TESTS: Siglent SDG5082 Function/Arbitrary waveform generator. (+Inside images.)

[rf-loop]

There is now new FW  (not yet available in Siglent public Download center)

More tests is coming soon after I  first look this new FW and if it change something in signal. This is signal generator so signal is most important thing in my scale of priority.

[grego]

Hey loop - do you think I could get ahold of that new firmware since it's not yet on the site?

[rf-loop]

Hey loop - do you think I could get ahold of that new firmware since it's not yet on the site?

I do not recommend. 5.01.01.08 have problem.
I have not yest tested if can go older version back. (perhaps not)

[grego]

Good to know.  Thanks.

I'm goign to be doing a use review on Youtube hopefully later today, at worst on Friday, of the 5082.

[rf-loop]

Good to know.  Thanks.

I'm goign to be doing a use review on Youtube hopefully later today, at worst on Friday, of the 5082.

Now I know that it can downgrade back to 7R1. (so it is good to keep also available if need)

I will look later more about this version 8 propable problem, after I have time for look it more.

[grego]

So I wanted to just throw a couple of screenshots up - I got my Agilent 6012 this afternoon and immediately hooked it up to the Siglent SDG5082 with a 50ohm feedthrough term (Siglent set to 50ohm out).  Rise time on the square wave is fantastic at under 7ns.  Also very little overshoot, coming in at around 0.8% on average.  Frequency is also pretty reasonable even at 80Mhz coming in with a standard deviation of around 200kHz (or about 0.25% of 80Mhz target).

More coming later.  If you're curious, yes the 6012 has a valid calibration at this time.

[rf-loop]

Frequency is also pretty reasonable even at 80Mhz coming in with a standard deviation of around 200kHz (or about 0.25% of 80Mhz target).

Do you really mean that 80MHz sinewave from SDG have 200kHz sdev. 
You have FM modulation on?

Which one now pull, horse or wagon.

So you have classified your oscilloscope and get information that oscilloscope measurement  have total lack of accuracy for this measurement (in this case 80MHz sinewave frequency stability)

Simply, with SDG you can test how much error this oscilloscope produce in this measurement.
200kHz sdev is so much that if you get this result with equipment what data really can trust in this measurement with enough accuracy you can immediately send your Siglent SDG back and tell it is failed, totally out of order.  it was sdev, and also, if it is peak to peak error it is so much that if signal generator p-p error for 80MHz sinewave is 200kHz it is just total garbage. Very bad made 80MHz simple LC oscillator is far better.

Instruments what give lot of data are nice but also it is good to  understand when they are not valid, decimal point is totally wrong place for this. (exept if want know this oscilloscope mesurement quality in this case)

[grego]

No, modulation was not on - devices had been warmed up for about 30 minutes and were humming along just fine.  I let the sine run for a bit, reset the stats on the scope and then let it acquire another set of points before taking the screenshot.

Please note - I did say my scope was a 6012 - e.g. a 100Mhz scope - so I was well past what I should be from a bandwidth perspective.  To truly test I'd use at least the 300Mhz version, which I don't have.  The measurements are completely valid - as long as you take into account the comments made.  I was not hiding anything and was quite upfront about the equipment I am using.

I'd hold off on bashing an Agilent MSO6000 with a valid calibration.  You do, however, have to take bandwidth into consideration.

rf-loop, you do great work here - I've really enjoyed your deep dive into the 5082 (hell, its one of the reasons I decided to risk buying one here in the states since I can't easily get it repaired, warranty or otherwise), but you get really antagonistic sometimes.  Relax.

[grego]

Separately I'll be posting a brief overview video on Youtube on this thing shortly.  Nothing special, just run through the basics.  Video work could be better but meh, it's an overview.

[rf-loop]

rf-loop, you do great work here - I've really enjoyed your deep dive into the 5082 (hell, its one of the reasons I decided to risk buying one here in the states since I can't easily get it repaired, warranty or otherwise), but you get really antagonistic sometimes.  Relax.

In this test image with scope there read
Freq current 80.3MHz
Freq mean 80.001MHz
Freq min 79.4MHz.
Freq max 80.6MHz
Std Dev 197.59kHz   

Then you tell:

  Frequency is also pretty reasonable even at 80Mhz coming in with a standard deviation of around 200kHz (or about 0.25% of 80Mhz target).

If signal have this max and min and this sdev (in freq) it is so total crap signal generator or better say it have some kind of failure. (but of course it is not)

I can not be without comment it or better say shoot it down. Becouse this is not my natural language I do not know how to write so that my words have still nice flowers. This is NOT at all any personal or any kind of angry or something like this.

