Siglent SDG1000X Waveform Generators

[tautech]

yes I discovered that hiZ changes only value of voltage.
I used a 50 Ohm termination

See last image in this post:
https://www.eevblog.com/forum/testgear/siglent-sdg1000x-waveform-generators/msg1449433/#msg1449433

[gianluigi]

yes I discovered that hiZ changes only value of voltage.
I used a 50 Ohm termination

See last image in this post:
https://www.eevblog.com/forum/testgear/siglent-sdg1000x-waveform-generators/msg1449433/#msg1449433

something is wrong in what I'm doing and I don't understand what!
Now also with 50 Ohm termination I see the first waveform

Trying also at 1 MHz the waveform is not good: I see ringing...
could be possible that termination is not good or the cable?
something is changed now

[rf-loop]

Good quality coaxial, example  M17/84-RG223 or similar with good quality Suhner connectors and good quality Tektronix 50ohm FeedThru terminator in scope input. Most frustrating thing in lab is bad cables, bad connectors bad crap this and that. They come expensive...
Just for this purpose here it can do well with nearly all normal good quality 50 ohm  rf coax.

One of most hated things are poor quality BNC connectors. 

But back to things here.
Here is typical one examples how SDG1000X Square looks.
In these old images used scope was SDS1202X-E
Yes cable was, if I remember right my normally used some M17/84-RG223 etc and some original Tek feedthru terminator.



If Siglent do not give roughly this result, something is wrong in Scope, Signal path or generator itself



30 MHz  My own test. SDG1032X/62X





60 MHz  My own test. SDG1032X/62X



Then just for fun Rigol DS1032Z,  25MHz, 50ohm termination and scope 100MHz Rigol (not my test)
Do not take seriously, this is not for any competition. Just for keep things somehow relative in price class.

[gianluigi]

Thank you
your explanation was very comprehensive.
The only thing that I don't explain is why leaving the hardware set up unchanged (cable, termination and oscilloscope) the displayed waveform changes when I change the load setting of the SDG (50 vs HI-imp).
I am convinced that it should be the same waveform, but only the indication of Vpp should change.
In any case I do other tests and update you.
Thanks to the availability

[gianluigi]

I made some tests:
I think (and I hope)  there is a bug in the firmware:
When it is in square wave and you switch off the channel, change setting of load (50 -> Hiz) and switch on the channel I have distorsion.
After that I switch off the channel, change setting of load (Hiz -> 50) and switch on the channel I have not distorsion.
If I set load from 50 to 60 it is ok, if I set load setting > 61 then I have distorsion.

It seems that when I get the distortion of the square wave another internal relay is also activated which does not eliminate the low pass filter used for the sinusoidal signal.
You can clearly hear the switching of an additional relay, when you set the different impedance and turn the channel off and on; when the impedance is not changed and the channel is switched off and on, the sound of the relays is lower.
Other times doing different steps it seems to freeze and selecting the square wave always comes out distorted.
Can anyone try his generator?
The fw is 1.01.01.33R1B6.
The sequence is:
- square wave 30 MHz 1 Vpp
- load 50 Ohm
- channel on -> good square wave
- channel off
- load Hiz
- channel on -> distorted square wave
- channel off
- load 50 Ohm
- channel on -> good square wave

[Orange]

There is definitely a difference.
I would say it's better (faster rise time) with HiZ
See pictures,

BTW taken with an Agilent MSO7054, 50 ohm terminated

This is on a real 1062, not an hacked 1032 

[tubularnut]

Can anyone try his generator?
The fw is 1.01.01.33R1B6.
The sequence is:
- square wave 30 MHz 1 Vpp
- load 50 Ohm
- channel on -> good square wave
- channel off
- load Hiz
- channel on -> distorted square wave
- channel off
- load 50 Ohm
- channel on -> good square wave

I'll give it a go on my real 1062 when I get home in a few hours.

[gianluigi]

There is definitely a difference.
I would say it's better (faster rise time) with HiZ
See pictures,

BTW taken with an Agilent MSO7054, 50 ohm terminated

This is on a real 1062, not an hacked 1032 

You got the same behavior I get.
90% of the time it is systematic, other times it stops with the distorted waveform.
If you try 60 MHz the difference is evident and the distortion is very bad!
Could you try?
thank you

[Hexley]

There is definitely a difference.
I would say it's better (faster rise time) with HiZ
See pictures,

BTW taken with an Agilent MSO7054, 50 ohm terminated

This is on a real 1062, not an hacked 1032 

You got the same behavior I get.
90% of the time it is systematic, other times it stops with the distorted waveform.
If you try 60 MHz the difference is evident and the distortion is very bad!
Could you try?
thank you

Here is a test of both 30MHz and 60 MHz. The generator is an upgraded SDG1062X (started life as as SDG1032X). The scope is a 500 MHz-modified DSOX3024A.

[Orange]

The thing is that at 60MHz the ringing of the LPF in the generator is showing its effect. It is a dilemma you have to face.

I use my generator for a lot of RF work, and have it always set to 50Ohms. This gives me the dBm option. And I use it mostly in sine mode.
However I now see that if you want to use the Square wave, it is better to use it in HiZ mode, you get better rise time and also higher amplitude.

