I really hope that this will turn into a real product one day. The cost of measurment equipment is holding back hobbyists from taking on RF and high speed digital projects. Affordable multi-Ghz oscilloscopes will be a big step forward for the community.
I have to say, Ted's approach to design reminds me of how I used to go at things. Build to learn, piece by piece, then put the puzzle together. Thanks for sharing Ted.
You can also put a rf frequency spectrum analyser circuit to reconstruct the actual waveform. By detecting amplitude of first harmonic and subsequent harmonics you can do inverse transform to actually reconstruct the wave. When a signal travels in a longer wire the real-time sampling does not stay accurate because of many factors including capacitance of line. So in practical world real-time sampling is always limited through probes even if you have terabytes of high speed ram, your instrument cannot create the actual waveform. You obviously loose amplitude as frequency increases.... there are so many factors that real-time sampling is limited by.
Thanks for the history of multi GHz sampling, unfortunately i came here thinking how is he going to store (data acquisition) the data over few miliseconds - he won't, its just sampling a repetitive waveform. An eye waveform is possible to achieve.
@hackaday, any chance of doing some content on connecting arduinos in series to double the MHZ sample rate possible? Using 3 for example as a master controlling 2 slaves for a more precise or higher sampling and perhaps with a pico also having the function of a digital logic analyser ?
Out of personal desires I’m very interested in this topic, have there been recent developments? (Wasted much time and effort with “premium” but actually bad-quality SFF cables for PCI Express 4.0 devices which uses 8 GHz)
yep, NLTLs are on my list to try out! Unfortunately, I haven't been able to simulate then successfully so far, so it looks like a lot of bench time will be involved.
Ted, maybe you monitor these comments. I'm trying to get into the Analog Devices ADCM582 fast comparator. The data sheet is clear as mud on pin connections. Then I may burn (multiple) chips. Is there any help on pin numbers to what connection?
So this designs based on comparators aren't able to capture an eye diagram of a random bit sequence signal, it only able to capture a waveform of a repeatable signal. Righ?
@@anatolyzapadinsky5995 you can't capture an eye diagram of a truly random signal, but commonly eye diagrams are tested using specified pseuduo-random bit sequences (PRBS) generated from LFSR's, which do repeat. These can be captured. To do this, the LFSR needs to be synchronized with the sample clock. A simple way is to have an FPGA generate the PRBS and also the sample clock. There are examples of eye diagrams captured this way at 25:15.
@@TeddyYapo I mean a random sequence with a separate clock channel or a clock recovered from the signal. The goal is to draw an eye diagram of a transmitter itself, such as HDMI video signal source for example. As I understand a fully functional sampling scope is capable of doing this with a separate trigger channel or a clock source. Are you evaluating a point voltage level by repeatedly comparing its value with different voltage levels through a binary search algorithm?
@@anatolyzapadinsky5995 yes, binary search is one way to do it; you can deal better with noise using oversampling. But, to your point, no, you can't generate eye diagrams without a coöperative transmitter that you can convince to output specific patterns if you only have a single comparator. With a window comparator made from two simultaneously clocked comparators, you could generate a type of eye diagram for your example, by sampling all time/voltage points to accumulate a probability density function. It would be quite slow, but would be capable of generating the equivalent of an eye diagram given many samples. I haven't built one like this, but it would be interesting.
All Jeri Ellsworth calls for now is money. Every few years she cries to the camera and says she is "passionate" about some project and every thirsty dude sends her cash. In her older videos she used a $5 dmm and now has $10k oscilloscopes yet none of her projects ever succeed.
@@michelfeinstein EEVBlog often goes into the details of those scopes. They even did a 70 GHz scope recently, you should watch it. Pro tip : keep a box of tissues handy.
Please Hackaday camera people -- when the speaker is talking about the slide, which is virtually all the time, PLEASE SHOW THE SLIDE!
they should take inspiration from cppcon
They prefer to capture the marvellous black background
I love when they use PIP (Big picture would be the slides and the smaller picture would be the speaker.) Too bad that didn't happen here.
I've mentioned this last year about this issue but nope... Feedback either here or email goes to a black hole.
slides: cdn.hackaday.io/files/1672927157420928/ted-yapo-supercon-2019.pdf
I really hope that this will turn into a real product one day. The cost of measurment equipment is holding back hobbyists from taking on RF and high speed digital projects. Affordable multi-Ghz oscilloscopes will be a big step forward for the community.
I have to say, Ted's approach to design reminds me of how I used to go at things. Build to learn, piece by piece, then put the puzzle together. Thanks for sharing Ted.
Awesome presentation! Thank you. Complex concepts explained in a straight-forward and practical way.
Wow! Just wow! :D
Makes me appreciate my Tektronix S-6 sampling head!
The stuff. Wow. The voice. Wow. Very cool.
Best project ever !
Great talk, this is how the state of the art advances for those of us who don't have a 100k laying around for test equipment. Thank you!
You know you can *rent* test equipment, right ?
@@TheNefastor You know you can rent a mansion and a Ferrari, right? Yet you drive a Ford Taurus, and live in your moms basement, why?
@@redresistor1330 where is this coming from ? The knowledge that I can't reach through the screen and punch your teeth out ?