But I need tell that this data do not tell about SDG5082 signal quality and there is not 200kHz sdev and not 1.2MHz max - min  freq error. It only tell that this oscilloscope measurement accuracy is not at all enough for test this sinewave frequency stability. It need  least 5 decade more resolution/accuracy.

Btw, one quik test what tell also easy something  about signal. If have good quality HF receiver. Take out some HF carrier, example 28.xx MHz.  Then, turn SSB on. Tune example 1kHz off. There need hear quite clean 1kHz audio. It can give some kind of mutual imagine about signal quality (or radio quality if it is not enough high)
 
Also if there is available special good quality sinewave and then ordinary mixer and then this device under test signal.

Take out exmple 80MHz from SDG and 80 001 000.00 Hz from this very good source. Now there can look this mixer output 1000Hz. and look how is its stability. Or it can mix more close... example to 100Hz.  Of course output have now both signals errors, so this external signal need be as good as possible.

I add note (for avoid more misunderstoodings): This answer and earlier answer is writed with  kind mind and it is very nice to see these test images. It is also useful becouse it clearly show what kind of limits there is or may be even with very good instruments - but if try measure things what they can not do.

Also it show that not only  SDG5082 what I have tested  have this risetime and  same shape of square wave. More data, more we know it is not only single individual unit.

[rf-loop]

Separately I'll be posting a brief overview video on Youtube on this thing shortly.  Nothing special, just run through the basics.  Video work could be better but meh, it's an overview.

Please do you have link for this? (I can not find it in youtube)

[grego]

Separately I'll be posting a brief overview video on Youtube on this thing shortly.  Nothing special, just run through the basics.  Video work could be better but meh, it's an overview.

Please do you have link for this? (I can not find it in youtube)

It'll be up shortly - it's just a use overview with a little bit of a dive into the square wave jitter, rise-time, etc.  Nothing major.  Waiting for youtube to do its thing then I'll post.  I even give you a shout out in it.

[grego]

BTW rf-loop - I need to run a couple more tests but I am beginning to think there's something wrong with the scope.  I've now checked at various frequencies on just a sine wave through my freq counter and the std dev is about (@35Mhz that I just tried) 0.162Hz.  On the Agilent it was 90kHz.  So something is up.  I was min/maxing on the Agilent (again @35Mhz) 34.7Mhz/35.1Mhz as I recall -- which doesn't jive with the freq counter.

I'm going to hook up another scope tomorrow and doublecheck but it looks like something is wonky in the scope.  All the other scope measurements are solid, but frequency seems to be off.

[rf-loop]

BTW rf-loop - I need to run a couple more tests but I am beginning to think there's something wrong with the scope.  I've now checked at various frequencies on just a sine wave through my freq counter and the std dev is about (@35Mhz that I just tried) 0.162Hz.  On the Agilent it was 90kHz.  So something is up.  I was min/maxing on the Agilent (again @35Mhz) 34.7Mhz/35.1Mhz as I recall -- which doesn't jive with the freq counter.

I'm going to hook up another scope tomorrow and doublecheck but it looks like something is wonky in the scope.  All the other scope measurements are solid, but frequency seems to be off.

 In measurement it need know what time it exactly cross levels beginning and ending of "cycle".  If oscilloscope is "ideal" machine without  noise and if it use direct sampled data, what is resolution. For easy thinking take example 50MHz. Its cycle time is 20ns.

In this cycle time, if there is 2GSa/s sampling speed (as in this case is) there is 40 samples for 50MHz. How accurate it can now tell how long is this cycle.  If there is 41 samples in some cycle. It is now 20.5ns = 48.78MHz
Of course now it may do something more. It may interpolate between true sampled points and estimate time when it cross same level what it use for start of mesured cycle.  It may give small amount better result. But of course this simple example is enough for tell that resolution/accuracy is far far  away from what need for characterize this sinewave fidelity in frequency/time.

I think your oscilloscope is working just ok. But it is wrong tool for just this. But good/very good for many other things.

How if look resolution for 50kHz instead of 50MHz. 50MHz we get 40 sample points 1/40 of cycle, but 50kHz 40000 sample points etc.

Your measurement with freq counter, result is ok and not conflict with my measurements, but now, there is also problem dependent of freq counter (time interval counter). It use still many many cycles (depends gate time) for one freq result so it is "average". Then agen, many many cycles for next freq result what is agen average. Then it do statistics with these individual results and give average, min, max, sdev etc...
But every sample in this calculation is average.

