EEVblog Electronics Community Forum
EEVblog => EEVblog Specific => Topic started by: EEVblog on July 06, 2018, 10:15:09 pm
-
Teardown and look at the new $1395 Siglent SVA1015X 1.5GHz Spectrum and Vector Network Analyser
Well, $2000 when you include the actual VNA option :-/
https://www.youtube.com/watch?v=HxBcQDooAYs (https://www.youtube.com/watch?v=HxBcQDooAYs)
-
Please hold while I fact check RF stuff with The Signal Path....
-
So if I estimate correctly, buying the VNA option and a Cal kit will make this puppy double the price. Not such a "low cost" VNA it becomes. As well as the function seems to be rudimentary (the useless FFT feature of rigol scopes comes to mind) and it is yet to be proved it was implemented correctly, such as the math behind it and stuff.
-
Teardown and look at the new $1395 Siglent SVA1015X 1.5GHz Spectrum and Vector Network Analyser
Well, $2000 when you include the actual VNA option :-/
Seems like they put in a Rubycon cap to make dave happy ^^
Right next to it was a lelon cap.
And the other Caps are different as well.
The gunk on PSU is for transport, though one would usually use that for the Coils and not just caps.
And I'd also not worry too much about the manufacturer but the Series...
As for Lelon low ESR caps, something like RXW or RZW would be nice.
As for the Processor board:
Doesn't that also save some layers on the Mainboard??
Like using 6-8 Layers on the module and 2 or 4 on the main PCB.
That would be my guess why they are doing it.
-
As for the Processor board:
Doesn't that also save some layers on the Mainboard??
Like using 6-8 Layers on the module and 2 or 4 on the main PCB.
That would be my guess why they are doing it.
Potentially, but I doubt that's the main or only reason.
-
Why are the signal traces exposed in only some parts of the RF section, and other parts are under soldermask? Is it so that they can be probed?
-
Why are the signal traces exposed in only some parts of the RF section, and other parts are under soldermask? Is it so that they can be probed?
No. It's because they are critical transmission lines, and it's easier to control the impedance of a PCB transmission line when it doesn't have solder mask (with it's relatively high variability) mucking up the equation.
-
Dave - when touring the user interface did you happen to notice if you can enter custom cal kit parameters? I see there was a greyed out ECAL option but don't recall seeing anywhere to enter your own parameters.
-
Dave - when touring the user interface did you happen to notice if you can enter custom cal kit parameters? I see there was a greyed out ECAL option but don't recall seeing anywhere to enter your own parameters.
Didn't notice anything, but wasn't deliberately looking for that. Not at the lab so can't check.
-
Why are the signal traces exposed in only some parts of the RF section, and other parts are under soldermask? Is it so that they can be probed?
No. It's because they are critical transmission lines, and it's easier to control the impedance of a PCB transmission line when it doesn't have solder mask (with it's relatively high variability) mucking up the equation.
But why is some of it under solder mask then?
-
But why is some of it under solder mask then?
That part was less critical to the performance, i.e. variability wouldn't have mattered as much.
Notice how the distributed element filters are all exposed, it's because they need controlled performance on those.
-
The soldermask has to do with transmission lines. A trace is just a trace for DC but not for RF. There has to be two "traces" one of them the groundplane, the other the trace. Everything in between, nearby or above it makes part of the line.A transmission line has a constant impedance. A soldermask is part of the transmission line but has more loss as air and is not so easy to control over a wide bandwidth. http://www.gsm-modem.de/M2M/m2m-faq/transmission-line/ (http://www.gsm-modem.de/M2M/m2m-faq/transmission-line/)
A SA + TG is a scalar analyser. I sources a signal and measures the insertion loss (the attenuation) There is some delay between the sourcing and measurement. Caused by the instrument and the DUT. For serious RF work a big no-no.
A VNA measures the sourced signal direct at the source and at the same time the result after the DUT. It needs 2 receivers for that. That gives the real attenuation and fase difference both at the same time. So the measured fase difference is only caused by the DUT. This is important but the why is a bit much to write in one post. A simpel example is a resonance. At the resonant frequency the fase jumps 180 degrees. So without phase info you are not sure is there is resonance and what is the exact frequency. You need phase info to see if something is capacitive or inductive (and you can calculate all kind of info from that) things that are impossible to do on a scalar analyser.
