Procom DPF 70/6 Duplexer Tuning

[sammy54322]

I've recently got a Procom DPF 70/6 5-7MHz split duplexer and I'm trying to tune it with a NanoVNA-H4 and I'm having some issues with it. This is a fairly popular duplexer here so having never bought/tuned one before I thought it was a safe choice.

I have managed to tune the high side as best as I can with the dynamic range of the NanoVNA, however, the low side is causing me some problems that -3.5dB loss can never be changed now matter how much the adjustments are made.

I have calibrated the VNA a few times now over some different frequency ranges covering the two that I'm tuning for. I have also tried to tune for a different pair of frequencies and had the same issue with about -4dB loss. The metal where the tuning screws are mounted is also a mounting bracket which seems to have been dropped and is bent but the tuning screw looks to be perpendicular to the threads it is in. One of the low side tuning screws is also a fair bit harder to turn than the other 5 so all I can think is this might be part of it.

If anyone can tell me if it being dropped or bumped in shipping could cause this sort of issue that can't be fixed with tuning that would be really helpful. I suspect this was cheap on eBay for a reason now and might need to be returned.

 

[prcguy]

I have a number of Procom duplexers and have probably tuned a good 30-40 of them. Their typical size flat pack notch only are some of the best I've used and I prefer them to Celwave and any other mfrs due to the smooth tuning and great lock mechanism on the tuning screws. They tune up easy with at least 75dB isolation and about 1.2dB insertion loss, maybe 1.5dB for a bad one. They also make a larger version that is rated higher power and those typically need a 10MHz split, otherwise the insertion loss is higher at 5MHz like 2dB or more. Is yours the typical 6 cavity flat pack with about 1" wide cavities or are they larger like 1.5" across?

Tuning any of these is a bit of an art and sometimes a single cavity can get tuned way off before you know it. In these cases you really need to take a step back, detune all cavities a bit and verify you know exactly where each notch is then stack them back over each other and maximize the total notch depth.

Your plots look wacky like you might have two notches over each other then one is out in left field. Take the time to detune it enough to identify each of the six notches and maybe tune to a 7MHz split by moving one side over 2MHz just to find them and line them all up. You should get a little less insertion loss at 7MHz. Then after you have them all identified and lined up walk the extended side back down 2MHz keeping track of all three notches on that side until they are on your desired freq and you have peaked them all to at least 75dB down.

There should be factory plots and I would recommend looking them up to compare with what you are getting. The factory plots and notches will be fairly symmetrical between hi and lo side with no glitches or unusual shapes across the passband. If you have anything that looks like your current plots you are still doing something wrong.

And the only time I've really seen a problem with a flat pack notch only is when someone broke the solder on a semi rigid coax jumper on the end of a cavity and I had to use a large heat gun and solder it back.

 

[sammy54322]

The one I have is the 6 cavity version with the 1" wide cavities. I've also made up cables for this instead of using a few adaptors on each end and I think that has helped a bit.

Thanks for the advice, I've followed what you said and I've got much better results now which I've attached here too. Moved all of them out of tune and worked on getting two aligned as best I can. The high side looks much better with just two aligned and there's one clearly off to the side. On the low side I have done the same and get similar results, however I do have an extra dip in loss, not sure what that is if you know? It doesn't seem to move about at all. It does look like I just didn't have a clue what I was doing and it works pretty well.

I'll need to keep spending more time on this and get the third cavity tuned now. I'll see if I can find the time to drive to a friend and use their test equipment and they'll show me as well. Didn't really want to send it off to a shop and have it done because there's no learning in that.

 

[prcguy]

The one I have is the 6 cavity version with the 1" wide cavities. I've also made up cables for this instead of using a few adaptors on each end and I think that has helped a bit.

Thanks for the advice, I've followed what you said and I've got much better results now which I've attached here too. Moved all of them out of tune and worked on getting two aligned as best I can. The high side looks much better with just two aligned and there's one clearly off to the side. On the low side I have done the same and get similar results, however I do have an extra dip in loss, not sure what that is if you know? It doesn't seem to move about at all. It does look like I just didn't have a clue what I was doing and it works pretty well.

I'll need to keep spending more time on this and get the third cavity tuned now. I'll see if I can find the time to drive to a friend and use their test equipment and they'll show me as well. Didn't really want to send it off to a shop and have it done because there's no learning in that.

