Siglent SDG6000X series 200-500 MHz AWG's

[Performa01]

With regard to amplitude accuracy and flatness, I want to remind you of this old reply #162, showing the response of my (original) SDG6052X and comparing it to an older high performance signal generator.

https://www.eevblog.com/forum/testgear/siglent-sdg6000-series-awg_s/msg2621457/#msg2621457

At the sweet spot of -30 dBm, I got a frequency response of 100 kHz - 500 MHz  +0.06 / -0.04 dB.
That's not only excellent, I'd call it even spectacular.

Yes in this individual unit it can say, it is spectacular.

Of course it’s just one individual sample, yet the measurement results indicate that the Siglent specs are no BS. I also expect that this particular sweet spot does not only exist on my particular unit (even though the numbers will of course not be the same), hence the info might be beneficial for other owners as well.

Btw, usually with RF work do not need this accuracy. If we look example older HP some "State Of Art" class RF generators specifications... often decimal point can shift right. Some time ago I have in my homeland "lab" some this kind of RF gen where also was high accuracy flatness/accuracy option. And it was not as tight as you measured SDG6kX.  But example one very important thing was that it did not leak RF so it was also suitable to RF isolated "silent room".

Yes, even the MG3633A has far worse specifications (+/-1 dB up to 1280 MHz) compared to the SDG6000X (+/-0.3 dB over the full frequency range), yet it performed slightly better than the SDS6000X most of the time – in the sub-range up to 500 MHz, that is. Up to its full bandwidth of 2.7 GHz, the error increases as expected and exceeds -0.45 dB at 0 dBm output. Still far better than specified.

Btw, how did you eliminate spurs and harmonics from this level measurement... (for avoid fun mess etc... of course in practice totally nonsense with these levels of harmonics and spurs, but your accuracy was amazing) /

You name it. I was amazed myself, how the SDG6052X and the venerable MG3633A both were almost spot-on even in terms of absolute level on my (some 40 years old) trusty Boonton 4210RF power meter.

Spurs do not exist in these waveform generators, because other than in traditional signal generators there are no mixing processes involved.

Look at the attached screenshot (taken some two years ago). It shows a 120 MHz signal and its harmonics up to 1 GHz. Absolutely no trace of a spurious signal there – and even though I don’t have a screenshot handy for this, I can still promise that there are no spurs above 1 GHz either.

The harmonics deserve a more serious consideration:

For output levels up to 0 dBm (as used here) and frequencies up to 200 MHz, the THD stays below 0.3 %, hence maximum possible amplitude error is < 0.03 dB.

At higher frequencies, THD increases quickly, especially above 350 MHz, where 0 dBm is about the maximum output level. Yet the possible error never exceeds 0.16 dB at 0 dBm output and it gets better at lower levels.

[rf-loop]

With regard to amplitude accuracy and flatness, I want to remind you of this old reply #162, showing the response of my (original) SDG6052X and comparing it to an older high performance signal generator.

https://www.eevblog.com/forum/testgear/siglent-sdg6000-series-awg_s/msg2621457/#msg2621457

At the sweet spot of -30 dBm, I got a frequency response of 100 kHz - 500 MHz  +0.06 / -0.04 dB.
That's not only excellent, I'd call it even spectacular.

Yes in this individual unit it can say, it is spectacular.

Of course it’s just one individual sample, yet the measurement results indicate that the Siglent specs are no BS. I also expect that this particular sweet spot does not only exist on my particular unit (even though the numbers will of course not be the same), hence the info might be beneficial for other owners as well.

Btw, usually with RF work do not need this accuracy. If we look example older HP some "State Of Art" class RF generators specifications... often decimal point can shift right. Some time ago I have in my homeland "lab" some this kind of RF gen where also was high accuracy flatness/accuracy option. And it was not as tight as you measured SDG6kX.  But example one very important thing was that it did not leak RF so it was also suitable to RF isolated "silent room".

