Siglent SDG1020 square signal "dirty"

[Mike Warren]

I received a SDG1020 yesterday. Same as others, channel 2 square wave is good, but channel 1 is so full of glitches that my scope won't trigger properly.

Software: 1.01.01.27
Hardware: 02-00-00-21-24
Serial: SDG0000113xxxx

[Tepe]

SDG 1020

Software Version: 1.01.01.27
Hardware Version: 0f-00-00-21-22
Serial Number: SDG00001120xxx

1 Hz square wave with 50% duty cycle on channel 1:

[rf-loop]

SDG 1020

Software Version: 1.01.01.27
Hardware Version: 0f-00-00-21-22
Serial Number: SDG00001120xxx

1 Hz square wave with 50% duty cycle on channel 1:

All we know that electronics is art and it also can be and need be fun.



Square risetime is <10ns (typically 6-8ns in this Siglent model)
In your image you perhaps have one sample every 2us. 2us is 2000ns.

(with 16k sampling buffer your Rigol  have 0.5MSa/s with 500us horizontal speed)
It can draw straight line between 2 sample points (just as your rising edge in zoomed window if there is good luck that real rising edge is just between these sample points)

You really can not see any glitches there due to this extremely low samplerate.

And for jitter. You are looking same rising edge where is your trigger. Yes you can test your oscilloscope trigger quality/jitter but not signal source timing jitter with this method.
This 0.5MSa/s sampling rate is too low for all this kind of purpose. But yes, it produce nice art draw on the screen.  For analyzing 1Hz square wave timing jitter: nearly impossible with your scope.
You need reliable trigger on rising edge and then go to look next riring edge! 1 second! How you do it if thewre need enough resolution for reliable jitter measurement. You have max 1M memory. With 1s period theoretical maximum is 1MSa/s. With 500ksa/s you have 2 second. Can you set trigger to rising adge and take next rising edge on the screen... and there you have 2us time resolution. Is it enough? Yes it is if think that specified jitter for Sqrwave is 0.1% of period (@ 1kHz typical)  But wait...  50ms/div and more slow your Rigol go to scroll mode!
So, game over for real jitter measurement with Rigol DS1000E for 1Hz square.


Btw.. you are looking 1Hz square and your input is AC coupled. (yes you can see this rising edge but if you turn to speed where you can see whole cycle... well you see only peaks up and down where is rising and falling edges. But yes with only this short time period it can do.


-----------------------

This your SDG1020 is old version.

I have not tested this version SDG1020 but I have tested this same version SDG1010 and 1025. They did not have this special CH1 square wave  problem where it out "burst" of pulses before real edges using low frequencies + high level of glitches near rising and falling starting  corners. (when these glitches duration is enough long and high they make siggle or several random burst of real flat top pulses.


Strange here is that these new glitches  with high level can see in different time position than this other "glitch" issue what is talked earlier an where PECL comparator hysteris setting was wrong (too narrow).  Typically this was common to both channels and around same. Also these special glitches looks reduced in these new versions. 


It looks more and more like it is in SDG1020 and if it is PCB version 02. (until find opposite information) and  they are made 2013.

I have tested several SDG1010 with this new PCB 02- version and no any sign about this issue.
Also new SDG1010 where is PCB version 02-: it looks that square corner glitches level is more low than units what I have tested before. (dependent of FW there may read HW 0f- or 00-)
PCB version is first two digits in HW version string.

-----------

But for low frequencies. WHY use square wave function at all? There is pulse function up to 5MHz.
After 5MHz up to 20 or 25MHz there can use Square function typically without problems.
Perhaps Siglent can shut off this Square function example under some some frequency... example 100kHz or something like it.   
Btw, with higher frequencies timing jitter is less than pulse mode what have always 8ns due to 125MHz clock and  simple machine principle.

Example. How much is 1MHz square wave timing jitter? (even with these units what have problem)
1ns or something like this from ideal edge time?

With pulse function 8ns

1kHz square wave (in good condition normal unit)
0.1% from cycle time   1us  = 1000ns  (and becouse not explained, it can not say if it is from ideal time point or peak-peak. (my measurements looks like from ideal time point)

1kHz pulse 8ns p-p (+ some small other kind of jitter if look more deep in practicde and theory)


----------------------------------------------

Open recommendation to Siglent:

Simple free solution what may solve many problems:

Please: In next FW update, shut off Square wave function example  <100kHz freq.
And then write new specs sheet.

Square wave: 100kHz - 5, 10, 20, 25MHz depending model.

