Products > Test Equipment
Building my own scope
<< < (19/32) > >>
Mechatrommer:

--- Quote from: tggzzz on October 22, 2022, 10:06:49 am ---
--- Quote from: py-bb on October 22, 2022, 03:03:02 am ---Say you want 1ghz bandwidth, you'd need to sample at at least 2ghz, this is where you run into problems, you can't just dump that into DRAM (you can get about ~100m/sec (100mhz) requests from DRAM, they're faster than this at sequential transfers once you've opened the page) - so you'll need to buffer there.

--- End quote ---

Not true. The bandwidth and sampling rate are independent.

Examples:

* SDR. Even the cheap dongles capture 1.5GHz waveforms, and their sampling rate is several orders of magnitude lower
* TDR. My 1970s Tek 1502 displays <140ps risetimes and its sampling rate is <100kS/s
* Scopes. One I used many decades ago is the HP54100A: 1GHz/350ps, 40MS/s
* X-Y plotters. In the mid 70s I saw an experimental sampling scope with >1MHz bandwidth that used an XY pen-plotter as an output device.There are many many other examples.

--- End quote ---
its funny you have a habit to get into disagreement with people and giving lecture philosophically about "in context" discussion, strawman fallacy etc (in other thread) without realizing you are the one you are lecturing about. i browse this thread quickly and nothing talk about sampling scope (which is a specialized scope), you cant bode plot with sampling scope. people talked about nyquist criteria (2.5X BW to be exact, not 2X) which is what we understand normally as what a normal scope do, and then you wandered off to suggest 1GHz BW with KSps scope, and how unimportant is trace length and propagation delay is in GHz domain. its just out of context, and even incorrect in some context... no need to further mess the already complicated subject (now i quote you fully so you have less chance to invent strawman argument on me ;D)

btw, i agree with tautech, OP is going into rabbit hole, or worse a worm or black hole. if you really want to waste your time you can dig i also created a thread like this to ask many years ago (while the spirit is high) and then i never made the first step, knowing more (or just a little bit) of what i actually dealing with... if you think you can beat the already cheap Rigol or Siglent and their in-house engineers, then imho you are delusional. lest if you want to talk about saving your pocket money. and later after that, i was thinking about the 10GSps or greater DSO segment where we dont have cheap option yet, maybe i can implement Rigol brilliant idea of interleaving 10x 1GSps ADC? but when i browsed digikey for the price, i closed the browser and never went back to the idea again. going fishing is much much better things to do for the sake of your brain and health.
pcprogrammer:

--- Quote from: Fungus on October 22, 2022, 10:12:24 am ---
--- Quote from: pcprogrammer on October 22, 2022, 06:13:25 am ---There are some threads on this forum with discussions about the needed sample rate. For a scope the opinions differ from 2.5 to 10 or more times for proper signal representation. Times 2 is the nyquist theorem for the ability to reconstruct a sine wave.

--- End quote ---

Times 2 can reconstruct any bandwidth limited signal, not just sine waves.

The main problems are:
a) Finding a bandwidth limited signal in real life
and
b) Using an infinitely wide filter to reconstruct it

--- End quote ---

You are correct, and I should have be more specific. When you only have two samples for a period of the signal you can only reconstruct it to a sine wave, as long as the samples are not taken on the zero crossing of course. And it is missing information about the max amplitude unless the samples are taken on the top and bottom of the wave.

To reconstruct a square wave you need much more samples to be able to do it to some extend. Fourier comes into play here. But as I wrote in another post, there are enough threads discussing this.
Messtechniker:
To start with, I'd keep it simple. Like an USB controlled
relay attenuator/amplifier box ahead of a standard sound card.
Once you have programmed time domain, XY, FFT, Bode and all the rest,
you might go for an extended frequency response by writing your own driver
having a 1 MHz+ sample rate for a specifically modded sound card.
In an even further step you might want a sound card generator with a similar USB controlled relay/amplifier output box.
So this would mostly be a software project.
Feeling up to it? Even this will not be a one-man-show. Or it will take a very long time
(10 yrs + ?) to implement.
Just an idea.
Fungus:

--- Quote from: pcprogrammer on October 22, 2022, 12:40:40 pm ---You are correct, and I should have be more specific. When you only have two samples for a period of the signal you can only reconstruct it to a sine wave, as long as the samples are not taken on the zero crossing of course. And it is missing information about the max amplitude unless the samples are taken on the top and bottom of the wave.

