Two oscillators resonate with the same crystal and one is offset. Possible?

[neazoi]

Is there any oscillator configuration where two independed oscillators can resonate with the same single crystal, but one of them is allowed to be offset by say 30Hz?

This way, if the frequency of the first oscillator changes by say 10Hz, the frequency of the second will change by the same ammount, i.e their difference will be always the same, if fed to a mixer.

Is that even possible with a syngle crystal or other weird means?

What I try to achieve, is to get the same difference out of a mixer, no matter of temperature changes etc, without using an oven.

This could be done by feeding the same oscillator to the RF and LO ports of a mixer, but then the difference would be zero, not 30Hz or so.

A possible way I am thinking is to feed the LO with an oscillator and then feed the RF with the same oscillator, but through a divider. This could produce a difference that would not change I think, but certainly not 30Hz.

 

[biff44]

sure. The crystal is just a BIG value inductance in a lot of oscillator type circuits. So you can tune one active device to one frequency, and the other device to another slightly different frequency.

or you can run one oscillator at a 3rd overtone frequency and divide down by 3 (the overtone is not exactly 3x the fundamental usually).

I would be nervous about blowing up the crystal, as they do not take that much maximum power.

 

[volker@muehlhaus]

Is there any oscillator configuration where two independed oscillators can resonate with the same single crystal, but one of them is allowed to be offset by say 30Hz?

30Hz? You will see injection locking and both resonators will move to the same frequency. This will happen even if the oscillators are coupled very loosely and each uses its own crystal.

 

[neazoi]

30Hz? You will see injection locking and both resonators will move to the same frequency. This will happen even if the oscillators are coupled very loosely and each uses its own crystal.

Hm... that is indeed a problem if one crystal is to be used. But there are cures for it. Look at this circuit https://www.ab4oj.com/test/imdtest/main.html I think this will do the isolation easy.

- - - Updated - - -

sure. The crystal is just a BIG value inductance in a lot of oscillator type circuits. So you can tune one active device to one frequency, and the other device to another slightly different frequency.

or you can run one oscillator at a 3rd overtone frequency and divide down by 3 (the overtone is not exactly 3x the fundamental usually).

I would be nervous about blowing up the crystal, as they do not take that much maximum power.

Very interesting... The fact that the harmonics are not exactly doubles or tripples etc is very important to your proposed way.
But, the amount of drift on the fundamental will be different from the harmonic. You can't have a stable output frequency by mixing the fundamental and the divided-down harmonic. OR can you??? (wondering)

Oh, I just realized that my idea for mixing with a divider is idiot... It is different from what you propose anyway.

 

[volker@muehlhaus]

Hm... that is indeed a problem if one crystal is to be used. But there are cures for it. Look at this circuit https://www.ab4oj.com/test/imdtest/main.html I think this will do the isolation easy.

I like your optimism.

But my experience with this topic is that it's really hard to solve. About 20 years ago we developed this frequency source (OEM work for Dressler) based on two oscillators (~1GHz and 1-1.2GHz) that are mixed down to cover a final frequency range of 100kHz to 200MHz. Two oscillators in separate shielded boxes, each PLL stabilized, but the small coupling via the mixer was enough for injection locking.

In theory we could have generated really low frequency output with this architecture, but in reality we had terrible issues with injection locking of the oscillators if we moved the frequencies too close. That's why the lower frequency limit is specified as 100kHz, not 10kHz.



View attachment synthie2.pdf

 

[FvM]

Using a crystal simultaneously for fundamental and 3rd harmonic can (theoretically) work, although I don't see a reasonable purpose for it. But I'm quite sure that sharing a single crystal by two fundamental oscillators is not possible.

 

[neazoi]

Using a crystal simultaneously for fundamental and 3rd harmonic can (theoretically) work, although I don't see a reasonable purpose for it. But I'm quite sure that sharing a single crystal by two fundamental oscillators is not possible.

I see...
The third harmonic will drift 3x times than the fundamental due to possible thermal changes, isn't that true?
So one could not actually do what said in post #2, if he wishes to obtain a stable output frequency by mixing.

"run one oscillator at a 3rd overtone frequency and divide down by 3 (the overtone is not exactly 3x the fundamental usually)."

If you try to mix the fundamental and the divided harmonic, you wont't get always the same difference at the mixer output, because the fundamental thermal drift amount will be different than that of the harmonic.

Is my thinking on this reasonable?

 

[Warpspeed]

If you want two very close frequencies that track exactly, the best solution is a modulation/demodulation scheme using balanced mixers.

The audio people have been doing this for years, with public address systems to prevent acoustic feedback. Basically you feed your audio into a magic box that shifts the entire audio spectrum up or down by about 2Hz.

