VCO Driving a LPF-Stopband refection problem

It is a pity that you can’t use a level 7 mixer with a 2dB pad between the filter and the VCO.

I am not sure what this "frequency pulling" effect is you are mentioning. Perhaps you could explain.

By the "first harmonic", I assume you mean the second harmonic in standard terminology. The fundamental would be the first overtone or the first harmonic according to standard terminology.

A VSWR of 40 means a reflection coefficient pretty close to unity. Then any reflected signal will get absorbed by the VCO (whose output reflection coefficient is not stated). Have you tried this system or are you trying to decide if it will work on paper before buying? Trying it should resolve the problem. Failing that, ask mini-circuits.

Suppose you take the reflection coefficient of the filter down from 1 to 0.5 by buying a better one. The VSWR has gone from 40 to 3, but the reflected signal only halved.

Depending on the bandwidth required, stability may be obtained by adjusting the line length between the VCO and the filter.

In response to "logbook", frequency pulling is a MiniCircuits spec for how much the output frequency of the VCO will shift as a function of load mismatch. Also, as you point out, I meant to say second harmonic.

I am laying out a prototype PCB at this point. I talked to MiniCircuits and they were not sure how much of a mismatch their VCO could tolerate. They said that " a VSWR of less than three should be OK". When I probed the underlying mechanism of why mismatch caused problems I did not get any coherent answer.

As to your better filter suggestion, any filter with with good stopband rejection will have a VSWR of at least 20 in the stopband. One colleague of mine suggested I use a diplexer to absorb the stopband signals. This would probably work, but would add to the circuit complexity.

Your suggestion of a 7dbm mixer is probably the best solution. Any feel for how low below the local oscillator spec I can go and still get linear mixing? Maybe 7dbm is enough for the 10dbm mixer I want to use?

Thanks for your suggestions

To GOTWW: Could you explain more on your suggestion to adjust line length? Do you mean the length of the PCB trace? How do you calculate the proper one to help the mismatch?

Direction-couple it into a terminator ( sorry for the new verb)

<nbucska@pcperipherals DOT com> subj: eng-tips

Neil,

Thanks for the explanation of the frequency pulling. I checked the mini-circuits glossary and it agrees with what you said. What I didn’t understand is where it comes into your thinking. Your load is constant at a given frequency so the frequency pulling affect is just a frequency calibration error, which is pretty uncertain in a VCO anyway. You said it wasn’t a problem so why am I still going on …

When you said that a filter necessarily has lousy VSWR outside its passband I felt uneasy about the factualness of this assertion. Take a look at the Mini Circuits application notes section on filters, the note on constant impedance filters. The net result is that it is possible to make (and presumably to also buy) filters with reasonable return loss outside the passband.

I was just thinking of the internal construction in terms of paralleling a low pass and a high pass filter with equal corner frequencies. The high pass drives a dummy load and this will absorb the unwanted HF, rather than reflecting it.

I can’t quite get my head around your desire to use a level 10 mixer with a 7dBm signal. Why not just use a level 7 mixer? At 3dB below the designed mixer level you should not get any problems, just 0.5dB-1dB conversion loss. (See mini-circuits app note "How to select the proper mixer").

To test how easy it is to make a more constant impedance filter, I simulated a 50 ohm 3 pole Butterworth low pass filter. Sure enough with ideal components the VSWR shoots off to infinity outside the passband. The point is that reactive filter components cannot absorb the signal so they have to reflect it. The finite VSWR of the real filters suggest losses in the actual components; not a good thing to rely on. I then put a series RC across the input to the filter, the resistor being 50 ohms. This high frequency load was set to the corner frequency of the filter, the idea being to absorb the HF signal. Sure enough the VSWR peaked at around 3 then came back down below 1.5. It is true the in-band filtering was no longer sharp, but that is just a question of how sharp you need it to be and what order of high pass filter you are prepared to use.

To "logbook",

The main reason I want to use a 10dbm mixer is that I happen to have a few of these in my parts stash.

Your suggestion of the series RC is basically the diplexer scheme I mentioned earlier. It should work but adds a little complexity. I am trying to keep the design as simple as possible. Maybe the right approach is for me to layout the PCB to allow the extra components, and then try the circuit first without the the RC loaded to the board.

I still would rather find an explanation for the mechanism that describes what happens in the VCO when there is a large reflection coefficient. Then I could decide more intelligently what the right approach should be.

Thanks again for your help.

By adjusting line length, the second harmonic input impedance can be adjusted from open to short circuit. It could be that the load pulling be the least sensitive in one quadrant of the smith chart.

To GOTWW:

Sounds interesting. Is there a reference text you can refer me to that elaborates on this? I'm a long time engineer but relatively new to RF.

Thanks

If you take a short-circuited length of transmission line 200ps long it will appear as an open circuit at 1.25GHz but a short circuit at 2.50GHz. This is a "quarter wave transformer".

It goes back to conditional or unconditional stability in high frequency amplifiers. I have seen amplifiers that are unstable looking into dc-open wg-coax adapters, yet are stabile if looking into dc-shorted wg-coax adapters. This may not apply to your VCO, but checking the load-pull effect versus 2nd harmonic reflection coefficient phase angle is an easy fix that is easy to test.