Yes, even the MG3633A has far worse specifications (+/-1 dB up to 1280 MHz) compared to the SDG6000X (+/-0.3 dB over the full frequency range), yet it performed slightly better than the SDS6000X most of the time – in the sub-range up to 500 MHz, that is. Up to its full bandwidth of 2.7 GHz, the error increases as expected and exceeds -0.45 dB at 0 dBm output. Still far better than specified.

Btw, how did you eliminate spurs and harmonics from this level measurement... (for avoid fun mess etc... of course in practice totally nonsense with these levels of harmonics and spurs, but your accuracy was amazing) /

You name it. I was amazed myself, how the SDG6052X and the venerable MG3633A both were almost spot-on even in terms of absolute level on my (some 40 years old) trusty Boonton 4210RF power meter.

Spurs do not exist in these waveform generators, because other than in traditional signal generators there are no mixing processes involved.

Look at the attached screenshot (taken some two years ago). It shows a 120 MHz signal and its harmonics up to 1 GHz. Absolutely no trace of a spurious signal there – and even though I don’t have a screenshot handy for this, I can still promise that there are no spurs above 1 GHz either.

The harmonics deserve a more serious consideration:

For output levels up to 0 dBm (as used here) and frequencies up to 200 MHz, the THD stays below 0.3 %, hence maximum possible amplitude error is < 0.03 dB.

At higher frequencies, THD increases quickly, especially above 350 MHz, where 0 dBm is about the maximum output level. Yet the possible error never exceeds 0.16 dB at 0 dBm output and it gets better at lower levels.

Yes, and if look my result, including cable, flatness is in same ballpark with your measurement. I believe, not know, in my used measurement some reason there is some level offset. If is or if not but flatness is well in specs and more... of course I also mesured upt to 500MHz and it keep flaness, so it do not continue dropping with this trend what can see my previous image between 150 - 200MHz... not at all... it even go back to up some amount later..

Naturally I know SDG6000X harmonics and non harmonics.
It can use as 500MHz RF generator but for this use need good external step attenuator. Bit sad its highest level is around 0dBm with higher frequencies what is lot of too low for many kind of things, example testing mixers, intercept points, compression points  etc..

[tautech]

if you want to design your own equipment, some special stuff that just dont exist,
and you like to use a case that is as identical as possible, and also with the same front back protective plastic and so on,
is that possible to get somewhere ?

If you hunt on Aliexpress a few enclosure manufacturers pop up. Just 2 minutes looking found these:
https://www.aliexpress.com/store/group/Iron-enclosure/2987020_514374388.html?spm=a2g0o.detail.0.0.67121a7662pcWU
https://www.aliexpress.com/store/group/Iron-case/3387014_512389135.html?spm=a2g0o.detail.0.0.7fc07eecmHM3Iw

[oz2cpu]

ha ha :-) not what i wanted..
i want a 99.9% look a like siglent case, and front, and protective plastics too,
so my own equipment can stand right next to, and look like it fit there.

getting the cheapest signal generator for 300$ and rip out the PCB is not that far from what I could imagine to do,
since I get the case exactly like i dream of, including front with buttons and display,
might want to hear if anyone allready reverse engineered the display interface ?
and even the buttons pcb interface ?
so it is easier to use as is, for own projects ?

[oz2cpu]

EDIT:
PROBLEM SOLVED : Thanks to TURBOTOM
got a new state file, recalled it, and now my unit live update the delay no matter what wave format,
wird side note : recalling my own previous stores state file, will NOT make the unit behave bad again
video is also updated with this info, so IF your unit behave odd, recall the settings file from TOM.

------

is there a Siglent Bug repport Feature request page for each product ?
I cant find it for SDG6022X

https://youtu.be/OXSKiE6Azmc

please tell me i do something wrong, delay between the two channels is not updating live,
anything else update live by the way.
is there a hack to have live update on delay ?

ch1 screen color is green, why is output terminal blue ?
ch2 screen color is yellow, and output terminal is correctly yellow.
is there a hack to correct the colors ?

