THD by bandwidth

[CopperCone]

How spectrally pure can a signal be generated in different bands?

I know audio can be very pure (i.e. the nearest harmonic is <160dB down with real nice oscillators, like using linear-optical feedback)

How does it go up?

I am particularly interested in the microwave band.. like

1-2 GHz
2-4 GHz
4-8 GHz
8-12 GHz
12-18 GHz
18-26.5 GHz

but the other bands are interesting too.. I don't know how to call them.

20KHz-190KHz ( biological ultrasonicsonic)
190-10MHz (high ultrasonic)
10-100MHz ()
100-300MHz ()
300-1000MHz ()


How pure can microwave generators get?

I also posted this in the regular forum, I will consolidate the responses later. This forum is kind of dead and this is general information .

[BigBoss]

The harmonics can be suppressed by filters.These filters are generally tracking filters for signal generators.But design a tracking filter is really really difficult at especially higher frequencies.
VCO is also important so that oscillator is designed with low level harmonic levels.In order to maintain this, the designed should optimize the oscillator so that the oscillator will work in a region  which the active device is driven by optimized one.It's more difficult than saying.
But more important than harmonic levels that is the Phase Noise.
The signal purity consists of these two figures.Harmonic Levels and Phase Noise.
Unfortunately, these cannot be obtained simultaneously except in a high-tech equipment.

[CopperCone]

Good information, but I would like to kind of compile a table that shows what the various technologies are capable of. I have seen these tables say, comparing wein bridge oscillators, to tuning forks, to dds, to wave shaping.. for lower frequencies,, but I wondered what the microwave region looks like

and of course it all depends on what you are interested in. for some applications harmonics are the most important.

or, are you saying that its basically, really really Dependant on what test equipment you buy?

[DaJMasta]

Some of the best equipment commonly available, like Keysight's PSG series, advertises -55dBc harmonics specs or better.  The problem is that while the signal can be filtered as well as any other range, there's problems in trying to get so much dynamic range and so much frequency span out of an instrument.  If you want a single pure tone at 10GHz, then you design your frequency source, then you filter out everything else and you have an extremely pure tone, well above such advertised specs, but if you want the same instrument to be able to do that at 5GHz, 1GHz, and 100MHz.... that's a tall order.

Part of the problem of making it difficult to do in reality is impedance matching of the load.  Even a tiny amount of reflected power or a tiny mismatch in impedance (maybe you wiggled your cable, maybe you didn't clean the connector as thoroughly as you needed to, maybe the aging of the capacitors in your matching network has skewed them a little off their initial characterization) will dramatically increase the power of tones outside of the one you want.  And while again, it is "easy" to match these sorts of things to a single frequency (not that it's simple, but that it can be done and done reliably), trying to get that over a wide frequency and amplitude range with common modulation options and stuff is quite a challenge.



And I really think you'll be hard pressed to do -160dB harmonics in the audio band without a laboratory setup or tons of averaging and processing.... even if your harmonics are that kept low, it's a tall order to get even 120dB of dynamic range digitized because of thermal noise, surface leakage, and other effects... so 4 orders of magnitude below that?  That's one hell of an analysis system that can verify that spec... probably cryo cooling is involved.

[CopperCone]

to answer your question about audio, here is what I was thinking of

http://www.janascard.cz/PDF/An%20ultra%20low%20distortion%20oscillator%20with%20THD%20below%20-140%20dB.pdf

140 db not 160.. but I want to say that forum member "GK" (from australia I believe) made something ~155db.

nevermind, his was 120. I guess 140 is the high for now (thought I'm sure it can be improved...)
https://web.archive.org/web/20151013065311/http://www.users.on.net/~glenk/thd/thd.htm


also, I guess I meant more like, oscillator topologies and related drifts. Like the base generator.. since all your filters will add tempco/other errors.

I know anything can be cleaned up.. but I was interested in sources, the core of what makes the signal. Like what requires the least cleaning.

Like with an audio source you can use
dds, tuning fork, crystal, wein bridge, negative resistance.. and you can catalog those as having some kind of distortion.

of course you can post filter them, but I am interested in the generation stage.

