Crystal Filter Response

[fuseless]

I have a 10.7 MHz crystal filter that needs to be impedance matched once placed in the IF strip. That's easy enough to do and I'm going to use L- networks on the input and output of the filter to do the job. I did the impedance matching on the Smith chart just for fun.

But how will I know when the matching networks are adjusted properly once placed between the first mixer output and the second mixer input?

 

[JimB]

But how will I know when the matching networks are adjusted properly once placed between the first mixer output and the second mixer input?

By plotting the frequency response of the system.

You can do this manually with a variable frequency signal source and a meter of some kind.
or
you could use a spectrum analyser and tracking generator.

JimB

 

[fuseless]

Yes but if the filter is in the IF strip, it isn't connected to a tracking generator with a swept frequency span. So the only signal I would see would be the IF frequency, 10.7 MHz in this case, right?

 

[JimB]

The matching circuits will affect the frequency response of the filter.

Using a spectrum analyser and tracking generator sweeping +/- 50kHz from 10.7MHz will enable you to see the response curve of the filter.

Inject the signal upstream of the input matching circuits, and measure the output downstream of the output matching circuits.

JimB

 

[fuseless]

I plan on building an active high impedance probe for the output as soon as I finish my main PCB. The probe will have a 1M Ohm input impedance and should be able to drive a 50 Ohm load(RF input of spectrum analyzer) easily. So that takes care of the output of the filter, no problem there. I should mention here that the output impedance of the first mixer and the input impedance of the second mixer are both 330 Ohms. Since the output impedance of the tracking generator is 50 Ohms and since the matching network is to transform the input impedance of the filter from 2581-j1040(3000//2pF) Ohms to 330 Ohms, won't the frequency response be effected by the 50 Ohm impedance of the tracking generator?

 

[Nigel Goodwin]

Is there any particular reason you feel the need for impedance matching?, it's fairly normal not to do so in IF chains as there's no long lengths of cable to cause issues, only short connections, low frequencies, and voltage transfer rather than power transfer is far more efficient. Power transfer is only theoretically capable of a maximum 50% efficiency.

 

[fuseless]

I'm not concerned about power transfer. I was under the impression that since the output of the first mixer is 330 Ohms, it needs to see 330 Ohms or the response of the filter will be distorted.

 

[Nigel Goodwin]

I'm not concerned about power transfer. I was under the impression that since the output of the first mixer is 330 Ohms, it needs to see 330 Ohms or the response of the filter will be distorted.

Less distorted if it feeds a high impedance, and less loss.

 

[unclejed613]

all you need to do is find the schematic for a late 1980s or 1990s FM receiver, and look at the IF section and see what was done there.... if i remember correctly they were designed pretty simple when they used crystal filters in the 10.7Mhz IF section... when i worked at NAD, we used an FM signal generator and a simple 60hz sine wave generator to generate the sweep signal, set the deviation to 200khz. we then fed the 60hz audio into the X channel of an oscope, and connected a RF sniffer (voltage doubling rectifier probe using 1N4148 diodes, and a couple of 30pF caps) to the circuit under test. set the X channel for full deflection and you have a 200khz wide window, and get a spectrum on the screen

 

[fuseless]

Gentlemen,
I might have been in slight error by saying that the response of the filter would be skewed because the output of the first mixer doesn't see its output impedance of 330 Ohms. What I learned, or think I learned on YouTube however was that the crystal filter needs to see its own terminating impedance looking out of the input and output ports respectively. This is the reason for the L matching networks.

Ok, this is what I'm going to do when the time comes. I'll use a spectrum analyzer with a tracking generator. I'll isolate the output of the first mixer from the tracking generator. I'll put the high impedance probe from the spectrum analyzer input across the output of the filter's L network to the second mixer input. Now, a 280 Ohm resistor in series with the tracking generator's 50 Ohm output impedance equals 330 Ohms, the first mixer's output impedance. With the output of the first mixer isolated from the tracking generator, I'll inject a 10.7 MHz signal with a 100 KHz span into the L network at the input of the filter. The input and output L networks are then adjusted until the proper response is obtained. In my case the response of the properly terminated crystal filter should be: 3dB passband = + - 7.5 KHz, Stopband = 18dB + - 25 KHz, Passband Ripple = .5dB, Insertion Loss = 2dB with a terminating impedance(Input, Output) of 3000//2pF or 2581-j1040 Ohms. Finally, after the proper response is obtained, the test equipment probes can be removed and the output of the first mixer can now be coupled in to the input network of the filter. Since the input of the input matching network was terminated with 330 Ohms(50 Ohms + 280 Ohms)when the network was adjusted, the impedance looking back from the input matching network should be the same 330 Ohms when the output of the first mixer is coupled back into the circuit with the input probe removed. In other words, the input matching network was designed to match the filter input impedance of 2581-j1040 Ohms to some impedance, in this case 330 Ohms, the first mixer's output impedance.