Siglent SSA3000X and SSA3000X-Plus Spectrum Analyzers

[nctnico]

In all T&M instruments, all what you see are errors and lies mixed with illusions of reality.

Too bad this text is too long for a T-shirt though.

[pherdie]

Re: New Firmware

Apparently no more system lock ups when manually adjusting the bandwidth!

Thanks.
Please tell us more.....previous firmware? Was it an ongoing problem?

I don't think I've seen this problem mentioned before. 

It was mentioned in a previous post and it occurred during Dave Jones' EEV Blog video review.

[TurboTom]

Since I found the 30kHz RBW limit on the SSA3X Analyzers (when the TG is active) a little awkward (see here: https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/msg975408/#msg975408), and while "rf-loop" correctly pointed out that this shouldn't be a problem, I wanted to dig a little bit deeper into the matter of the tracking generator in this machine.

What made me really wonder is that I wasn't able to get a proper parallel resonance dip with a very accurate 1MHz crystal, whatever I tried -- there was always some kind of "hills and valleys" where the dip should be visible. Also the series resonance peak appeares a little "wiggly", see the screenshots in the above link.

Other (less accurate) crystals measure fine without any visible artefacts. The difference is that the mentioned "high accuracy crystal" has a very small frequency span between series and parallel resonance, it's just about 950Hz whereas more common crystals have spacings more like several tens of kilohertz. The reading cannot be an inherent characteristic of the crystal since on my other SA (Rigol 815TG) and also on my 8753C VNA it measures fine. So the problem has to be related to the way the TG in the SSA3X works.

As I reported in the contribution linked above, I found the TG of the Siglent analyzer to produce some "wobble" in the frequency while scanning, and I wanted to undestand better why it does so and what it means. With the TG signal viewed on the scope, it appeard the wobble gets worse when the video bandwidth (VBW) is increased (resolution bandwidth of the analyzer anyways preset at 30kHz for the small spans). So I hooked up the Siglent's TG output to my Rigol DSA to get an idea of the spectrum and guess I was surprised what I found: It seems the increments are not the same over the whole VBW range but get bigger with incresing VBW setting. Moreover, it seems there is some kind or relationship with the "frequency wobble" and the scan of the SA input (it's difficult for me to explain but otherwise it would be impossible for the "hills and valleys" of the very narrow bandwidth crystal to appear). Please see the attached screenshot of the TG scan shape of my "slightly tuned" SSA3X. The basic settings are always the same (center frequency 1MHz, span 3kHz, TG 0dBm, 30kHz RBW, sweep time approx. 12s, changing the "frequency step" has no effect on the TG scan pattern. The scans on the Rigol were done with "max hold" trace, summed up over approx. 20 minutes each. The cyan trace resembles 10Hz VBW, the magenta one 100Hz and the yellow one 1kHz VBW on the Siglent.

While the cyan trace should be able reproduce the characteristics of the high precision crystal fairly well, it actually causes a lot of artefacts while increaseing VBW settings just "smears" the curve. When I use the TG in the Siglent and the Rigol SA to display the spectrum (sweep time about 24s (Siglent) / 300ms (Rigol)), I need only about three scans to get a "half-way decent" spectrum of the Crystal while the Siglent alone simply generates artefacts. For comparison, a trace of the Rigol TG/SA is also shown.

I also noticed a small bug in the recent SSA3X firmware (7.07): When the device is configured to start with the last configuration when powering up, and it has been powered down with the TG enabled, the TG will be enabled upon power-up but the TG control light will stay off. I'm not sure if the TG should be enabled after power-up at all even if the Analyzer had been shut down with the TG on. But at least the TG indicator should be lit if it is turned on.

Otherwise, the SSA3X appears to be a decent device so far, yet I would appreciate if Siglent had a closer loog at the TG sweeping and if it may be possible to simply depend TG frequency step size on the sweep time so the user could decide to have a slow/accurate measurement or a faster but not as accurate one, especially since RBW is limited when using the TG.

Cheers,
Tom

[rf-loop]

Broadly speaking agree.


