50 MHz Bessel Filter

Thread Starter

cb83

Joined Aug 9, 2007
24
I want to design a Low Pass Bessel Filter with a cutoff of 50MHz. I know how to design a simple active one with op-amps. The signal to be filtered is coming on 50 ohm BNC coax.

My question is how do I handle impedance matching? Can I just take the center conductor on the coax and connect it to a standard Bessel filter input and pass the 'outer' conductor straight through or do I need to do something to impedance match?

Also will op-amp designs be noisy? If anyone has good information on analog filter design methods other than using op-amps I would be very interested. Information on Passive filter design would be interesting as well.
 

Papabravo

Joined Feb 24, 2006
21,159
I would suggest the ARRL Handbook for information on passive RF Filters. Practically speaking a 50 MHz. passband will be challenging, I don't think it can be done. The matching impeadance will be in parallel with the center conductor and the shield. The problem will be finding a match that looks like 50 Ohms across DC-50 MHz.
 

Thread Starter

cb83

Joined Aug 9, 2007
24
Does that go for all filters, Butterworth, Chebyshev etc? I know these can be purchased with 50 ohm impedance. Do they just have very complex impedance matching circuits?
I have a very good passive 40MHz Butterworth LPF that I am currently using. I didn't design it though. I need something with linear phase, so I'd like to see if I could replace it with a Bessel Filter.
 

Papabravo

Joined Feb 24, 2006
21,159
Does that go for all filters, Butterworth, Chebyshev etc? I know these can be purchased with 50 ohm impedance. Do they just have very complex impedance matching circuits?
I have a very good passive 40MHz Butterworth LPF that I am currently using. I didn't design it though. I need something with linear phase, so I'd like to see if I could replace it with a Bessel Filter.
Yes, all filters have implementation problems at higher frequencies. In any case as you go up into the VHF range (>30 MHz.) the physical construction of filters becomes more difficult because the parameters of a component change. Inductors start to look like capacitors because of the capacitence between the windings, and capacitors start to look like inductors, and stripline techniques take over. If you have the budget I would suggest you acquire a network analyzer so you can see the actual behavior of these things.

Again I would be interested in the actual response of those purchased filters over that wide a range. Have you actually measured the insertion loss at very low frequencies like at DC?
 

Thread Starter

cb83

Joined Aug 9, 2007
24
The BW filter I told you about has very little insertion loss for DC values. While I don't have a network analyzer to test everything, the results over the entire range have been great for our application.

The insertion loss is very low at DC. The spec is 0.2 dB I believe, and while I haven't done any real accurate tests myself, I can see that it's definately less than 1dB. TTE is the company that makes them.
 

Thread Starter

cb83

Joined Aug 9, 2007
24
50 ohm BNC coaxial cable is meant for applications using frequencies of 50Mhz or even higher. All of these high frequency applications have to impedance match to the 50 ohm cable somehow. If it can't be done with a simple resistor in parallel then how is it done?
 

RiJoRI

Joined Aug 15, 2007
536
I am going to show my glaring ignorance here, in hopes of lessening it.*
Would it be feasible to put multiple bandpass filters in parallel, with each filter optimized as well as possible for its band? So f'rinstance, while the 0-10K filter does very poorly at 10M, the 900K-15M filter would be doing very well.

--Rich

*Actually, I hope to be declared a genius, but experience shows otherwise! :(
 

Thread Starter

cb83

Joined Aug 9, 2007
24
That's a good thought, but wouldn't work for my application. I would have to somehow add the 3 filtered signals back together before the A/D converter and if the frequency responses of the 3 filters overlapped at all that would cause problems, I think.
 

Papabravo

Joined Feb 24, 2006
21,159
How the impedance match happens depends on what the cable is connected to. The output impedance of the driver also needs to match the characteristic impedance of the cable. At the receiving end the type of amplifier you are using is very important. Are you using an IC or do you have a discrete amplifier?

BTW Check out AppCAD from Agilent. It's a piece of freeware you may find helpful.
http://www.hp.woodshot.com/
 
Top