--- End quote ---

Reconstruction only works at less than Nyquist.

If it's 99.99999% of Nyquist the sampler will go in and out of phase with the signal and still reconstruct correctly if the signal+filter are infinitely wide.

(which they won't ever be in real life, obviously)

py-bb:

--- Quote from: tggzzz on October 22, 2022, 11:40:02 am ---
--- Quote from: py-bb on October 22, 2022, 11:15:00 am ---
--- Quote from: tggzzz on October 22, 2022, 10:18:27 am ---
--- Quote from: py-bb on October 22, 2022, 08:36:03 am ---
--- Quote from: pcprogrammer on October 22, 2022, 08:21:47 am ---Not going to start a discussion about nyquist. There are plenty around here on the forum.


--- Quote from: py-bb on October 22, 2022, 07:49:00 am ---As for the 2nd one, why? I am sure at some point separate ADCs becomes sensible, but when and for what? I imagine really really high frequencies because multiplexing the sample values (rather than ADCs) is easier and gives good results.

But either way you've gotta demux the samples.

--- End quote ---

Well most likely money. High speed ADC's are expensive and using two that are capable of half the speed might cost way less than the single high speed one.

Take the FNIRSI-1013D and 1014D, these use two AD9288 chips. One for each channel, and this chip has 2 ADC's in it. This way they have 200MSa/s, but do need calibration to equalize the readings between the two.

The Hantek DSO2000 series use a ADC08D500 chip. This is a 2 channel 500MSa/s ADC where the two ADC can be connected to the same signal inside the chip. So a single channel can do 1GSa/s. Have not looked into the software of this one, but assume it also needs some calibration.

These are the ones I know of, but there will be others.

Using a single ADC to sample multiple channels can of course also be done, but it lowers the sample rate instead of increasing it.

For really high frequencies (> GHz) it requires a lot more knowledge to get it right with multiple ADC's, and this is not within the realm of what I know.

--- End quote ---

If you're going that high you need knowledge of microwave propagation... which also doesn't help OP.

--- End quote ---

Not true.

Here's an example that doesn't - and can use a pen plotter as an output device http://www.redrok.com/Circuits_1GHz-samplig-Oscilloscope-Front-End.pdf http://www.redrok.com/sampscope.htm

Shame that author measures time intervals in mohs (e.g. 50nS)

--- End quote ---

I'm not convinced it'll be any good, trust me at ~ghz speeds things are a lot less nice.

--- End quote ---

If you had bothered to understand the techniques in that circuit, you would see that the high frequencies are limited to a very small part of the board. They are entirely tractable; see the old scopes I referred to earlier.

Besides, zig zag traces aren't principally used to avoid "fibres in the circuit board", and there are standard ways of reducing how the weave affects the spatial variation of impedance.


--- Quote ---At ~4ghz digital stuff you need to zig-zag traces to avoid problems from the fibres in the circuit board. Width and bends in "traces" matter - it's really non-trivial.

--- End quote ---

Here's an example of a 4GHz sampling circuit where the weave in the PCB is clearly visible (Tek 1502, from half a century ago).



The only interesting point is the notch next to the CR1732 sampling diodes, used to counteract the effect of the diode's capacitance.


--- Quote ---Also that link is analogue.

--- End quote ---

Irrelevant. All circuits are analogue - except arguably for photon counting and femto-amp circuits (and Josephson Junction, but I don't know enough about those).

You could easily add a very slow ADC and it would work unchanged. For example, you could use an Arduino's ADC if you wanted to show the result on a computer screen.  Alternatively ( if you used skill and imagination to circumvent a lack of DC response) you  could use a PC Sound card as an ADC :)


--- Quote ---If you don't believe me BTW, I encourage you to try it (seriously, love to know how you get on) PROVIDED you have some equipment to know if it's working well or not. I'm sure it'll do something! The question is it it accurately reflecting that something.

--- End quote ---

That's valid for each and every approach, therefore doesn't distinguish between techniques.

--- End quote ---

In the picture you linked to a scope you see those fat copper traces - waveguides, because it had to be designed with great care to be useful.

That they succeeded doesn't mean it's trivial or easy.
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod