically 10ms to 20 seconds say) into how long it takes to sweep say 1GHz at 1kHz RBW? Since FFTs on IFs are used it is not as simple as a time per point I presume?
3.) I already have an RF generator (Hameg HM8135) and noise sources are available cheap on ebay so I'm wondering if I can get away with not getting a tracking generator - a lot of SAs for sale second hand don't have them. The question is, is it practical to have the RF generator trigger the SA (or the other way around) and hope the two can track each other or is that impossible and some sort of multi-sweep approach where the SA sweeps very slowly and the RF generator sweeps multiple times and peak detect is used is required?
As I say, I perhaps should have put these questions in the Beginners section. Any answers - particularly practical ones on the lines of other people's experience with Rigols/Siglents etc would be great.
-
1) No, phase noise depends on the internal local oscillator generation circuitry. This has to be designed to sweep over a large range and this makes it difficult to optimize for really low phase noise. Big spectrum analyzers tend to use tuned yttrium garnet oscillators to get better phase noise.
2) Yes in the old days of classical detector spectrum analyzers this put a limit on RBW versus sweep speed. FFT based ones are now mostly limited by processing power and the performance of the IF ADC. This of course widely varies between different models and brands.
3) Yes you can use a seperate signal generator but it will be very slow. When they are not synronised you need to have one of them do the full sweep before moving on to the next point(Ether do a full sweel on the SA before the signal generator does a step forwards, or the signal gen doing a sweep by the time the spectrum analyzer does one step). It works in a pinch but its anoying. A real tracking generator can track perfectly and so it runs as fast as any normal spectrum analyzer operation.
Sometimes you can build your own tracking generator if the spectrum analyzer provides a LO output.
-
1) No, phase noise depends on the internal local oscillator generation circuitry. This has to be designed to sweep over a large range and this makes it difficult to optimize for really low phase noise. Big spectrum analyzers tend to use tuned yttrium garnet oscillators to get better phase noise.
2) Yes in the old days of classical detector spectrum analyzers this put a limit on RBW versus sweep speed. FFT based ones are now mostly limited by processing power and the performance of the IF ADC. This of course widely varies between different models and brands.
3) Yes you can use a seperate signal generator but it will be very slow. When they are not synronised you need to have one of them do the full sweep before moving on to the next point(Ether do a full sweel on the SA before the signal generator does a step forwards, or the signal gen doing a sweep by the time the spectrum analyzer does one step). It works in a pinch but its anoying. A real tracking generator can track perfectly and so it runs as fast as any normal spectrum analyzer operation.
Sometimes you can build your own tracking generator if the spectrum analyzer provides a LO output.
Thank you for your very clear answers.
-
You have to ask yourself if a network analyser is more suitable. With a tracking generator you can only see the amplitude. In order to develop a circuit you will also want to see the phase. I think this is the reason so few spectrum analysers have a tracking generator: the use is limited. In my experience a noise source isn't very useful either because the output level is already quite close to the noise floor of the spectrum analyser. Changing the RBW doesn't help because the noise from the source scales along with the bandwidth.
-
Especially if you can sweep the generator's frequency off the SA's sweep trigger, you can build up a bode plot pretty easily on a RF gen/SA combination without the built in TG, but integrated tracking generators make it easier to setup, normalize, and quicker to configure in most cases. You also have the option of using a wideband noise source for your bode plot generation, provided you have one that can output a pretty level spread in the spectrum you're concerned with. As said, though, TGs don't offer phase measurement so they're not able to measure everything you may want, so while they're useful and convenient, I wouldn't consider them a must-include.
I think it is possible to improve phase noise with an external reference in some situations, but really only when the reference oscillator is noisier than the PLL, which I doubt is the case in most modern SAs. It's as described, the PLL contributes most to phase noise when you're referenced to a low phase noise oscillator especially when you're tuning the PLL to sweep quickly, and YTOs are a common alternative to get good phase noise and fast sweep speed.
Regardless of architecture, the sweep speed will decrease with narrower and narrower RBWs, but different instruments will act differently, it's very design dependent. Probably the best way to know is to see it in action, I've had a portable SA that took 3s+ for a full span sweep at 3MHz but like 35 seconds for a 1Hz RBW sweep over 100Hz whereas a benchtop SA that could do a full span (double the bandwidth) sweep in 200ms and a 1Hz RBW sweep over 100Hz span in more than a minute.... while the narrower RBW will always be slower for the same number of points, it can vary by how much lower and at what ratio to other sweep speeds depending on how the instrument is designed.
-
You have to ask yourself if a network analyser is more suitable. With a tracking generator you can only see the amplitude. In order to develop a circuit you will also want to see the phase. I think this is the reason so few spectrum analysers have a tracking generator: the use is limited. In my experience a noise source isn't very useful either because the output level is already quite close to the noise floor of the spectrum analyser. Changing the RBW doesn't help because the noise from the source scales along with the bandwidth.
I agree that network analysers are the best way of measuring impedance - at least for small signal. My first job as a graduate electronic engineer was doing research on GaAs MMICs and one of my jobs was measuring them with an HP8510 VNA system - it cost about 10 to 20 times the cost of my flat at the time (early eighties). They have since got much cheaper but they are still expensive for hobbyists. (The one port additions to Spectrum Analysers are perhaps the cheapest option.)
I don't really need to do such measurements at present but I didn't want to buy a SA without a TG and then found I regretted it a few years down the road - but perhaps at that stage VNAs will be available at reasonable hobbyist prices anyway.