If you think about it, you can then massively increase the system gain where the microphone picks up direct energy from the surrounding loudspeakers, without getting howl round.

You could also do this just as easily at radio frequencies.
Start off with an rf source at X Mhz, then generate a frequency at X mhz plus 30 Hz (or minus 30Hz) in the exact same way.

Its done by driving two balanced mixers with ninety degrees out of phase signals, with 30Hz and the carrier frequency.
Its exactly like generating single sideband by the phasing method, where the carrier and opposite sideband are suppressed.

You then demodulate that the reverse way, again with ninety degree phase shifted signals, so you get another carrier wave offset by 30 Hz.
Which sideband you use decides if the frequency shift is up or down.

It works very well with plain discrete frequencies, and as the audio guys are doing, it also works with broad spectrum modulation such as natural audio.

In the age of fast digital counters, generating two exactly ninety degree out of phase signals is trivial. Its even easier at 30 Hz. Some pretty good balanced mixer modules are available as well.

Its definitely the way to go with this, you can tune your rf carrier over a wide range and generate a constant 30 Hz offset second carrier.

 

[neazoi]

You could also do this just as easily at radio frequencies.
Start off with an rf source at X Mhz, then generate a frequency at X mhz plus 30 Hz (or minus 30Hz) in the exact same way.

Its done by driving two balanced mixers with ninety degrees out of phase signals, with 30Hz and the carrier frequency.
Its exactly like generating single sideband by the phasing method, where the carrier and opposite sideband are suppressed.

I am amazed it can be done this way, If this is the case why they don't use this technique to generate reference oscillators without the need for ovens or gps?

Will the I/Q channels in this quadtrature mixer you describe be appart by some frequency? I thought that only the phase changes in an I/Q mixer, not the frequency of the generated signals!!

I may be loosing something, please describe this technique in more detail.

 

[volker@muehlhaus]

I am amazed it can be done this way, If this is the case why they don't use this technique to generate reference oscillators without the need for ovens or gps?

Isn't that a completely different topic? Creating some frequency offset tells nothing about the accuracy of the carrier frequency.

 

[Warpspeed]

Will the I/Q channels in this quadtrature mixer you describe be appart by some frequency? I thought that only the phase changes in an I/Q mixer, not the frequency of the generated signals!!

This is one very well known way to generate single sideband radio signals. Google "single sideband phasing method".

A single sideband radio signal has no carrier wave. If you have a 10mHz SSB transmitter and you whistle into the microphone at 1Khz, the transmission is a single carrier at 10,001 Mhz if its upper sideband. If lower sideband, it will be at 9.999 Mhz.

When you simply amplitude modulate a carrier wave, the output consists of three frequencies. The original carrier, Carrier + modulating frequency, and carrier - modulating frequency.
New frequencies are indeed generated.

If you use two modulators, driven out of phase, and the outputs combined, one sideband adds, with complete cancellation of the other (if the process is perfect).
Balanced mixers null out the carrier, (if the process is perfect).

By combining these methods you can get rid of both the unwanted sideband and the original carrier, and be left with only a single new frequency.

 

[neazoi]

Isn't that a completely different topic? Creating some frequency offset tells nothing about the accuracy of the carrier frequency.

If you create a frequency offset that varies the same rate as the original frequency, then if you mix these two signals, their difference will always be the same no matter of the stability of the original. This is because both the original and the offset are allways apart the same amount, no matter of temperature variations.
Am I right?

- - - Updated - - -

This is one very well known way to generate single sideband radio signals. Google "single sideband phasing method".

A single sideband radio signal has no carrier wave. If you have a 10mHz SSB transmitter and you whistle into the microphone at 1Khz, the transmission is a single carrier at 10,001 Mhz if its upper sideband. If lower sideband, it will be at 9.999 Mhz.

When you simply amplitude modulate a carrier wave, the output consists of three frequencies. The original carrier, Carrier + modulating frequency, and carrier - modulating frequency.
New frequencies are indeed generated.

If you use two modulators, driven out of phase, and the outputs combined, one sideband adds, with complete cancellation of the other (if the process is perfect).
Balanced mixers null out the carrier, (if the process is perfect).

By combining these methods you can get rid of both the unwanted sideband and the original carrier, and be left with only a single new frequency.

Yes Tony, I know how this works.
But I am a bit confused, because you mix two topologies of ssb generation.
First you refer to the phasing method with the polyphase audio filter. Right, this should shift the audio phase but not the frequency.
However, afterwards you refer to the balanced mixer, which produces a DSB-SC signal and then the sideband can be filtered by a filter, which is the filter method of ssb generation.