[TurboTom]

The color issue is a simple one: util->system->page2->UI style

The pulse delay between the two channels is something i wasn't able to reproduce. I'm running F/W 6.01.01.36 on my SDG6kX and as along as the instrument is running in phase locked mode (util->page2->phase mode->locked), an entry to the individual channel parameter's "delay" field has direct effect on the output. To test this, I had all other channel couplings (util->CH copy coupling->CH coupling) set to "off". Whatever I tested, I wasn't able to reproduce the "delayed effect" of the entry that requires toggling channels off/on as you demonstrated. What firmware are you running?

[tautech]

The color issue is a simple one: util->system->page2->UI style

The pulse delay between the two channels is something i wasn't able to reproduce. I'm running F/W 6.01.01.36 on my SDG6kX and as along as the instrument is running in phase locked mode (util->page2->phase mode->locked), an entry to the individual channel parameter's "delay" field has direct effect on the output. To test this, I had all other channel couplings (util->CH copy coupling->CH coupling) set to "off". Whatever I tested, I wasn't able to reproduce the "delayed effect" of the entry that requires toggling channels off/on as you demonstrated. What firmware are you running?

V1.36

thanks a lot for the latest fw tautech
just installed it, my brand new SDG6022X came with 6.01.01.35R5B1

oz2cpu is still learning the capabilities of these units. 
The Sys and Output menus take a little time to become familiar with.

[oz2cpu]

THANKS for a super forum, and your help.

>The color issue is a simple one: util->system->page2->UI style

YES that worked, ok that is funny why would you like a bright and clear green
over a weak and hard to define kind of blue that looks a little bit like nothing :-)
i am clearly turning old and color blind, ha ha..

 (util->page2->phase mode->locked)
WAS locked
tried all other ways, NO solution

 (util->CH copy coupling->CH coupling)
WAS Channel Copy
Tried Track Off, same issue
Tried Track On, then Channel 2 become a copy of ch1, and you cant change any parameters of ch2
tried all other ways, NO solution to this issue, but I learned a few other cool things
the amplityde lock together and the frequency lock, very usefull features

>oz2cpu is still learning the capabilities of these units
Absolutly :-)
still have not found a solution, my unit will only update the ch-ch delay if I hit one of the output off-on

NEWS
CH1 SQUARE WAVE, I trigger on that one, so dont change its delay,
if i change ch2 to SINE or RAMP, the CH2 delay become LIVE
if I change it back to SQUARE or PULSE, the delay change will only update on the output if i click off-on
this you must be able to recreate.. and it is now affected by any other settings

[TurboTom]

I've got an idea what you may try: I saved my AWG's state that I used to generate the "immediately adjustable" delayed pulses to an XML file and attached an archive of that to this message. If you unzip it and copy the file "DELAY-TST1.XML" to a thumb drive, you could import it to your AWG using store/recall->file type (state), navigate to the drive, highlight the file and select "recall". This will put your AWG exactly into the same configuration that I used to test the delay behaviour. You may want to save your AWG's current state to another file before so you don't lose your own preferences. If you like, publish our saved instrument setup here, I'ld like to understand what actually causes your observed behaviour.

[oz2cpu]

> I'ld like to understand what actually causes your observed behaviour.

Me too, THANKS alot for trying to help here.
now it gets very strange, I saved my state FIRST, like you suggested,
double checked it failed the delay update as explained.

Recall your state setting, it works perfectly live updates, it is in 2 x pulse mode,
ok fine i change one and two to squarewave, and still, live update delay..

now the wierd : I recall my own setting, to get back to see if my unit is still bad,
BINGO now it works, live update,
and all my settings are exactly like i left it.
so that means you mostlikely cant use my setting here attached, for anything it all

I assume, recalling your file, cleaned something

[DL2XY]

I could not reproduce the issue oz2cpu reported. But in trying to do i found severe inconsistancys regarding delay/PM settings behavior:

SDG6052  F/W 6.01.01.36

Settings:
I am using the sync output (set to ch1)  to ext. trigger input of scope for timing reference.
Phase Mode is locked.
Channel Coupling is off.

