### Author Topic: Siglent SDG1000X Waveform Generators  (Read 28857 times)

0 Members and 2 Guests are viewing this topic.

#### TurboTom

• Frequent Contributor
• Posts: 907
• Country:
##### Re: Siglent SDG1000X Waveform Generators
« Reply #150 on: October 23, 2020, 09:49:37 pm »
@Johnny B Good -

I'm sorry if it sounds a little rude, but in none of your screenshots I can see a "properly" sampled Sinc pulse. The Sinc function has to be symmetrical, as soon as there's even the slightest asymmetry visible, it's likely that samples had been skipped.

I'm not sure if you are really familiar with basic sampling theory: Theoretically, in order to reproduce a sine signal, two samples are sufficient for each period of that sine. Since this is not possible practically, the maximum frequency that you will find in real-world generators, is 1/2.5 of the sampling frequency, i.e. the highest frequency that an AWG running at 250MSa/s will be able to generate is about 100MHz (Rigol DG992 for example). A Sinc signal from -32Pi to +32Pi will have 32 "hills and valleys" in total, so the highest frequency that this signal contains is 32 times the "repetition frequency" (22 as you reported in case of the SDG2000X). If you repeat a signal like this at 10MHz, the highest frequency component this signal contains is about 320MHz (vs. 220MHz in your case). There's no way the FY6600 nor the SDG2122X will reproduce this. What happens is called "aliasing", i.e. not every sample point in the waveform buffer gets "played back" and the reconstruction filter forms of the remaining sample points what you see on the screen of your scope. This may actually resemble a Sinc signal quite well, but it's not what's stored in the AWG's memory. It also leads to the "smearing" of the trace if there's a low-frequency interference between the sampling frequency and the highest frequency component of the Sinc signal to be reproduced.

You may argue that Siglent specifies for their SDG2000X generators 1.2GSa/s, and you are right. But if you look more closely at the spec sheet of that generator, you will quickly get disappointed: The 1.2GSa/s refers to the internal sampling rate of the AD9122 DAC. But the DAC has got an internal interpolation engine and Siglent "drives" it only at 300MSa/s while the interpolation makes signals that are within the specified frequency range appear (more ro less) like they had been sampled at 1.2GHz. So basically, the FY6600 and the SDG2000X aren't that far apart from each other. The DDS and "TrueArb" sampling scheme also only differ in details, you will find that the "reproduction rate" of the TrueArb scheme is another factor four slower than the DDS scheme (75MSa/s vs. 300MSa/s). This is necessary to reproduce slopes by adjusting their slew rate in order to apparently shift their position so they lie in between the "true" sampling points, thus reducing phase jitter, but this comes at the cost of bandwidth.

All this means that none of the manufacturers can do magic. All of them have funny names for their performance enhancement (TruArb, SiFi II,...) that describes the same principle.

I also don't exactly understand what you mean with that the Sinc signal is a difficult to reproduce waveform. If you stay within the mathematically possible frequency range, it's no more difficult than any other waveform. Moreover, an ARB generator is better suited to produce sensor signals for example that are otherwise difficult to obtain in the lab, i.e. to design a post-processing electronics. And these signals usually aren't that fast. I haven't heard of anyone requiring a 20MHz Sinc signal, but maybe I'm just ignorant  .

Anyway, I hope this information may give you some deeper insight into the subject.

Cheers,
Thomas

The following users thanked this post: rf-loop, Johnny B Good

#### Johnny B Good

• Frequent Contributor
• Posts: 443
• Country:
##### Re: Siglent SDG1000X Waveform Generators
« Reply #151 on: October 24, 2020, 03:05:06 am »
@Johnny B Good -

I'm sorry if it sounds a little rude, but in none of your screenshots I can see a "properly" sampled Sinc pulse. The Sinc function has to be symmetrical, as soon as there's even the slightest asymmetry visible, it's likely that samples had been skipped.

