Show us your square wave

[Gerhard_dk4xp]

<    https://kh6htv.com/pspl-app-notes/   >

Most interesting reading.  When the founder/owner of Pico Secod Puls Labs retired,
he sold the farm to Tek.
First thing that TEK did was to axe the app notes from the PSPL web site.
Probably too many independent comparisons.   

Gerhard

[the Chris]

My pulse generator from Leo Bodnar arrived yesterday, so I would like to contribute with my slow-but-I-am-loving-it EDUX1052A. (...)

I wonder where that aberration came from; it is much greater than I would expect considering the much faster settling time of the pulse generator compared to a 60 MHz oscilloscope. My guess is that the oscilloscope does *not* have a Gaussian or single pole response.

Yes, I did not expect to see this kind of overshoot on a scope with 50MHz bandwidth. For reference, this is the same pulser on the EXR104 I am using at work. This time without inline terminator in native 50R mode.

[David Hess]

Yes, I did not expect to see this kind of overshoot on a scope with 50MHz bandwidth. For reference, this is the same pulser on the EXR104 I am using at work. This time without inline terminator in native 50R mode.

Now you get to ask yourself why there is preshoot.

[joeqsmith]

GigaWave with Tektronix S-52 tunnel diode head (<25ps Tr)

https://www.sjl-instruments.com/

[hpw]

GigaWave with Tektronix S-52 tunnel diode head (<25ps Tr)

https://www.sjl-instruments.com/

Nice, what I would like to see the real time behavior, as 20..80% or 10..90% rise time changes over time

On a LeCroy easy to graph over time as rise f(t).

[SJL-Instruments]

GigaWave with Tektronix S-52 tunnel diode head (<25ps Tr)

https://www.sjl-instruments.com/

Nice, what I would like to see the real time behavior, as 20..80% or 10..90% rise time changes over time

On a LeCroy easy to graph over time as rise f(t).

Thanks for the suggestion  We'll add the ability to graph any measurement vs. time to the software. (Fastest update rate would be 1 datapoint per sweep.)

[Aldo22]

- kind of curious to see what is the lowest price device is that can produce a really good looking square wave at 1, 5, 10, 20, and 30 MHz?

A candidate for “lowest price” (~$150 scope with AWG).
50MHz from the built-in AWG of the DSO2000.
AWF with coax and 50Ω LR.

[BillyO]

Lowest cost candidate: Home made fast risetime oscillator

Cost: Less than $10
Risetime: 1.36nS  (actually more like 1nS taking into account the risetime of the scope)

Shown on a 600MHz scope (to show all the ringing).  It looks really nice on a 200MHz or 100MHz scope!

[David Hess]

Shown on a 600MHz scope (to show all the ringing).  It looks really nice on a 200MHz or 100MHz scope!

You should reconsider your oscilloscope choice if you want to see the ringing.  That is not ringing unless ringing can happen *before* the edge.

I can explain it, but I would like to see what others think is going on.

[BillyO]

Well, it is "ringing" before the edge because of the propagation skew between the different inverters?


Edit:
But, wait, you don't think it's because of that, do you?  That's good because like others that have suggested that, you'd be wrong.

No, you think it's being caused by the scope.  Aliasing because, in your opinion, it's a "shitty Chinese scope", right?  Well, if that's what you think then you are still wrong.

If that's not what you think, then I'd love to hear what you think it is.

[MarkL]

I think there might be a small amount of actual ringing after the edge because the ringing before and after the edge is not perfectly symmetric wrt to the edge, but most of it looks like Gibbs artifacts to me.

  https://www.signalintegrityjournal.com/articles/2175-how-to-avoid-gibbs-ringing-artifacts-in-measurements

[BillyO]

I think there might be a small amount of actual ringing after the edge because the ringing before and after the edge is not perfectly symmetric wrt to the edge, but most of it looks like Gibbs artifacts to me.

  https://www.signalintegrityjournal.com/articles/2175-how-to-avoid-gibbs-ringing-artifacts-in-measurements

Good guess, but not in this case.  With Gibbs artifacts, the effect get's worse with faster rise times.  This screen shot is the same scope with the same Bodnar pulser as was used in that article.