What is wrong here. Nothing or much of. It depends what we want exatly really know.
If we want know every single sequential cycle frequency and result what is min, max and sdev etc...  well we are troubles how to do it. Directly with time interval counter? I do not know time interval counter what can do it for example 50MHz or just this 80MHz.  I have very poor counter HP/Agilent 53131A (good but poor for just this). Its time interval resolution is 500ps. (so just nothing for this)  If it have 50ps... still nothing... 5ps... nothing, how about 500fs or 50fs. And even if enough resolution, can it timestamp every single sequential cycles, or if can not then example every tenth or hudreds cycle.. or random but enough fast... then collect 10000 these and calculate.  I like to have this machine.

But then, if we have enough good quality signal available... we can use mixer for convert down this freq what we are testing. If there is available "ideal" signal exactly 10 000 0100Hz and then signal under test is "10MHz. After mix there is  clean 100Hz signal (if 10MHz under test is perfect). It have now these errors what this 10MHz signal under test have. (if 10.0001Hz signal is "ideal") If 10MHz signal under test jumps 1Hz our 100Hz signal jumps 1Hz. Not difficult to detect.  Now there is 100Hz stream where we need detect all time errors.

[grego]

Video overview is up:

[rf-loop]

I have made some more tests with spectrum analyzer.

Here in these pictures signal from SDG5082 is of course sinewave. Level is in all pictures 0dBm and naturally 50ohm. Signal go around 2.5m in coaxial to HP8568B spectrum input. (some amount crap cable but no matter in this test)
Note about spectrum analyzer. Freq related some calibrations are not perfect but in this case this do not mean "anything" just "cosmetics" in numbers.

I have used 10Hz RBW and 100Hz RBW.
In some image signal is 20dB over top reference line! (note this as you think trace levels - as told signal is always 0dBm level)

Note for this displayed traces. In all pictures mode have been MAX hold. So this level is more like peak noise level than avarage noise level!

Many times in specs is use average noise levels or displayed average noise levels or something like it. With human eye it may somehow imagine where is average.

Then, most specs today give  dBc/Hz
Oh well, it is nice but it is not what you see.

How to convert displayed noise level for dBc/Hz.
Ok, example. If carrier (this c) is level 0dBm  and then  example 1kHz we can see trace level is
-50dBm. We can say its level is -50dBc (dB related to carrier level) Then RBW (resolution bandwidth) is 100Hz.  It is 100 times more wide than 1Hz RBW. Ok, it is 20dB.  So this -50dBc with 100Hz filter responds -70dBc/Hz  (70dB below carrier if RBW is 1Hz)
Why I use 100Hz RBW is just becouse my timetable was busy and narrrow RBW sweep takes forever.


Image A


80MHz  carrier is just left and there is 1kHz span  and  10Hz RBW
Reference level is same as signal (0dBm)
note  for dBc/Hz -10dB due to 10Hz RBW
(if you think reference level is -10dB this give right level for dBc/Hz)



Image B


10MHz  carrier is just left and there is 1kHz span  and  10Hz RBW
Reference level is same as signal (0dBm)
note  for dBc/Hz -10dB due to 10Hz RBW
(if you think reference level is -10dB this give right level for dBc/Hz)



Image C


1MHz  carrier is just left and there is 20kHz span  and  100Hz RBW
Reference level is -20dB under carrier level.
note  for dBc/Hz -20dB due to 100Hz RBW
(if you think reference level is -40dB this give right level for dBc/Hz)



Image D


50MHz  carrier is just left and there is 200kHz span  and  100Hz RBW
Reference level is -20dB under carrier level.
note  for dBc/Hz -20dB due to 100Hz RBW
(if you think reference level is -40dB this give right level for dBc/Hz)

[rf-loop]

These previous tests with spectrum analyzer are not real phase noise measurements - of course.
So, need be careful when doing  conclusions.

Siglent specifications (service manual, undefined version but 2013 - and also in other documents):

Phase noise 100kHz Offset, –116dBc / Hz (typical value)

Image D show 50MHz carrier and 200kHz USB.
100kHz from carrier, trace is between -100 and -110dBc/Hz (displayed between -80 - 90dBc with 100Hz RBW). But now, it need tightly remember what I told before = HP spectrum MAX HOLD  was used so from this trace can not estimate signal average noise level.