I'm not sure if you are really familiar with basic sampling theory: Theoretically, in order to reproduce a sine signal, two samples are sufficient for each period of that sine. Since this is not possible practically, the maximum frequency that you will find in real-world generators, is 1/2.5 of the sampling frequency, i.e. the highest frequency that an AWG running at 250MSa/s will be able to generate is about 100MHz (Rigol DG992 for example). A Sinc signal from -32Pi to +32Pi will have 32 "hills and valleys" in total, so the highest frequency that this signal contains is 32 times the "repetition frequency" (22 as you reported in case of the SDG2000X). If you repeat a signal like this at 10MHz, the highest frequency component this signal contains is about 320MHz (vs. 220MHz in your case). There's no way the FY6600 nor the SDG2122X will reproduce this. What happens is called "aliasing", i.e. not every sample point in the waveform buffer gets "played back" and the reconstruction filter forms of the remaining sample points what you see on the screen of your scope. This may actually resemble a Sinc signal quite well, but it's not what's stored in the AWG's memory. It also leads to the "smearing" of the trace if there's a low-frequency interference between the sampling frequency and the highest frequency component of the Sinc signal to be reproduced.

You may argue that Siglent specifies for their SDG2000X generators 1.2GSa/s, and you are right. But if you look more closely at the spec sheet of that generator, you will quickly get disappointed: The 1.2GSa/s refers to the internal sampling rate of the AD9122 DAC. But the DAC has got an internal interpolation engine and Siglent "drives" it only at 300MSa/s while the interpolation makes signals that are within the specified frequency range appear (more ro less) like they had been sampled at 1.2GHz. So basically, the FY6600 and the SDG2000X aren't that far apart from each other. The DDS and "TrueArb" sampling scheme also only differ in details, you will find that the "reproduction rate" of the TrueArb scheme is another factor four slower than the DDS scheme (75MSa/s vs. 300MSa/s). This is necessary to reproduce slopes by adjusting their slew rate in order to apparently shift their position so they lie in between the "true" sampling points, thus reducing phase jitter, but this comes at the cost of bandwidth.

All this means that none of the manufacturers can do magic. All of them have funny names for their performance enhancement (TruArb, SiFi II,...) that describes the same principle.

I also don't exactly understand what you mean with that the Sinc signal is a difficult to reproduce waveform. If you stay within the mathematically possible frequency range, it's no more difficult than any other waveform. Moreover, an ARB generator is better suited to produce sensor signals for example that are otherwise difficult to obtain in the lab, i.e. to design a post-processing electronics. And these signals usually aren't that fast. I haven't heard of anyone requiring a 20MHz Sinc signal, but maybe I'm just ignorant  .

Anyway, I hope this information may give you some deeper insight into the subject.

Cheers,
Thomas

Thanks for responding so quickly.

As it happens, I am acquainted with sampling theory so quite understand your argument. I was also aware of the 300MSa/s and 75MSa/s limits in this "1.2GSa/s" AWG model. However, what I wasn't aware of was there being a specific number of "wiggles" in a properly formed Sinc pulse. I'd just assumed that the ten in total peaks of the Feeltech version and the twenty in total of peaks in the Siglent one (I just set the frequency of both to 1MHz and counted them again, including the tiny pip in the middle) were just approximations to a mathematical function involving an infinite series of coefficients. I hadn't realised there was meant to be a specific 32 waves in total to recreate an accurate facsimile of a Sinc pulse.

In the light of that information (please excuse my ignorance and my ASS-U-MEd conclusions ), it seems that both Siglent and Feeltech have been cheating (Feeltech doubly so). That rather begs the question as to why Siglent in this case chose to set the limit at 20MHz rather than say 6MHz or even 12MHz at a push. The SDG1Kx models use a 6MHz limit and even this is double what a 150MSa/s rate could deliver.

The appeal of the Sinc pulse in the Feeltech was purely down to the fact that it offered a uniquely distinctive jitter free waveshape pulse with nice steep edges so suited for the task of monitoring a slowly drifting sine or square wave from my gpsdo projects. I wasn't really concerned about the squiggles in between (quality or their quantity).

Well, the fact that even Siglent felt impelled to cheat on the quantity of squiggles suggests they were trying to overcome a difficulty in producing them at  the highest frequency possible, given all the constraints. Presumably they must have decided that a close enough facsimile to a genuine full fat Sinc pulse would suffice most, if not all, users' needs. Just how exact does a Sinc pulse need to be, considering that Feeltech had cheated twice as much?

Presumably, a user who demands nothing less than a full fat Sinc pulse below the 3.75MHz limit of a 300MSa/s DAC could arbitrarily create or import  their own. Perhaps Siglent felt that their ersatz example would suit 90% of users and Feeltech had decided most of their customers wouldn't notice the difference anyway (and any that did could create one, after all, it is an AWG after all).

As for the question of anyone wanting a 20MHz Sinc pulse, I wonder who would want or need a 1MHz ECG?