Still, this thing is not really performing at this high frequency. The more expensive SDS2042X is even worse in rise time, and only goes to 25MHz....
Oh I tested also a hacked 1032X, I get the same pictures as on my 1062X



 

[gianluigi]

Since I'm not the only one having this behavior, I think it's not a hardware defect.
I am more convinced it may be a firmware bug also because now I can no longer generate the distortion, even by doing the procedure listed by me.
Now the problem occurs both in Hiz and 50 exceeding the threshold of 1Vpp at 50 Ohm.
I don't know now 

[gianluigi]

Since I'm not the only one having this behavior, I think it's not a hardware defect.
I am more convinced it may be a firmware bug also because now I can no longer generate the distortion, even by doing the procedure listed by me.
Now the problem occurs both in Hiz and 50 exceeding the threshold of 1Vpp at 50 Ohm.
I don't know now 

I understand what the reason is!
The change in load setting only results in coincidence.
The real reason why the square wave sometimes comes out "clean" or distorted is due to the voltage range that the generator uses.
In fact, exceeding the value of 1Vpp by 1 mV even with load set at 50 Ohm, distortion occurs.
If you set the output to 995 Vpp 50 Ohm and then set Hiz there is no distortion.
Obviously, the variation of the waveform (distortion) due to the change in gain of the output stage occurs for the threshold value of 1 Vpp and via software when the output is activated there is a different setting of the output gain due to different approximations in the calculations.

[tautech]

Since I'm not the only one having this behavior, I think it's not a hardware defect.
I am more convinced it may be a firmware bug also because now I can no longer generate the distortion, even by doing the procedure listed by me.
Now the problem occurs both in Hiz and 50 exceeding the threshold of 1Vpp at 50 Ohm.
I don't know now 

I understand what the reason is!
The change in load setting only results in coincidence.
The real reason why the square wave sometimes comes out "clean" or distorted is due to the voltage range that the generator uses.
In fact, exceeding the value of 1Vpp by 1 mV even with load set at 50 Ohm, distortion occurs.
If you set the output to 995 Vpp 50 Ohm and then set Hiz there is no distortion.
Obviously, the variation of the waveform (distortion) due to the change in gain of the output stage occurs for the threshold value of 1 Vpp and via software when the output is activated there is a different setting of the output gain due to different approximations in the calculations.

Hmm.
That reasoning doesn't align with the datasheet spec.

Can you try again after returning your unit to Default settings ?
Utility>System>Set to Default

[Hexley]

This behavior shows up in some other use cases.

First, set up the instrument to drive an oscilloscope with 50 ohm input termination.

• Set the instrument for 25 MHz square wave output, 50 ohm load, 500 mVpp. Enable modulation by pressing the Mod key. Select PM, and choose some nominal setting for the modulation wave, such as 1 Hz and 1 degree deviation. Now enable/disable modulation with the Mod key, and note how the waveform instantly changes from properly damped to the underdamped shape whenever modulation is enabled. This is true for any modulation type; PM is used for this example since there is no question of it changing the amplitude of the modulated wave.  The behavior can be seen down to 20 mVpp output signals.
• Alternatively, enable sweep by pressing the Sweep key. Again select nominal parameters for the sweep setting, such as start frequency = 24.995MHz, and stop frequency = 25.000 MHz. The waveform again changes instantly from properly damped to underdamped as the Sweep key is toggled. Again, this is true even for very small output levels.

[rf-loop]

Since I'm not the only one having this behavior, I think it's not a hardware defect.
I am more convinced it may be a firmware bug also because now I can no longer generate the distortion, even by doing the procedure listed by me.
Now the problem occurs both in Hiz and 50 exceeding the threshold of 1Vpp at 50 Ohm.
I don't know now 

I understand what the reason is!
The change in load setting only results in coincidence.
The real reason why the square wave sometimes comes out "clean" or distorted is due to the voltage range that the generator uses.
In fact, exceeding the value of 1Vpp by 1 mV even with load set at 50 Ohm, distortion occurs.
If you set the output to 995 Vpp 50 Ohm and then set Hiz there is no distortion.
Obviously, the variation of the waveform (distortion) due to the change in gain of the output stage occurs for the threshold value of 1 Vpp and via software when the output is activated there is a different setting of the output gain due to different approximations in the calculations.

Independent of load value selected (what can be what ever between 50 ohm to HiZ) we need just think final amplifier level. There is small changes when it is working in different output voltage bands. I do not remember these thresholds where system change level band. Between these ranges there is small differences in edge response. I have tested it and it is nominal to this and not failure/error. As oscilloscopes also generators result is sum of errors. There is not ideal machines instead of school books.
Now it depends of course what kind of compromises tghere have done in circuit. All can do nearly - nearly - perfect instrument if have infinite amount of bank notes for it. Even I can. Question is how cheap can done instrument what still is acceptable in performance in its price segment.
 
With all these anomalies it well meet its specs. What other it need meet.

After then users use these and when users use they add own errors when they measure these with not ideal scope and signal connected with not ideal matching. Also SDG output true impedance, what is ALWAYS nominal 50ohm DC impedance +/- something and after then some reactances.

Just need understand real practice and reality. All can be better .. but...

Now sum together all error marginals (scope, sdg, signal pathway between, in your known environment) and make worst case calculus and best case calculus and for all parameters and think between these is "window" where inside need stay.  After then look if your result is inside this "window". if this match without breaking window borders all is ok. If still not ok, then you have wrong instruments for your needs. Period. Real world is not perfect.

But also then, if results do not stay inside this previously used "window" thebn need be alarmed and then need look what is out of order and perhaps need repair etc.

Remember, what ever External Load set in SDG settings its output impedance do not change.
So if select

Case 1. 50ohm and set voltage level out example 1Vpp it is 1Vpp  IF external load is perfectly 50ohm and only then. 
Case 2.  if then set HiZ and also external load is HiZ and set output is 1V, it is 1V IF external load is perfectly HiZ..

But in Case 1. SDG internal end aplifier output is 2V
and in Case 2. SDG internal end amplifier output is 1V 

But in both cases, using right Load, user see with oscilloscope this 1V.

Now all understand that SDG works with different internal output voiltage and this may lead to bit different square or other wave shape. Specially if these two voltages are in differen internal voltage band  (you can listen these relays change)

And as just told, it is NOT perfect ideal machine. So there is differences and no need wonder. Least I do not wonder. Even when I think it can do perhaps better without high rise in retail price.
But products must also be in its performance / price segment and not competite example with same manufacturer more expensive segment products, what it partially now done,  if we look some small part of functions separately. Example more expensive SDG2000X can not do this kind of square wave. It do not at all have this.
 
