Show us your square wave

[tggzzz]

Btw., what is the output impedance of those G14s?

If by G14 you mean LVC1G14, then for the purposes of generating a square wave, start with 7ohms at Vcc=5V

[iMo]

Yep, asking as people are attaching resistors (to the G14s outputs) of round multiplies of 50ohm.
Also, what about frequency compensation of the output due to the scope's input capacitance?
Like around 12pF in parallel with the output resistors, or something (depends on the input capacitance) like that?
 

[rolfdegen]

My ATtiny_Pulser V3.0

The new circuit is soldered and finished. I had hoped for a slightly better result. But the rise time remained the same as in the old circuit. I measure a RiseTime of 780-800ps on my Rigol 5104 (350MHz hack).

RiseTime


New ATtinyPulser_3.0 circuit

[Martin72]

I measure a RiseTime of 780-800ps on my Rigol 5104 (350MHz hack).

At some point, the bandwidth will be exhausted, preventing faster rise times from being measured/displayed.

[rolfdegen]

Yes, I believe that too. If you use the measured 800ps to calculate the scope's cutoff frequency, the formula is 0.35/800ps = 437MHz. That's more than the stated 350MHz 
Leo Bodner's pulse generator should then produce similar results on my Rigol 5104 (350 MHz hack). I'm curious...

[Martin72]

Bodnar Pulser on MSO5000:

https://www.eevblog.com/forum/blog/new-rigol-scope/msg2446644/#msg2446644

750ps...

[rolfdegen]

Yep, asking as people are attaching resistors (to the G14s outputs) of round multiplies of 50ohm.
Also, what about frequency compensation of the output due to the scope's input capacitance?
Like around 12pF in parallel with the output resistors, or something (depends on the input capacitance) like that?

MSO5000 series has..
Input Impedance 1 MΩ±1% and input Capacitance 17 pF ± 3 pF
I soldered a 22pF smd capacitor to the pulser output against GND. Here's the measurement result..

ATtinyPulser_3.0 with 22pF to GND on the output. Yellow line with 22pF. Red without cap

[tggzzz]

Yep, asking as people are attaching resistors (to the G14s outputs) of round multiplies of 50ohm.
Also, what about frequency compensation of the output due to the scope's input capacitance?
Like around 12pF in parallel with the output resistors, or something (depends on the input capacitance) like that?

Before you think about "frequency compensation", you have to decide and define your objectives.

[iMo]

With the output resistor(s) you create a low pass.. So you have to compensate that somehow.. Perhaps the parasitic capacitance is enough already in Rolf's design above so it works better without.

[Renaud F5ZR]

My show of the LA314 rise time above was more for the use of the HMC363S8G to generate 100 pS transitions cheaply.
I bought it around 20€ with Mouser. Of course it's not Leo Bodnar!
Renaud

[iMo]

..ATtinyPulser_3.0 with 22pF to GND on the output. Yellow line with 22pF. Red without cap
(Attachment Link)

If you want to experiment - I would put a "gimmick" capacitor in parallel with the output resistors and slowly decrease its capacitance while watching the rising edge.
Gimmick capacitor - for example two twisted 0.15-0.2mm diameter enameled wires, twisted in a length of aprox 15mm == aprox 15pF. Decreasing its capacitance with cutting off its twisted end (while your pulser is switched off).

[rolfdegen]

Thanks for your tip  It is difficult to solder on my circuit board. I think I'll wait for the Leo Bodner Pulser and use it to measure my rise time.

[tggzzz]

Before fettling too much, what are you trying to demonstrate?
A fast edge?
The scope's response to a fast edge?
Sub-optimalities of different construction techniques.

If you just want a fast edge, consider using tunnel diodes or step recovery diodes plus resonant circuits

[gf]

I do have an Adalm Pluto SDR that can be used as an improvised spectrum analyzer up to 6GHz, but it is still not high enough I guess.

With a cut-off at 6GHz, you won't get estimates less than roughly 100ps, even if the true edge is faster.

