General > General Technical Chat
Frequency Divider for older Oscilloscopes??
bdunham7:
Just to be clear, I'm not claiming Jorge is wrong or that what he says he made didn't exist, just that I don't understand how it would work very well given the description. I hope he is patient enough to continue the conversation and isn't put off by language barriers. He mentioned some TTL logic, so perhaps something like a 7474? which would imply some trigger logic. Making a workable trigger, ramp and sample/hold circuit that operates into the UHF region does not seem a trivial task.
Jorge Ginsberg:
--- Quote ---I agree with Bdunham. To down-convert a 100.000MHz signal to 10KHz, you need to sample at 99.990MHz, not at 100.000MHz. If you want to automate this so that the output frequency is always 1/10000 of the input frequency, at the very least you'll need a good VCO for the sampling clock that covers the input range you are interested in, and to PLL lock it to 9999/10000 of the input frequency (which is a whole other can of worms if the input is not a nice quasi-sinusoid).
The requirements for the sample and hold are also rather stringent - it may well be easier to implement as a pair of carefully matched track and holds that alternate acquiring the signal, and a fast analog switch to select the one currently holding.
--- End quote ---
No...no...no... and NO.
--- Quote ---OK, but it seems clear that your sampling rate is just slightly slower than the input signal, not equal to it, right? How does the circuit determine and set the sampling rate?
--- End quote ---
No...no...no... and NO.
What changes is not the frequency of sampling. What changes is the TIME at which the sampling occurs. In the first cycle, (for example) it occurs 1 picosecond after the wave starts (zero crossing from bottom to top). The second sample is taken 2 picoseconds after the zero crossing, from bottom to top.
The third sample is taken 3 picoseconds after the zero crossing, from bottom to top.
The fourth sample is taken 4 picoseconds after the zero crossing, from bottom to top.
The fifth sample is taken 5 picoseconds after the zero crossing, from bottom to top.
And so on until the cycle is complete.
And if you use 10000 samples to reconstitute the original waveform, and you have taken only one sample for each cycle, then the reconstituted signal will have a frequency 10000 times lower and you will be able to observe it on a cheap oscilloscope.
The sampled voltage is charged into a capacitor. And on that capacitor a perfect sine wave is obtained, and a faithful copy of the sampled signal.
No A/D converters are needed. No need for counters. No need for memories. The method and the circuit are much simpler than you suppose. But you must use your wits to figure it out. That's the beauty of electronics design: the ingenuity you use to solve a complicated problem with a simple solution.
The most remarkable thing about all this talk is that
GlennSprigg , the original author of the query, has not even shown up here......
bdunham7:
--- Quote from: Jorge Ginsberg on July 18, 2021, 11:21:10 pm ---The method and the circuit are much simpler than you suppose. But you must use your wits to figure it out. That's the beauty of electronics design: the ingenuity you use to solve a complicated problem with a simple solution.
--- End quote ---
OK, let's reason through it so at least we agree on the steps needed. Then I'll think about how hard it would be to build.
--- Quote ---What changes is not the frequency of sampling. What changes is the TIME at which the sampling occurs. In the first cycle, (for example) it occurs 1 picosecond after the wave starts (zero crossing from bottom to top). The second sample is taken 2 picoseconds after the zero crossing, from bottom to top.
--- End quote ---
In many things there are different ways of describing the same thing. Often two very descriptions will boil down to the same thing. In this case, at a given signal (say our 100MHz) sampling as you describe will result in a sampling frequency just under the signal frequency, just by a different method. In this case the method is important, but the result may be considered in different ways.
--- Quote ---And if you use 10000 samples to reconstitute the original waveform, and you have taken only one sample for each cycle, then the reconstituted signal will have a frequency 10000 times lower and you will be able to observe it on a cheap oscilloscope. The sampled voltage is charged into a capacitor. And on that capacitor a perfect sine wave is obtained, and a faithful copy of the sampled signal.
No A/D converters are needed. No need for counters. No need for memories.
--- End quote ---
OK, so to the meat of the problem. Someone correct me if I err. These are the issues, more or less in order. Assume a 100MHz input signal to be displayed at 10kHz.
1. You need a way to trigger on the zero crossing (or anywhere, I suppose) of the input signal.
2. You are sampling at 100MSa/s using an offset that cycles at 10kHz and has a range sufficient to cover one period.
3. If you want to display this as single period 10kHz signal on a scope, the scope should have a sweep time of 100 microseconds.