One of the most important and critical things for a VNA is calibration. On an SNA you can come a long way with simple normalisation. The result of a VNA is totally depending upon calibration and so on the used call kit. A good call kit comes with data. You need to feed that data to the VNA so he knows what you use for calibration. The details are very complex to explain. See it like this. Suppose the VNA calibration kit load is 60 ohm and 1 uH (complete bogus values). I connect it to the VNA and do a calibration run without telling him the specs. If I now connect a perfect 50 ohm with zero inductance the VNA will tell me the resistance is 40 ohm and capacitive (these are not correct values but it makes more clear what calibration is about) Call kits from R&S or Keysight cost more as the Siglent. You can make your own but need a calibrated VNA to extract the parameters. A good call kit has those documented but a VNA is only usable if you can enter those values. Old VNA could not do that so you needed an almost perfect call kit (even more expensive) and things like line strechters to make sure the signalpath to the DUT was as long as that to the reference receiver.
I just repaired a R&S 4GHz VNA. It is something like 30-40 kilo, huge in size and only to get the powersupply out I had to remove over 80 srews. Everything is covered with metal. All interconnections are rigid coax (low loss and more stable impedance as movable coax cable).
See here for the way that is build:
https://youtu.be/slErj2toXKo (https://youtu.be/slErj2toXKo)
-
That part was less critical to the performance, i.e. variability wouldn't have mattered as much.
Notice how the distributed element filters are all exposed, it's because they need controlled performance on those.
Yep, I understand that the dielectric constant matters for the filters, but I guess I’m just a bit confused why those straight sections in the photo are unmasked whereas there are large chunks that are under mask; the straight sections aren’t filters and they don’t seem to be anything other than standard transmission line. The only reason I can think of is to allow for probing during testing. If it does indeed make a huge difference to loss/variability, then why not totally do away with the solder mask along the trace and expose it through the whole signal chain? It’s gold plated after all, so corrosion shouldn’t be an issue.
-
Hi everyone! Can someone give me a hint where I can find the high resolution pictures, that David mentioned?
-
Hi everyone! Can someone give me a hint where I can find the high resolution pictures, that David mentioned?
Hey!
Like usual, they are on his FLICKR page -
https://www.flickr.com/photos/eevblog/albums/72157692982928880 (https://www.flickr.com/photos/eevblog/albums/72157692982928880)
Cheers, Phil!
PS. Thank you Dave for a awesome video! These spectrum /vector analyzers are so great candidates to be torn down)) 8)
-
Pricing of options is pretty disappointing. Also, you cannot use it for 2.4GHz stuff.
I guess it could be a good investment for someone working in the ISM band, Lora, 868MHz stuff and others, without the VNA option, if the firmware can be hacked.
-
It looks to me like the board material is different between the two instruments. By eye, FR-epoxy vs teflon. I wonder if some of the discrete component filters are built they way they are because of the difference in the dialectic properties of the board.
-
It looks to me like the board material is different between the two instruments. By eye, FR-epoxy vs teflon. I wonder if some of the discrete component filters are built they way they are because of the difference in the dialectic properties of the board.
The discrete components are more likely the result of the lower maximum frequency compared to the spectrum analyser. But yes, it is too bad it can't reach beyond 2.5 GHz where all the modern communication standards sit. That makes the SVA1015 obsolete straight away.
-
It looks to me like the board material is different between the two instruments. By eye, FR-epoxy vs teflon. I wonder if some of the discrete component filters are built they way they are because of the difference in the dialectic properties of the board.
Dave pointed out many times in the video, that the discrete element filters are bigger in the SVA1000, because the frequency is lower. So you need a bigger cap and inductor. So imagine, if all the filters need to be physically bigger, it makes sense to replace that "big" discrete element with an 0402, instead of reorganizing the entire board.
-
So if I estimate correctly, buying the VNA option and a Cal kit will make this puppy double the price. Not such a "low cost" VNA it becomes. As well as the function seems to be rudimentary (the useless FFT feature of rigol scopes comes to mind) and it is yet to be proved it was implemented correctly, such as the math behind it and stuff.
Yeah Bud when you add all the options in it does pump up the price dramatically however it's functionality does cover a lot of bases and it will be interesting to see how well it does them all. Even the touch screen is new in Siglents larger equipment as we've only seen it in SDG***2X (AWG) models.