John

The following users thanked this post: Figaro

#### TurboTom

• Frequent Contributor
• Posts: 907
• Country:
##### Re: Siglent SDG1000X Waveform Generators
« Reply #152 on: October 24, 2020, 07:56:45 am »
Yes, tachycardia on steroids...

Yet, I may have explained the situation with the Sinc pulse somewhat inaccurately. The number of "wiggles" is not fixed, it's rather the decision of the instrument manufacturer. If you need a longer sample, you can easily generate it as a custom arbitrary. It's just the representation of the mathematical formula f(x)=Sin(x)/x The Sin(x) generates the periods (wiggles) while the 1/x is resposible for the hyperbolic amplitude drop and the phase change and funny shape around zero. Actually, at exactly zero, this formula isn't defined since you would divide by zero. But it's possible to calculate the limit lim(x->0) by calculating the quotient of the derivatives d/dx(Sin(x)) = Cos(x) (which is 1 @ x=0) and d/dx(x) = 1, so you're dividing 1/1=1, which is the value the Sinc function approaches for x -> 0. IIRC, this is called l'Hospital's rule...long time since I was at school and univerity...

Anyway, the interval that's used to reproduce the function in an AWG is more or less arbitrary. Btw, the Sinc function is used in digital oscilloscopes to mathematically interpolate a waveform with highest frequency contents only a little below nyquist, i.e. that's only a little more than two samples per period, so there are also practical applications of that "waveform". The details of this are best to be found in books on sampling theory.

The following users thanked this post: Johnny B Good

#### rf-loop

• Super Contributor
• Posts: 3354
• Country:
• Born with DLL21 in hand
##### Re: Siglent SDG1000X Waveform Generators
« Reply #153 on: October 24, 2020, 08:20:56 am »

As for the question of anyone wanting a 20MHz Sinc pulse, I wonder who would want or need a 1MHz ECG?

John

Who need. No one.
Who "need". Kids or kids like grown ups just as for playing fun with scope and generator. Whole hobby may be just playing with test instruments alone like with Nintendo. They are main thing in hobby in some cases. Some cases peoples hobby is electronics  and so on and test and measurement instruments are only tools, just like screw driver and pliers etc.  When these are needed for real work for salary, mostly they are just tools. Who there need any ready made simplest ECG wfm what is only some how "looks like" joke copied from some basic school book, yes we can say there is artifical or better say artist painted QRST parts but after then it is all what it have. If someone need test some system he need very very different and not monotonic wfms for simulate real ECG for test some circuits for it, who think these jokes are useful in any real works but just for kids playing fun and for salesmans.
One purpose I know. Teaching. Example for teaching how to use and adjust oscilloscope for different things and signals. Example exercise lessons for simple trigger settings with different signals for look just wanted details. This can be one piece of whole set of different signals in basic training.

Sidenote(s) to some random readers who just read bit this and bit that and is confused by sampling speed, jitter, ARB detals disappearing and DDS one clock jitter and ARB frequency:

I take example about Siglent SDG1000X (because it is in topic)
ARB waveforms in DDS mode when fixed memory length is in use.
SDG1000X have 150Msa/s and ARB memory length is 16384  14 bit "words" for DAC.

One memory length is one ARB waveform  (later awfm).

All know working simplified principle is that it reads ARB memory using this fixed 150Msa/s speed  and  value goes to DAC input.

Now we can read data sheet. It tell that maximum ARB frequency is 6MHz! ARB memory lenght is 16k. It mean that it can repeat this one ARB memory waveform  6 million times in second (6M awfm/s).

If we think it can repeat whole ARB memory wfm with all its possible details using 6MHz ARB frequency -- this is just totally impossible. Why.

Because awfm length is 16k and it is sampled using 150Msa/s speed.
If we sample every single awfm memory position and we get out 6M awfm/s it need 6M*16384=98304Msa/s = 98.304Gsa/s
All know it is impossible. But data sheet do not tell it directly. User need understand how these work.

Lets look other way. What is maximal awfm/s what can do so that memory length is 16384 and clock is 150M and every data points in memory can take to DAC. 150/16384=0,0091552734375.
It mean that 9.1552734375k, bit over 9k awfm/s is maximum. Without loosing details maximum ARB frequency is 9.155 kHz !  Not 6 MHz of course.

Data sheet tell:
Arbitrary Wave characteristics
Frequency min 1uHz and  max 6MHz.

