Generator sine to spikes converter

[neazoi]

Hi, I have a 30khz to 180MHz generator. Is there a simple circuit that can convert its sine output to spikes, so I can use them to calibrate old scopes graticules in horizontal?

 

[barry]

zero-crossing detector, to start with. Define “spike”.

 

[neazoi]

zero-crossing detector, to start with. Define “spike”.

I have to convert the sine from the generator to short duration pulses, say one pulse every peak of the sine. The rate of these pulses would depend on the sine frequency that way. These spikes will be used as markers in the time domain of old scopes that lack markers.
We are taking about frequency as high as 1ghz and as low as 1mhz

 

[betwixt]

If the generator only goes to 180MHz you don't need 1GHz spikes, especially as time markers. If the 'old' scope can really display at 1GHz you can show several cycles between lower frequency markers.

It doesn't matter where on the sine wave you pick to generate the marker but the zero crossing will be by far the easiest to detect and it gives you two pulses per cycle. If you want peak detection you need to use a comparator and be sure exactly what voltage the peak is at. You can make a tracking comparator with a peak detector and a divider so it compares at slightly under the peak but it makes things somewhat more complicated.

If you are attempting to measure frequency using a signal generator to compare the waveform timing, a far simpler method is to feed the generator into the X input of the scope and the unknown into the Y input then tune the generator for a static image (like a Lissajous). You can then read the frequency directly from the signal generator. It will also work on harmonics of the generator.

Brian.

 

[KlausST]

Hi,

again: define "spike".
Voltage, waveform, timing, how often..

You say: "short" .. which can mean anything
you say "every peak": positive peak only, or negative peak, too?
Btw: the peak of a sine is the rather difficult to detect, noise will cause largest error on output.
Better use zero cross.

Klaus

 

[neazoi]

If the generator only goes to 180MHz you don't need 1GHz spikes, especially as time markers. If the 'old' scope can really display at 1GHz you can show several cycles between lower frequency markers.

It doesn't matter where on the sine wave you pick to generate the marker but the zero crossing will be by far the easiest to detect and it gives you two pulses per cycle. If you want peak detection you need to use a comparator and be sure exactly what voltage the peak is at. You can make a tracking comparator with a peak detector and a divider so it compares at slightly under the peak but it makes things somewhat more complicated.

If you are attempting to measure frequency using a signal generator to compare the waveform timing, a far simpler method is to feed the generator into the X input of the scope and the unknown into the Y input then tune the generator for a static image (like a Lissajous). You can then read the frequency directly from the signal generator. It will also work on harmonics of the generator.

Brian.

Exactly, the whole point is to make some markers for these old scopes and spectrum analyzers and make them more usable again.

Here is a primary design I have come up with. I am using exactly the comparison method you mention with a calibrated generator for indirect power and frequency measurements (although I do have also an accurate frequency counter).
*I am not sure if my 1M input counter will interfere with the 50R measurement by the way if connected like this?

The device simply switches between one input and the other and selects either the scope or SA output. DUT is provided for filters characterization.

On the spectrum analyzer markers are easier. On scope, amplitude comparisons are easier, but frequency comparisons are more difficult (taking into account the different phases of the test and generator signals.)

I was actually thinking not measuring the frequency on the time domain, but introducing arbitrary markers between 2 random points of the waveform, to measure their time difference. So I was thinking of a generator with much higher frequency than the test signal, and a sine to pulse converter. By varying the frequency of the generator I vary the pulse repetition rate. Matching to the test waveform can be done by delaying the test waveform.

But now I am thinking that even a simple square generator could do the same thing easier, you just take as markers the vertical lines of the square.

 

[BradtheRad]

A book suggests getting accurate timing from the shortwave radio tuned to the time marker stations WWV, CHU, etc. WWV frequencies 2.5 / 5 / 10 / 15 / 20 MHz.
Experimentally 25 MHz.

CHU (Canada) 3.333 / 7.333 / 14.667MHz.

The broadcast frequency itself is one source of calibration.
Furthermore each station sounds a click every second.
You might need a sensitive antenna & receiver to pick up these stations. Also good propagation conditions.

 

[dick_freebird]

Many signal generators and spectrum analyzers have a 10MHz ref
port on the back, to serve the rack a common timebase if that is
what's wanted.

Many PLL eval boards want a 10MHz reference as they are meant
for bench evaluation.

You might go this way, look for a wide frequency range one (not
one rigged for a cell band only) and get yourself a programmable
timebase (probably with buffered output) for cheap-ish.

Of course 60MHz function generators capable of all this cost
about $39 fresh off the boat by way of eBay. So depending on
what your time is worth, your signal of interest really looks like
and your wallet holds, this might not be one of your better
make/buy "make" calls.