Of course SDG1000X can not what many things SDG2kX can but square/rectangle it beat it.

It was also small surprise Siglent add TrueArb feature tiny version to SDG1000X just in one previous FW update. 
Also for users who use it as rf generator they time ago change level handling when AM modulation. Now if you turn AM mod on or off or change modulation depth it keep constant carrier level as all serious conventional rf generators have done tens of years. 
Many things can do inside FW but some are in HW circuit and thats it. Just like this small changes in square shape over whole voltage range. it change and it need accept. If it is broken, out of normal, then need repair or factory cal depending problem.

Times ago I have very expensive (and heavy, made like tank and inside pure gold more than little) HP programmable pulse generator.  And also its pulse edge change some amount when pulse level change and more when it move to some next level voltage band. Also overshoot change some amount and sure it was not due my scope. Also signal pathway from generator to scope was best possible Huber Suhner cable what short cable price is perhaps ten of SDG1000X.  But this expensive generator pulse was still inside specs. Not at all ideal perfect but, inside specs.  If I try use grade C specs instrument for grade B needs I am wrong not instruments. Opposite way is  -  expensive but easy.


Important is now just look if OP SDG is normal ok or if it have problem or if user have problem in his thinking or his other things used for testings. Nearly all is possible.

[gianluigi]

you are completely right.
Before now I did not know this model and was perplexed by the behavior of the load setup change. Since the instrument is new, I thought it was faulty but obviously the reason is the change in the gain range of the output stage.
Thanks for your support.

[rf-loop]

you are completely right.
Before now I did not know this model and was perplexed by the behavior of the load setup change. Since the instrument is new, I thought it was faulty but obviously the reason is the change in the gain range of the output stage.
Thanks for your support.

Just this what you have done is very important part to do when who ever get some new instrument. It is important, many times used, phrase "Know your instrument" for successful using in many kind of variable purposes. It is important to know limits and performance with  some own selfmade tests etc, in different situations so that instrument can not pull user to more or less hidden trap.

[Johnny B Good]

you are completely right.
Before now I did not know this model and was perplexed by the behavior of the load setup change. Since the instrument is new, I thought it was faulty but obviously the reason is the change in the gain range of the output stage.
Thanks for your support.

Just this what you have done is very important part to do when who ever get some new instrument. It is important, many times used, phrase "Know your instrument" for successful using in many kind of variable purposes. It is important to know limits and performance with  some own selfmade tests etc, in different situations so that instrument can not pull user to more or less hidden trap.

 I couldn't agree more!  Indeed, I'd made this very point in the long running FY6600 thread some time back (and, it turns out, not without good reason).

 Having recently purchased a SDG1032X from Telonic Instruments Ltd, I've made the rather surprising discovery that the humble FY6600-60M I'd bought almost two years ago, is in many ways superior to this Siglent generator (especially now that it has a 10MHz CQE branded - as used by Symmetricom - OCXO, stable to within +/- 100ppt, which can be injection locked to an external 10MHz reference).

 As is typically the case with published specifications, the usual "lying by omission" (the favoured tactic of the ad man) means you have to probe very deep to elicit, for example, that the quoted 300ns jitter figure is an rms value rather than a pk-pk one which typically corresponds to a pk-pk value around the 800 to 900ns mark which compares very unfavourably against the observed 200 to 250ns pk-pk jitter on sine and Sinc pulse waveforms generated by the FY6600. Also, the Sinc pulse generated by the Siglent at the 6MHz limit wriggles and squirms to the point of being almost completely unusable versus that of the Feeltech which shows only the slightest hint of this effect at 12MHz (watch the attached movie files depicting this effect and that of the Flicker noise jitter from its internal reference clock).

 With only 8 digits of frequency display resolution, I haven't found any way to offset a 10MHz signal by Siglent's promised 1uHz resolution, unlike the FY6600's (verified) promise of such resolution of 1uHz over the full DC to 60MHz frequency range of the FY6600-60M model courtesy of the 14 digit resolution display which seems OTT until you upgrade to an OCXO with an external reference clock optional extra (injection locked rather than phase locked in my case) with the potential to let you meaningfully utilise all 14 of those digits of precision.

 I chose the SDG1000X over the more expensive SDG2000X series simply on account of its superior square wave performance since this seemed more useful for my needs but, having witnessed the surprising limitations and the severe flicker noise like jitter on its internal clock reference which clears up when you use an external 10MHz reference, I'm now beginning to regret not spending the extra cash on an SDG2042X.

 I don't expect miracles of stability and accuracy out of the cheap 50ppm smd XO chip Siglent had decided to inflict upon their SDG1000X models since the external clock reference socket takes care of that requirement assuming the +1.43mHz offset at 10MHz isn't an issue.

 That, BTW, corresponds to a gain of one second per 45 years. That may seem pretty good but it's shite compared to the minuscule rounding error in that FY6600's DDS processing which requires some 1.8 million years to accumulate a one second error (well over 4 orders of magnitude more accurate!).

 I'm guessing that bogus +1.43mHz offset must have been due to my triggering from the RFS waveform at that time. It isn't locked to anything (natch. ) and so must have been adrift of the GPSDO by -1.43mH at the time - it isn't mounted into an enclosure where I can stabilise the baseplate temperature (and ultimately compensate for barometric pressure changes) to make it worth adjusting it to within 50uHz again.

 Anyway, the use of a cheap smd XO and the horrible Flicker noise like jitter leaves me wondering whether this a systemic issue that's simply gone unnoticed or just a case of bad luck on my part for landing up with a "Lemon" that just happened to be afflicted by a faulty XO chip.