Square and Pulse Waveforms:

You can shift delay on both channels independently as it should be. Even negative values are possible.
Adding additional phase modulation to ch2 works as expected.
But trying phase modulation to ch1  things get wierd:
The previous value of delay is reset to zero (in relation to sync output) but stays displayed as before.
Trying to change ch1 delay will change ch2!
From now on there is a permanent offset in delay according to the value active at invocation of modulation, even if turning off modulation.
Only way to reset this offset is restarting the unit.

Sine Waveform
Delay on ch1 works, but delay of ch2 stay at zero, instead ch1 will be changed.
Additional PM on ch1 does not work at all.
Additional PM on ch2 works, but changing ch2 delay will effect ch1.

Ramp and Arb. Waveform
Delay on ch2 works, but delay of ch1 stay at zero, instead ch2 will be changed.
Additional PM on ch1 works, but changing ch1 delay will effect ch2.
Additional PM on ch2 works as expected, delay ch2 works too.

Overall it seems that the Sync output is not generated on the ch1 timebase but on delayed/modulated signal itself.
That could be the reason for all that confusion.
But it also depends on history, the missing PM on Sine ch1 worked again after ramp test only changing both ch waveform to sine.

[TurboTom]

Those are the things that I've been complaining about virtually since when I received the instrument. A lot has been fixed by Siglent's firmware updates meanwhile, but as you dig deeper into the machine, you will find quirks and inconsitencies behind every corner. I guess that's the consequence of Siglent re-using the code of their previous AWGs (SDG1000X and SDG2000X), modifying it to work on the Zynq platform and adapting it to the substantially uprated hardware.

I guess in hindsight, it would have been a more decent move to start software development for this machine with a clean sheet and not having to deal with "old sins" and half-baked fixes of previous implementations. Especially, since the touch screen functionality had already been added as a plug-on to the original UI, long before the SDG6000X was in the pipeline.

Since the SDG6000X is clearly aimed at the professional users, doing without the frequency counter function and utilizing the free BNC connection (since the enclosure and hence the available real estate for connectors has been established long time ago) for another auxiliary in/out port would have been a wise choice. It's actually ridiculous to only have a single port for modulation, trigger, sync and what not function for BOTH channels on an instrument like that! I guess even most hobbyists will rarely turn on the frequency counter function except maybe for curiosity's sake. The whole implementation of the F/C on these instruments makes me believe that it was added more or less as an alibi since ce"R"tain contenders had that fearure as well (though on those, it's actually well implemented and useful, and they still don't lack in AUX connectivity).

Whatsoever, it is what it is (got better with every firmware release, the latest one making a big difference for me, getting the instrument's initialization consistent and reliable) and I don't expect to see a completely revamped firmware. All we can do is pestering Siglent with our findings, wishes and ideas for improvements. At least, it seems that once in a while they listen to this, probably to a large extent to Tautech's intervention.
Regarding this, a big "thanks" to him  .

[TurboTom]

Found another quirk:

After @hpw's contribution regarding his findings concerning frequency modulating a square wave signal with the SDG2000X, I decided to have a close look at the SDG6000X. No, modulation doesn't drop out after some time of operation, but at low frequency deviations, there seems to be a fade-in/fade-out effect. I used the following settings:

Waverform: Square
Frequency: 10MHz
Duty Cycle: 50%
Output amplitude: -20dBm (probably irrelevant)

Frequency modulation
Modulation waveform: Square (also works with sine, ramp...)
Modulation frequency: 1kHz
Frequency deviation: 15Hz (the lower the frequency deviation, the stronger the effect is, at > 50Hz it's almost gone)

If you listen to this signal with a narrow band FM receiver, you can hear the modulation fade in and out at a fraction of a Hertz. Viewed on an SA (span 10kHz, RBW <= 100Hz), you can also see the sidebands rising and falling at the same rate. There's also some variation in amplitude of the individual sidebands vs. each other visible with this periodicity.