4. You need a method of offsetting samples of the input frequency from 0 to 10 nanoseconds, corresponding to one period.
5. To display the samples on the scope, they have to be correlated so that their positions in real time on the scope are proportional to the sample offset time. So the +5nS sample must appear at the 50uS position.
6. Since there is no storage, the samples have to occur in real time with the two values correlated. Thus you need a circuit that 'knows' where it is on the 100uS scale and offsets the sample by the corresponding amount.
So you need a trigger that operates on the 100MHz input signal, a 10kHz ramp signal followed by an offset delay circuit that corresponds to the ramp and then operates the sample circuit. You can trigger the display scope on the 10kHz ramp or on the sampled output. Anything else?
--- Quote ---The most remarkable thing about all this talk is that
GlennSprigg , the original author of the query, has not even shown up here......
--- End quote ---
Time zones. We're all over the world and people pop in and out.
Jorge Ginsberg:
--- Quote ---OK, so to the meat of the problem. Someone correct me if I err. These are the issues, more or less in order. Assume a 100MHz input signal to be displayed at 10kHz.
1. You need a way to trigger on the zero crossing (or anywhere, I suppose) of the input signal.
--- End quote ---
That's right.
And any oscilloscope has a trigger circuit. There is no difficulty in that.
--- Quote ---2. You are sampling at 100MSa/s using an offset that cycles at 10kHz and has a range sufficient to cover one period.
--- End quote ---
I'm afraid I don't understand what you're saying
--- Quote ---3. If you want to display this as single period 10kHz signal on a scope, the scope should have a sweep time of 100 microseconds
--- End quote ---
.
Once you have the reconstituted signal you send it to the oscilloscope of your choice. To the oscilloscope, that reconstituted signal is just like any other signal it receives. The oscilloscope is not obliged to know that you are feeding it a signal that was the product of another sampled signal... In fact, you are doing all this to be able to see VHF signals with your cheap oscilloscope...
--- Quote ---4. You need a method of offsetting samples of the input frequency from 0 to 10 nanoseconds, corresponding to one period.
--- End quote ---
from 0 to 10 ns... from 0 to 5 ns... from 0 to 30 ns... Whatever you need. This is varied with a potentiometer, preferably a 10 turns one.
--- Quote ---5. To display the samples on the scope, they have to be correlated so that their positions in real time on the scope are proportional to the sample offset time. So the +5nS sample must appear at the 50uS position.
--- End quote ---
You are not feeding any sampled signal to the oscilloscope. Your oscilloscope is receiving an already reconstituted signal and your oscilloscope doesn't care where that signal comes from nor does it care how it was created.
--- Quote ---6. Since there is no storage, the samples have to occur in real time with the two values correlated. Thus you need a circuit that 'knows' where it is on the 100uS scale and offsets the sample by the corresponding amount.
--- End quote ---
No correlation with anything is necessary. The circuit simply takes the samples one after the other, period after period, and a capacitor takes care of integrating everything and reconstituting the signal.
What you must find out now is how to achieve that each sample is displaced with respect to the previous one and that this time is not modified, because if it is modified, the waveform will change and that is what we do not want to happen.
--- Quote ---So you need a trigger that operates on the 100MHz input signal, a 10kHz ramp signal followed by an offset delay circuit that corresponds to the ramp and then operates the sample circuit. You can trigger the display scope on the 10kHz ramp or on the sampled output. Anything else?
--- End quote ---
No. Nothing else.
Put your ingenuity to work and see if you can figure out how to achieve all that in a simple way.
Enjoy the challenge !
bdunham7:
--- Quote from: Jorge Ginsberg on July 19, 2021, 02:55:15 am ---
And any oscilloscope has a trigger circuit. There is no difficulty in that.
No correlation with anything is necessary.
Put your ingenuity to work and see if you can figure out how to achieve all that in a simple way.
--- End quote ---
OK, I can see generally what is needed. Making a reliable trigger may not be an insurmountable challenge, but picosecond jitter.... I suppose it doesn't have to be that good to nominally work.
Correlation will occur whether explicitly designed for or just an implicit result of things 'boiling down' to what they actually are. The oscilloscope timing just serves conceptually as a framework, I understand the circuit doesn't depend on it in any way.
I'll let GlennSprigg pick this up if he is still interested.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version