There is no any need for special square wave function under 100kHz.
For under 100kHz there can use Pulse function.

[Wuerstchenhund]

It looks more and more like it is in SDG1020 and if it is PCB version 02. (until find opposite information) and  they are made 2013.

This does not only affect generators made in 2013. My SDG1020 has been bought in December 2012 and suffers from the same problem.

Simple free solution what may solve many problems:

Please: In next FW update, shut off Square wave function example  <100kHz freq.
And then write new specs sheet.

Great idea, artificially limit the functionality and fiddle the specs to cover what is a major design omission which should have never passed any quality control (assuming there is one in Siglent, that is, which I doubt).

The appropriate way to react to this would to provide affected owners with a free board replacement with a fixed new board revision, offer a partial refund, or allow them to trade-in the affected generators for a different model at a hugely discounted price. In fact, this could be a good opportunity for a Chinese manufacturer to shine and show that customer satisfaction is actually important to them, and that they are more than one of many slave shops putting out cheap crap (I'm not holding my breath, though).

Your suggestion goes exactly in the opposite direction, a very good way to burn a brand name for good in this industry.

For under 100kHz there can use Pulse function.

Well, no, as at least on my unit square wave is still shit at 100kHz.

[rf-loop]

Your suggestion goes exactly in the opposite direction,

It (recommendation) was my try for some kind of sarcasm... but it is difficult.

Your unit HW  version is 02-  ?
Made 2012.
Before I have seen only 2013 made HW 02- 


And your problem is exactly this CH2 new more bad problem or is it  this old glitch issue what typically was in both channels.?

And simply, with this principle can not easy produce clean variable duty square, or circuit need be more complex. 
(my opinion is that these are two different internal problem in units)

I have not yet analysed it in circuit. I'm waiting some comment first from Siglent engineers.
 

[rf-loop]

..... on my unit square wave is still shit at 100kHz.

Here is pictures from some test with Siglent SDG (unit what works normally).

1.6V pp  Square, 50% duty, 100kHz
first rising corner. Near corner before rising, acceptable level of "glitches".
Oscilloscope trigger from this same rising edge what is in display.
Cycle time is 10us.
There can not see any this kind of signs about severe glitches what can name as "issue"

Next same setting in SDG
Oscilloscope trigger is rising edge one cycle before.
(there is also cumulated oscilloscope trigger jitter (what is not markable here))

It can see that cycle-cycle jitter time is well in specs and not any kind of "issue".

In both images infinite prsistence and watching time 15 minutes.
(btw, for rms jitter calculation there is normal used 10000 events and many manufacturers give rms jitter... of course rms jitter may be "nice number".)

Of course this kind of square is shit  but then if look all functions, signal quality in this class of equipments and building quality and then take also price. What is better with all features in this price class if do not look old second hand professional equipments but look new units.

But this is BAD they have shipped some units what have really problem. This is bad. Really bad.
So, it is better to buy from sellers who really know what they sell and not only nice boxes.
Factory need better QC and not only that units looks like work but also they need watch this kind of problems. But then... also as long as chinese manufacturers are as they are... distributors need do control. Example I do not sell any individual unit what I have not chcked. (still there may be some rare problems sometimes)

[Tepe]

1 kHz square wave, 4 Vpp from channel 1:


The same, but with DC coupling:


An attempt at duplicating Xyphro's first screenshot with a non-50% duty cycle and 877.2 Hz, 1 Vpp:


And finally 50 Hz, 600 mVpp like what rf-loop posted earlier:

[commongrounder]

Yeah.  I got curious and had to go back and check my SDG1005 square wave at 100khz, 50%, too.  I set up like rf-loop for the "glitches" at the trigger point and the jitter at the one cycle down from the trigger.  I set the trigger threshold above the jitter peaks to get as stable a trigger as possible on just the "real" square wave rising edge.  What do people thing about the periodic nature of the glitching?  There seems to be about 2.3ns between the peaks, 435mhz.  Does that relate harmonically to any clock frequency in the generator?  I am noticing that the height of the glitches varies, depending on how warmed up it is, warmer=less.  They do not ever go away entirely, however.

[commongrounder]

Another quick note:  Changing the frequency does not seem to affect the spacing of the glitches (except for how far away they are from the base of the rising edge) until crossing over certain frequencies. Crossing below 25khz changes the spacing to around 1.8ns.  The wave shape is the same, but is much more cluttered.  Given time, the persistent buildup on the scope screen is filled in.  The amplitude of these is incredible!:

 

[MasterOfNone]

I’ve also got an SDG105 in my home lab, but it doesn’t have the glitch problem on any of the channels. It is an older model so it looks like rf-loop is right about the problem being introduced recently:
SW 1.01.01.27
HW 0f-00-00-21-22
SN SDG00004110xxx
Since the glitches are cause by the comparator hysteresis, does it happen on the falling edge as well?