@@TheNefastor Ok ya four-eyed nerd. You look so tough in your profile photo, lmao. The only thing you punch is nuts against your tonsils 😁😂
very very cool, thank you
You can also put a rf frequency spectrum analyser circuit to reconstruct the actual waveform. By detecting amplitude of first harmonic and subsequent harmonics you can do inverse transform to actually reconstruct the wave. When a signal travels in a longer wire the real-time sampling does not stay accurate because of many factors including capacitance of line. So in practical world real-time sampling is always limited through probes even if you have terabytes of high speed ram, your instrument cannot create the actual waveform. You obviously loose amplitude as frequency increases.... there are so many factors that real-time sampling is limited by.
Oh I also forgot to mention speed of differential signal which is a whole new complexity in analysis of signal
well done
Yesss! We need this!
For god's sake. Why is the video interlaced? Was it filmed on VHS or what?
Great talk. Thanks!
In 2019 we still had working interlaced cameras?
if the input signal be a 1khz repeatitive pulse but very short (
Awesome
Thanks for the history of multi GHz sampling, unfortunately i came here thinking how is he going to store (data acquisition) the data over few miliseconds - he won't, its just sampling a repetitive waveform. An eye waveform is possible to achieve.
One guy clicked the dislike button by mistake .... :-/
Would totally hope this was a zero dislike video on there.
Jitter in the periodic input signal could be a problem for that method, but nonetheless interesting.
@hackaday, any chance of doing some content on connecting arduinos in series to double the MHZ sample rate possible? Using 3 for example as a master controlling 2 slaves for a more precise or higher sampling and perhaps with a pico also having the function of a digital logic analyser ?
I just want to know how Ted makes those cool black and white schematics
Out of personal desires I’m very interested in this topic, have there been recent developments?
(Wasted much time and effort with “premium” but actually bad-quality SFF cables for PCI Express 4.0 devices which uses 8 GHz)
Nice work. Have you looked into shock lines for fast-edge generation?
yep, NLTLs are on my list to try out! Unfortunately, I haven't been able to simulate then successfully so far, so it looks like a lot of bench time will be involved.
Shouldn't GHz sampling of repetitive signals also be possible with an HackRF? If not why?
Very cool
Any link to follow this project?
Already found it: hackaday.io/project/167292-8-ghz-sampling-oscilloscope
And Ted's twitter: twitter.com/tedyapo
Ted, maybe you monitor these comments. I'm trying to get into the Analog Devices ADCM582 fast comparator. The data sheet is clear as mud on pin connections. Then I may burn (multiple) chips. Is there any help on pin numbers to what connection?
Nice
Hey Ted, what (tool/language/package) are you using to plot your graphs?
python/matplotlib
@@TeddyYapo cheers mate, beautiful presentation.
There are a lot of interlacing artifacts in the video.
Here's a link to the slide deck for anyone interested: cdn.hackaday.io/files/1672927157420928/ted-yapo-supercon-2019.pdf
So this designs based on comparators aren't able to capture an eye diagram of a random bit sequence signal, it only able to capture a waveform of a repeatable signal. Righ?
@@anatolyzapadinsky5995 you can't capture an eye diagram of a truly random signal, but commonly eye diagrams are tested using specified pseuduo-random bit sequences (PRBS) generated from LFSR's, which do repeat. These can be captured. To do this, the LFSR needs to be synchronized with the sample clock. A simple way is to have an FPGA generate the PRBS and also the sample clock. There are examples of eye diagrams captured this way at 25:15.
@@TeddyYapo I mean a random sequence with a separate clock channel or a clock recovered from the signal. The goal is to draw an eye diagram of a transmitter itself, such as HDMI video signal source for example. As I understand a fully functional sampling scope is capable of doing this with a separate trigger channel or a clock source. Are you evaluating a point voltage level by repeatedly comparing its value with different voltage levels through a binary search algorithm?
@@anatolyzapadinsky5995 yes, binary search is one way to do it; you can deal better with noise using oversampling. But, to your point, no, you can't generate eye diagrams without a coöperative transmitter that you can convince to output specific patterns if you only have a single comparator. With a window comparator made from two simultaneously clocked comparators, you could generate a type of eye diagram for your example, by sampling all time/voltage points to accumulate a probability density function. It would be quite slow, but would be capable of generating the equivalent of an eye diagram given many samples. I haven't built one like this, but it would be interesting.
Hello
if the input signal be a 1khz repeatitive pulse but very short (
talk starts at 1:25
I donno but sounds like something CERN might find interesting.
Was that Jeri Ellsworth that called out dead bug?
All Jeri Ellsworth calls for now is money. Every few years she cries to the camera and says she is "passionate" about some project and every thirsty dude sends her cash. In her older videos she used a $5 dmm and now has $10k oscilloscopes yet none of her projects ever succeed.
@ungratefulmetalpansy Just saying what every smart person is thinking.
But how are you going to read high-speed digital communication with no real time sampling, as the bits will always change?
You can't. You'll still need a scope that costs as much as a house to do that.
Yeah, thought so, and a DIY Open-Source project focused on this is so much cost prohibitive as well?
@@michelfeinstein judging by reviews of real time GHz scopes and their custom silicon ADC's, I think so.
@@TheNefastor yeah, those custom ADCs must be the thing :/
@@michelfeinstein EEVBlog often goes into the details of those scopes. They even did a 70 GHz scope recently, you should watch it. Pro tip : keep a box of tissues handy.
Hi
The SMA connectors are not a good match. There must be a better brand.