I have also made some investigations and it is partially ok  but then there is also need for  improvement if we want SA-TG what can really use for DUT what have very steep and deep edges.  SA RBW is not so important (exept that it can use for hide problems in TG but also depending how  SA work. In full analog sweeping SA things are different)

Most important, mandatory,  is TG signal good quality(1).
If stepping sweep, there really can not be glitches in level or freq jitter - phase noise. Of course also SA side is important but this is other thing (when TG is perfect, then need also look SA side how it works for more or less perfect result)

1) Simply talking: If run pure fixed sine to example filter steep edge still get fixed level out. But then start "jittering" this frequency and it  is of course translated to AM modulation (note AM LSB and USB). Using sweep with RG and SA we just want see this AM modulation related to perfect sweep. But we do not want see AM modulation what is result from other freq change - jittering. More steep edge filter etc is our DUT more important is TG quality.

Now First: This example is how it need  look, this we need, Siglent! (exept RBW setting and truth that this take 1000 second)
(later how it looks with SSA TG)

This is made using External (step)Sweeping HP8644B generator instead of internal TG

And also this example do not at all have very deep and very steep level change with small freg delta
(60dB with <<10ppm  change in frequency is perhaps this kind of steep what I mean, is it for work with this kind of machine...all kind things can dream..)

[rf-loop]

...

(later how it looks with SSA TG)

Here..




Here first image is one simple example with more steep changes (Note 600Hz/div) .  TG jitter together with RBW 30kHz start coming in and set limits...
Trace B. Same DUT ((Xtal) sweeped using Extrernal Sweep and others with SSA TG. 
Trace C is just for show where is bottom when input is open and when look traces A and D.





Second image is using other Xtal (not so steep) as DUT. Here response is not so steep and no problem visible.
Problems start as shown in first image where is much more steep response as TurboTom has previously been well presented.

Note 2500Hz/div. 

TurboTom tests and also my tests can show what kind of limits is its usability. If you work out from these limits all is ok. If user reach these limits he can still use it but the interpretation of the results requires knowledge and experience with the device. If go more deep over limits it can say, game is over. Every equipment have limits. If can not live with these limits then need other kind of equipment.
If there come improvement then it come. But this is true now and with this user need live now.

[jjoonathan]

Speaking of knowing limitations, I keep hearing that the sweep rate isn't quite as good for wide spans as the DSA815 (which reports 50ms SWT for SPAN=1.5GHz RBW=1MHz but due to crippled update rate only squeezes in 4 or 5 sweeps a second). I can see the slower sweep time in the eevblog 3000x review video, but I believe this was with sweep type set to "accuracy" rather than "speed" -- what is SWT (and update rate) of SPAN=1.5GHz RBW=1MHz on the "speed" setting? Can you manually adjust it to be even faster?

I'd love to swap out my 815 and get the better BW, noise floor, and max update rate of the 3000x, but one of my major use-cases of the 815 is for "wide span surveys" and I'd hate to take a step back on that front without knowing what I'm getting into.

[TurboTom]

"Standard" sweep times for wide band scans are worse than with the Rigol, yet the preset sweep time can be overridden and reduced ("uncal" will be displayed in the upper left corner of the screen). Sweep times and screen update rates appear to follow some funny relation, standard sweep time for 0...1.5GHz (to resemble the DSA815) in "speed" setting is 144ms and screen update frequency is approx. 3.1 Hz (??). Sweep time can be reduced to 24ms which results in approx. 12.6Hz screen update rate, albeit at considerable more noise and artefacts. A somwhat acceptable trade-off between accuracy and fast reaction appears to be the combination of the fastest sweep time with a low count trace averaging (5...10 sweeps).

Narrow bandwidth scans are lighning-fast compared to the Rigol due to the FFT algorithm (if it suits the job).

I also did some more testing of the TG over the 3.2G range and found it to be catastrophically bad above approx. 2.5GHz. There's actually a dip at around 3GHz that reaches -5dBm after the device has warmed up (initially it's about 1dB better). Unfortunately, I haven't got a calibrated RF generator that covers the range up to >3GHz. All that I've got that's half-way suitable at all is one of these cheap Chinese USB synthesized sweep generators that run from 35MHz to 4.4GHz. I wanted to understand if the problem with the very ugly TG spectrum is actually the TG or the SA section (probably a little of both...). so I fed the signal of the "el-cheapo" sweep generator into the SA and set the sweep time very slow so i could get a more or less proper spectrum with a "max hold" trace over some 30 minutes. This is the yellow trace in the attached screenshot. The other three traces are of the internal TG of the siglent, yet with different connection cables:

The yellow trace with the sweep generator and the purple trace were taken with a high quality Suhner semi-rigid cable of approx. 20cm length (N to SMA / N to SMA-M plus quality SMA-F to N adapter).