Image C shof 1MHz carrier and 20kHz USB.
10kHz from carrier, trace is between -110 and -120dBc/Hz (displayed between -90 and -100dBc/100Hz)

Images A and B show 1kHz USB from 80MHz and 10MHz carrier.
If look both traces curve and compare these example between 100 to 200Hz from carrier.
Image A (80MHz) display trace is nearly between -40 and -50dBc
Image B (10MHz) display trace is nearly between -57 and -67dBc
(10Hz RBW so subtract 10dB if want think dBc/Hz)

With this measurement, I think there is not markable conflict with specifications.
But I hope Siglent (and others) give more detailed specifications.
What it tell to customer if write "Phase noise 100kHz Offset, –116dBc / Hz (typical value)"
what is typical value meaning. How much variations typically. What is "typical" frequency. 1MHz? 33MHz? 157MHz? or what?
What are quaranteed values for different frequencies.

Personally I do not like at all  these "typical" values what means that manufacturer is not responsible if this given value is right or wrong.

[rf-loop]

Here some amount better phasenoise measurements.
EDIT: later I do tests agen. There is some problem in my measurements.
It looks like results are better and partially far better than now measured.
Before I have new data I keep this picture as "worst case" until other data owerwrite it.


(other thing is area below -110dBc between 10kHz - 100kHz)


Used equipment HP-8568B
Becouse some small freq inaccuracies, for every scan I have first looked what HP cursor show about frequency using 30Hz RBW and then used it in well knownKE5FX software.
In all measurements SDG output so that HP see it as -10dBm
Last two traces are just for reference, measured without any signal.

All phasenoises are
< -100dBc/Hz offset 20kHz
< -120dBc/Hz offset 200kHz

In Siglent specifications only given value is:
100kHz Offset, –116dBc / Hz (typical value) 
(what means typical?)
in my measurement between -112 and -115dBc/Hz  offset 100kHz

It need note that this old HP8568B calibration certificate is expired long ago.


80MHz
40MHz
20MHz
10MHz
5MHz
1MHz
1MHz no signal
80MHz no signal


test work error, image removed

This measurement have undefined error! Phase noise  values may be much better.

Also note: In this measurement, with these settings,  HP limit can think if you draw around -113dBc level horizontal line from 1kHz to around 90kHz. Then slope down and somewhere around -130dBc or some amount better.  100Hz have limit around -100dBc.  Traces without signal show only base noise level with this setting! They are not bottom base levels for phasenoise. (and together with correction factor in PNB software.) 

[jpb]

Here some amount better phasenoise measurements.
Used equipment HP-8568B
Becouse some small freq inaccuracies, for every scan I have first looked what HP cursor show about frequency using 30Hz RBW and then used it in well knownKE5FX software.
In all measurements SDG output so that HP see it as -10dBm
Last two traces are just for reference, measured without any signal.

Thank you for those. The specs give -116dBc/Hz at 100kHz as a typical value which is in agreement with your measurements for the 10MHz curve (Rigol state their specs at 10MHz while Siglent don't give a specific frequency but I guess that 10MHz is reasonable).

I'm no expert, but the values below 10kHz offset seem quite high (comparing with Agilent and Tabor specs which again are "typical" but do give several values rather than just one point).

Presumably if a good external reference is used the phase noise is reduced? (Someone on these forums did some measurements on the Rigol DG4062 with and without an external reference).

[rf-loop]

Here some amount better phasenoise measurements.
Used equipment HP-8568B
Becouse some small freq inaccuracies, for every scan I have first looked what HP cursor show about frequency using 30Hz RBW and then used it in well knownKE5FX software.
In all measurements SDG output so that HP see it as -10dBm
Last two traces are just for reference, measured without any signal.

Thank you for those. The specs give -116dBc/Hz at 100kHz as a typical value which is in agreement with your measurements for the 10MHz curve (Rigol state their specs at 10MHz while Siglent don't give a specific frequency but I guess that 10MHz is reasonable).

I'm no expert, but the values below 10kHz offset seem quite high (comparing with Agilent and Tabor specs which again are "typical" but do give several values rather than just one point).

Presumably if a good external reference is used the phase noise is reduced? (Someone on these forums did some measurements on the Rigol DG4062 with and without an external reference).

It need note that this my measurement with HP8568B give perhaps too bad results.
I have compared same measurements using my HP8644B for signal, and example HP53131A Option 010 Owen 10MHz  output (what is typically very good) and results are so that I need suspect this HP8568B give too bad results specially between 10kHz to 200kHz under 100dBc levels.
(specially test with HP8644B give sign that something is wrong under 100-110dBc and specially offset > 10kHz.)

Add, now also looked something using HP8901B and 5386A and this whole phasenoise test can read as garbage. I do not know yet what is wrong but it is sure, it is not SDG phasenoise.
It is difficult to compare equipments data sheets becouse specs are so different and ambiguous.