[vk6zgo]

Well, it is "ringing" before the edge because of the propagation skew between the different inverters?


Edit:
But, wait, you don't think it's because of that, do you?  That's good because like others that have suggested that, you'd be wrong.

No, you think it's being caused by the scope.  Aliasing because, in your opinion, it's a "shitty Chinese scope", right?  Well, if that's what you think then you are still wrong.

If that's not what you think, then I'd love to hear what you think it is.

A square wave can be regarded as a fundamental & all the odd harmonics of that fundamental.
Real world networks suffer from the phenomenon of "Group Delay", where the fundamental & harmonics are displaced slightly in time.

[BillyO]

A square wave can be regarded as a fundamental & all the odd harmonics of that fundamental.
Real world networks suffer from the phenomenon of "Group Delay", where the fundamental & harmonics are displaced slightly in time.

I'm not sure group delay would produce what we are seeing in the pink trace.  If somehow it can, then it's all happening within the 74AC14 used to create the waveform.

[T3sl4co1l]

Looks like sinc interpolation to me -- zoom in to better show the edge (and hopefully get a faster [equivalent, if applicable] sampling rate), and try with and without interpolation, or show vectors/dots.

Ed: note that the sample rate is only 2G/s: for an edge rate ~1ns, what does this suggest about the number of points in the rise itself?

Tim

[BillyO]

Looks like sinc interpolation to me..

Hmm, okay but I'd like to draw you attention to the traces I posted in reply #486.  Same sampling rate and a pulser with an risetime of < 40ps.  No "ring" before the edge.

[David Hess]

A square wave can be regarded as a fundamental & all the odd harmonics of that fundamental.
Real world networks suffer from the phenomenon of "Group Delay", where the fundamental & harmonics are displaced slightly in time.

Some analog oscilloscopes suffer from that because of their delay line.  The delay line produces group delay which would limit performance except that a phase compensation network is included to remove it.  The faster Tektronix 24xx series of analog oscilloscopes will show a small amount of preshoot when a fast enough edge is applied.

I'm not sure group delay would produce what we are seeing in the pink trace.  If somehow it can, then it's all happening within the 74AC14 used to create the waveform.

Group delay could do it, but I do not know how a simple 74AC14 oscillator could produce group delay.  If a bunch of sections are in parallel for higher speed, could some sections be switching before the others?  Maybe the paths connecting the sections are too unequal?

I have used parallel AC gates before and never had that happen, but I do not remember specifically using AC Schmitt triggers.  I have used HC and other Schmitt triggers that way, but they must be too slow to create this problem.

Looks like sinc interpolation to me -- zoom in to better show the edge (and hopefully get a faster [equivalent, if applicable] sampling rate), and try with and without interpolation, or show vectors/dots.

It does to me also, although I am fuzzy as to how this would differ from the Gibbs phenomena when the number of samples is limited.  They both seems to describe the same problem.

When I was evaluating the now old Tektronix MSO5000 series with a 350 picosecond edge source, which was significantly slower than the oscilloscope bandwidth, I could produce exactly this type of display by enabling the software bandwidth limiting, which lead me directly to the Gibbs phenomena.  The Tektronix application engineers supervising us could not explain it either.  The hardware bandwidth limiting with the same cutoff frequency produced an analog type display without preshoot.  If my understanding is correct, then Sinc interpolation should do exactly the same thing when the Nyquist frequency is not far above the edge frequency.

[magic]

What's the frequency of this ringing, 300MHz or so? Note the slow timebase of 50ns. I think it's too slow for filter artifact on a 600MHz DSO.

It should be genuine ringing and I have seen enough of it in similar generators that I'm surprised this one performs so well, despite being built with THT components. What chip is this? You expect there to be ringing because the chip is a capacitor connected to the power supply through inductors and it draws sharp pulses of current which excite the resonance.