The blue trace has been recorded with a DIY 30cm RG316 SMA-M to SMA-M cable and two quality SMA-F to N adapters.

Just out of interest, I used a cheap 80cm RG58 BNC cable and two BNC to N adapters for the green trace.

The comparison between the yellow and the purple traces (these are the only relevant ones) suggests that it's very probable that the calibration problem rather lies in the TG than the SA section since there aren't any obvious identical features between these two traces. So it will be necessary to have a look at the internal calibration data to see if it's possible to "tune" the TG to better match the 0dB line.

Cheers,
Thomas


Edit -- typo(s)

[jjoonathan]

I actually probed the DSA815 VCO at one point (well, its three VCOs, which have charming red LEDs under the RF shield to tell when each is active), and it wasn't lying about its sweep speed, just conveniently neglecting to mention the dead time -- I suspect the Siglent is the same way. The real question is why. I suspect it has something to do with the PLL, but I don't know enough about PLLs to speculate. It would be a real shame for an instrument with such modest compute requirements to be bottlenecked by computation in this day and age.

[TurboTom]

Okay - just had a look at the TG calibration, and - unfortunately - it's a joke... There's a file "cali_tg" that's just 135 bytes long. It contains the complete calibration data for the TG and looks like this (on my machine):

0 599300 4 599300 1244300 5 1244300 1324300 4 1324300 1569300 3 1569300 2834300 4 2834300 2884300 3 2884300 2964300 1 2964300 3200000 0

There are always three consequent values that belong togeter which define the frequency span segment and the attenuation value for that segment, i.e. the first segment spans from 0 to 599.3MHz and is attenuated by 4dB and so on. Actually, the band pass filters of the 1st LO switch over at input frequencies of 599.3MHz and 1569.3MHz. At these frequencies, there are also obviously calibration band edges configured.

I thought to myself "super -- let's change the calibration bands a little so they better match the spectrum of the particular TG in my machine", so initially I prepared a calibration file with all 0dB attenuation just to see the "natural" response of the TG and already got the first setback: There's a deep ditch of the TG at around 3GHz which (even unattenuated) drops down to approx. -5dBm when the machine has warmed up. When cold, it's about 2dB better. So this problem cannot be solved by mere software calibration, it would require a review of the TG hardware. My next attempt to calibrate the TG otherwise was also more or less a complete failure: It seems the calibration frequency bands are fixed and cannot be changed to slightly other values. So all I could do was modify the attenuation values of the existing bands a little which brought some minor improvement but no real change.

Unfortunately, I didn't record a trace with the original calibration so I can only show the TG without calibration (yellow trace) and with my new, modified calibration (magenta). Please keep in mind that the dip around 3GHz will get worse (by almost 2dB) when the instrument warms up, the attached screenshot was taken just minutes after powering up.

Room for improvement... but wait a moment, I bought a 2.1GHz spectrum analyzer and over this frequency range, the TG works okay. Everything else is my own problem... 

Cheers,
Thomas



Edit: Attached the wrong screenshot...

[Pinkus]

Just an idea: At everything > 2 Ghz  cables and connectors can play an important role.  Did you check them at a known system to be sure about their impact to the displayef response?

[rf-loop]

Measured TG output directly  from connector. (Power sensor N directly in TG out N and all temperatures well stabilzed)
(This method of course include all RF what is coming out)

Reference level 0dBm @50MHz

TG level set 0dBm. TG in zero span mode and manually stepped from around 1MHz to 3200MHz.
Sensor calibrated to 0dBm @50MHz and for other freq used sensor correction table.

TG 50MHz out between 0 and  -0.1 dBm

from 1 to 3200MHz there is level variations mostly less than < +/-1dB and then  pair +/-2dB  and then one clear  most high and low:
-3.0 - -3.1dB @ 2996MHz 
+2.9 - +3.0dB @ 2468MHz

because my power sensors cal are obsolete I did same measurement with two sensor, 8481A and 8482A so I can trust result bit better. Both give so near same that in practice it can say "same".