We'll know soon enough as they are being shipped all over the place right now and there'll be some findings posted in the dedicated thread for these fairly soon.
https://www.eevblog.com/forum/testgear/siglent-sva1015x-1-5ghz-spectrum-vector-network-analyzer-(coming)/ (https://www.eevblog.com/forum/testgear/siglent-sva1015x-1-5ghz-spectrum-vector-network-analyzer-(coming)/)
-
But yes, it is too bad it can't reach beyond 2.5 GHz where all the modern communication standards sit. That makes the SVA1015 obsolete straight away.
Not obsolete, it just has a narrower target market. I can imagine plenty of uses for a 1.5GHz VNA.
-
This would only be good enough for pre compliance work if the UUT maximum clock frequency was less than 200 MHz, otherwise the testing has to go up to 5 times the max clock speed.
-
Hey!
Like usual, they are on his FLICKR page -
https://www.flickr.com/photos/eevblog/albums/72157692982928880 (https://www.flickr.com/photos/eevblog/albums/72157692982928880)
Thanks for the link!
One thing I'm not shure about: To measure the S11 parameter, somehow the reflected signal has to be measured. But simple multiplexing, like sending the testsignal in, then switching and sending the reflected signal to the receiver could not work. Sending and receiving must be carried out simultaneously. Because of that, there has to be a directional coupler. In the picture of the tracking generator, I marked the component in which all three signal pathes ends. But isnt that device way to small for such an coupler?
-
Hey!
Like usual, they are on his FLICKR page -
https://www.flickr.com/photos/eevblog/albums/72157692982928880 (https://www.flickr.com/photos/eevblog/albums/72157692982928880)
Thanks for the link!
One thing I'm not shure about: To measure the S11 parameter, somehow the reflected signal has to be measured. But simple multiplexing, like sending the testsignal in, then switching and sending the reflected signal to the receiver could not work. Sending and receiving must be carried out simultaneously. Because of that, there has to be a directional coupler. In the picture of the tracking generator, I marked the component in which all three signal pathes ends. But isnt that device way to small for such an coupler?
That does appear to be the directional coupler. Minicircuits has similar transformer based directional couplers with the Siglent's spec'd VNA frequency range (10MHz - 1.5GHz).
What you have marked as an amplifier in the path to the small connector near the top of the board is actually a forward/reverse switch (PE42553). Note the symmetrical DC blocking caps with RF traces going to either side of the package from the coupler.
The other part you have marked as an amplifier (in the tracking gen path) is actually a 7 bit digital step attenuator, also from Peregrine Semi (PE43711).
Edit:
Another note about why some of the RF traces are masked and some are not. In addition to getting more accurate impedance, removing the mask also reduces the loss of the TL segments, so it makes sense to remove it over the relatively long straight sections. It's not such an issue for short sections, which is why it's not removed everywhere.
-
I would expect a directional bridge under 1500 MHz
About the soldermask over traces , Signal path has a video he shows an attenuator (for higher frequencies) The theory is related to the reason you do not cover traces.
The discrete filterers is more easy for low frequencies. I made a 25 MHz to 2 GHz sweepgenerator in several bands. For the lowest bands I used caps and inductors. Then "stripline" (a mix of pcb as caps and wire as inductor?), the highest range ( 2 GHz LPF is "microstrip") http://www.pa4tim.nl/?p=2662 (http://www.pa4tim.nl/?p=2662)
-
Check out the gold plated edges, surprised I don't see this more often. It should be capable in a normal PCB process (plated slot) with no added cost and superior to via stitching alone right?
I think the cost of board space being "free" is slightly exaggerated though. Not only do you have the controlled impedance board material (not as high end as SSA3021X material but still), you also have the cost of machined aluminum to contain it. On a low end product like this, adding $20-40 of cost would be significant.
-
Check out the gold plated edges, surprised I don't see this more often. It should be capable in a normal PCB process (plated slot) with no added cost and superior to via stitching alone right?
I think the cost of board space being "free" is slightly exaggerated though. Not only do you have the controlled impedance board material (not as high end as SSA3021X material but still), you also have the cost of machined aluminum to contain it. On a low end product like this, adding $20-40 of cost would be significant.
It's done like this for a reason......there's 2 models the western markets won't see: SSA1015X and SSA1010X-C so it makes sense that any addition HW the SVA has is not replicated in the those SSA models.