I compared this to the DG800 but didn't find anything like this here -- modulation is "clean as a whistle". Seems like Siglent is using too low accuracy math in their modulation engine. 

That's not a big deal since apparently, this bug bites only in very rare occasion with specific settings, but nevertheless one should keep it in mind when modulating non-sinewave carriers.

Edit: Correction of a misleading term

[hpw]

Well, the SDG2000 series are almost the SDG6000 series... some more nice features for heavy bucks 

So one more to check on SDG6000 please...

I checked the PM/sine modulation as:

. base freq 12khz
. PM freq. sine & 1kHz

and then lowered the phase as low as it gets as 0.000'1°    ..

But below may 0.001° there was NO change on the side peaks levels visible. Checked with a audio ADC as 24bits. For each decade of phase a -20dB should be visible...

While did not find any specifications on that matter 

Thanks for testing

Hp

 

[Le_Bassiste]

Since the SDG6000X is clearly aimed at the professional users, ....

maybe it was aimed at being that, but one thing for sure it is not: an instrument for professional use.
if they could at least manage to make that parameter knob function smoothly w/o any hiccups, _and_ make frequency transitions smooth while manually sweeping, _and_ the amplitude less noisy,...
"als tiger gesprungen, als bettvorleger gelandet" /rant

[TurboTom]

@hpw:

I tried to replicate your tests but since I lack the mentioned, high-resolution audio gear that you utilized, I had to change the carrier frequency so I can comfortably measure it with my spectrum analyzer. The sweet spot regarding S/N ratio appeared to be around 1MHz, so that's what I uesd. These are my findings with a sine carrier of 1MHz, 0dBm, 1kHz sine modulation signal:
Below 0.01° phase shift, the sidebands @ 1kHz offset stay at ~ -84dBm, right down to 0° where they disappear in the noise floor all of a sudden.
Increasing the phase shift by 0.001° increments, not every increment has an effect, and only at 0.1° phase shift, the first order side bands reach the approx. correct level of -61.5dBm.

Moreover, as a reaction to @blue's reports on his SDG2000X findings, today I took a closer look at the PSU of my SDG6000X due to the hot running MLCCs parallel to the switching MOSFETs. Other than I initially thought, the MLCC at the low-side switch wasn't cracked but there was just a speck of dust on it. I removed the PSU module altogether and took a closer look at both sides:





Obviously, the build quality is good and there's nothing to complain about the choice of components. All of the elecrolytics are Rubycon brand.
The controller is an ST L6599 resonant mode half-bridge specimen with integral gate drivers. A PFC circuit isn't utilized in the PSU. In order to understand better what's going on here, I decided to run the PSU separate from the AWG with a variable power supply and a DC load at its 6.5V output (that's the one used for regulation of the converter and designed for the highest output current). Here are some thermographs, taken at 115V:





It's quite peculiar to find that the MLCCs at the switches stay rather cool while a resistor in the current sensing circuitry get noticeably warm. Moreover, the three dropper resistors for the initial startup supply run at an acceptable temperature of some 70°C.

At 230V supply voltage, things change considerably:





The elevated temperature of both MLCCs at the switches as well as the unhealthy temperature of the dropper resistors lead me to the impertinent assumption that the PSU has been designed to be operated at a 115V mains and had only more or less "by coincidence" been found to be running off 230V as well. Anybody heard of the manufacturer "Unicorn" as revealed by the sticker at the solder side?

At a load current of 3A, the first smoothing cap at the scondary gets pretty warm as well. I may have already been overloading this output with the chosen settings, though.



Since the individual component temperatures appear to be as they are "by design", I decided to leave everything as it is and only take care of it if something should fail.

But since the AWG was apart anyway, I tore it down a little further to take some more photos of the things that make it tick...





Here the CPU module is shown. It contains a Xilinx Zynq 7010 ARM Cortex A9 / FPGA combo SoC, a total of 256MB DDR3 RAM and 256MB of NAND Flash memory. There's also other peripheral stuff present like power supplies, a LAN PHY, several clock generators and two USB transceivers. Altogether, this appears to be a pretty capable system core.