The idea was to look at the spectrum of the pulses with the SA, and to write down the amplitude of each harmonic.

With 10MHz square you had to write down 300 amplitudes of odd harmonics up to 6Ghz - that's hard work if you cannot automate it

Eventually, starting from the amplitude of each harmonic, it should be possible to apply a reverse FFT in order to reconstruct the time domain pulses, and measure the edges raise/falling time on the reconstructed signal in the time domain (TD reconstruction being possible by assuming the phase - which is not measured by the SA - is always according to an input signal with perfect edges).

I tried that (using simulated ideal SA measurements), but without phase it is not as easy as you think. Eventually I got a quite nice reconstruction if the square wave has a rational duty cycle M/N, where M and N are positive integers, M < N, N >= 2, and N is not too large (where "not too large" depends on the slowest rise time you want to measure and the square wave frequency). With these rational duty cycles, I was able to extract a single impulse response from the time domain signal. But for some (arbitrary) duty cycle values, the impulse responses overlap, making separation in the time domain impossible. Smoothing with a window function helps to some extent (but increases the estimate rise time). Using a lower square-wave frequency should also reduce the set of duty cycles that don't work well (but at 1MHz you'd have to write down 3000 measurements and you'd need 75dB dynamic range). I have not yet investigated the sensitivity to noise or measurement errors.

EDIT:

Addedd example diagrams of my reconstruction attempts
1) when the duty cycle happens to be a "good" one
2) and for comparison, with a "bad" duty cycle
The simulated "ground truth" edge was shaped with a 3rd-order ellipic filter with 1Ghz cutoff.

[ The horizontal offset between ground truth and reconstructed edge seems to be a consequence of the extra cutoff at 6Ghz (since you can't provide measurements beyond 6Ghz). It becomes much smaller when I keep the frequency content >= 6Ghz. I don't consider it a big problem, as it does not affect the shape very much.  ]

Btw, if you want to continue this discussion, I also suggest to open a separate thread, as it related, but not really on-topic either.

[rolfdegen]

Hello everyone. I want to measure the jitter of my two ATtiny_Pulsers with the MSO5104. One ATtiny_Pulser has a crystal oscillator, the other a chip oscillator. How exactly do I do this?

[tggzzz]

Hello everyone. I want to measure the jitter of my two ATtiny_Pulsers with the MSO5104. One ATtiny_Pulser has a crystal oscillator, the other a chip oscillator. How exactly do I do this?

Can I suggest starting a separate thread on that topic. Many people ignore this thread, but some might be interested in that different topic.

[Martin72]

Today, my BNC to BNC adapters (plug type) arrived, providing a good opportunity to test them on the demo board.
The pictures show a “direct” connection with the adapter, then a classic connection via a 1 m long BNC cable, and finally with a 1:10 1M probe head.

[rolfdegen]

Leo Bodnar's pulser arrived today. I took a few comparison measurements with my pulser and Leo's pulser...
MSO5104 scope with 350MHz hack and 50R termination at the scope input

Leo Bodner Pulser RiseTime


RiseTime ATtiny_Pulser 3.0


Leo Bodner Pulser Eye and Jitter


ATtiny_Pulser 3.0 Eye and Jitter

[RoGeorge]

Nice! 
Thanks for the comparison of the results.

(So, the undershoot right before the raising edge was not a digital artifact of the oscilloscope, after all.  The undershoot is not visible with the Leo's pulser, I was wrong blaming the oscilloscope and the sinx/x )

[gf]

(So, the undershoot right before the raising edge was not a digital artifact of the oscilloscope, after all.  The undershoot is not visible with the Leo's pulser, I was wrong blaming the oscilloscope and the sinx/x )

Since the sample rate is > 15x the analog bandwidth, it is unlikely that the sin(x)/x interpolation will become the bandwidth-limiting factor in the end. The high sample rate (and high sample rate to bandwidth ratio) is the strength of the MSO5000.