[rf-loop]

Since the glitches are cause by the comparator hysteresis, does it happen on the falling edge as well?

Yes, this glitch problem what is related to PECL hysteresis (and PECL comparator itself other things)  is around same in falling edge but of course now upside down.

[commongrounder]

It is an exact inverse of the positive going.  The same action is occurring on both transitions.
"Newer is Better" is becoming one of the most hollow sounding catch phrases around, these days.

[Tepe]

For comparison with rf-loop's test. 100 kHz, 50% duty cycle, 1.6 Vpp, channel 1, one cycle to the right of the trigger:


edit: Just for fun, a comparison with a Thurlby-Thandar TG210 at ~100 kHz. There is less jitter but the rise time is longer (but well within the <100 ns specified).

[commongrounder]

Tepe's Channel 1 output looks like my Channel 2 output.  Better.  Channel 2 has less drive capability, but it can certainly be used more reliably for square waves if one understands the limitations.

[Mike Warren]

I have been communicating with Jie Wang at Siglent and he advised the best cure for this problem is to fit a 47pF capacitor in parallel with R194. I've just done that and I'm now happy with the performance.

Jie also said that changing the value of R87 to 20K could be done instead of adding the capacitor, but the capacitor was the preferred fix.

Don't do both mods, just one or the other.

I found Siglent to be very responsive. I received replies to my emails within 24 hours.

[bmwnomad]

Thanks Mike!

I can confirm that putting the 47pf cap in parallel with R194 fixes the square wave on channel 1.  I haven't done any extensive tests, just 1khz, 4vpp, 50% duty, but so far the glitches are gone.

SDG1025
SW 1.01.01.27
HW 02-00-00-21-24

Steve

[Mike Warren]

R194

Oops! Fixed.

[pickle9000]

Siglent just earned a bonus point, maybe even two. First they talk to a customer then suggest a fix. I sure wish I could see more of this kind of stuff.

[Fennec]

The same crap with mine, but on both channels. No difference between Ch1 & Ch2      

4Vpp,  50%, 1kHz, no load (my T-adapter saw this "signal" and must be went away to the dark side of Moon. Can't find it.)


300mV !! 

SDG1020
SW. 1.01.01.27
HW 0f-00-00-21-23
SDG00003120xxx

[Mike Warren]

The same crap with mine, but on both channels. No difference between Ch1 & Ch2      

4Vpp,  50%, 1kHz, no load (my T-adapter saw this "signal" and must be went away to the dark side of Moon. Can't find it.)


300mV !! 

SDG1020
SW. 1.01.01.27
HW 0f-00-00-21-23
SDG00003120xxx

That doesn't look like the glitch problem being discussed here. It just looks like a lot of noise.  I notice you have older hardware, which based on posts here doesn't seem to suffer from the glitch problem.

These signal generators aren't particularly quiet, so what you are seeing might be perfectly normal; at least 22 dB down.

[commongrounder]

Thank you Mike Warren, for working with Siglent on this.  Despite my reluctance to open up a brand new unit, I am going to proceed with the capacitor mod, since it is a recommended solution directly from Siglent.  I expect I will see the same improvement on my SDG1005.  I have a little more space to work in because there is no fan to get in the way on the 5mhz version.

[bmwnomad]

Yeah, I had to remove the fan to give me clearance to get in there with the tweezers and soldering iron. 

Steve

[rf-loop]

The same crap with mine, but on both channels. No difference between Ch1 & Ch2      

4Vpp,  50%, 1kHz, no load (my T-adapter saw this "signal" and must be went away to the dark side of Moon. Can't find it.)


300mV !! 

SDG1020
SW. 1.01.01.27
HW 0f-00-00-21-23
SDG00003120xxx

It have also handled here.
http://siglent.freeforums.org/old-sdg1000-possible-improvement-for-square-wave-quality-t27.html

and more here
https://www.eevblog.com/forum/testgear/siglent-sdg1020-arbitrary-fungen-first-impressions-and-a-possible-signal-glitch/

Your square wave corner noise is <10% from signal p-p

If you want more clean square for under 5MHz
Just do not push "Square" but push "Pulse"  what you can read as "Pulse and sguare wave functions"

With "pulse" you can of course do square wave but there is also more features and it have always around 8ns p-p timing jitter what is only related to 125MHz sampling speed.