 I've attached a couple of short movie clips you should find particularly interesting, demonstrating the clock jitter and the shape shifting Sinc pulse with a picture showing the required +11 or 12 Hz offset to get within one Hz of the RFS and GPSDO reference frequencies in order to reveal the internal clock jitter.

 CH1 shows the 10MHz sine output from the FY6600's second channel which is locked to the GPSDO. This is the trigger source. CH2 displays the sine wave output from the SDG1032X. CH3 is showing the free running RFS with CH4 displaying the FY6600's 10MHz Sinc pulse on its second channel (MVI_3031.mkv - 5 second clip).

 In the second, 1 minute long movie clip (MVI_3039.mkv), I've turned the odd channels off to de-clutter the display. The Sinc pulse that isn't wriggling is the 12MHz FY6600 one, the other 6MHz Sinc pulse comes from the SDG1032X.

============================================================================================
[EDIT 2020-10-04]

 I've just repeated that last test and can no longer recreate that "squirmy" effect on the Sinc pulse  The only thing I did notice was a rather dramatic drop in amplitude when setting it to the 6MHz maximum frequency limit.

============================================================================================
 So, my question is this:- Is the clock jitter a faulty smd XO chip or the result of Siglent trying to reduce their BoM costs by using the cheapest of smd XOs they could lay their hands on with (just like Feeltech) no consideration for the consequences?

 It would be interesting to see whether this jitter effect afflicts all SDG1000Xs in general or whether it's merely an unfortunate manufacturing defect in the unit that was shipped out to me.

Regards, John

[Johnny B Good]

Here's another 20 seconds movie clip using a 10MHz square wave from the SDG1032X demonstrating the weird jitter from its internal frequency reference clock, plus a picture of the generators' settings.

 John

[tautech]

Question on Discord WRT adding data bits onto a waveform.
SDG1032X ch1 2V p-p 50 Hz sinewave.
ch2 3V p-p 200us pulse.
Phases locked.

Wave Combine result:

[Johnny B Good]

you are completely right.
Before now I did not know this model and was perplexed by the behavior of the load setup change. Since the instrument is new, I thought it was faulty but obviously the reason is the change in the gain range of the output stage.
Thanks for your support.

Just this what you have done is very important part to do when who ever get some new instrument. It is important, many times used, phrase "Know your instrument" for successful using in many kind of variable purposes. It is important to know limits and performance with  some own selfmade tests etc, in different situations so that instrument can not pull user to more or less hidden trap.

 I couldn't agree more!  Indeed, I'd made this very point in the long running FY6600 thread some time back (and, it turns out, not without good reason).

 Having recently purchased a SDG1032X from Telonic Instruments Ltd, I've made the rather surprising discovery that the humble FY6600-60M I'd bought almost two years ago, is in many ways superior to this Siglent generator (especially now that it has a 10MHz CQE branded - as used by Symmetricom - OCXO, stable to within +/- 100ppt, which can be injection locked to an external 10MHz reference).

 As is typically the case with published specifications, the usual "lying by omission" (the favoured tactic of the ad man) means you have to probe very deep to elicit, for example, that the quoted 300ns jitter figure is an rms value rather than a pk-pk one which typically corresponds to a pk-pk value around the 800 to 900ns mark which compares very unfavourably against the observed 200 to 250ns pk-pk jitter on sine and Sinc pulse waveforms generated by the FY6600. Also, the Sinc pulse generated by the Siglent at the 6MHz limit wriggles and squirms to the point of being almost completely unusable versus that of the Feeltech which shows only the slightest hint of this effect at 12MHz (watch the attached movie files depicting this effect and that of the Flicker noise jitter from its internal reference clock).

 With only 8 digits of frequency display resolution, I haven't found any way to offset a 10MHz signal by Siglent's promised 1uHz resolution, unlike the FY6600's (verified) promise of such resolution of 1uHz over the full DC to 60MHz frequency range of the FY6600-60M model courtesy of the 14 digit resolution display which seems OTT until you upgrade to an OCXO with an external reference clock optional extra (injection locked rather than phase locked in my case) with the potential to let you meaningfully utilise all 14 of those digits of precision.

 I chose the SDG1000X over the more expensive SDG2000X series simply on account of its superior square wave performance since this seemed more useful for my needs but, having witnessed the surprising limitations and the severe flicker noise like jitter on its internal clock reference which clears up when you use an external 10MHz reference, I'm now beginning to regret not spending the extra cash on an SDG2042X.

 I don't expect miracles of stability and accuracy out of the cheap 50ppm smd XO chip Siglent had decided to inflict upon their SDG1000X models since the external clock reference socket takes care of that requirement assuming the +1.43mHz offset at 10MHz isn't an issue.

 That, BTW, corresponds to a gain of one second per 45 years. That may seem pretty good but it's shite compared to the minuscule rounding error in that FY6600's DDS processing which requires some 1.8 million years to accumulate a one second error (well over 4 orders of magnitude more accurate!).

 I'm guessing that bogus +1.43mHz offset must have been due to my triggering from the RFS waveform at that time. It isn't locked to anything (natch. ) and so must have been adrift of the GPSDO by -1.43mH at the time - it isn't mounted into an enclosure where I can stabilise the baseplate temperature (and ultimately compensate for barometric pressure changes) to make it worth adjusting it to within 50uHz again.

 Anyway, the use of a cheap smd XO and the horrible Flicker noise like jitter leaves me wondering whether this a systemic issue that's simply gone unnoticed or just a case of bad luck on my part for landing up with a "Lemon" that just happened to be afflicted by a faulty XO chip.

 I've attached a couple of short movie clips you should find particularly interesting, demonstrating the clock jitter and the shape shifting Sinc pulse with a picture showing the required +11 or 12 Hz offset to get within one Hz of the RFS and GPSDO reference frequencies in order to reveal the internal clock jitter.