...tbc in next post...

[TurboTom]

The main board looks pretty impressive as well. But one thing that immeadiately catches the eye are the unpopulated footprints for the sample memory around the FPGA. These positions must have been populated in earlier versions of the AWG as shown in Shariar's review/teardown video.



Even though the Xilinx Kintex7 FPGA that Siglent utilized here is "quite a beast", the available block RAM memory of 4.86Mbit results in only about 150kSamples of directly available storage per channel. I wonder how Siglent manages to provide the specified 20MSamples for arbitraries per channel?? Is the digital link between the SoC and the Kintex fast enough to provide the data rate? I also finally get an idea why eliminating glitches when switching parameters seems to be so difficult: There's simply not enough fast sample memory available to provide effective double buffering / buffer interleaving! Appears like the bean counters at Siglent won the contest against the engineers (is it different anywhere these days??).



But it gets worse: Above you can see a shot of the output section at an awkward angle, showing that there's provisions to install two output drivers per channel. One is installed, the other, well, "economized". The THS3491 utilized here is specified by the manufacturer at a maximum long-term output current of ~100mA to prevent electromigration (which will probably lead to parameter deterioration and eventually failure). The "small" problem is, that the paralleled output resistor branches form a total resistance of 50 ohms (obviously) which will result in DC currents of up to 200mA out of a single drive if the generator is used as a current source. Twice the manufacturer's absolute maximum ratings, while there's space for the second buffer that would completely eliminate this problem! Shame! That's definitely not a professional approach to use components.



One more small issue that could be improved is the mounting of the fan: Four very long machine screws hold it in place, the nuts pulling and eventually warping the fan casing towads the mounting surface. Smaller screws and the nuts placed between the two flanges would have bypassed this issue altogether, as show in the above photo (albeit with the original, long screws).

So what's my conclusion from these findings, combined with the experience of the very dragy firmware updates? Plainly, this AWG is shamelessly overpriced for what it delivers. Siglent cuts corners everywhere and appears to sort things only where / when the pressure gets too high. I don't see any substantial room for significant future inprovements or extensions by firmware update, Siglent just castrated the hardware to a point absolutely necessary for basic operation. The whole series (SDG 1000X / 2000X / 6000X) appears to have reached a dead end and I wouldn't be surprised if we will see a new platform to be released some time soon.

Rigol's entry level AWG (DG800) outperforms even this "wannabe flagship" AWG in almost every concern except bandwidth. It's really embarrassing for Siglent to offer such poor products on the AWG market while their scopes and SA / VNA product appear to be pretty competitive. I'd say it's time to change heads in certain departments...

[JohnG]

Rigol's entry level AWG (DG800) outperforms even this "wannabe flagship" AWG in almost every concern except bandwidth. It's really embarrassing for Siglent to offer such poor products on the AWG market while their scopes and SA / VNA product appear to be pretty competitive. I'd say it's time to change heads in certain departments...

This is a nearly meaningless comparison. There is little point to buying this generator if you don't need the bandwidth. I would love to find another generator with this capability, including bandwidth, for this price. The closest Rigol generator (DG5352) can do only 350 MHz, and cost is about 2.5x higher than the SDG6052x. And, the Rigol has 500 ps (or more) jitter according to their data sheet! This makes it basically useless as a high bandwidth generator, and all the features in the world can't fix that jitter.

A useful bit of information would a better generator than the Siglent at a comparable price, or even a somewhat higher price. I have looked and not found one even close. But, please don't take a primary specification like bandwidth and throw it away to make the comparison.

Also, the thermal camera measurements are not useful for determining operating temperature when the cover is off. The actual temperatures are highly likely to be affected by this. They are probably cooler with the cover on, but they could be hotter. I don't know. Nobody does because the cover almost always has a large impact on airflow, which has a large impact on temperature. I know thermal camera pictures look fantastic, and thermal cameras are very useful. That does not make them correct when you have modified the thermal system.

