Function generator square wave rise time

[MarkusAJ]

An FY6900 only needs <7ns and the one built into my cheap Hantek scope <5ns (screenshot, 20MHz).
No idea why this is the case.

Hmm, that makes you think. What kind of FY6900 do you have, I take it the 100MHz version?

The screenshot is from the Hantek DSO2000 AWG.
I don't have the FY6900, the 7ns is from the specifications.

I just wanted to give two examples, that rise times do not seem to be directly related to price or overall quality.

@hp3310a
I have the FY6900 60 MHz, I bought mine in 2020 for $120 from Amazon, it is branded "DOMINTY".
I made for you few measurements using the Siglent 2504X+ scope and attached screenshots of 100Hz, 1kHz, 10kHz, 100kHz, 1MHz, 10MHz, 20MHz, and 25MHz  2.5V(p-2-p) square waves.

The FY6900 was connected to the scope with 1ft. RG58 cable and input was 50Ohm terminated.
Scope setup: sin(x)/x interpolation OFF, acquisition: normal, bandwidth limit: OFF.

I attached also screenshot with FFT of 1kHz sinusoidal wave, it will give some information about harmonic distortions.
FFT: window: frequency span: 500Hz - 30kHz, flattop, length: 2Mpts, ERES 3 bits enchantment.

Edited: I just noticed that mistakenly I connected the FFT (F2) directly to C1 and not to the ERES3  (F1), sorry.

[edpalmer42]

Now, since you are a timenut, what bandwidth is good enough for the 1pps/1Hz output from a GPSDO?

5Hz? 10Hz? 100Hz? Surely the 99th harmonic will look good enough. Won't it?

No, for a Time-Nut, there is no limit.  I want all the bandwidth!  But, for this application, it makes more sense to look at the time domain, i.e. risetime rather than the frequency domain, i.e. bandwidth.  To get the best measurements, you need to minimize jitter which means the fastest possible risetime.  Is that what you wanted me to say? 

Your discussion with bdunham7 brought up a few questions that caused me to make some measurements.  The results are interesting.

1.  Start with a Trimble Thunderbolt GPSDO.  Risetime on the 1 PPS signal is ~2 ns as measured on a 1 GHz LeCroy 9384L scope.
2.  Use a Fluke PM6681 Timer/Counter/Analyzer to make 100 measurements of the period of the Thunderbolt 1 PPS signal and then display the Std. Dev. of the measurements.  The PM6681 has a resolution of 50 ps.
3.  Make the measurements with and without the PM6681's 100 KHz low pass input filter.  With the filter:  ~0.5 ns.  Without the filter:  ~0.1 ns.

So even a 100 KHz filter results in unacceptable degradation of the measurements.

To summarize, for some situations, LIKE THE OP's INITIAL QUESTION, it's easier to understand what's happening by using the frequency domain i.e. bandwidth.  For other situations, like the measurements above, the time domain offers better clarity.

[BillyO]

The FY6900 was connected to the scope with 1ft. RG58 cable and input was 50Ohm terminated.

<5ns is pretty impressive.  I was deciding between the FY6900 and the PSG90XX and chose the PSG9080.  It's risetime measures around 9ns, so not a killer WRT square waves, but sufficient for what I need this unit for considering I have better AWGs.  I mainly chose it for the UI.  The keypad was the deciding factor.

But, yeah, <5ns is very nice for the price you paid.

[bdunham7]

I have the FY6900 60 MHz, I bought mine in 2020 for $120 from Amazon, it is branded "DOMINTY".
I made for you few measurements using the Siglent 2504X+ scope and attached screenshots of 100Hz, 1kHz, 10kHz, 100kHz, 1MHz, 10MHz, 20MHz, and 25MHz  2.5V(p-2-p) square waves.

One of the issues with the FY6900 and its ilk is that if the frequency you select is not an even submultiple of the sample clock, then it will use a modified edge to get the middle of the transistion to be at the correct time.  It does this in an ad hoc manner, not consistently like better AWG designs.  This results in mostly faster but inconsistent rise times between different periods.  You can see this in your numbers, even though the pictures look good--at least the low frequency ones. 

Here is a rise and fall time screenshot of a 10.0001kHz square wave selected on an FY6900.  Note that the displayed frequency is off a bit, the FY6900 clock isn't all that great.

[bdunham7]

So even a 100 KHz filter results in unacceptable degradation of the measurements.

That's what I was driving at--a slowed rise time measured in microseconds gives you an error measured in picoseconds.  But we've gone far enough off topic, IMO.

[hp3310a]

@hp3310a
I have the FY6900 60 MHz, I bought mine in 2020 for $120 from Amazon, it is branded "DOMINTY".
I made for you few measurements using the Siglent 2504X+ scope and attached screenshots of 100Hz, 1kHz, 10kHz, 100kHz, 1MHz, 10MHz, 20MHz, and 25MHz  2.5V(p-2-p) square waves.

The FY6900 was connected to the scope with 1ft. RG58 cable and input was 50Ohm terminated.
Scope setup: sin(x)/x interpolation OFF, acquisition: normal, bandwidth limit: OFF.

