Making an oscilloscope is relatively easy, while making a very fast oscilloscope is hard. There’s a trick that converts a mundane instrument into a very fast one, it’s been around since the 1950s, and [CuriousMarc] has a video explaining it with an instrument from the 1960s. The diode sampler is the electronic equivalent of a stroboscope, capturing parts of multiple cycle of a waveform to give a much-slowed-down representation of it on the screen. How it works is both extremely simple, and also exceptionally clever as some genius-level high-speed tricks are used to push it to the limit. We’ve put the video below the break.
[Marc] has a Keysight 100 MHz ‘scope and the sampler allows him to use it to show 4 GHz. Inside the instrument is a pair of sample-and-hold circuits using fast diodes as RF switches, triggered by very low-rise-time short pulses. Clever tricks abound, such as using the diode pair to cancel out pulse leakage finding its way back to the source. To complete this black magic, an RF-tuned stub is utilized to help filter the pulses and further remove slower components.
It’s slightly amusing to note that the Keysight 100 MHz ‘scope is now “slow” while the early sampling ‘scopes had their “fast” capabilities in that range. The same technique is still used today, in fact, you probably have one on your bench.
The sampler he’s showing us is an accessory for another instrument we’ve previously shown you his work with.
Good stuff, Ive been looking at adding some AtoD on an existing instrument project
for electro-chemistry where the electrode noise spectra could be of interest as in
might betray some aspect useful re metallurgical properties. Timely post, thanks :-)
My Tek 1502 observes 50ps risetimes with a ~10kHz sampling rate. (yes, p and k). Lovely instrument; just sold another :)
My seriously weird and wonderful Philips PM3400 has a 200ps risetime (1.7GHz) and has signal path “audio” transistors (BC107, fT+200Mhz). There are three 2GHz fT transistors in the trigger.
For the next trick, get youngsters to think how they could create a divide-by-10 circuit using only 3 transistors. (They can cheat by looking at the Tek 184, which also has HV transformers made from bits of bent wire)
I’m curious about the answer, as I often deal with discrete frequency dividers for clock design. If this uses resonant circuits though, it’s cheating :-)
If the 184 is anything like it’s vacuum tube predecessor, the 181, it uses a monostable multivibrator with an RC circuit that keeps it from being triggered for a certain amount of time. If not adjusted correctly, it’s a divide by 9 or divide by 11 counter. So yeah, it’s ‘cheating’ but still pretty clever ;)
I’ve seen this type of “divider” already and yes, it’s not the most reliable system. Sadly.
As long as it doesn’t involve a 555.
Please
Woah, I’ve never seen that Philips scope before. I’ve got a bunch of Tektronix 3S/3T plugins working, but I need a faster pulser & sampling head to get below 100pS. Absolutely wild what they were able to get working back then, and thank goodness they did such a great job documenting it so we can keep these old beasts running!
It depends on what other passives are allowed. The transistors alone won’t do it.
If I were to guess, it’s an analog design that fills up a capacitor in a staircase ramp and triggers after n-pulses when the voltage gets high enough.
That’s right.
Each input pulse dumps a glug of charge onto a capacitor. After sufficient glugs the voltage has risen high enough that the output triggers, which removes the charge on the capacitor and pulses the next stage.
Many years ago, I was given a Tek sampling plug-in to calibrate.
I had no idea what it did, but had the proper test equipment and calibration manual. So, I was “blindly” following the procedure not knowing what to expect. It may have been on Day 3 when the proper “Eye pattern” appeared on the screen. Oh! That’s what it is supposed to do!
Incidently I was just looking for something yesterday I could hook up to a signal path where I suspect ~1 ns ringing on, but I only have a 200 MHz BW scope. Anyone know how to tackle this? Preferably with some commercial stuff, although not too pricey… :)
High pass filter, diode, small capacitor, monitor capacitor with your current scope, if it charges you must have some signal present thats higher frequency than your chosen filter. Don’t know how practical that would actually be, but could be tested quickly.
The problem with that approach is that every time the actual signal changes the capacitor will be charged; not just due to ringing.
Obviously DC uncoupling either by the filter or other means, but by nature a high pass filter should be DC uncoupled, so long as the signal of interest changes below the filter Freq this is a valid way to achieve this. A solid example: in audio(particularly car audio) where you want to run your subwoofer flat out up to clipping for an event, clipping can be detected by using a small speaker and an inline capacitor (as a high pass filter) if the speaker sounds your getting a harmonic of the clipped 40-60hz wave, dial it back a bit :)
Anyone have a link to a write up, hopefully with diagrams?
Check the TekWiki here: https://w140.com/tekwiki/wiki/Sampling_oscilloscope
Lots of good pointers.
Absolutely fantastic! I have been looking for something like this for so long! This looks like the electronic equivalent of an autocorrelator used to measure fast laser pulses.
My wife was formerly into needle point, so “sampler” initially had me thinking of something else.
B^)
Then there are those boxes of chocolates
Oh yeah!
Equivalent Time analog sampling techniques are wild. HP pioneered a lot of it, but Tektronix really ran with it in the 1960s, and has a lot of great resources. I actually had no idea it was used for counters (but makes sense), I’ve been tearing through a pile of Tektronix sampling plugins & sampling heads over the past year and have gotten a bunch of them working.
Analog sampling was also how they got the first digital readout scopes working in the 1960s – The Tektronix 567, and 568/230s being arguably the first digital readout scopes ever.
Get lost in the Tek Wiki if you’ve got an afternoon to blow.
This is interesting. I would like to measure PCIe bus – but i don’t need to decode bits – only thing i need is basically eye diagrams – that should be achievable with sampling oscilloscope with persistence. What i found so far are 2 projects on the internet where people try to implement it with variable delay lines. The other option would probably be used equipment market (ebay) but i really don’t know what to look for. I already own a DSO (100 MHz) so maybe some external sampler? Are there such things? On a budget? And yes, my other problem after solving the sampling would be differential probes.
It’s also just bandpass sampling? The short pulses “expose” the waveform for a tiny bit each, but at different points in the repeating signal’s phase?
Like a wagon wheel in a movie, or a stroboscope?
https://hackaday.com/2017/03/23/say-it-with-me-aliasing/
Tunnel diodes are part of the secret sauce. I see lots of Soviet “tunnel diodes”on eBay and wonder how can they be so cheap? Are there affordable and real tunnel diodes?
Tunnel diodes are *not* necessary for sampling. There are none in the Philips PM3410, for example, nor in http://www.redrok.com/sampscope.htm
Tunnel diodes can be used in a trigger circuit (e.g. Tek 4×5 scopes) or to generate a step waveform (e.g. Tek 1502 TDR).