Cheers,
John

[TurboTom]

I understand your point. But the SDG6000X is advertised as a versatile, general purpose AWG with high bandwidth, low jitter and fast slopes. Needing higher frequency signals once in a while, I purchased it some two and a half years ago as an instrument to replace all the other, lower spec'd AWGs that I've got. Does it fit into this frame? Maybe with the latest firmware , it's getting clos(er), but it's (still) not my go-to instrument when doing a quick test that doesn't require its bandwidth. Its U/I IMO is way too uncomfortable and appears rather old fashioned. And as reported before, there are still quirks and small bugs to be found everywhere. Prior to the recent F/W update, its operation was simply not predictable, directly after powering up, it could very well produce a waveform that had little to do with what was displayed on the screen. Fortunately, these problems now appear to have been sorted.

Even though its predominant feature is the high bandwidth, I think it still has to be able to operate as a "mainstream" AWG, and it's got to do that well.

I'm completely aware that the thermographs of the PSU ouside the casing tell little about the component temperatures when enclosed and vented. I also mentioned that in a contribution somewhere further above. But are MLCCs supposed to run hot at all? Especially the comparison between the two supply voltages is interesting, maybe not so much for the temperatures displayed but for the relative figures.

And my main critics is that - in test gear - using components in a way that they can be operated way beyond the manufacturer's absolute maximum ratings under certain circumstances, simply is a no-go.

[JohnG]

MLCCs can definitely heat up and often do in switch mode power supply applications. They can carry some serious ripple current, and the dissipation factor for high capacitance parts (X7R or similar) runs around 1%.  It is also common for "universal input" power supplies to run hotter on 240V input versus 120V. Sometimes the parts that get hot will change when you change the input voltage.

Also, one cannot expect things to run properly if you remove the cover of a fan cooled system. It is not a reasonable expectation that one would run a piece of test equipment with the cover off, or a system with the cooling system defeated, unless the system was designed for that. So to show pictures of parts getting hot with the cover off, and calling out that the temperatures are high, implies to many that the design is poor. It might be or might not, but the conclusion cannot be made, nor should it be implied, by these thermal camera measurements.

I can understand being disappointed in problems with the generator. It seems like there are a lot of bugs. But, an equivalent generator from a top tier test equipment company (Keysight, R&S, etc.) costs many times more, and it is still not certain that it won't have hidden quirks, bugs, or outright design flaws. I have been burned by this on some really expensive equipment.

I have a fair amount of Siglent equipment. It does me well for the money I paid. But, they are still barely a Tier II test equipment maker, like Rigol. I wish they would get this right, because they have so much potential. But then, they will raise their prices, too.

Cheers,
John

[2N3055]

Tom,

Thanks for detailed analysis...

b.r,
Sinisa

[TurboTom]

Some more testing with EasyWave X proved that Siglent somehow must have managed to store arbitraries inside the CPU module's DDR3 RAM up to a size of 20MSa per channel and transfering this into the Kintex FPGA at a rate to support 2 * 300MSa/s which equals 1.2GB/s averaged. At this load, the UI gets a little sluggish but the generator manages to continuously output the arbitrary signals. So no worries here, the AWG performs as specified. If I'm not mistaken, there are a total of 16 LVDS high-speed lines running from the Zynq module to the Kintex, so physically the required bandwidth wouldn't be a problem.

Considering this, I wonder why the four 1Gbit DDR3 RAMs have been grouped around the FPGA in earlier models. Maybe Siglent planned to provide even higher sampling rates for arbitraries or more space for samples. It's sad to see that obviously this intention had been dropped somewhere along the line...

[egonotto]

Hallo,

@TurboTom which hardware version have you?

Shahriar has a SDG6052X. Perhaps it is yet a difference between SDG6022X and SDG6052X?

Best regards
egonotto

[hpw]

I understand your point. But the SDG6000X is advertised as a versatile, general purpose AWG with high bandwidth, low jitter and fast slopes....

