Harmonic filtering

Hi all,

I would like to have a word of advise from you all. I want to design a harmonic filter, that suppresses 2nd, 3rd and 4th harmonic but leaves base frequency untouched. The technology is microstrip, frequency 5GHz. Any ideas are welcome...

Low pass filter is a simply way.

Stepped-impedance low-pass filter would be a good choice. What's the harmonic suppression requirement? Max passband ripple allowed?

Build a Bandpass Filter with corner frequencies 4.99GHz and 5.01GHz. I recommend you to avoid lowpass. Depends on application filter is going to be applied it is wisely you may want to consider is it going to be Butterworth or Chebyshev.

Now you want us to design the filter in microstrip for you? Give any information detailed like: Er,Thickness of microstrip,Tagent Loss etc. which are relevant and I will synthesize a filter for you. I can even post a gerber file for you to mill it.

Hi,

ADS can also design the filter for you. Use the "Smart Component" option and the design guide.

Also, if you are talking about harmonics generated by a transistor, using bias tees comprising quarterwave lines and radial stub for the DC bias will help remove the even harmonics. Eg, 1/4W for 1GHz is 1/2W for 2GHz so what is a open cct for 1GHz is a short for 2GHz.

I know how to design filters, that is not the problem.

When designing a bandpass filter, it always is recurrent at 3*f0, 5*f0 etc. Yet I want to suppress all harmonics from a fixed oscillator. So I am in need of an easy way to filter harmonics - every one.

Besides, a very smallband bandpass filter would require very strict technology tolerances - not quite possible on low budget.

Suppression needed: preferrably >30dB
Bandpass loss: as low as possible

You build a bandpass filter you make sure all 3*f0 5*f0 are gone by attenuating them. This what filters do. Pass frequency you want and attentuate range you does not want.

To design a filter is not a problem in lumped elements. Everyday I think people make software to design them in lumped elements. The problem arises in implementing into in transmission line. microstrip for example. This where Richard transforms, Kuroda identities come in play.

Well, what is your low budget microstrip material? Does it have an Er,Width,Height,Tagent Loss,Dielectric Loss, etc?

You take them and synthesize with filter you designed from what came from Kuroda transforms.

I will state it clearer:

20MHz bandwidth at 5GHz = 0.4% relative bandwidth.
* what filter order would you need? 10? 15?
* what pcb manufacturer can be sure that he does much better than this 0.4% tolerance (reasonable physical size) without trimming? And how about tolerance on dielectric constant?

Okay, so let's stick to the low order bandpass filter then. I would need a bandpass filter and a low pass filter in cascade. So far so clear. But you should know than these designs cannot simply be cascaded. They only work properly in a 50 ohm environment. So very strange things might happen when cascading the two of them.

Lumped elements: They don't work properly at these freq's, because of parasitics. And even if you could account for this, the component tolerance would make your design a hell. Microstrip clearly is the way to go.

What I need is a simple microstrip filter. A simple structure to suppress harmonics.

PS: "low budget" material is rogers 4003. What I referred to was that I cannot afford to spend unnecessary money when not needed.

First remember the oscillator harmonics are low relative to the fundamental. You can simulate it, but let say for the 2nd harmonic it might be -15dBc and even better for so narrow bandwidth VCO.
You found several idea such as BSF or LPF and I think it is a good idea to simulate them all. Now becouse it is narrow bandwidth think of a simple filter to filter the odd harmonics seperatly then filter the 2nd also. If you use for example an Lamda/4 open stub for the 3rd harmonic it will reject the odds, and a short at the fundamenyal so the 2nd will be rejected.

I do not understand why do you want to cacade lowpass after bandpass?

**Of course we do not design them in lumped elements but in schematics we do, then we transform them in transmission line (your case microstrip).

But first you need to shop what microstrip you will use. There are parameters, needed before deciding, I have included a picture we use to synthesize it. The supplier of microstrip substrate must make them available to you.

Mentioned rogers 4003, but I can find many of them with er 4.3 or 3.8, which one is it? Be specific.

Now seriously: how would you implement a parallel tank in microstrip of 10.3pH and 98pF in parallel? That is an impedance of 0.32Ohm. On RO4003 3.38 material, a 3 Ohm line is 54mm wide and a quarter wavelength is 8.2mm long. Now that wouldn't work out huh! That's your present for making a fifth order filter ...

If you want this small bandwidth, you need a much higher order to relaxate on these kind of specifications. But even then you must be sure all your resonators are at the right frequency. And that's only possible by trimming.

Now I don't want to go further into this. Your solution is mathematically okay. But I doubt if you have a lot of experience on making a practical filter design yourself.

You propably are right, It would not work in lengths you just mentioned.

try to use defect ground structure to suppress 3rd and 5th harmonica, for the more higher order harmonica , the microstrip circuits will malfunction at those frequencies.
I think low pass will be better than band pass when considering insertion loss.