Why use Square function if need "square wave" between 500uHz to 5MHz and specially below 100-200kHz.


Here is one reason why NOT use Square function:
 
Square: jitter in specs: 0.1% cycle time.  (@1kHz typical)

1Hz cycle is 1000000000ns
Square function:  square wave cycle time jitter may be  something like 1000000ns
Pulse function:  square wave cycle time jitter may be 8ns peak to peak.
(and corner clitches...  try find)

Just use Pulse function and use Square function only if really need it in special cases.
(with higher frequencies it have less time jitter than 8ns p-p)

http://siglent.freeforums.org/sdg1000-square-wave-t43.html

Also today I get recommendation directly from Siglent for this special CH1 problem where it have extremely high level glitches and even full level random pulses and high time jitter.   1st solution for this is add 47pF capacitor as Mike Warren have told and also he have made nice pictures for this. Nice work!  Tnx!

Notes for work.  All know it but still: Remember ESD protection!
SDG 1000 series service manual is also good for read.. It is available directly from Siglent sides (whoever can downlod and no need any registration)

(Old units with not this problem but where both channels have some amount too high level gliches before rising and falling edges, there is old Siglent recommended change for comparator hysteris setting resistor value.)

[Mike Warren]

Yeah, I had to remove the fan to give me clearance to get in there with the tweezers and soldering iron. 

You just need to hold your tongue at the right angle.
 

[rf-loop]

SDG1020

CH2 parasitic oscillation problem in some situations.

But then CH2 have other issue in unit what I have tested..

Square, ramp, sine etc. All fails but looks like only if use 50ohm and highes voltage band. Specially it can most easy look with square or pulse. After signal rise it looks like go flat but just after corner it sart "oscillate".

Oh well, if you look it with slow oscilloscope it looks just ok or like only "noise level in signal top is more than bottom. Take fast oscilloscope and ... oops.
Signal top (not bottom) have high level some kind of oscillation with high frequencies. Something like over 200MHz!

This is first time I have tested SDG1020. (becouse Teledyne do not have this version), they have 1010, 1025 and 1050 only.  (2012, 2022, 2052)

What is this problem, I do not know anything more now.
(I do not sell this 1020 model at all before whole this issue is solved)

Please, who have this issue in SDG1020 model:
HW version, FW version and whole serial number exept last 3 digits  as xxx.
(I think posibility that this issue is in some manufacturing lot.)

This oscillation problem is solved by Siglent:

(due to Holiday in China, there is now open question how they support these who have this problem. If you have this problem, please contact your Siglent distributor / seller for get after sales customer care.)

CH2 have two end amplifiers.  One (a) for low voltage levels and one (b) for high levels.
(with high levels both of these are in use. (a is preamplifier for b))


Problem is in some units CH2 high level end emplifier (b).

This  amplifier (b) is Texas Instruments OPA695 (it is used in some units).
It need change to Intersil EL5166ISZ
(also Siglent have used CLC1606 in some units and this also looks like ok
EDIT/ADD: in problem case also this need change to Intersil EL5166ISZ  )

I have tested this repair and it solve this problem as 1-0

OPA695 is problem only in this CH2 high level end amplifier. (if there is OPA695 in other circuits it is not problem)

There do not need any kind of other component changes if change this amplifier.

How to test if there is this problem.

CH2 sine or square.
Output 50 ohm
Frequency depends your oscilloscope. Try example with 10kHz
Level 2Vp-p   (shift 0V)
Connect  >1m 50ohm coaxial cable to oscilloscope input using 50ohm "simple" termination. 
(this test NEED that there is oscilloscope input capasitive reactance together with 50ohm termination, If 50 ohm termoination is wide freq band enough pure 50ohm impedance  this problem is perhaps not visible)

Now look this signal top. (use peak mode)

If top looks very "noisy" (trace is "fat") then zoom to this fat area. You may see there high frequency oscillation (if square wave, it may start littlebit after signal rising edge)
If you change level between 1.00V and 1.01V you can hear also inside SDG internal relay voice becouse here it change to upper voltage band. Lower Voltage range  is clean but upper Voltage range is problem.  In my tests maximum oscillating level is somewhere half way this highest voltage band where this amplifier is added to signal pathway.

If you want see this oscillation signal you need 1GSa/s or more wiith this horizontal speed what you are using and oscilloscope analog front end need be enough good for far over 200MHz, in other case you may see only "some kind of noise".  in real world this is nice sinewave.