 CH1 shows the 10MHz sine output from the FY6600's second channel which is locked to the GPSDO. This is the trigger source. CH2 displays the sine wave output from the SDG1032X. CH3 is showing the free running RFS with CH4 displaying the FY6600's 10MHz Sinc pulse on its second channel (MVI_3031.mkv - 5 second clip).

 In the second, 1 minute long movie clip (MVI_3039.mkv), I've turned the odd channels off to de-clutter the display. The Sinc pulse that isn't wriggling is the 12MHz FY6600 one, the other 6MHz Sinc pulse comes from the SDG1032X.

============================================================================================
[EDIT 2020-10-04]

 I've just repeated that last test and can no longer recreate that "squirmy" effect on the Sinc pulse  The only thing I did notice was a rather dramatic drop in amplitude when setting it to the 6MHz maximum frequency limit.

============================================================================================
 So, my question is this:- Is the clock jitter a faulty smd XO chip or the result of Siglent trying to reduce their BoM costs by using the cheapest of smd XOs they could lay their hands on with (just like Feeltech) no consideration for the consequences?

 It would be interesting to see whether this jitter effect afflicts all SDG1000Xs in general or whether it's merely an unfortunate manufacturing defect in the unit that was shipped out to me.

Regards, John

 This is by way of a (relatively) quick update on the situation regarding the SDG1032X as described in my previous post quoted above.

 You may be pleased to note that I won't be posting any more video clip attachments demonstrating the rather curious clock jitter issue and shortcomings relative to a cheap toy AWG of its arbitrary waveforms (notably that of the Sinc pulse waveform limited to a maximum frequency of just 6MHz which plummets by a bad 6 to 10dB compared to the 5MHz output level).

 I had come to conclude that the horrible clock jitter was simply due to a fault in or around the smd xo chip and have returned it to Telonic for a full refund. Telonic accepted my request without argument and even sent a courier to collect it and I have no qualms about doing any future business with them the next time I'll be looking for more T&M kit.

 The delay involved in reaching this point seems to lie with Siglenteu, leaving Telonic as much a victim as I was. It seems this particular defect is extremely rare (or more likely a sneaky type of fault that isn't so readily revealed using a cursory waveform check with a 'scope), hence the apparent misunderstanding of the problem by Siglenteu's "technicians" who obviously lack the required experience to understand the big "Clue by Four" video clip evidence passed onto them by Telonic.

 I did quite a bit of searching to find similar issues strongly hinting at this oscillator instability but none I could point at and say, "Here's another example of the jittery oscillator issue!". For instance, one example involved a Rigol (what else  ) MSO which quite clearly to me was a timebase jitter issue that could only be another example of the same unstable clock reference I was seeing. The fact that it was effecting the timebase in an item of test equipment made it all the more difficult to demonstrate what was actually going on (in this case, the MSO would have to opened up and the circuitry probed using another properly functioning 'scope to pin down the culprit).

 With my request for either a full refund or replacement with an SDG2042X with my paying the price difference going ignored whilst Telonic were so preoccupied with their dealings with Siglenteu that they forgot to keep me informed, I decided to have done with the whole business and insist on a straightforward refund and go to my previous supplier for a swift delivery of the SDG2042X I'd hoped would suffice as a sufficient upgrade as to finally allow me to retire my much abused FY6600-60M.

 You might be surprised to learn that I still might have to wait a while before I can actually retire that FY6600 since there are still some serious issues to be addressed by a thoroughly revamped firmware update. It's far more than just a user hostile UI that has to be re-written but also code in the FPGA to correct inexcusable shortcomings in the generation of Arbitrary waveforms, most notably the Sinc pulse which only looks good at the (apparently magic) 10MHz frequency setting (but which my SDS2504X Plus's CCJ measurement stats still show a +/-171ns jitter regardless) turning into a scrambled jittery mess just 1Hz either side of this magic 10MHz spot frequency (but curiously, no change to the CCJ stats).

 Meanwhile, the FY6600-60M's Sinc pulse carries on demonstrating a -0.15ns and +0.15ns CCJ stat regardless of frequency setting around the 10MHz mark (I haven't bothered checking out its 15 and 20MHz stats - just the +/-4Hz neighbourhood of the magic 10MHz that seems to disguise this CCJ issue in the SDG2042X).

 Now, in view of the fact that the hardware in a 1.2GSa/s 16 bit AWG must be at least an order of magnitude better in both build quality and specification over the hardware used in a cheap plastic cased 250MSa/s 14bit AWG, the problems can surely only be a matter of piss poor firmware. I mean, how else can you cripple an order of magnitude better hardware into something that falls short in so many aspects compared to a cheap "Toy AWG" like Feeltech's FY6600-60M?

 It's only the fact that a firmware upgrade is a doable fix that stops me from immediately declaring it as "Unfit for purpose" and claiming a full refund to either give up on the whole idea of upgrading to an improved replacement for my 'Toy AWG' or spend yet more wads of cash on an "A Brand" alternative which, the way things have been going, might yet prove to have its own surprising shortcomings.

 So here I am, contemplating whether it's worth hanging onto the SDG2042X or simply giving up and send it back for a full refund. The problem is there aren't really any viable alternatives in the same sort of price bracket (Rigol are most definitely not an alternative imo).

 I'm not fishing for buying advice here. I know researching the market is likely to be a long drawn out ball ache of a process to find a best value for money solution that's likely to still involve a compromise or three (one of which is almost certainly going to involve the state of my bank balance).