[Pinkus]

So the TG is only in a +-2dB range up to approx. 2.7 Ghz.
Fine for me....  but good to know.

[rf-loop]

So the TG is only in a +-2dB range up to approx. 2.7 Ghz.
Fine for me....  but good to know.

No, you do not know after my test. This do not tell your SSA! It tell my SSA in my test condition.
Every SSA is different!  This is why there is (afaik) personal correction table for every individual SSA.

Also sidenote.
Power meter measure total RF power out from TG N connector. It is NOT frequency selective!

[pascal_sweden]

If the spurious response comes from the device internals itself, isn't there an easy way to characterize the spurious response for a spectrum analyzer under no load, and calibrate, adjust, and eliminate it from the readings?

Or would that spurious response be unstable over time, depending on external parameters, such as electro-magnetic noise, interference, reflection, ambient temperature?

For comparison: Are all spectrum analyzers subject to this? How does the Rigol DSA185-TG perform on this front?

Is it too costly to make a spectrum analyzer with zero spurious response, or is it just impossible by design, based on the topology and build up of the various stages in the spectrum analyzer?

Are there different variations possible in the build up of the various stages in the spectrum analyzer,
and are some variations less subject to spurious response compared to others?

Is the build up of the stages in most spectrum analyzers the same or very similar, or are there 2 or more fundamental variations and approaches that are completely different from each other, and that have a completely different internal spurious response by design?

[rf-loop]

Is it too costly to make a spectrum analyzer with zero spurious response, or is it just impossible by design, based on the topology and build up of the various stages in the spectrum analyzer?

Not only too costly but also impossible.  Just as impossible as doing pure sinewave or ideal rectangle wave - or what ever ideal. School books with nice ideal sine and square are just so out from reality what can. In reality there is no pure sine wave, no exact time, no exact voltage, no ideal mixer, no ideal filter, no ideal amplifier...  you want more. As told previously, all what you see on oscilloscope or spectrum analyzer display are collection of errors mixed with undefined truth.

Too often I quote what Keyshit (former Agilent, former Hewlett-Packard) tell us.
Here from Tektronix.
Why they all talk about spurious, residuals etc. Rohde&Schwarz, Tektronix, Keysight, Anritzu, Advantest, and so on...


From "Fast, Low Level Spurious Search with
Tektronix Real-Time Signal Analyzers"

Considering the Spurious Performance of
Spectrum Analyzers

Measuring low level spurs requires care even when the
spectrum analyzer’s Displayed Average Noise Level has
sufficient margin to perform the required measurements.
All spectrum analyzers create artifacts or spurs that can
appear at low levels. Some spurs are created internally by the
Spectrum Analyzer’s circuitry. Others are generated inside the
instrument as a result of input signal interactions with internal
signals and non-linear circuit behavior. These unwanted
signals, typically related to harmonics of the input signal, are
highly dependent on the maximum signal level present in the
input, even when the large signal lies outside the displayed
span.


Residuals
Residual Spurious responses are internally generated spurs
that exist in all spectrum analyzers and are independent of
any input signal. These unwanted signal components are
the result of imperfect isolation between the various signal
paths inside the spectrum analyzer and can come from digital
clocks, local oscillators or switching power supplies. Spectrum
analyzers are regularly used to measure spurs far below their
Residual Spurious specifications. One technique to account
for these residual signals is to measure them by taking a spur
sweep with the input terminated. The resulting list of spur
locations and levels can be tabulated and then removed from
the subsequent measurement results. Residual Spurious
are expressed as an absolute power level, specified in dBm.
This means they do not change level with any input. The
specification for the RSA6120B is -90 dBm for frequencies
from 40 MHz to 200MHz, and -95 dBm (-110 dBm Typical)
from 200 MHz to 20 GHz. These signals must be well
understood to ensure they are not mistakenly included in the
results as they may not be from the DUT.


