EEVBlog #805 - Siglent SDG2122X Arb Generator Teardown

[nowlan]

Is the ground cable that was on a spade plug up to code? Thought you needed shake proof washers etc.

[apis]

"Latest EEVblog Video" on the website is still #804

[DanielS]

This exact model of sticker is often used in some of the cheapest, lowest quality products China has to offer.

Those cheap stickers, or even cheaper plain paper ones, can also be found in more expensive equipment. Their only purpose is to keep track of which units have passed through QC stations. They do not tell you anything about what the individual stickers stand for. For all you know, the last "QC" sticker you see on the housing could very well be someone checking that all the chassis screws are present and tight.

[ivan747]

Could someone explain why this instrument only have 20Mhz arbitrary wave output, when its DAC is running at 1230Mhz, please?
I feel confused. 

20MHz is how fast it can repeat whole Arb memory (it also depends how many Arb memory points waveform have)

Exactly. If you fill up the memory, it can repeat itself 20 million times per second, but what you have on that memory can have higher frequency components in it.

You can also fill only a small part of the memory, say 1/2 of the memory, then it will repeat at 40MHz, if I am not mistaken. Is that right, rf-loop?

[LapTop006]

From the hardware point of view it looks like a very capable instrument. All thanks to the advances in telecom technology . Most of these high speed ADCs and DACs that we see nowadays in the budget gear has its origins in developments in RF for 3G/4G or core network (fibre optic, DWDM etc) equipment.

Origins yes, but quite old ones. 1g equipment (~1.2g often base rate) entered mass deployment right around 2000 (with carrier core elements of 2.5g OC48 a few years prior), 10g was in mass deployment maybe 2010, and we're just now seeing the wide deployment of 100g equipment (in a 4x25g form) in network cores.

The latest DWDM equipment is now doing 400g/wavelength, going to 16, 64 and even higher QAM rates driving 20+ terabit per fibre pair.

Certainly some of it is from the mobile world where chips and miniaturization of LTE are starting to offer us really nice platforms for SDR at quite amazingly low prices.

[rf-loop]

Could someone explain why this instrument only have 20Mhz arbitrary wave output, when its DAC is running at 1230Mhz, please?
I feel confused. 

20MHz is how fast it can repeat whole Arb memory (it also depends how many Arb memory points waveform have)

Exactly. If you fill up the memory, it can repeat itself 20 million times per second, but what you have on that memory can have higher frequency components in it.


You can also fill only a small part of the memory, say 1/2 of the memory, then it will repeat at 40MHz, if I am not mistaken. Is that right, rf-loop?

My quess is: No. I do not know exactly how just this Siglent SDG2000X works in details. I do not have it (yet).

I think user can select only max 20MHz repeating frequency for stored arbitrary waveform.

It is "True Arb" technology  and if it works so that it do not "jump over" memory points lets think.

( this  is not Siglent or any other named generator. Just basic thinking:
If need handle every single Arb memory points in memory and Arb repeating frequency is 20MHz. Hmm... if Arb memory read speed is 100Mpoints/s how many memory points we can read. 20MHz cycle time is 50ns. We can read only 5 memory points in one turn.. (5x10ns=50ns)  If we can read arb memory points using 1Gpoints/s then with 20MHz repeating frequency we can read  50 memory points.)
 

I do not know at all what is exactly maximum speed what Siglent SDG2kX can process Arb memory and how it is arranged for amounts of waveform points etc.  I hjave understood they use "True Arb" for imply that it do not jump over any single memory points (?)

Siglent web side:

?  High-performance sampling system with 1.2GSa/s sampling rate and 16-bit vertical resolution. No detail in your waveforms will be lost
?  Innovative TrueArb technology, based on a point-by-point architecture, supports any 8pts~8Mpts Arb waveform with a sampling rate in range of 1?Sa/s~75MSa/s

This is one "good to read" paper
http://www.alliedelec.com/images/Products/mkt/pb/Tektronix/Pdf/understanding_signalgeneration.pdf

I hope Siglent do better material for show exactly how it works and what are limits exactly. Whole principle exactly with all information with numbers how it works and how it do not work.  It is very extremely different what is example in SDG800, 1000 and 5000 generators. 
 
Example with every SDG1000 can produce 50MHz sine out using Arb, There Arb max repeating frequency is 5MHz. There you can do 10 cycle sine to arb memory and run it using 5MHz and out come 50MHz.  But if there is short details in arb memory it may (randomly!) jump over it because this model have fixed arb memory lenght and fixed clock. Missing details is one disadvantage what user need know.  When it have 125MSa/s it mean it have 8ns step interval.  It have 16kpoint Arb memory. If handle every Arb memory points without missing points then how much time it need. 8ns x16000 = 128000ns = 128us.  Whole Arb memory every single points can then handle only if repeating frequency is under 7.8 kHz !  If go over this Arb frequency it need start skip over memory points and due to fixed clock in every turn different memory points are skipped. (depending frequency. There can also find special frequencies where skipping and waveform iare "in synch".  This is disadvantage in all fixed clock fixed memory lenght "simple principle Arb generators" independent of manufacturer name.)

SDG2000X is very very different and it have more advanced principle. It have variable Arb memory lenght and variable Arb memory "stepping" speed and then this Arb repeating frequency what have limit value max 20MHz.  And this is not only in Siglent, it is and have been long time, in many bit more advanced generators. But perhaps with different details combination.

All is adjusted so that there is "No detail in your waveforms will be lost."  And it is perhaps much much more important than maximum frequency what it can produce.
It need think for what we use and need arbitrary waveform function.