Rigol DG4000  60MHz model around USD 900
Typical (0 dBm, 10 kHz deviation)
10 MHz: ?-115 dBc/Hz

Agilent 33250A (used somehow trusted condition around USD 3000
Phase noise (30 kHz band)
10 MHz <-65 dBc (typical)      (Siglent SDG5082 around  <62dBc)
80 MHz <-47 dBc (typical)      (Siglent SDG5082 around  <44dBc)

Agilent 33500B (max 30MHz)  Around USD 3300
Phase noise (SSB) (typ) (what frequency?)
1 kHz offset: -105 dBc/Hz
10 kHz offset: -115 dBc/Hz
100 kHz offset: -125 dBc/Hz

Tabor WS-8102    Saelig price USD 5400  eBay USD 4500
SSB Phase noise (10kHz offset)
1MHz <-115 dBc
10MHz <-100 dBc
100MHz <80 dBc

Siglent SDG1000 series (25MHz < USD 400)
Phase noise 10kHz Offset? –108 dBc/Hz (typical value) (What frequency?)

Siglent SDG5000 series (80MHz in eBay around USD 700)
Phase noise 100kHz Offset, –116dBc/Hz (typical value)(What frequency?)

[jpb]

Yes, it is difficult to compare data sheets as they (deliberately?) give their numbers under different conditions.

The Tabor I was looking at is the WW5061/2 as I can get that at a low price. The figures for that are for an 8 point sine wave at maximum clock which I calculate as 6.25MHz (50/8):
100Hz <-103dBc/Hz
1kHz <-110dBc/Hz
10kHz <-118dBc/Hz
100kHz <-124dBc/Hz
1MHz <-135dBc/Hz

The list price for the WW5061/2 is very high but it has been around a long time and is available at a discount. Unfortunately it has a maximum sample rate of only 50MS/s (limited by firmware - the hardware is capable of 125MS/s).

[rf-loop]

My last phase noise measurement with spectrum analyzer have bad error!
I do not yet know what is wrong but when I "listen" SDG using HP 3586A and HP 8901B it give sign that  phase noise from SDG can not be this level specially between measured 100Hz - 10kHz USB.  Error is not "cosmetic" -  can not say now more than "error is big"

[rf-loop]

This test annul the previous phase noise tests in this context what I have made.

SDG5082  phase noise measured using HP8568B spectrum analyzer and KE5FX phasenoise tool.
SDG signal (Ch1) 0dBm, 80, 40, 20, 10 and 1MHz,
Spectrum analyzer input attenuator 10dB
Clip level 20dB (in KE5FX phasenoise tool)

Dark blue line is made with exactly same settings and levels using Agilent 8644B signal generator.
Output level 0dBm, freq 10MHz. Line is hand draw as accurate as reasonable. This is only as bottom  limit reference line and valid only with this individual spectrum analyzer.
Note for 1MHz carrier. Range is 200Hz to 1MHz. Near 1MHz offset carrier 2nd harmonic visible and that is why red 1MHz curve jump up there due to quite wide RBW used near 1MHz offset.



Note: spectrum analyzer limit around -115dBc


In attachment image:
HP3114 function generator.
10MHz sine, level 0dBm
just for fun.

[rf-loop]

Some "finding".

If CH1 and CH2 is on and then shut off SDG. (start with last setting selected)

It turns on and settings are same but...

If there was selected dBm for channels level it change to display voltages peak to peak.
(same level still of course)  I like more if it keep also level display units what I have selected before.
Now how I return to display dBm if I do not want change level and I do not exactly remember last dBm value.

There need be feature to change displayed units without change settings. If I have selected dBm example 0dBm, how I can change it display volts.
I hope they add this in later FW.


If power up, of course it powers up with channels off,  but internally it still produce this waveform what was selected. Internally it continue generating still this waveform with previous settings.
If now measure output there is still some (but very low level) leakage to output.

If now turn channel on (of course now selected signal come out with selected level)
and  then shut  channel off it stops whole channel and now if measure this channel output, there is not at all this small level leakage.

If do some very sensitive works it need take care that really shut off not used channel and not only output relay.

If shut off when channel is really totally shutted off and then power up, it is shutted off totally.

If shut off when channel is on. Then power up. Only output  is shut off but still channel signal on internally.

In sensitive works it may also affect due to small leakage from channel to channel internally.

I hope they change this in FW. Always if channel output light is off this channel need be totally off and not only output off also after power up after channels have been on when power shutted off.




 

[happyday]

I have test the result that at 50ohm,0dbm and 1KHz square wave

with my SDS1102CM, and I found it's Vpp=896.0mV
I think maybe you check whether it is correct