 Just for added interest, even though mention of the SDG2042X is slightly off topic in this thread, I've attached some screen shots to give you a little more detail on the CCJ stats and the peculiar jitter of the 2042's Sinc pulse waveform compared to that of the Feeltech's at 10 and 10.000001MHz

 CH1 is sinewave from the 2042
 CH2 is the RFS
 CH3 is the Sinc pulse from the 2042
 CH4 is the Sinc pulse from the FY6600 from which I'm triggering the 'scope

 I've disabled CH1 CH2 in the last four images to clarify the Sinc pulse waveforms. The first two of these images show the waveforms at exactly 10MHz (both generators locked to my GPSDO) and the last two are at 10.000001MHz to demonstrate the jitter effect which reveals itself only on the 2042's pulse - the FY6600's never so much as bats an eyelid in response to this 1Hz offset and this remains true for the 2, 3 and 4Hz offsets I tried.

John

[rf-loop]

you are completely right.
Before now I did not know this model and was perplexed by the behavior of the load setup change. Since the instrument is new, I thought it was faulty but obviously the reason is the change in the gain range of the output stage.
Thanks for your support.

Just this what you have done is very important part to do when who ever get some new instrument. It is important, many times used, phrase "Know your instrument" for successful using in many kind of variable purposes. It is important to know limits and performance with  some own selfmade tests etc, in different situations so that instrument can not pull user to more or less hidden trap.

 I couldn't agree more!  Indeed, I'd made this very point in the long running FY6600 thread some time back (and, it turns out, not without good reason).

 Having recently purchased a SDG1032X from Telonic Instruments Ltd, I've made the rather surprising discovery that the humble FY6600-60M I'd bought almost two years ago, is in many ways superior to this Siglent generator (especially now that it has a 10MHz CQE branded - as used by Symmetricom - OCXO, stable to within +/- 100ppt, which can be injection locked to an external 10MHz reference).

 As is typically the case with published specifications, the usual "lying by omission" (the favoured tactic of the ad man) means you have to probe very deep to elicit, for example, that the quoted 300ns jitter figure is an rms value rather than a pk-pk one which typically corresponds to a pk-pk value around the 800 to 900ns mark which compares very unfavourably against the observed 200 to 250ns pk-pk jitter on sine and Sinc pulse waveforms generated by the FY6600. Also, the Sinc pulse generated by the Siglent at the 6MHz limit wriggles and squirms to the point of being almost completely unusable versus that of the Feeltech which shows only the slightest hint of this effect at 12MHz (watch the attached movie files depicting this effect and that of the Flicker noise jitter from its internal reference clock).

 With only 8 digits of frequency display resolution, I haven't found any way to offset a 10MHz signal by Siglent's promised 1uHz resolution, unlike the FY6600's (verified) promise of such resolution of 1uHz over the full DC to 60MHz frequency range of the FY6600-60M model courtesy of the 14 digit resolution display which seems OTT until you upgrade to an OCXO with an external reference clock optional extra (injection locked rather than phase locked in my case) with the potential to let you meaningfully utilise all 14 of those digits of precision.

 I chose the SDG1000X over the more expensive SDG2000X series simply on account of its superior square wave performance since this seemed more useful for my needs but, having witnessed the surprising limitations and the severe flicker noise like jitter on its internal clock reference which clears up when you use an external 10MHz reference, I'm now beginning to regret not spending the extra cash on an SDG2042X.

 I don't expect miracles of stability and accuracy out of the cheap 50ppm smd XO chip Siglent had decided to inflict upon their SDG1000X models since the external clock reference socket takes care of that requirement assuming the +1.43mHz offset at 10MHz isn't an issue.

 That, BTW, corresponds to a gain of one second per 45 years. That may seem pretty good but it's shite compared to the minuscule rounding error in that FY6600's DDS processing which requires some 1.8 million years to accumulate a one second error (well over 4 orders of magnitude more accurate!).

 I'm guessing that bogus +1.43mHz offset must have been due to my triggering from the RFS waveform at that time. It isn't locked to anything (natch. ) and so must have been adrift of the GPSDO by -1.43mH at the time - it isn't mounted into an enclosure where I can stabilise the baseplate temperature (and ultimately compensate for barometric pressure changes) to make it worth adjusting it to within 50uHz again.

 Anyway, the use of a cheap smd XO and the horrible Flicker noise like jitter leaves me wondering whether this a systemic issue that's simply gone unnoticed or just a case of bad luck on my part for landing up with a "Lemon" that just happened to be afflicted by a faulty XO chip.

 I've attached a couple of short movie clips you should find particularly interesting, demonstrating the clock jitter and the shape shifting Sinc pulse with a picture showing the required +11 or 12 Hz offset to get within one Hz of the RFS and GPSDO reference frequencies in order to reveal the internal clock jitter.

 CH1 shows the 10MHz sine output from the FY6600's second channel which is locked to the GPSDO. This is the trigger source. CH2 displays the sine wave output from the SDG1032X. CH3 is showing the free running RFS with CH4 displaying the FY6600's 10MHz Sinc pulse on its second channel (MVI_3031.mkv - 5 second clip).

 In the second, 1 minute long movie clip (MVI_3039.mkv), I've turned the odd channels off to de-clutter the display. The Sinc pulse that isn't wriggling is the 12MHz FY6600 one, the other 6MHz Sinc pulse comes from the SDG1032X.

============================================================================================
[EDIT 2020-10-04]

 I've just repeated that last test and can no longer recreate that "squirmy" effect on the Sinc pulse  The only thing I did notice was a rather dramatic drop in amplitude when setting it to the 6MHz maximum frequency limit.

============================================================================================
 So, my question is this:- Is the clock jitter a faulty smd XO chip or the result of Siglent trying to reduce their BoM costs by using the cheapest of smd XOs they could lay their hands on with (just like Feeltech) no consideration for the consequences?

 It would be interesting to see whether this jitter effect afflicts all SDG1000Xs in general or whether it's merely an unfortunate manufacturing defect in the unit that was shipped out to me.

Regards, John

 This is by way of a (relatively) quick update on the situation regarding the SDG1032X as described in my previous post quoted above.