But just ago I show that max is tiny bit over 9kHz.
Yes, this IS maximum  arbitrary waveform repeating frequency in SDG1000X model when memory length is fixed and sampling clock is fixed. 16k memory, 150M sampling speed)

Do they lie.
No.

How it is done. (now this is extrmely simplified)

What happen now if you set ARB speed for 6MHz aka 6M awfm/s.
With 150MSa/s it can do. "Only" need jump over many many points in memory and take just 25 samples. (150/6=25)
Now if there are lot of details in waveform in ARB memory they are just ignored, just jump over like lazy fox jumps. We loose details.  (of course it is bit more complex because memory length is 16384 and divide by.... and so on... so there need some what then can see as one clock jitter in final result)

Wait a minute, somewhere I just read that someone have told to my friend who tell to me that ARB can produce even 30MHz  or over sinewave, not so nice looking but still somehow nice wave.

Think other way previously told things...

Still bit rounding corners and simplified.
Just told that using 6MHz ARB max freg can read 25 data points when 150Msa/s clock.
Ok. Lets draw 10 cycle sinewave in ARB memory. Then run it with 6MHz ARB freq. Now it can read again 25 samples, 2.5 sample per one cycle.
Yes there is some filters and result may look some kind of "sine".  But if you  you run it using 3MHz or  draw 5 cycles sine and run it using 6MHz you get 30MHz out... now it looks bit better... and if you change frequency, with some freq it can even look quite nice... as also can do with 10 cycle sine in arb memory.

If we need every position in arb memory is used and forwarded to DAC,  SDG1000X can max up to 9.155kHz ARB frequency when memory length is 16k andsampling is 150Msa/s.  This is hard fact.
Every AWG have this limit as long as it is fixed clock with some memory length DDS

After we accept some memory positions loose (jump over) then it can repeat more awfm/s (higher ARB freq).

Siglent have selected max limit as 6MHz what is ok for very simple low detail ARB waveforms. This selection for limit is not based to some truth how it must be. If Siglent set limit to 1MHz all are screaming that neighbourg have higher ARB freq. Also it can set for 10MHz, even more, it can set even for 30MHz. If it is 30MHz all are crying it can not repeat acceptable nearly any other wfm than nearly sinewave or square. 6MHz is just compromise and nothing more, it is not some kind of technical fact limit. For simple awfm it is ok and for complex awfm with dense details it is far too fast for details.
User is master and just user need know his equipment and how to use it for his application. This is why user need know basic simple fundamentals and some principles how just his instruments works. Except if he is using these only for playing fun for loose free times.

Depending user selected ARB freq it can also vary what memory position is jump over and some other things so that average frequency what user have selected is correct. So mostly we can see 6.666... ns jitter in edges. But it is difficult in practice predict what detail is loosed exactly until you are working only up to trusted 9kHz ARB freq.

For roughly same result, if we reduce ARB memory to 8k  we can rise it to 12MHz and if we reduce memory to 4k we can set limit to 24MHz and result is same. So if you see noname generator what have same clock speed and example you see there higher ARB freq, think carefully what it mean. Example FY6900 have 250Msa/s, 8k memory and 10MHz ARB max. With its max ARB freq it can read same 25 data points from ARB memory as Siglent and all other data points are jump over, where jump over details and variations is depending freq adjustment and producing one clock, more or less semi random, jitter -  in FY6xxx 4ns and in Siglent 1000X 6.7ns.

In Siglent TrueArb (and many others similar mode with different names) mode there is not this one clock jitter due to variable sampling clock and ARB memory length.

Originally SDG1000X series have only fixed length fixed clock DDS mode.
Not anymore.

Some time ago Siglent add "True Arb" mode also in SDG1000X series. It was software update, just not remember FW version when it happen.

In "TrueArb" mode things are very different.
Now we do not need  jump over and periodically change data point what we move to DAC for keep average frequency.
In this mode there is three (main) things what change.  One is of course ARB freq as also in DDS mode. But now also memory reading clock change and more, also memory lenghth can change. When stay in its limits user can be sure no detail is loosed *no jump over data points) and no one clock period jumping depending ARB freq.
But this model TrueArb have limits. 1u - 30Msa/s  and 2 - 16384 pts. sampling speed is always limited to speed where do not need jump over memory position so user know that all details are included. In this mode cycle to cycle (awfm to awfm) jitter is specified as 2 pts,  300ps RMS @ 20.1MSa/s  when direct DDS mode have specified as 6.7ns peak to peak

There is example earth quake wfm in ARB...  who can think 6MHz repeating this one quake period where is lot of details is somehow clever... as also ECG as also lot of other wfms including also sinc.  These just do not work with maximum ARB freq 6MHz.  Users need know even some basics about things.