-
There is not that much additional HW needed for the VNA mode. It gets even less if you look what could be left unpopulated, still keeping the same board. I don't think the savings on a smaller CPU are that large.
So I can understand that they might have dropped the SSA version. So maybe the lower BW version, without the generator might still make some sense.
-
But it does raise the question as to when the 3GHz SVA1030X is coming.... :popcorn:
-
But yes, it is too bad it can't reach beyond 2.5 GHz where all the modern communication standards sit. That makes the SVA1015 obsolete straight away.
Not obsolete, it just has a narrower target market. I can imagine plenty of uses for a 1.5GHz VNA.
Like what? Because it is a 50 Ohm device you can't use it for systems with an impedance very different to 50 Ohm without losing accuracy. So this does limit it to 'radio stuff'.
When it comes to network analysers there are basically two types: the HF 50 Ohm types which are primarily useful for HF (radio) work and the LF types which also have 1M Ohm inputs and various methods to do the analysis to offer a wide range of useable impedances. The LF network analysers are much more useful as a general purpose tool. Siglent should make one of those (just like the Omicron Bode 100 you reviewed recently).
-
But yes, it is too bad it can't reach beyond 2.5 GHz where all the modern communication standards sit. That makes the SVA1015 obsolete straight away.
Not obsolete, it just has a narrower target market. I can imagine plenty of uses for a 1.5GHz VNA.
Like what? Because it is a 50 Ohm device you can't use it for systems with an impedance very different to 50 Ohm without losing accuracy. So this does limit it to 'radio stuff'.
When it comes to network analysers there are basically two types: the HF 50 Ohm types which are primarily useful for HF (radio) work and the LF types which also have 1M Ohm inputs and various methods to do the analysis to offer a wide range of useable impedances. The LF network analysers are much more useful as a general purpose tool. Siglent should make one of those (just like the Omicron 100 you reviewed recently).
So you think corrections can't be applied for alternative usages ? ::)
From the manual:
2.1.3.8 Correction
Correct the displayed amplitude to compensate for gains or losses from external devices such
as antennas and cables. When using this function, you can view the correction data table and
save or load the current correction data. When amplitude correction is enabled, both the trace
and related measurement results will be mathematically corrected. Positive correction values
are added to the measured values. Negative (-) correction values are subtracted from the
measured values.
1. RF Input
Set the input impedance for numeric voltage-to-power conversions. To measure a 75 Ω device,
you should use a 75 Ω to 50 Ω adapters to connect the analyzer with the system-under-test
and then set the input impedance to 75 Ω.
2. Apply Correction
Enable or disable amplitude corrections. Default is Off. The analyzer provides four correction
factors that can be created and edited separately, but they can be applied independently in any
combination.
-
@Tautech: I know you want to make it look good but there are limits to Siglent equipment. Laws of physics and so on.
-
Like what? Because it is a 50 Ohm device you can't use it for systems with an impedance very different to 50 Ohm without losing accuracy. So this does limit it to 'radio stuff'.
When it comes to network analysers there are basically two types: the HF 50 Ohm types which are primarily useful for HF (radio) work and the LF types which also have 1M Ohm inputs and various methods to do the analysis to offer a wide range of useable impedances. The LF network analysers are much more useful as a general purpose tool. Siglent should make one of those (just like the Omicron Bode 100 you reviewed recently).
You are looking to narrow minded.
You are talking about impedance analysers used for gain/phase measurements. Just because an instrument is called an analyser and measures phase does not mean it is a VNA. There are vector and impedance analysers and some of them are both.
Impedance analysers often can output much more as a VNA, and I think they can handle much higher input too
VNAs like the siglent can only do s parameter measurements (and often TDR) and that is always 50 or 75 ohm.
There are network analysers that can do both functions, like the Omicron (limitted to only 50 MHz) or the Keysight E5061B-3L5 (a LF analyser up to 3 GHz ! )
The siglent is a hobby SA+VNA and 1,5 GHz is very useful for HAM use. I have a SDR-kits VNWA and that can do also impedance analyses, that is, if you use a RF-IV adapter (the software supports it) I have that too but seldom use it because it is only usable to something like 100 MHz. Not interesting for ham/hobby use
-
Dave pointed out many times in the video, that the discrete element filters are bigger in the SVA1000, because the frequency is lower.