I attached also screenshot with FFT of 1kHz sinusoidal wave, it will give some information about harmonic distortions.
FFT: window: frequency span: 500Hz - 30kHz, flattop, length: 2Mpts, ERES 3 bits enchantment.

Edited: I just noticed that mistakenly I connected the FFT (F2) directly to C1 and not to the ERES3  (F1), sorry.

Thanks a lot for taking the time to put this up.

One of the issues with the FY6900 and its ilk is that if the frequency you select is not an even submultiple of the sample clock, then it will use a modified edge to get the middle of the transistion to be at the correct time.  It does this in an ad hoc manner, not consistently like better AWG designs.  This results in mostly faster but inconsistent rise times between different periods.  You can see this in your numbers, even though the pictures look good--at least the low frequency ones. 

Here is a rise and fall time screenshot of a 10.0001kHz square wave selected on an FY6900.  Note that the displayed frequency is off a bit, the FY6900 clock isn't all that great.

Also thanks for that! Never in my wildest dreams would I think of things like that, so great to know.

This discussion is very interesting to follow. Turns out there are a lot more "moving parts" to consider in a frequency generator than I thought.

[tautech]

Turns out there are a lot more "moving parts" to consider in a frequency generator than I thought.

And uses !

Back when I fixed CRO's for a hobby having a pulse generator to use instead of a Mark space generator would have been very useful and when I got into this game my 1st but only half decent AWG was the quite ordinary SDG1010 with only a single higher amplitude output. 

Since then several have stuck for a while and now settled on an unhacked SDG6022X and a RF gen for any HF work.

However customer needs always interest me and the most recent unusual was the need to burst 3 short pulses at a specified rate with SDG1032X.
Had them bluffed but it was quite easy with the pulsing channel burst for 3 cycles and triggered at the required repetition rate by the 2nd channel where we left it self triggering whereas we could also manually or externally trigger it.
After the little usage lesson they were on their way wagging their tails like happy pooches. 

[Aldo22]

However customer needs always interest me and the most recent unusual was the need to burst 3 short pulses at a specified rate with SDG1032X.
Had them bluffed but it was quite easy with the pulsing channel burst for 3 cycles and triggered at the required repetition rate by the 2nd channel where we left it self triggering whereas we could also manually or externally trigger it.
After the little usage lesson they were on their way wagging their tails like happy pooches. 

Many people do not know the capabilities of their  (even very limited)  tools.
What you describe can also be done with a poor FY3200S, you just have to find out how to do it. 

[tggzzz]

Now, since you are a timenut, what bandwidth is good enough for the 1pps/1Hz output from a GPSDO?

5Hz? 10Hz? 100Hz? Surely the 99th harmonic will look good enough. Won't it?

No, for a Time-Nut, there is no limit.  I want all the bandwidth!  But, for this application, it makes more sense to look at the time domain, i.e. risetime rather than the frequency domain, i.e. bandwidth.  To get the best measurements, you need to minimize jitter which means the fastest possible risetime.  Is that what you wanted me to say? 

Your discussion with bdunham7 brought up a few questions that caused me to make some measurements.  The results are interesting.

1.  Start with a Trimble Thunderbolt GPSDO.  Risetime on the 1 PPS signal is ~2 ns as measured on a 1 GHz LeCroy 9384L scope.
2.  Use a Fluke PM6681 Timer/Counter/Analyzer to make 100 measurements of the period of the Thunderbolt 1 PPS signal and then display the Std. Dev. of the measurements.  The PM6681 has a resolution of 50 ps.
3.  Make the measurements with and without the PM6681's 100 KHz low pass input filter.  With the filter:  ~0.5 ns.  Without the filter:  ~0.1 ns.

So even a 100 KHz filter results in unacceptable degradation of the measurements.

To summarize, for some situations, LIKE THE OP's INITIAL QUESTION, it's easier to understand what's happening by using the frequency domain i.e. bandwidth.  For other situations, like the measurements above, the time domain offers better clarity.

Bingo! We've got there, finally.

Note that when my response[1] corrected your statement, there was no definition of "this application". I've forgotten whether the OP has subsequently added a specific application.

[1] reply #16 https://www.eevblog.com/forum/testgear/function-generator-square-wave-rise-time/msg5849241/#msg5849241

[DaneLaw]

@hp3310a: If you simply want to see higher frequency square waves, you can use an Si5351.
You can't set the amplitude, offset or duty cycle directly, but it might be enough for simple tests.
https://learn.adafruit.com/adafruit-si5351-clock-generator-breakout

Got a few of those..Si5153B with PLL, VXCO.  https://www.eevblog.com/forum/testgear/adjustable-8ch-clock-generator-2-5k-200mhz-0-5ppm-(si5351bpll)
Doesn't cost a lot (20 to 40$ incl. VAT) these TypeC units fitted with a small screen where you individually can adjust the variables on the 8 channels (2.5k to 200MHz) and 5-way phase difference ability, and 7 channel division reference-Hz-sett & clock  [some time since I looked at it]

[BillyO]

If the OP  needs REALY nice square waves then I'd suggest this: https://www.leobodnar.com/shop/index.php?main_page=product_info&cPath=124&products_id=406&zenid=7e46e2ec849efabc0612f5d4abf56be6

[joeqsmith]

...
Some sig gens may have a special square wave or pulse mode that has faster rise times, but it is going to be difficult to get a clean, crisp 20MHz square wave especially if you want a universal device that can give you arbitrary frequencies and periods.

