Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Phase noise after comb generator

Not open for further replies.


Advanced Member level 1
May 13, 2004
Reaction score
Trophy points
Heart of Europe
Activity points
low phase noise comb generator

We knwo that a divider /N reduces the phase noise of signal by 20*log(N).

Suppose we have a crystal oscillator, and apply a comb generator, times M. Is the phase noise of the M-th harmonic 20*log(M) higher then?

Any references on that?

comb generator harmonic output of 13 dbm

You are correct. The easiest way is to do the math. Write out the cos(wt + phasenoise) and take the nth power. wt and phasenoise gets multiplied by n.


    Points: 2
    Helpful Answer Positive Rating
comb generator phase noise

Thanks flatulent,

How do you design a frequency multiplier for MW frequencies using discrete
technology, I would like to multiply a 120MHz clock to st about 4GHz. Feasible or not?

74ac phase noise

Comb generators were used before PLL methods reached the microwave frequencies. One disadvantage of combs is getting rid of all of the spectral lines you do not want. Another is the very low power conversion efficiency (signal level of what you want/signal level of driving signal) If you do not mind these problems, a step recovery diode is the usual way to go. There are some books and utility programs around which help in the design of such things.

One slightly improved method is to get to the goal in two steps. First multiply 120 up to around 500 and filter it and amplify it before multiplying again up to the final frequency.

This is correct. SRD can go up to 20GHz. For PLDRO using the SPD that includes the SRD. Last years there are several papers show good results using FET transistor.

Hello Radiohead and all the others.
I think you should be able to get this 4 GHz reasonably easy.
I would propose the following (I already did it and it works):

1-First use any fast digital cmos IC like inverters available in the market now, something like 74AC familly and make a 120MHz square wave out of your signal.

2-Try using standard lumped element filter to extract some harmonic arroung 300, 400 or even 500 MHz. Try to get output power arroung 100mW if possible (odBm).

3-Now use one of those ICs available from:
like ERA3 and amplify this signal to about 10dBm

4-Use a second chip ERA3. It will already generate lots of harmonics at the output.

5- Use a microstrip band-pass filter to extract the 4 GHz harmonic.

The phase noise using this method is in general much lower than using plls.
I have used this method to generate a phase reference and it was much more stable than 2 professional generators!!!
Hope it helps


    Points: 2
    Helpful Answer Positive Rating
In my opinion, use PLL would be much better: smaller size, better out-band phase noise than Mr.sinatra's method. With a well designed PLL, there are less spur than frequency multiplication. At 4GHz, the componets are easily be found like Analog Device ADF4113 PLL chip and Hittite VCOs. Try it...

Hello zcr35 and all the others.
zcr35, I understand your point.
In my case I compared with 2 professional generators and the generators were by far worse.
But maybe those new chips can do better. I remember I had checked the documentation of oe of those generators at the time I made that phase reference and I found out that it was very complicated and let's say "burocratic". The control loop of the whole generator was probably adding quite a lot of noise.....
I think those new chips sholud be much simpler,. Probably generating much less noise in the internal control loop and maybe providing better phase noise performance, etc....
I think it is also a good option to buy some of those chips and give them a try.
But I still believe it is important to compare the 2 methods....

radiohead said:
Thanks flatulent,

How do you design a frequency multiplier for MW frequencies using discrete
technology, I would like to multiply a 120MHz clock to st about 4GHz. Feasible or not?
but as there are several ways of acheiving 4 GHz you have to tell us your need. With a multiplication factor of ~80 -->> 38 dB of increased noise wrt to your 120 MHz oscillator. If you have a good oscillator to start with you will have a noise floor around -100 dBm at 4 GHz (assuming noisefloor for base osc. at -140 dBm) Not to shabby, but you can pretty easy do better by using a PLL. Simple VCO's have 10-20 dB better noise floor but you will have worse close in noise... If you go for a mixed scheme using either a multi loop PLL or use PLL's to clean up in stages you should be able to have good phase noise both close-in and low noise floor, but it will be a lot more complicated than just multiply and filter.

Lots of good reading here: **broken link removed**


I wonder Sinatra's approach is valid for ALL cases.

Could Sinatra tell us what is the multiplication factor in your design, and what professional generators do you use to make the comparsion?

I think both sides (Mr. Sinatra and PLL sides) are right. Everything depends on the actual application.

HEllo all.
In my case I was going from 100MHz or 500MHz to around 4GHz.
The multiplication factor was not so high ( 8 times using 500 MHz at input). Input and output signal levels around 0dBm.
As filters at 4GHz I had played with 2 types, one microstrip coupled line (designed by me), 6th order chebyshef, and another comercial filter, interdigital, probably 10th order (I don't remember) and with response like chebyshev, very low insertion loss.
The performance of both types of filters were similar and I remember that the microstrip one had higher loss and worse attenuation in higher frequencies but still usable.
And also another important remark, I was using low phase noise generator to provide my input signal. I didn't need to design the oscillator myself.
Unfortunately is t is too long ago and I don't remember the models of the 2 generators I had used, but the brand was HP. One of them could go to more than 10GHz, the other one something around 3GHz.
Hope it helps.
Greetings to all.

Dear Sinatra,

Can you kindly provide us phase noise mask of your solution? :)
Better is not the best. HP produces eqipment for different applications, so it isn't a fact that generators you use were phase noise optimized.

Best Regards,

Several here have mentioned a SRD for the comb generation. This will work well but the phase noise of the amplifier to drive the SRD will be a factor, in fact it is probably much higher than the crystal. In general, a BJT will have less phase noise than a FET type amp. Also the desired comb signal will be attenuated and need to be filtered. Post amplification of the 4GHz signal will add more noise.

If power supply is clean, post amplification add noise within tolerance.

Tony, it is not the post amplifier per se that causes the phase noise. It is the fact that the conversion loss of a comb generator is quite high (maybe 30 dB), so the power level of a single comb tooth might be very low. kTbF noise + NF of post amp means that your far out phase noise will degrade.

Another phenomenon is if you use a comb generator to drive a mixer as the effect the mixer becomes a sampler and samples the noise at +/- IF frequency at EACH comb tooth, vastly increasing the mixers effective noise figure.

I have seen comb generators used successfully. PTS makes a line of synthesizers that depend on a comb generator in the frequency expander portion. Never the less, I do not trust them fully!
Not open for further replies.

Part and Inventory Search

Welcome to