But, as told, Siglent need write better specification so that we can read exactly what it do and what it do nnot do. How it works, exactly. This is important so that buyer do not disappoint.  If you promise gold and buyer find brass all are disappointed and angry. But you have his money and you are still happy, you win. If you do it fast enough you can win many....  But, try repeat it... no way.  On the chinese street corners this happends thousends of times every day  and you can change street corner and repeat, you can change city, and repeat... but here, international markets you can not run to ther corner, other city... so, this is different.

If you promise brass and buyer find he also get brass he is satisfied.   If you promise gold and buyer get gold, agen  he is satisfied.  In both cases you can continue and continue win.  If one is not satisfied he tell it to thousends, ten thousends, to whope world of peoples. You have just one shot... If one is satisfied perhaps he do not talk anything..  Good enough deep and open information about product, it is ONLY way to do and continue.  Please learn how HP build they brand... start reading example HP journals from history. How much they tell and explain lot of things.. well explained some new equipment working principles including also very deep details. Why they do not afraid copycats. They show all schematics, they show even part of design calculations and theories and things how they have developed something. Why they do not afraid copycats.

Everyone can copy schematic, everyone can solder and do printed circuits, all can do. 
What can not copy? Lets assume that price is still limit.  Answer is: Quality. This can not copy (so that price is cheap enough for copycat business)

Who is first chinese company what really realise this and do it. I'm waiting.

[sergey]

Dave, very nice video, thanks!

Got a question to the audience tho. It seems 10MHz reference input is fed to the main board via ribbon cable. Is it gonna to cause some parasitic leakage and hence have negative effect on the instrument stability?

[f4eru]

Could someone explain why this instrument only have 20Mhz arbitrary wave output, when its DAC is running at 1230Mhz, please?
I feel confused. 

Because you need a lot higher a sample frequency to render your waveforms properly.
Also, it's uncommon to use an arb at higher frequencies. Usually you need an RF generator (extremely clean sine + modulation only) on the higher bands

[Dr. Frank]

Thanks Dave, for that interesting teardown video.
Well explained, and just the right length.

These Q.A. / Q.C. stickers on the back.. in contrast to that, many solder joints , especially at resistors, (@15:00min) they look really awful, they wouldn't have passed the AOI in our assembly plant.
The solder meniscus does not look smooth and clean, more  looking irregular, and not fully wetting the surfaces, like manually soldered.
Was it an inhomogeneous solder profile, or too low solder temperature?

That will for sure become a severe quality problem after a few years of operation.

Frank

[analogNewbie]

The firmware from siglent website for SDG2000X has high entropy. It seems that they encrypted the firmware by some algorithm. If someone wants to check out the frequency unlocking code, one should buy one and dump the flash. I think from 40MHz to 120MHz the only difference is inside the software. Hope they dont change the output opamp to reduce the cost.

[billfernandez]

I've been looking through the user manual and datasheet and thought I 'd share a few notes about the SDF2000X series:

MAXIMUM OUTPUT FREQUENCIES:

sinewave up to 120MHz.
squarewave up to 25MHz.
ramp ( incl. triangle) up to 1MHz.
pulse up to 25MHz.
Noise up to 120MHz bandwidth.
Arbitrary waveforms up to 20MHz.

(The specs don't say whether the non-sine max frequencies are lower than shown above if it's not a 120MHz model.)

MISC NOTES:

o Modulation input 50KHz max.

o Sync output = 500nS pulse.  When the frequency of the waveform is greater than 1MHz, there is no sync signal output.

o The base unit can be upgraded to higher frequencies by entering a license code.

o The manual says "SDG2000X supports 40MHz, 80MHz and 120 MHz bandwidth for now." (what does "for now" mean?)

o Zout = 50 Ohms.  If you want the Vout readout to be correct, you must tell the unit the impedance of your load (over a 50 Ohm to 100K Ohm range).

o The 200MHz frequency counter seems to resolve to 1Hz, and is independent of the two signal generator outputs.  So all three can be used at the same time.

o Digital values are drawn from ARB memory at a basic clock rate of 300MHz. These values are fed to an interpolator that runs at four times that rate (1.2GHz) and which generates interpolated values between the old and new value.  Thus for each "real" value drawn out of ARB memory four values are presented (in turn) to the DAC, resulting in greater Spurious-Free Dynamic Range (SFDR).

[tautech]

I've been looking through the user manual and datasheet and thought I 'd share a few notes about the SDF2000X series:.............................

Welcome to the forum

Just pointing out your typo ^
Should be SDG

[TheSpruceMoose]

Got a question to the audience tho. It seems 10MHz reference input is fed to the main board via ribbon cable. Is it gonna to cause some parasitic leakage and hence have negative effect on the instrument stability?

I actually had the same reaction at first when I saw that ribbon cable!

It seems like you could have some capacitive loss, but if your reference signal doesn't have any content at other frequencies, I think it wouldn't affect the overall quality of the signal (i.e., an RC filter at one frequency is just an attenuator)...  At 10 MHz, the wavelength is ~3m, and I think that you don't really have to worry about reflections or other weird high-speed stuff until the line is longer than ~10% of the wavelength (30cm (1ft) here).  Maybe that's different when it's a really high-quality signal?  I have really, really limited knowledge and experience, so I could be exceptionally wrong :/

The frequency counter input is on the same ribbon cable, and it can supposedly count up to 200 MHz, so I *really* wonder how much that would be affected...