Well have the SG2082x and run into a ch1 & ch2 level issue setup issue, may you have the same on SDG6k models?

. CH1 as square wave for master clock up to 24MHz and min & max levels as for digital levels (SDG2000 do not have digital level TTL...HCMOS options so in & max level to be used)

. CH2 setup you get the same min & max level where level settings as 1..4Vpp 

. CH2 so change as a sine wave as Vpp & offset= 0... OK

now the min & max level settings gets also the level & offset settings = 

In other words, as on SDG2K model, the level setup for each channel is NOT INDEPENDENT... 

Hp
 

[TurboTom]

@egonotto:
You are correct in so far as I've got a different hardware version (SDY8.007.146D) than Shahriar reviewed (SDY8.007.146B). I compared the PCBs as accurately as possible but the differences appear to be small. All that I could really find is the presence of the two footprints for the jumpers J14 and J15 which are actually meant for installing mounting lugs to use a different kind of heat sink on the FPGA (the configuration with the torsional wire spring for attachment) while Siglent utilized a plastic clip version on my unit which is completely adequate as well.

And then, instead of using adhesive for the output driver (THS3491) heat sink as in the "B" revision, Siglent now added a soldered sheet metal heat sink that is coupled to the output driver(s) via heat conduction through the PCB. The footprint for this has also been added in the "D" revision.

Btw, Shahriar was slightly mistaken in his description of the output amplifier: The component that he identified as a voltage regulator and "not being of much interest" is actually another THS3491 that is used as a driver for lower output levels, thus bypassing the high(er) power output stage and eliminating the noise it generates.

Signal offset voltages aren't generated by the main (high-speed) DAC so its resolution is preserved for the signal. Instead, there's a separate offset DAC (here, the small MSOP-8 chip between U34 and U45, an AD5660CRMZ-1) that is used to generate a whole buch of analog control voltages from its single output that are sampled via analog multiplexers (U34, U43, 74HC4051) and buffered by capacitive "hold"-circuits made up of CMOS Opamps (U32, U35, U45 -- TP1272).

This offset signal is now amplified by the TP1272 installed in the output section and added resistively to the main DAC output by the two "big" SMD resistors of 200ohm (offset) and 61.9ohm (signal), which, combined (paralleled) will end up at a signal impedance of approx. 50 ohms. At rather low selected output levels (combined output/offset range up to approx. +-2V hi-Z), this is the direct output signal, at higher levels, the additional amplifier is routed into the signal path. At (considerably) lower output levels, attenuators are added by the bunch of relays in the output sections. This permits the cleanest, lowest noise output signal possible. Such a configuration (separate signal/offset branches almost right to the analog outputs) can be found in most high(er) quality AWGs.

Regarding the output levels of 40Vpp that Shahriar mentioned in his review, I believe he must have erred on the high side -- I don't think the "B" revision of the AWG had been so much more capable than the currently sold revision (which maxes out at 20Vpp high-Z). It also would require a completely different "animal" as the output amplifier.

Your point regarding the AWG model I've got (SDG6022X) may be valid. Since probably few individuals purchased the SDG6052X version, we may not see another teardown of such a unit any time soon to verify they are configured identically (i.e. the dedicated sample memory unpopulated as well). But just by comparison of the specs (the figures for the arbitraries of the three models are identical), I don't believe we would find a hardware difference. Moreover, the PCB label tells "SDG6000X..." which may substantiate the assumption that there's actually only a single hardware configuration in production.

@hpw:
I copied your configuration and found -- if I understood correctly -- the same behaviour. But this only affects the U/I, the signal output is not changed. I.e. the choice to enter the amplitude as Max/Min values or as Amplitude/Offset combination is a global one -- it's common for both channels. But the instrument will internally convert the figures for the "idle" channel and won't change its output characteristics, and that's what is relevant IMO. Of course, it would be nice to have the entry scheme independent for both channels, but this once again is just another example for the overall dated appearance of the user interface.