[dmills]

The thing about RF is that usually we don't really care about harmonics from an oscillator because for any given frequency they are relatively easy to clean up,  FAR more interesting are things like phase noise which cannot easily be cleaned up, and ADEV and this is where you see the money being spent when it comes to single frequency sources. 

73 Dan.

[G0HZU]

Several years ago I designed a few microwave test oscillators to test my 22GHz HP8566B spectrum analyser on the first few ranges. For example, for range 2 I designed an oscillator up at about 4.4GHz and this gave -40dBc or better for all harmonics and the idea was to make a basic/cheap oscillator that would have similar or better phase noise than the HP8566B at (say) 100kHz offset. This was to prove that my analyser was fairly healthy (in terms of stability and phase noise) as I didn't have a low phase noise sig gen in those days that could generate signals up at many GHz.  The oscillator was just a printed resonator with a couple of capacitors and a low cost ($3?) MMIC amp from Mini Circuits. It cost maybe $5 total in parts and gave similar/better phase noise than the HP8566B up at 4.4GHz.

[BigBoss]

Good information, but I would like to kind of compile a table that shows what the various technologies are capable of.
and of course it all depends on what you are interested in. for some applications harmonics are the most important.

or, are you saying that its basically, really really Dependant on what test equipment you buy?

There is no such comparison table for Microwave signal sources.I mean nobody says "this type oscillator is better than that type in term of harmonics" etc.
Every design has its own specification and feature.It very depends on the frequency,oscillator topology and of course designer's skills and experiences.
While one is obtaining a kind of harmonic level with a topology, another one may catch a lower harmonics with another topology.
In fact RF and Microwave oscillator design is an art.However even-tough the principles of the oscillators are well established there are still some "unknowns" about them such as chaotic behaviors,multiple resonance states etc.
For testing of oscillators you should have at least a Spectrum Analyzer with-if it's possible-Phase Noise measurement feature.Zero span is very necessary to observe the settling time of the oscillator.

RF and Microwave Oscillators are principally

Hartley, Colpitts,Clapp,Clapp-Gouriet,Negative Resistance,Traveling Wave,Magnetron,Klystron,YIG Based,Di-Electric Based,Coaxial Resonator Based, etc. oscillators.
There are some academic researches but those above are practically proven ones.

[CopperCone]

I'm sure you can make a table, I guess microwave is just less studied at this point to be able to do so.

Page 4 of this document has one for lower frequencies.

http://www.ti.com/lit/an/snoa665c/snoa665c.pdf

It looks like we would need to break up the various oscillator technologies mentioned by bigboss by bandwidth..

by the way, bigboss, you can still make generalizations and spreads that "gunn diode signal generators in the ___ band will typically have parameters between ___ ___ _ _"

having this sort of information available makes a design easier to approach.

All elements you add to a signal chain lessen its stability. Each of those things can drift. You are typically getting amplitude and phase drift by adding filters.

For instance, if you take a 0.2ppm drift per kelvin DAC and put a filter in front of it, it turns into a 90ppm capacitor drift (foil capacitor) per kelvin .. or a 20ppm for a ceramic.
And you have a more complicated filter, you get more drift.

Aim small miss small.

Phase noise is only important if you are interested in it I guess its imporant for some reason or another for conventional use of microwaves but that does not mean we should not have the ability to choose an option that balances amplitude. phase and frequency drift, + harmonics generation.

I am not necessarily interested in the conventional uses, this is of academic interest.''



 BAND                                      HF VHF UHF L S C X Ku K Ka V W mm

Hartley,
Colpitts,
Clapp,
Clapp-Gouriet,
Negative Resistance,
Traveling Wave,
Magnetron,
Klystron,
YIG Based,
Di-Electric Based
,Coaxial Resonator Based


This is even difficult at audio frequencies.

perhaps slowly with crowd sourcing and people with experience we can slowly fill this list in. I guess I am just not interested in phase noise. Having this information available may lead to insights. We should not ignore a parameter because filters are possible/practical. This implies our thoughts are lead by economics and industry rather then science.