... +  offset, amplitude, markers, slew rate ....

If OP just needs a fixed logic level, maybe just buffer the output.

[ozkarah]

Siglent SDG2000X series can do 8ns rise/fall time for square wave signal.

But, using ARB Mode (DDS - Duty 50) can go as low as 4ns. However, the signal is not a perfect square wave in DDS mode.

[bdunham7]

But, using ARB Mode (DDS - Duty 50) can go as low as 4ns. However, the signal is not a perfect square wave in DDS mode.

You'll also have the same DDS issue that the FY6900 has--unless you pick certain specific frequencies, you'll have a lot of jitter.  And even if you pick a magic frequency, you might still get strange occasional jitter due to how the clock correction works.

... +  offset, amplitude, markers, slew rate ....

If OP just needs a fixed logic level, maybe just buffer the output.

Yes, all those things too.  The OP seemed mostly interested in getting better performance out of an AWG, but clean fast rise times are tough to do when you expect to just be able to dial in anything from 2mV_p-p to 20V_p-p.  This shows 5GV/s or 5kV/µs slew rate. 

[tautech]

But, using ARB Mode (DDS - Duty 50) can go as low as 4ns. However, the signal is not a perfect square wave in DDS mode.

You'll also have the same DDS issue that the FY6900 has--unless you pick certain specific frequencies, you'll have a lot of jitter.  And even if you pick a magic frequency, you might still get strange occasional jitter due to how the clock correction works.

Is that FY6900 jitter ? 
Is anything 10 MHz referenced or just free running ?

[bdunham7]

Is that FY6900 jitter ? 
Is anything 10 MHz referenced or just free running ?

Sorry, I didn't make that clear!  That is the SDG2122X using the stored ARB Duty50 waveform at 9.9MHz.  Essentially the same setup as ozkarah posted except I changed the frequency from 10MHz to 9.9 and increased the amplitude to show the slew rate capability.  And I used a 10x probe so my scope/probe rise time is 1.5ns at best.

[gf]

Sorry, I didn't make that clear!  That is the SDG2122X using the stored ARB Duty50 waveform at 9.9MHz.

Using which TrueArb interpolation mode?
(0-order, linear, sinc, sinc27, sinc13)
Zero-order hold is, of course, expected to jitter and linear interpolation is not expected to provide the best image/alias rejection either. But does it also happen with the sincXX interpolation modes?

Edit: Looking at the datasheet, it seems that the SDG2000X does not support the sincXX modes, but only the 6000 and 7000 and the new 3000 models do?

Edit: Or do you talk about traditional DDS mode (not TrueArb)?

[bdunham7]

Using which TrueArb interpolation mode?

None of the above, this was DDS.  TrueArb only does 75MSa/s and the Duty50 stored arb waveform has 16384 points and that translates to ~4.5kHz and about a 13ns rise time.

[Mahagam]

And I used a 10x probe so my scope/probe rise time is 1.5ns at best.

Use BNC cable and 50 Ohm scope internal termination.

[gf]

Using which TrueArb interpolation mode?

None of the above, this was DDS.  TrueArb only does 75MSa/s

Oh, only 1/4 of the 300MSa/s DDS/DAC sample rate
[ Edit: To avoid confusion: I just mean the input sample rate of the DAC, not its oversampling rate. ]

I compared the SDG2000X/6000X/7000A datasheets, and interestingly it seems that only the SDG7000A supports TrueArb sample rates up to the full DAC sample rate.

Also, neither the SDG6000X datasheet nor the User Manual mentions sincXX interpolation, although there was apparently a firmware upgrade that introduced it. Since the docs seem to be outdated, I wonder if this SG6000X firmware upgrade also coincidentally introduced TrueArb sample rates up to 1.2Gsample/s? Does anybody know?

[ I speculate that the sincXX interpolation modes and the support of TrueArb sample rates up to the full DAC sample may be related, because as long as only linear interpolation is supported, I see a good reason for limiting the TrueArb sample rates. ]

Edit: According to
https://www.eevblog.com/forum/testgear/new-siglent-sdg3000x-arb-waveform-generator/?action=dlattach;attach=2515697
the forthcoming SDG3000X also appears to support both, sincXX interpolation modes and up to full 600MSa/s TrueArb sample rate.

[gf]

But, using ARB Mode (DDS - Duty 50) can go as low as 4ns. However, the signal is not a perfect square wave in DDS mode.

You'll also have the same DDS issue that the FY6900 has--unless you pick certain specific frequencies, you'll have a lot of jitter.  And even if you pick a magic frequency, you might still get strange occasional jitter due to how the clock correction works.

Try the attached custom DDS waveforms instead [ I hope I got the file format right ].
At frequencies <= 10MHz, the 2nd one should not jitter, and the first one should not jitter that much.