 

[BradtheRad]

Capacitor conducts brief spike whenever sine wave peaks. Current creates a voltage across R.

 

[neazoi]

Capacitor conducts brief spike whenever sine wave peaks. Current creates a voltage across R.

View attachment 184672

Interesting and dead easy. One spike per cycle! I wonder how well will this work on say 200mhz with components scaling maybe?

 

[KlausST]

Hi,

looks good so far.
But etter check the circuit including some realistic noise and amplitude fluctuation.

Is 38mV enough? from the description I rather expected 5V or so.

Klaus

 

[neazoi]

Hi,

looks good so far.
But etter check the circuit including some realistic noise and amplitude fluctuation.

Is 38mV enough? from the description I rather expected 5V or so.

Klaus

38mV is enough. The generator has +20dbm of drive power and the output spikes will be used as time markers, by changing the generator frequency and delay. Their amplitude does not matter much since I can reposition the trace to any amplitude of the test signal I want to time-measure. I wonder if this circuit will work at 200MHz and what component variations are appropriate for that frequency.

 

[BradtheRad]

A larger Farad value suits slower frequencies. I almost recommended a large value electrolytic type too for high frequencies but I've seen experts here tell how some types don't behave the same at high versus low frequencies.

The principle is that the capacitor charges to whatever maximum voltage is the waveform's peak, and then periodically carries a brief burst at each subsequent maximum.

 

[neazoi]

A larger Farad value suits slower frequencies. I almost recommended a large value electrolytic type too for high frequencies but I've seen experts here tell how some types don't behave the same at high versus low frequencies.

The principle is that the capacitor charges to whatever maximum voltage is the waveform's peak, and then periodically carries a brief burst at each subsequent maximum.

It didn't work well in LTspice.
But this circuit of mine did. This is an unbiased class C amplifier. It works at least as well from 10Hz to 200MHz. It is still usable on 1GHz.
Enjoy

 

[betwixt]

I doubt that would work well in real life. It isn't class C and the bias will be created by the input capacitor charging through the B-E junction of the transistor so the source impedance would be important. It would need at least some DC return path across the input for the bias to develop. At higher frequencies the parasitic capacitances in the transistor and the inductance of the capacitors would be sufficient to stop it working.

Brian.

 

[FvM]

Why not look for various existing frequency marker designs?
They typically use differentiated square wave or step recovery diodes as "spike" generator.

 

[neazoi]

I doubt that would work well in real life. It isn't class C and the bias will be created by the input capacitor charging through the B-E junction of the transistor so the source impedance would be important. It would need at least some DC return path across the input for the bias to develop. At higher frequencies the parasitic capacitances in the transistor and the inductance of the capacitors would be sufficient to stop it working.

Brian.

The actual circuit I have used before, does have a resistor from the collector to the base to bias it. Without the resistor simulation shows even better results, that's all. I haven't tested this above 30MHz in real life.

--- Updated Aug 27, 2023 ---

Why not look for various existing frequency marker designs?
They typically use differentiated square wave or step recovery diodes as "spike" generator.

I am not sure it is a frequency marker I need. What is needed is not to produce fixed harmonics of a low frequency signal but convert a variable generator signal to spikes. The spikes distance will be controlled by the frequency variation of the generator. Since the frequency is known (in a synthesized generator) the time difference between 2 spikes is known, hence they can be used as markers.

Yesterday, I tried to do it without any circuit. It worked.

What I did, was to use an old storage CRT scope and store the waveform to be measured on the CRT.
Then feed the generator waveform to it, but do not increase the scope trace brightness but make the trace only barely visible, so that it cannot be stored into the CRT.

Without changing the TIME/DIV, I varied the frequency of the generator in conjunction with the vertical position of the scope generator trace and its timing delay of the scope internal delay line, so as to point the generator sinewave peaks to the points to be measured on the stored signal.

A trick one can do to make these generator sinewave peaks point more accurately to the test signal, is to increase their amplitude, or to decrease the mVolts/DIV much. This stretches the generator signal way out of the CRT vertical range. Repositioning the trace so that only the sine peaks are displayed (and the rest out of the CRT) this creates more sharp tips (sinewave peaks) on the CRT, to pinpoint the test signal points of interest.

There is a limit of how sharp a tip point can be, which depends on the low signal handling of non-sampling plugins or the dot density of the sampling plugins.
But this technique is very usable and it does not depend on the scope to be calibrated or dual input plugins.

If a non-storage scope is to be used, a way to vary the phase of one of the two waveforms is needed for this to work I think. Maybe you can think of another technique that fits that case.

 

[BigBoss]

A high speed Comparator can also be used. If the threshold level is adapted to near peak level of the sinus wave, the comparator will create a good pulses around peak values of the sinus waveform.