 You may be pleased to note that I won't be posting any more video clip attachments demonstrating the rather curious clock jitter issue and shortcomings relative to a cheap toy AWG of its arbitrary waveforms (notably that of the Sinc pulse waveform limited to a maximum frequency of just 6MHz which plummets by a bad 6 to 10dB compared to the 5MHz output level).

 I had come to conclude that the horrible clock jitter was simply due to a fault in or around the smd xo chip and have returned it to Telonic for a full refund. Telonic accepted my request without argument and even sent a courier to collect it and I have no qualms about doing any future business with them the next time I'll be looking for more T&M kit.

 The delay involved in reaching this point seems to lie with Siglenteu, leaving Telonic as much a victim as I was. It seems this particular defect is extremely rare (or more likely a sneaky type of fault that isn't so readily revealed using a cursory waveform check with a 'scope), hence the apparent misunderstanding of the problem by Siglenteu's "technicians" who obviously lack the required experience to understand the big "Clue by Four" video clip evidence passed onto them by Telonic.

 I did quite a bit of searching to find similar issues strongly hinting at this oscillator instability but none I could point at and say, "Here's another example of the jittery oscillator issue!". For instance, one example involved a Rigol (what else  ) MSO which quite clearly to me was a timebase jitter issue that could only be another example of the same unstable clock reference I was seeing. The fact that it was effecting the timebase in an item of test equipment made it all the more difficult to demonstrate what was actually going on (in this case, the MSO would have to opened up and the circuitry probed using another properly functioning 'scope to pin down the culprit).

 With my request for either a full refund or replacement with an SDG2042X with my paying the price difference going ignored whilst Telonic were so preoccupied with their dealings with Siglenteu that they forgot to keep me informed, I decided to have done with the whole business and insist on a straightforward refund and go to my previous supplier for a swift delivery of the SDG2042X I'd hoped would suffice as a sufficient upgrade as to finally allow me to retire my much abused FY6600-60M.

 You might be surprised to learn that I still might have to wait a while before I can actually retire that FY6600 since there are still some serious issues to be addressed by a thoroughly revamped firmware update. It's far more than just a user hostile UI that has to be re-written but also code in the FPGA to correct inexcusable shortcomings in the generation of Arbitrary waveforms, most notably the Sinc pulse which only looks good at the (apparently magic) 10MHz frequency setting (but which my SDS2504X Plus's CCJ measurement stats still show a +/-171ns jitter regardless) turning into a scrambled jittery mess just 1Hz either side of this magic 10MHz spot frequency (but curiously, no change to the CCJ stats).

 Meanwhile, the FY6600-60M's Sinc pulse carries on demonstrating a -0.15ns and +0.15ns CCJ stat regardless of frequency setting around the 10MHz mark (I haven't bothered checking out its 15 and 20MHz stats - just the +/-4Hz neighbourhood of the magic 10MHz that seems to disguise this CCJ issue in the SDG2042X).

 Now, in view of the fact that the hardware in a 1.2GSa/s 16 bit AWG must be at least an order of magnitude better in both build quality and specification over the hardware used in a cheap plastic cased 250MSa/s 14bit AWG, the problems can surely only be a matter of piss poor firmware. I mean, how else can you cripple an order of magnitude better hardware into something that falls short in so many aspects compared to a cheap "Toy AWG" like Feeltech's FY6600-60M?

 It's only the fact that a firmware upgrade is a doable fix that stops me from immediately declaring it as "Unfit for purpose" and claiming a full refund to either give up on the whole idea of upgrading to an improved replacement for my 'Toy AWG' or spend yet more wads of cash on an "A Brand" alternative which, the way things have been going, might yet prove to have its own surprising shortcomings.

 So here I am, contemplating whether it's worth hanging onto the SDG2042X or simply giving up and send it back for a full refund. The problem is there aren't really any viable alternatives in the same sort of price bracket (Rigol are most definitely not an alternative imo).

 I'm not fishing for buying advice here. I know researching the market is likely to be a long drawn out ball ache of a process to find a best value for money solution that's likely to still involve a compromise or three (one of which is almost certainly going to involve the state of my bank balance).

 Just for added interest, even though mention of the SDG2042X is slightly off topic in this thread, I've attached some screen shots to give you a little more detail on the CCJ stats and the peculiar jitter of the 2042's Sinc pulse waveform compared to that of the Feeltech's at 10 and 10.000001MHz

 CH1 is sinewave from the 2042
 CH2 is the RFS
 CH3 is the Sinc pulse from the 2042
 CH4 is the Sinc pulse from the FY6600 from which I'm triggering the 'scope

 I've disabled CH1 CH2 in the last four images to clarify the Sinc pulse waveforms. The first two of these images show the waveforms at exactly 10MHz (both generators locked to my GPSDO) and the last two are at 10.000001MHz to demonstrate the jitter effect which reveals itself only on the 2042's pulse - the FY6600's never so much as bats an eyelid in response to this 1Hz offset and this remains true for the 2, 3 and 4Hz offsets I tried.

John

Can you explain what is your named Siglenteu. Do you mean exactly  https://www.siglenteu.com   or some other.

As everyone can see in every one your last "test" image trigger come from CH4 what is FY6600 signal.
I think it was not "accidentally" without any meaning. But yes, Im not so interested about this fooling kids game.
But you can not fool all of us.

 

[TurboTom]

@Johnny B Good -

when doing tests with arbitraries, it's always a good idea to look at the waveform at low frequencies first. I've got a 250MSa/s AWG (among others...) that reproduces a Sinc (Sin(x)*x^-1) waveform with a period from -32Pi to +32Pi, i.e. consisting of a total of 32 "wiggles". It becomes completely obvious that such a signal cannot be reproduced at a period frequency of 10MHz, the underlying sine wave would be at 320MHz, almos three times Nyquist! I found this generator to reproduce the Sinc signal nicely up to approx. 1MHz with slight unsymmetry up to 2MHz and identifyable as "wannabe" Sinc maybe up to 2.5...3MHz. Above that, it loses "wiggles" due to aliasing and bandwidth limitation in the reconstruction filter.

