sum two signal with different frequency

[paramis]

I have two pulse with freq 20MHz and 200Mhz.how can i get signal with freq 220Mhz from 20MHz and 200Mhz.HOW can I sum that?

 

[barry]

You are confusing summing with mixing. If you "sum" a 200 Mhz signal with a 20 MHz signal, you get a signal that has components of 200 MHz and 20 MHz, you don't get 220MHz!!!

Are you proposing a system where you put 20 and 200MHz pulses in and get 220 MHz pulses out?

If they were sinusoids, you could use a mixer and low pass filter, but because they are pulses, it's a LOT more difficult.

 

[paramis]

Hi barry
yes.I want a system where put 20 and 200MHz pulses in and get 220 MHz pulses out.

 

[FvM]

You didn't particularly mention it, but I guess, the "220 MHz" output should have a constant pulse rate (no or low jitter)? In this case, a PLL is required. A special designed phase detector could trace the number of edges of both signals.

 

[barry]

You didn't particularly mention it, but I guess, the "220 MHz" output should have a constant pulse rate (no or low jitter)? In this case, a PLL is required. A special designed phase detector could trace the number of edges of both signals.

Sounds messy, FvM. What happens when the edges of the two signals line up?

Paramis, can you elaborate on your system a bit? Do you always have exactly 20 and 200? If so, why do you need to sum them? If not, what's the bandwidth of the signals? Are they phase-locked?

 

[FvM]

What happens when the edges of the two signals line up?

I imagine, they'll be counted separately and the results added. Possible in principle, but not necessarily serving a reasonable purpose.

 

[paramis]

Hi FvM can u explain more about phase detector in PLL?

 

[Mityan]

If they were sinusoids, you could use a mixer and low pass filter, but because they are pulses, it's a LOT more difficult.

Nothing difficult. Square wave is just a combination of sinusoids (spectrum looks like sin(x)/x).
Mixer, or multiplier produce a lot of different components with the frequencies +/-n*f1 +/- m*f2.
Anyway, if the inputs are sine or square.
Applying a Bandpass filter for 220 MHz you remove all other components and leave the single sine you need.
To convert it to square use amplifier + hard limiter.

 

[barry]

After further review, I think you're correct, a mixer/bandpass just might work with square-wave inputs. However, there are still some unanswered questions, e.g., bandwidth of inputs, etc.

 

[FvM]

After further review, I think you're correct, a mixer/bandpass just might work with square-wave inputs.

Yes, but the solution hasn't to do with digital design. It requires a high Q, accurately adjusted filter.

 

[Mityan]

If he generates 200 MHz square wave in digital form, let him just generate 220 MHz instead and that is all, why doesnt he do it?

 

[barry]

If he generates 200 MHz square wave in digital form, let him just generate 220 MHz instead and that is all, why doesnt he do it?

I agree. That's why I asked that in post #5. Paramis is not very forthcoming with his real requirements.

 

[Tunelabguy]

Nothing difficult. Square wave is just a combination of sinusoids (spectrum looks like sin(x)/x).
Mixer, or multiplier produce a lot of different components with the frequencies +/-n*f1 +/- m*f2.
Anyway, if the inputs are sine or square.
Applying a Bandpass filter for 220 MHz you remove all other components and leave the single sine you need.
To convert it to square use amplifier + hard limiter.

It is normal in a mixer to have one of the signals be essentially a square wave (switching the other signal). So maybe if one of the signals is bandpass filtered to make it mostly a sine wave then you could follow traditional heterodyne methods.

 

[Mityan]

It is normal in a mixer to have one of the signals be essentially a square wave (switching the other signal). So maybe if one of the signals is bandpass filtered to make it mostly a sine wave then you could follow traditional heterodyne methods.

Anyway the result of mixing will be a sum of lot of components, so it is unnessesary to convert both input signals to sine prior to mixing. Just a single good bandpass at output I think.

 

[kripacharya]

Anyway the result of mixing will be a sum of lot of components, so it is unnessesary to convert both input signals to sine prior to mixing. Just a single good bandpass at output I think.

i disagree. it is clear that the asker has in mind two different rectangular inputs of freq (say) 20Mhz and 200MHz, and he wants an output which is a rectangular signal output whose frequency is the "sum" of these, and hence at 220MHz. Lets not second-guess the original requirement.

As it turns out, despite the seeming simplicity of the requirement, this is a non-trivial problem statement. AFAIK none of the solutions presented so far have been able to do this.

The best option seems to be one which simply does edge detection of each signal, mono-shots this for about 1-2nS, and OR's the result. Along the lines of what Barry and FvM have been thinking.

as long as the signals are not precisely in sync, this will produce a better result than all the band-pass filtering, sin(x)/x calculations and mixer methods.

Just adding my 2-bits ;-)

 

[Tunelabguy]

The best option seems to be one which simply does edge detection of each signal, mono-shots this for about 1-2nS, and OR's the result. Along the lines of what Barry and FvM have been thinking.

as long as the signals are not precisely in sync, this will produce a better result than all the band-pass filtering, ...

This is about the worst solution. It produces an irregular pulse train that could be anywhere between 200,000,000 and 220,000,000 pulses per second, depending on the relative phase of the two input signals and their precise frequencies.. The most likely count is 216,000,000. (I leave that as an exercise for the reader, assuming 1nS mono-shots.)

Mixing, bandpass filtering, and limiting is straightforward and will give precise results, as long as the input signals have reasonable frequency stability.

 

[kripacharya]

This is about the worst solution. It produces an irregular pulse train that could be anywhere between 200,000,000 and 220,000,000 pulses per second

who said anything about a requirement for a regular pulse train ? and i did say the 2 signals should NOT be in sync - implying non-aligned transitions.
thanks for the vote of confidence btw.

 

[Tunelabguy]

who said anything about a requirement for a regular pulse train ? and i did say the 2 signals should NOT be in sync - implying non-aligned transitions.
thanks for the vote of confidence btw.

Even if the two signals are not in sync you will still get about 216,000,000 pulses per second out because about one in five of the 20 MHz pulses will just happen by random chance to fall close enough to a 200 MHz pulse so that it will not be counted separately. The only way your scheme could work is if the two signals were in fact in sync, but offset in phase just enough so that all the 20 MHz pulses missed all the 200 MHz pulses - a very unlikely scenario.

 

[kripacharya]

Even if the two signals are not in sync you will still get about 216,000,000 pulses per second out because about one in five of the 20 MHz pulses will just happen by random chance to fall close enough to a 200 MHz pulse so that it will not be counted separately. The only way your scheme could work is if the two signals were in fact in sync, but offset in phase just enough so that all the 20 MHz pulses missed all the 200 MHz pulses - a very unlikely scenario.

since the OP is not commenting, neither will i.
all da best !