Spurious with Signal
Spurious with Signal Present or signal-related spurs are the
result of unintended interactions between the input signal
and the various internal clocks and local oscillators that are
part of the spectrum analyzer’s circuitry. Most signal-related
spurs are caused by non-linear behavior in the spectrum
analyzer’s circuitry and are highly dependent on the levels
of signals present at the spectrum analyzer input. There are
several types of signal-related spurs which are often specified
separately. They include image rejection, harmonics, third
order intermodulation, second order intermodulation, etc.
The specifications for signal-related spurs are usually in terms
of dB below the input signal level or dBc. Signal related
spurs specifications are especially relevant if a low level spur
search must be made in the presence of a high level signal.
Measuring spurs in a transmitter output, for example, may
require the measurement of spurs at the -120 dBm level as in
our example while the transmitters intended output signal has
a power of several watts. In these cases, it might be required
to filter out the transmitter’s signal (notch filter) to make sure
that its level does not exceed the input level specified in the
analyzer’s spurious specifications.
All spectrum analyzers publish spurious with signal
specifications that vary with acquisition BW and input
frequency at a specified level of input signal. The RSA6120B,
for example, performance varies from -78 dBc to -70 dBc
depending settings with a maximum signal level of -25 dBm
after RF attenuation. The option 51 preamplifier, when used,
would typically achieve similar performance with a maximum
input signal at approximately a 30 dB lower level.

Harmonics
These unwanted signals can appear whenever the analyzer
is tuned to N times the frequency of a signal present at the
input (N is an integer). The most relevant is the 2nd-harmonic
specification (N=2). The RSA6100B Series specifies 80 dBc
harmonics for -25 dBm input signals with no RF attenuation
and preamplifier off. The Option 51 preamplifier, when used,
would typically achieve similar performance with a maximum
input signal at -55 dBm with no RF attenuation.
Predicting the spurious behavior with Input is usually more
difficult than harmonics or residual spurious. A detailed
analysis requires knowledge of the frequency conversion
stages internal to the Spectrum Analyzer (Local Oscillators,
IF frequencies, ADC clocks, etc.) These kinds of spurs will
be present at frequencies related the mixing of internal
frequencies and harmonics of the input. For example, if one of
the local oscillators in a spectrum analyzer is at 9 GHz, then
signals harmonically related to the combination of the input
signal and 9 GHz could show up as spurs.

How to recognize internally generated spurious what are not input related.

It is simple (mostly).

Terminate input and look if it is there.

Thumb rule: If you have signal connected. Change attenuator example 5 or 10dB. If it is your real signal its level do not change or change is minimimal. If it is spur what is there also if there is not input its level change around same amount as you change attenuator. (because this "signal" does not walk through the attenuation)

Type this phrase to Google: spectrum analyzer residual response

....and start reading.

Here SSA3000X most important normal input independent internally generated  "spurs"   and also one internally (input independent) generated spur compared to real input signal and changing attenuator level. There can see how real signal behave and iiis behave when change attenuator.




Correct: recognize

[pascal_sweden]

Thanks for your detailed feedback! Really nice explanations and screenshots.

The screen composition of the Siglent is very well laid out, and looks very professional!

BTW: I noticed you are based in Finland. Greetings neighbour!
Only spurious response in spectrum analyzers, and no spurious response in ice hockey right?

[tautech]

I won't profess to be any sort of SA expert, so some shots of my fiddlings using a simple 25mm diameter 50 ohm loop "sniffing" various emissions from a Samsung tablet



Loop



I'm quite impressed with the file management UI, you can maybe see some of it in the poor pic above but you can even capture a screenshot of it. 





More fiddling to come. 

[pascal_sweden]

What about the spurious response (both internal spurs without signals connected, or spurs as a result of the interaction with a connected signal) in an analog spectrum analyzer versus the spurious response in a digital spectrum analyzer?

Maybe analog spectrum analyzers are better in this field? Or are they worse?

How is the analog spectrum analyzer HP 8560E doing in this field, in comparison with the digital spectrum analyzer Siglent SSA3021X?  Any clear winner to identify in terms of low to zero spurious response? (both internal spurs without signals connected, or spurs as a result of the interaction with a connected signal)

If the spurious response comes from the device internals itself, isn't there an easy way to characterize the spurious response for a spectrum analyzer under no load, and calibrate, adjust, and eliminate it from the readings?

Or would that spurious response be unstable over time, depending on external parameters, such as electro-magnetic noise, interference, reflection, ambient temperature?

For comparison: Are all spectrum analyzers subject to this? How does the Rigol DSA185-TG perform on this front?

Is it too costly to make a spectrum analyzer with zero spurious response, or is it just impossible by design, based on the topology and build up of the various stages in the spectrum analyzer?