Your plots still look disturbing and you should have closer to 75dB notches and cleaner more predictable traces going in and out of the notches. Are you terminating the unused port with a good 50 ohm load? I usually have the VNA generator into the antenna port and the hi or low side going into the measure port with the unused port terminated tuning just the hi or low side independently. I might use a wide sweep like you have initially to find the notches or wherever the duplexer is tuned, then I switch to just a few MHz wide sweep and concintrate on just the hi or low side to maximize the notch depth. Then I might re-cal and look at about a 7MHz sweep for a 5MHz split looking for distortions or glitches in the trace before final touch up.

I've found using test cables cut to 1/2 wavelengh or multiples for the band including velocity factor can give more reliable results, especially if the device under test can vary a lot from 50 ohms. Or terminating the DUT with 6dB pads to force a good resistive match at the device. But a NanoVNA may not have enough dynamic range to make up for 6 or 12dB loss in the test setup.

Whenever possible I also use final cables between the receiver and transmitter to the duplexer cut to 1/2 wavelength with velocity factor and that seems to lessen problems in the system over using a random length. Using 1/2 wavelength multiples will replicate the impedance where a random or 1/4 wave multiple can become an impedance transformer if the load is not 50 ohms or is really reactive, etc.

 

[sammy54322]

Your plots still look disturbing and you should have closer to 75dB notches and cleaner more predictable traces going in and out of the notches. Are you terminating the unused port with a good 50 ohm load? I usually have the VNA generator into the antenna port and the hi or low side going into the measure port with the unused port terminated tuning just the hi or low side independently. I might use a wide sweep like you have initially to find the notches or wherever the duplexer is tuned, then I switch to just a few MHz wide sweep and concintrate on just the hi or low side to maximize the notch depth. Then I might re-cal and look at about a 7MHz sweep for a 5MHz split looking for distortions or glitches in the trace before final touch up.

I've found using test cables cut to 1/2 wavelengh or multiples for the band including velocity factor can give more reliable results, especially if the device under test can vary a lot from 50 ohms. Or terminating the DUT with 6dB pads to force a good resistive match at the device. But a NanoVNA may not have enough dynamic range to make up for 6 or 12dB loss in the test setup.

Whenever possible I also use final cables between the receiver and transmitter to the duplexer cut to 1/2 wavelength with velocity factor and that seems to lessen problems in the system over using a random length. Using 1/2 wavelength multiples will replicate the impedance where a random or 1/4 wave multiple can become an impedance transformer if the load is not 50 ohms or is really reactive, etc.

Yep, the other port is terminated and I've checked the load and it seems to be good, used another one and the same result.

I thought it looked easy enough on the videos and reading online but like everything it'll need a bit of practice and like you say there is an art to tuning these. I've been busy with work but I'll be back to trying to get it right on days where I've got a bit of time and I'll see about going to see a friend that has the right tools and, most importantly, the experience.

I have managed to find out that the duplexer itself isn't damaged so I'm not in a rush now. I'll be taking the stuff I've been using as well when I go to get it tuned so they can tell me what I'm doing wrong and I'll post the final results here in the hopes that it can help someone else if they have the same issue.

Thanks for the help on this though, just seems to be above my skills at the moment.

 

[prcguy]

Yep, the other port is terminated and I've checked the load and it seems to be good, used another one and the same result.

I thought it looked easy enough on the videos and reading online but like everything it'll need a bit of practice and like you say there is an art to tuning these. I've been busy with work but I'll be back to trying to get it right on days where I've got a bit of time and I'll see about going to see a friend that has the right tools and, most importantly, the experience.

I have managed to find out that the duplexer itself isn't damaged so I'm not in a rush now. I'll be taking the stuff I've been using as well when I go to get it tuned so they can tell me what I'm doing wrong and I'll post the final results here in the hopes that it can help someone else if they have the same issue.

Thanks for the help on this though, just seems to be above my skills at the moment.

I've retuned a number of these that others were having trouble with and I usually detune them to see all three notches per side on slightly different frequencies, then carefully tune them where they belong a little at a time into one big notch on the desired freq. The end result gets you the minimum 75dB notch and clean looking skirts around the notches and minimal insertion loss.