The "pre-ringing" is probably ringing caused by the oscillator gate switching, and then you see the edge one t_pd later followed by more violent ringing caused by the line driving gate(s) switching.

These generators perform better if built with SMD components.

edit
case in point: SOT23 package and a 400MHz analog scope:
https://www.mikrocontroller.net/topic/217945#2205244

[David Hess]

What's the frequency of this ringing, 300MHz or so? Note the slow timebase of 50ns. I think it's too slow for filter artifact on a 600MHz DSO.

Both examples are at 2 GS/s, which apparently is the limit for this DSO despite the high claimed bandwidth.  It looks like the number of listed points is correct for the time/div setting, if that is what points means.  I am dubious about the bandwidth claim until I see it measured at different sensitivities with different amplitudes because this may be another case of a modern DSO variable full power bandwidth limited by slew rate.

It would help to see both measurements made at the same time/div.  Could the DSO filtering be executed on the display record, so changing the time/div would do all kinds of weird things with a fast transition time edge?

It should be genuine ringing and I have seen enough of it in similar generators that I'm surprised this one performs so well, despite being built with THT components. What chip is this? You expect there to be ringing because the chip is a capacitor connected to the power supply through inductors and it draws sharp pulses of current which excite the resonance.

The "pre-ringing" is probably ringing caused by the oscillator gate switching, and then you see the edge one t_pd later followed by more violent ringing caused by the line driving gate(s) switching.

Expanding the display of the preshoot should help.  If the preshoot starts by a gate delay, then maybe it is the ringing of the first stage coupled into the supply of the output buffer?  I do not see any output series resistors but they could be on the other side of the printed circuit board.

These generators perform better if built with SMD components.

The construction looks pretty good, but I agree.  I would not use that circuit at all if I did not want ringing, although it does fine with slower parts like HC, AS, and FAST.  AS and FAST have the advantage over CMOS of faster operation with slower edge rates so they had fewer problems with construction.

[BillyO]

Okay, here are some more shots of the home made pulser.

I have used the same channel are with the Bodnar to eliminate that as a variable.


Same as the pink trace




Same as the Bodnar




Detail.  Seems the oscillation is @ approximately 400MHz

[BillyO]

A few slides from the sweep used to determine BW.  The generator is not mine, so I can't easily do this at multiple sensitivities.  However, I am not the only one that has found this.  The scope itself is spec'd to be 500MHz, but like most of these goes a little beyond that.  Sweep started @ 600mV  (6 divisions).

These results are supported by the risetime found using the Bodnar.

[BillyO]

Here are some more picture of the home made pulser, the schematic and the board layout.

[Dave_g8]

Hi Bill,
As previously mentioned, I suspect the noise may be from other elements switching in the 74AC14 chip.
U1 Pin 2 and U1 Pin 12 will be changing just before the final output.
I am not sure if you are using a Schmitt type device, if not, the slow edge on U1 Pin 1 may also lead to high frequency oscillations as the signal passes through the threshold voltage.
Regards, Dave

[BillyO]

Here is sequence through the circuit.  Probes used are Tektronix P6205s through an 1103 power supply.  Yellow trace is directly connected via an adapter to the output of the pulser.

The blue trace is used to probe various points using a spring ground.

The first screen shot is at pin 2.  The output of the oscillator.

The second is at the output of the buffer (pin 12)

The 3rd is at the output of one of the drive inverters before the series resistor (pin 6)

You can see how the anomaly builds as the signal goes through the system.

[BillyO]

Hi Bill,
As previously mentioned, I suspect the noise may be from other elements switching in the 74AC14 chip.
U1 Pin 2 and U1 Pin 12 will be changing just before the final output.
I am not sure if you are using a Schmitt type device, if not, the slow edge on U1 Pin 1 may also lead to high frequency oscillations as the signal passes through the threshold voltage.
Regards, Dave

Yes, the 74AC14 is a Schmitt hex inverter.  And I agree.  It's not a problem with the scope, or phase noise, group delay, sinc interpolation or Gibbs artifiacts.
 It's simply switching noise created in the chip itself.