Please have a look at the Sinc signal of your FY6600 AWG at low frequency -- I'm sure it looks pretty different from what you get at 10MHz.

Moreover, as @rf-loop pointed out, triggering on one generator output while observing another generator's signal and to judge from the "smeared" signal of the latter that this generator is at fault, is simply incorrect. Each of the two generators could be the "culprit" unless it's clearly an amplitude thing. And as explained above, in this case even your test signal was not adequate or even valid to do this kind of measurement. Know your gear, know its limitations.

[Johnny B Good]

@Johnny B Good -

when doing tests with arbitraries, it's always a good idea to look at the waveform at low frequencies first. I've got a 250MSa/s AWG (among others...) that reproduces a Sinc (Sin(x)*x^-1) waveform with a period from -32Pi to +32Pi, i.e. consisting of a total of 32 "wiggles". It becomes completely obvious that such a signal cannot be reproduced at a period frequency of 10MHz, the underlying sine wave would be at 320MHz, almos three times Nyquist! I found this generator to reproduce the Sinc signal nicely up to approx. 1MHz with slight unsymmetry up to 2MHz and identifyable as "wannabe" Sinc maybe up to 2.5...3MHz. Above that, it loses "wiggles" due to aliasing and bandwidth limitation in the reconstruction filter.

Please have a look at the Sinc signal of your FY6600 AWG at low frequency -- I'm sure it looks pretty different from what you get at 10MHz.

Moreover, as @rf-loop pointed out, triggering on one generator output while observing another generator's signal and to judge from the "smeared" signal of the latter that this generator is at fault, is simply incorrect. Each of the two generators could be the "culprit" unless it's clearly an amplitude thing. And as explained above, in this case even your test signal was not adequate or even valid to do this kind of measurement. Know your gear, know its limitations.

@TurboTom and @rf-loop,

 Many thanks to the both of you...  

 To answer @rf-loop's siglenteu question, yes, I was referring to siglenteu.com, Siglent's official European distributor, to distinguish it from the cybersquatting siglent.eu.com which is just another dealer who happens to sell Siglent kit.

TBH, I was so hung up over the fact that the FY6600's Sinc pulse with its lack of the infamous 4ns DAC clock jitter that afflicts all the other pulse waveforms that have at least one rapid transition, makes it extremely useful for triggering or use as a datum point for observing slow frequency drifts against things like DIY GPSDO reference clock outputs that I was rather put out to see the same couldn't be said of the Siglent's Sinc pulse (at least as far as fine tuning it to match whatever I was tracking).

 At 1MHz the '66 shows about the same number of squiggles as it did at 10MHz but you see twice as many (22 in total) between the peaks on the '42's Sinc pulse. I'm no mathematics wizard but even I can recognise a function that involves an infinite series which can only be approximated anyway. I guess how well an AWG can approximate a Sinc pulse depends on how much hardware resources you have available and where you want to make compromises.

 JOOI, I decided to run a sequence of CCJ stats, starting at the 20MHz upper limit of the 2042 (the 6600 has no such limit, leaving the users to decide for themselves just how far they can usefully increase the frequency). I worked my way down in 1MHz steps after skipping the 18MHz step. You'll notice some of the steps had several screenshots devoted to them on account of the CCJ errors on CH3 (the 42's pulse) due to the severe amplitude jitter (I suspect dedicating the full vertical resolution in 10 bit mode would probably address this but I'm not planning on spending any more time on this if I can possibly avoid it).

 Bearing in mind what had been said about the choice of trigger source, I started with the '42 as the sine wave trigger signal on CH1 for the first image, using the '66 for the next two before going back to the '42 for images 04 through to 18 inclusive, switching back to the '66 for images 19 to 24, finishing by going back to the '42 for the final image.

 I saw very little difference between the two triggering sources which shouldn't be too surprising considering they were both locked to the GPSDO reference (possible in the case of the 6600 simply as a result of its OCXO upgrade with add on injection locking circuit). The 2042 was used for the vast majority of these CCJ test runs as described and the 6600's sine wave output used to demonstrate what little difference the choice of triggering source actually made.

 The first image shows the 42's Sinc pulse in its best light by fortuitously changing the frequency to the 20MHz limit at just the right moment (it took several tries). The second demonstrates the same situation using the 6600 as the triggering signal. Still using the 6600 to trigger the 'scope, I switched between the 19 and 20MHz as before until I caught it "with its pants around its ankles" (the other extreme of its amplitude jitter range). And, no it isn't a "Single triggered snapshot", that's exactly how it appears in a repetitively refreshed live view. I then switched triggering back to the 2042 to get another shot of this collapsed Sinc pulse in image 04 before continuing onto image 18. I used the 6600 in the next 5 images before switching back to the 2042 in the final image.

 Images 19 to 23 demonstrate the CCJ errors at 11MHz on the 2042's Sinc pulse with the last two images showing the reason why. It took me an hour and a half just to capture those screenshots alone but I thought they'd help demonstrate my bemusement at the startling difference between what Feeltech/FeelElec and Siglent had been striving for in their own approximations of a Sinc pulse waveform.

 On the basis that the Sinc pulse is a tricky waveform to produce with any realism at the high end of the frequency range, It looks like I'll be hanging onto the 2042. After all, Sinc pulse waveforms aren't the be all and end all of the arbitrary wave zoo.

 Once again, my thanks to you both for your input. I can now set aside what I suspect would be a futile quest for a better AWG in a similarly affordable price bracket and be happy (once Siglent have revamped the UI into something more worthy of its place in their otherwise very fine AWG product line).

John

[Johnny B Good]

 Oops! It looks like EEVBlog's server has FSCKed again!

 Here's the final 5 images which the server had ignored that I'd attached to the previous post.

John