If you use the secong way, you indeed have two signals with difference 2x the audio, BUT if the audio frequency changes their difference changes. So the LSB and the USB do not track each other.
Unless I am missing something?

 

[Warpspeed]

The mixing (modulation) process always creates two mirror image sidebands.
These could each be a discrete single frequency, or a mixture of continually changing frequencies and amplitudes such as normal audio.

The only requirement to generate single sideband this way, is to drive each mixer with signals that are ninety degrees phase shifted.

This is known as the phasing method of generating SSB. No frequency selective filters are required.


Its slightly more difficult with audio, but only because phase shifting an audio signal by precisely ninety degrees is not so simple.

Radio amateurs have been doing it this way for about fifty years.

With two fixed single frequencies (such as you are planning to use) the phase shifting part is trivial.
Usually a two stage twisted ring counter (Johnson counter) clocked at x4 will create four outputs all in EXACT phase quadrature.

Two of those counters, plus a pair of balanced mixers, and a resistor network for combining at the output are all that you require. Thus creating the third very clean output frequency with any residual rubbish typically down about -35db to -40db.

 

[betwixt]

I'm siding with FvM on this one.

Certainly there are all kinds of mixing systems that produce low frequencies or small offsets by combining two similar frequencies but I can't see any way of producing two non-harmonically related close frequencies from a single crystal oscillator simultaneously.

It would be asking the crystal to oscillate at two rates at the same time. The only way I can see to do it is to derive two new signals from the single oscillator, divide them differently then mix the products together. Two PLLs running from the same reference with different divisors.

Brian.

 

[volker@muehlhaus]

If you create a frequency offset that varies the same rate as the original frequency, then if you mix these two signals, their difference will always be the same no matter of the stability of the original. This is because both the original and the offset are allways apart the same amount, no matter of temperature variations.

Sure, but what's the point?

You have carrier frequency f0 with some accuracy +/- df0.
You have offset frequency f1 with some accuracy +/- df1.

Now you create signals at f0 and f0+f1, and then mix it down to have (f0+f1) - f0 = f1 again. True, the error in df0 cancels out, and you again have f1 with the accuracy +/-df1. But what's the point here?

 

[neazoi]

I'm siding with FvM on this one.

Certainly there are all kinds of mixing systems that produce low frequencies or small offsets by combining two similar frequencies but I can't see any way of producing two non-harmonically related close frequencies from a single crystal oscillator simultaneously.

It would be asking the crystal to oscillate at two rates at the same time. The only way I can see to do it is to derive two new signals from the single oscillator, divide them differently then mix the products together. Two PLLs running from the same reference with different divisors.

Brian.

The problem is that harmonics do not vary linearly with the fundamental, so it would be very hard for one to calculate the required harmonic and the division ratio for equal final products offsets. In fact these could never be perfectly tracking, but the error could be minimized.

What about these crystal filters? I mean these hc-49 case crystals that have 3 pins, like some ceramic filters. I think that these have two crystals inside right? But if they are closely matched in temperature (if they are), they could be used in two independent oscillators
I know, however these are two crystals, which is not related to the original post, already answered for a single crystal, but how well will these track each other?

 

[Warpspeed]

The only way I can see to do it is to derive two new signals from the single oscillator, divide them differently then mix the products together. Two PLLs running from the same reference with different divisors.

That would certainly work.

And I agree with both of you that a crystal only oscillates at one frequency, either its fundamental, or forced overtone.

Another approach to all this might be direct digital synthesis.
As it is now possible to get a complete DDS system on a chip, a pair of those could be programmed to track with a fixed frequency offset.

 

[neazoi]

Another approach to all this might be direct digital synthesis.
As it is now possible to get a complete DDS system on a chip, a pair of those could be programmed to track with a fixed frequency offset.

A DDS is as stable as the reference oscillator is, I think. I may be wrong though.

If we could create two frequencies that track exactly each other independent of temperature variations, then a reference signal could be produced. It is this tracking that I am trying to investigate here, which is in fact very hard as I see.

 

[volker@muehlhaus]

If we could create two frequencies that track exactly each other independent of temperature variations, then a reference signal could be produced.

You are trying to cheat physics.

With your mixer concept, you have just moved the frequency uncertainty from the carrier source to the offset source. Now that offset source must be stable.

 

[Warpspeed]

DDS is as stable as the source, its direct frequency division through a phase increment latch, plus a digital to analog conversion in cases where a sine wave output is desired.

Two DDS dividers running from the same source would track exactly.
If the obtainable DDS frequency increments were 1 Hz, you could generate two frequencies 1Hz apart very easily.