I know he said that, but I am not sure I understand why it matters. Both instruments he compared operate down to 9 kHz. So it is the upper limit, not the lower limit that is different between the two. Unless the IF frequency range is really different, that I didn't hear mentioned, then the microstrip distributed element filter layout could have been about the same and cover 9 ... 1,500 kHz the same way it does in the SSA3032X
-
Dave pointed out many times in the video, that the discrete element filters are bigger in the SVA1000, because the frequency is lower.
I know he said that, but I am not sure I understand why it matters. Both instruments he compared operate down to 9 kHz. So it is the upper limit, not the lower limit that is different between the two. Unless the IF frequency range is really different, that I didn't hear mentioned, then the microstrip distributed element filter layout could have been about the same and cover 9 ... 1,500 kHz the same way it does in the SSA3032X
I'm sure the IF frequency is different between the two devices.
The SSA has a first IF around 4GHz.
OTOH, the mixer in the SVA tops out at 4.5GHz. To have the same IF frequency as the SSA, it would need to run at 5.5GHz which is unlikely. So that is a strong hint that the IF is now running at a significantly lower frequency. Which would of course dictate that any first IF filters need to change at least.
So we know the input frequency is different, the 1st LO frequency range is different, and the 1st IF is different. Same story with the TG. So it is not a surprise that those filters have changed.
-
So, try as I may, I have searched online, read the user manual and looked for anything I can on google - I can't find exactly what the bloody AM/FM modulation analyzer feature will give me! Can it measure peak and average deviation etc for example?
How are we meant to know if the feature is worth shelling out $$$ for, if we can't even find basic information on what the thing does!?!!!!!!
:-- :--
-
So, try as I may, I have searched online, read the user manual and looked for anything I can on google - I can't find exactly what the bloody AM/FM modulation analyzer feature will give me! Can it measure peak and average deviation etc for example?
How are we meant to know if the feature is worth shelling out $$$ for, if we can't even find basic information on what the thing does!?!!!!!!
:-- :--
I have tautech's SVA here at the moment, I'll have a play tomorrow after work and post a pic - please remind me if I forget!
-
Here is a pic of the demod in action:
FM radio station at about 100MHz - both SA view and FM demod view.
I have to say that the fast www interface to this makes grabbing screenshots a breeze.
There is a screenshot button, just click it and it saves a nice small png. No need to stuff around with USB sticks.
VNC works as well - no password.
It works well with a direct ethernet cable connection with static IP's set on both ends. The interface to set the static IP on the SVA isn't very good though.
I really like this thing. Even more so that the SSA. It's a wee ripper.
-
Here is a pic of the demod in action:
FM radio station at about 100MHz - both SA view and FM demod view.
So, deviation according to modulation analyzer max 25KHz? :-//
SA screen shows what you expect from an FM broadcast station, 100-150Khz.
-
Here is a pic of the demod in action:
FM radio station at about 100MHz - both SA view and FM demod view.
So, deviation according to modulation analyzer max 25KHz? :-//
SA screen shows what you expect from an FM broadcast station, 100-150Khz.
Hmm, yes.
Firstly, the Carrier Freq reset itself to 100MHz when changing modes, so it wasn't actually on a station frequency in the second pic. That looks like a bug to me.
Secondly I don't know what the deviation is showing. It isn't the absolute occupied bandwidth which is about 180 kHz
I've seen it up to 50kHz, but for this station it seems a bit lower - max about 42 kHz.
The demodulated trace above shows +/- 80kHz roughly taking the max/min values.
I suspect it is the result of some calculation.
-
One of the first things that struck me was a button battery on the PCB - inside the sealed enclosure.
So its a return to replace a dead button cell???
I wonder how much that costs (ex. the shipping costs)?
-
One of the first things that struck me was a button battery on the PCB - inside the sealed enclosure.
So its a return to replace a dead button cell???
I wonder how much that costs (ex. the shipping costs)?
What, this one ?
https://youtu.be/HxBcQDooAYs?t=830
You must've been looking at something else.
-
Yep. That one.
(https://vogon.net/bat.png)
Unless I missed an access panel on the cover it looks like you would need to break the seal to replace it?
-
Yep. That one.
(https://vogon.net/bat.png)
Unless I missed an access panel on the cover it looks like you would need to break the seal to replace it?
And based on my PC bios backup battery experience the SVA will be well past the 3 year warranty period before the battery fails. No issue for anyone breaking the tamper seal then !