It is good question to ask for what purpose users typically use or need ARB and with what speed and other paramaters and I do not mean what is nice to have for some niche rare use or just for "nice to have" because some others have.
Total 196 pre programmed wfm in memory. Just for checkbox feature in sales brochure and nothing else in real life. Of course nice for just playing fun. Also with other tools can play fun, even with hammer and saw. Some play fun with nintendo. Some play fun with scope and other instruments.
And because neighbour company have more perhaps other need also add more. Perhaps some year we can see some company advertise they have 10000 ready made ARB wfm in memory, ready for use. Even 196 is some difficult. You need lot of time to find if some there meet your needs. How you select them. In menu they have some names. Some wfm names do not tell nearly anything and you need just select and look. How often you want do it, finding if there is something what you just now need. Who need these.
« Last Edit: October 24, 2020, 08:31:17 am by rf-loop »
If practice and theory is not equal it tells that used application of theory is wrong or the theory itself is wrong.
-
Harmony OS

The following users thanked this post: TurboTom, 2N3055, ahope, Johnny B Good

#### Johnny B Good

• Frequent Contributor
• Posts: 443
• Country:
##### Re: Siglent SDG1000X Waveform Generators
« Reply #154 on: October 27, 2020, 11:48:52 pm »
@rf-loop

Thank you very much for that clear and concise explanation. I suppose it really all boils down to the frequency content embedded within the awfm sample and the maximum speed limit on clocking all the vital data points out if all 16K of them need to be preserved.

This reminds me of the quote "It's turtles all the way down!" often used to describe the cosmological regression problem.

https://en.wikipedia.org/wiki/Turtles_all_the_way_down

In this case, I've thought up a rephrasing of the turtles one, "It's sine waves all the way up!"

That's true of "perfect square waves" of course and nearly true for many other complex awfms. In short, the limitation is down to the complexity of the awfm generating a complex mix of sine waves - every wave form can be analysed into a mix, complex or simple, of sine waves. That's just the way things work in the real world.

With that 'fact of life' in mind, it's little wonder that the higher specification AWG kit becomes so very expensive so very bleedin' fast.

John

#### munxx

• Newbie
• Posts: 1
• Country:
##### Re: Siglent SDG1000X Waveform Generators
« Reply #155 on: November 03, 2020, 06:18:34 pm »
i have downloaded sdg1000x ads file but i am unable to flash it to my AWG...are there any additional steps to use it?

Hi all,

I had the same issue with the error message after mounting with rw privilege (I did read the instructions carefully).

I tried to reflash version 1.01.01.33R1B6 but bricked it. Probably wasnt a good idea to have the telnet ADS in the root of the thumbdrive used to do the reflash.

So I had to shamefully contact Siglent Europe who answered me on the same day, a sunday (!) and got back to me with the u-boot repair kit the next morning. Nice experience.

This you can not flash at all. You do all like FW update with this file and SDG tell that it fails... do not care...do not shut off... just telnet and do what ever inside SDG system. If you do not know what to do there inside system and how... then please study first...yes it may take time to find all or if you find right place there can perhaps find all very detailed full instructions (it is like spring festival gift package what include all you need) what can use like blind (after you have telnet working) but be careful... it do not forgive any kind of typemistakes.   Do not try..oops...try...oops iteration... because this method may lead you to real problems and nearly like "destroy" your SDG

And seriously, do NOT downgrade to semidanger factory original or "eevblog version" FW ....01P22 even if you find it somewhere because if bad happen and it goes to boot problem... it do not support USB recovery method.
does anyone tried the mod SDG1000X-MOD-Part-II in order to change maximum frequencies of waveforms. I can't remount root partition in rw to change NSP_limit_data.xml. I have error on:
/ # mount -o remount,rw /
/ # cd /usr/bin/siglent/config

Hi all,

I had the same issue with the error message after mounting with rw privilege (I did read the instructions carefully).

I tried to reflash version 1.01.01.33R1B6 but bricked it. Probably wasnt a good idea to have the telnet ADS in the root of the thumbdrive used to do the reflash.

So I had to shamefully contact Siglent Europe who answered me on the same day, a sunday (!) and got back to me with the u-boot repair kit the next morning. Nice experience.

The following users thanked this post: rf-loop

Smf