-
Especially if you can sweep the generator's frequency off the SA's sweep trigger, you can build up a bode plot pretty easily on a RF gen/SA combination without the built in TG, but integrated tracking generators make it easier to setup, normalize, and quicker to configure in most cases. You also have the option of using a wideband noise source for your bode plot generation, provided you have one that can output a pretty level spread in the spectrum you're concerned with. As said, though, TGs don't offer phase measurement so they're not able to measure everything you may want, so while they're useful and convenient, I wouldn't consider them a must-include.
I think it is possible to improve phase noise with an external reference in some situations, but really only when the reference oscillator is noisier than the PLL, which I doubt is the case in most modern SAs. It's as described, the PLL contributes most to phase noise when you're referenced to a low phase noise oscillator especially when you're tuning the PLL to sweep quickly, and YTOs are a common alternative to get good phase noise and fast sweep speed.
Regardless of architecture, the sweep speed will decrease with narrower and narrower RBWs, but different instruments will act differently, it's very design dependent. Probably the best way to know is to see it in action, I've had a portable SA that took 3s+ for a full span sweep at 3MHz but like 35 seconds for a 1Hz RBW sweep over 100Hz whereas a benchtop SA that could do a full span (double the bandwidth) sweep in 200ms and a 1Hz RBW sweep over 100Hz span in more than a minute.... while the narrower RBW will always be slower for the same number of points, it can vary by how much lower and at what ratio to other sweep speeds depending on how the instrument is designed.
Thanks for the further information.
My position is I generally want a SA for doing direct measurements rather than S-parameter type measurements but I don't want to get one without a TG and then find I'm stuck if I want to do such a measurement. The option to do the measurement albeit slowly and inconveniently is sufficient since it won't be something I'll be doing a lot of.
In some cases a TG only adds a bit to the cost and is worth getting but often there are well specced SAs available without TGs and I don't want to exclude the option of getting one of these - especially as I already have a signal generator.
-
My position is I generally want a SA for doing direct measurements rather than S-parameter type measurements but I don't want to get one without a TG and then find I'm stuck if I want to do such a measurement. The option to do the measurement albeit slowly and inconveniently is sufficient since it won't be something I'll be doing a lot of.
In some cases a TG only adds a bit to the cost and is worth getting but often there are well specced SAs available without TGs and I don't want to exclude the option of getting one of these - especially as I already have a signal generator.
Well you can have the best of both worlds with some of the newer gear such as the Siglent SVA models......SA/VNA combo.
The SVA1015X has been out for a little while and apparently it will soon have a big brother SVA3032X:
https://www.eevblog.com/forum/testgear/new-siglent-products-sds2000x-e-sva3032x-and-a-dc-load/msg1988441/#msg1988441 (https://www.eevblog.com/forum/testgear/new-siglent-products-sds2000x-e-sva3032x-and-a-dc-load/msg1988441/#msg1988441)
There's zip about this new SVA model thus far so we don't even have an idea of when it's due to be released.
-
The new combined one-port VNA and SA models from Siglent and R&S are interesting but there are trades off to be made in terms of cost or SA performance for a given cost.
Comparing the SVA1015X to the SSA3021X in Dave's tear downs (which I must admit I haven't looked at in detail) I get the impression that the SVA1015X has lower specs than the SSA3021X when used just as a SA in terms of DANL and amplitude precision as well as bandwidth.
Also looking at Rigol, and I suspect Siglent might go the same way, they seem to be shifting from selling very cheap SAs that hobbyists can afford to concentrating on much more expensive models for business. As a business model it makes sense for them if they can provide the support that business needs and even if they can't they will probably still try to move more upmarket. I suspect that the SVA3032X will be a big jump in price over the SVA1015X model.
-
The new combined one-port VNA and SA models from Siglent and R&S are interesting but there are trades off to be made in terms of cost or SA performance for a given cost.
Comparing the SVA1015X to the SSA3021X in Dave's tear downs (which I must admit I haven't looked at in detail) I get the impression that the SVA1015X has lower specs than the SSA3021X when used just as a SA in terms of DANL and amplitude precision as well as bandwidth.
AFAIK that is true for every network analyser not matter what the brand or price is. You can't seem to have a very good spectrum analyser and a very good network analyser in one device.
-
The new combined one-port VNA and SA models from Siglent and R&S are interesting but there are trades off to be made in terms of cost or SA performance for a given cost.
Comparing the SVA1015X to the SSA3021X in Dave's tear downs (which I must admit I haven't looked at in detail) I get the impression that the SVA1015X has lower specs than the SSA3021X when used just as a SA in terms of DANL and amplitude precision as well as bandwidth.
AFAIK that is true for every network analyser not matter what the brand or price is. You can't seem to have a very good spectrum analyser and a very good network analyser in one device.
You can get some pretty amazing combined devices but the price is not kind.
-
The new combined one-port VNA and SA models from Siglent and R&S are interesting but there are trades off to be made in terms of cost or SA performance for a given cost.
Comparing the SVA1015X to the SSA3021X in Dave's tear downs (which I must admit I haven't looked at in detail) I get the impression that the SVA1015X has lower specs than the SSA3021X when used just as a SA in terms of DANL and amplitude precision as well as bandwidth.
AFAIK that is true for every network analyser not matter what the brand or price is. You can't seem to have a very good spectrum analyser and a very good network analyser in one device.
You can get some pretty amazing combined devices but the price is not kind.
Currently.
SVA1015X price = SSA3021X price but the SVA is a much more capable device but at a lower BW.
The coming SVA3032X (and maybe a 2.1 GHz model) could make these technologies affordable for many.
-
The SVA3032X will need to have much better spec's to make it compare to any of the big brands. However it will likely be priced well for what you do get.