[Andie]

No squares, only rising edges from various generators, shown by various sampling oscilloscopes. Starting with a follow-up to this project from several years ago: https://www.eevblog.com/forum/projects/mechanical-vs-avalanche-pulse-generator/msg1013573/#msg1013573

The avalanche generator evolved and now outputs an edge with a rise time of about 55ps, shown by an IBZ Electronics DS800 sampling oscilloscope:




Experimenting with russian tunnel diodes was somewhat successful, but then I got a Tektronix 152-0177-01 tunnel diode, with which the edges got even faster (290ps):





Later I bought a Tektronix 284 tunnel diode pulse generator, which initially wasn't working. I found out that bending the output connector changed the behaviour. After firstly unsuccessful troubleshooting, I disassembled the airline assembly completely, cleaned every part, while being almost sure, that afterwards this thing will never output anything again. But it worked exactly as before! Nothing changed. After one or two more disassemblies I am quite sure, that the tunnel diode has some mechanical issue, which requires permanent mechanical stress in a specific direction. After some tweaking I got the airline assembly to work reliably, but the whole device must be handled very carefully. The edge (rise time about 74ps):




In 2018 I ordered a Leo Bodnar 30ps pulser and tried it on the Tektronix CSA803C with SD-26 sampling head (20GHz). The oscilloscope measured a rise time of 34.5ps:




Using a SD-32 sampling head (50GHz) measured a rise time of 29.4ps:




Measuring the Leo Bodnar 30ps pulser with the IBZ Electronics DS800 oscilloscope resulted in a rise time of 34.6ps:

[MF-jockey]

Inspired by this thread, I also wanted a fast pulse generator. I liked the circuit by user tggzzz with parallel 74LVC1G14 and I acquired ten 74LVC1G14 integrated circuits.

Using a CAD program, I created a design with the smallest possible length differences between the three signal paths at the output (I calculated approximately 3.5 mm, e.g. Folnia's design has length differences of 15 mm, which means runtime differences of 45 ps). Then I milled the traces by hand into a piece of copper-clad material.
Because I wired the three inputs, the ground plane remained largely closed and I was able to solder the capacitors on the side (not on top).

After soldering, a correction with an additional resistor at the generator was necessary to ensure stable operation.

A test on my 100 MHz oscilloscope (I didn't have a faster one in 2024) yielded a rise time of 1.56 ns for the DS7102V, corresponding to a cutoff frequency of 224 MHz. I was naturally pleased with this, but I didn't know the rise time of my generator (I estimated around 270 ps).

Now, in a comparative measurement with a 150ps-generator, I determined rise time values ​​of 218-251ps.

[BillyO]

Now, in a comparative measurement with a 150ps-generator, I determined rise time values ​​of 218-251ps.

But the scope measurement function is showing t_r> 650ps. 

Is the scope worng?

[MF-jockey]

No, the scope is too slow with its cutoff frequency of 635 MHz and distorts the pulse.

For more precise measurements, an oscilloscope with a bandwidth of several GHz is needed.

[MF-jockey]

I recently acquired a used Siglent SDS5034X. To verify its cutoff frequency during an upgrade, I needed a reliably faster pulse generator. A steep-edged generator was also desired for the high-frequency calibration of the probes. A Bodnar pulser was too expensive for me, so I ordered a "150ps risetime pulse generator" with a 10 MHz clock for under €40. Unfortunately, I couldn't find any information about the driver used, a 6-pin SMD circuit.

I soldered suitable contacts for probe tips directly onto the generator.



A test directly at the inputs of the SDS5034X (after the upgrade) yielded the expected rise times. I determined a cutoff frequency of 584 MHz for C1, assuming a 150ps risetime for the generator. A sweep using LiteVNA even yielded cutoff frequencies of 635 MHz for all channels.

The 150ps generator is sensitive to capacitive loads. When loaded with a P7300 probe and simultaneously connected to the SDS5054X, the pulse was significantly distorted, and to some extent also with the P2200. This hardly occurred with my other probes (P6169A, P6500B, PP510, SP3050A, T3100, TT-HV-150).

My homemade generator similarly collapses under high capacitive loads.