Are there different variations possible in the build up of the various stages in the spectrum analyzer,
and are some variations less subject to spurious response compared to others?

Is the build up of the stages in most spectrum analyzers the same or very similar, or are there 2 or more fundamental variations and approaches that are completely different from each other, and that have a completely different internal spurious response by design?

[tautech]

What about the spurious response (both internal spurs without signals connected, or spurs as a result of the interaction with a connected signal) in an analog spectrum analyzer versus the spurious response in a digital spectrum analyzer?

Maybe analog spectrum analyzers are better in this field? Or are they worse?

How is the analog spectrum analyzer HP 8560E doing in this field, in comparison with the digital spectrum analyzer Siglent SSA3021X?  Any clear winner to identify in terms of low to zero spurious response? (both internal spurs without signals connected, or spurs as a result of the interaction with a connected signal)

As suggested study replies #298 and #314.

[Performa01]

What about the spurious response (both internal spurs without signals connected, or spurs as a result of the interaction with a connected signal) in an analog spectrum analyzer versus the spurious response in a digital spectrum analyzer?

Maybe analog spectrum analyzers are better in this field? Or are they worse?
...

The difference between ‘analog’ and ‘digital’ spectrum analyzers is only in the final IF processing, where analog SAs basically have (quartz crystal) filter banks, whereas digital ones utilize a high resolution ADC and digital signal processor.

The RF part is basically always the same, i.e. some swept multi-conversion superheterodyne receiver - and this is where the spurs come from. It is mainly the non-ideal characteristics of amplifiers, mixers, filters and LO synthesizer(s) causing the spurs (both residual and input signal related ones).

So it is not a question of digital vs. analog, but the design of the RF part of the instrument. The Siglent SSA3000X appears pretty decent and I would be surprised to find any vintage SA that performs significantly better in terms of residual spurs. It might be something different when it comes to signal related spurs, particularly intermodulation products, where I assume that the SSA3000X is as good as the majority of SAs, but of course cannot compete with the few instruments optimized for high dynamic range like R&S FSEA30, probably not even with an old HP 8560 as this instrument also appears to perform above average in terms of 3rd order dynamic range.

Btw. The HP8560 are ‘digital’ as well, at least for RBWs of 100Hz and below.

[tautech]

More fiddlings 



  My ~3 yr old Siglent SDG1010 non-external referenced AWG appears to be a whole 20 Hz out.   

AWG settings 10 MHz (max) @ 200 mV p-p.

[tautech]

So how accurate is a new SDG2042X ?



That's better, only 2 Hz off without an external reference. 

[hendorog]

That's better, only 2 Hz off without an external reference. 

Impressive! At 40 MHz thats 50ppb isn't it?

20Hz out in the previous post is 2ppm by the same logic.
I don't know if thats the correct way to calculate it, the beauty of EEVBlog is there is always someone who knows.

[tautech]

That's better, only 2 Hz off without an external reference. 

Impressive! At 40 MHz thats 50ppb isn't it?

20Hz out in the previous post is 2ppm by the same logic.
I don't know if thats the correct way to calculate it, the beauty of EEVBlog is there is always someone who knows.

Not having a frequency counter I wasn't sure what to expect but I did know there are many that use the SDG1k series externally referenced and some even rework/replace the internal reference for better accuracy. There's quite a bit of discussion about that in the SDG1000 and 800 thread.
Being nosey and needing to become more familiar with SA's I just had to check. 

However there's an Auto Cal setting in the SSA3kX I had turned off. 
After getting into that part of the UI I hit the Help for some info on it.
It first runs after 30 mins then at intervals after that.......If it's been done right the frequency of Auto Cal should be related to ambient temp variations.........well, IMO.

More fiddling needed. 

[tautech]

Spurred on by the above results and gaining a little more confidence with basic use it was time to look at another sine wave output, this time from the inbuilt AWG in a SDS2000X DSO.

Autoset


Yep, that's what is obtainable with just one push but look at the resolution bandwidth, no way I could see what I was looking for.

Autoset Zoom


Better but the frequency Span and sweep values are still too large.

Manual


A few tweaks later in manual mode revealed what I was looking for.
108 Hz out. 

But from the SDS2kX datasheet:
SDS2000X series AWG Frequency Accuracy ±50 ppm