900 MHz duplexer tuning

[Mogman]

So the guy I am putting together a 900 MHz Quantar for brought me his duplexer to tune.
It is a 4 can Cellwave but I cannot make out the model, the original frequencies were at 950 MHz and if I am seeing correctly there was only 4 MHz separation but the numbers are hard to make out.
I have only tuned the RX side so far and I am getting 108 Db of rejection but I am seeing 2 Db loss on the RX frequency, I have never tuned a 900 MHz duplexer but if it was a VHF I would think 2 Db is pretty high.
So comments are welcome!
THANKS!!
EDIT, sorry I am tuning it with a 25 MHz offset 902.xxx RX and 927.xxx TX

 

[prcguy]

I typically remove the harness and tune each cavity individually so you can see the exact tuned freq with loss and the notch freq with depth. More often than not when you reassemble the harness the duplexer will be tuned ok and you can't improve it much if any. If you tune it as a complete assy you can loose track of a notch and it can add insertion loss.

Once you have tuned each individual cavity out of the harness you can try and touch it up as a complete assy.

 

[kf8yk]

Original tune of 950ish MHz with a 4 meg split indicates this came from a Part 101 microwave link. You might want to try swapping the TX & RX ports as depending on which end of the mw link this was used on it could be built for high in/low out or low in/high out.

The other concern is power handling; in the original application it only needed to handle a few watts, the Quantar can do quite a bit more.

 

[Mogman]

Original tune of 950ish MHz with a 4 meg split indicates this came from a Part 101 microwave link. You might want to try swapping the TX & RX ports as depending on which end of the mw link this was used on it could be built for high in/low out or low in/high out.

The other concern is power handling; in the original application it only needed to handle a few watts, the Quantar can do quite a bit more.

They had it on a MSF5000 repeater for years, IIRC that repeater was making more than 100W..
Can anyone tell me what I should expect?

 

[prcguy]

They had it on a MSF5000 repeater for years, IIRC that repeater was making more than 100W..
Can anyone tell me what I should expect?

The duplexer ports may have been labeled TX and RX but as mentioned you do need to keep the high and low sides high freq and low freq for your new pair. Swapping those can cause problems with how the notches tune and fall into place. Otherwise in my opinion going from 950MHz range to the 902-927MHz range should not be a problem as its a small percent of BW.

I have a pair of TX-RX brand duplexers factory tuned in a 930MHz commercial band and the specs mentioned a 3.6MHz split. They tuned up nicely in the 900 amateur band and work great with very low insertion loss and high isolation at a 25MHz amateur split and are now fed by 100W Quantars. There are duplexers by Celwave and others made for cell phone service and a 45MHz split where you may have to derate some specs for a narrower 25MHz amateur split.

I also picked up a couple of broad band window filter duplexers by Telewave and Wacom (cheap!) and retuned for 900 amateur. These turned out great with high isolation and I can move the repeaters anywhere in the 1MHz BW 902 and 927 range without retuning. Perfect for an emergency frequency agile repeater. There are a lot of options out there for 900 duplexers in case yours doesn't tune up right.

 

[Mogman]

I typically remove the harness and tune each cavity individually so you can see the exact tuned freq with loss and the notch freq with depth. More often than not when you reassemble the harness the duplexer will be tuned ok and you can't improve it much if any. If you tune it as a complete assy you can loose track of a notch and it can add insertion loss.

Once you have tuned each individual cavity out of the harness you can try and touch it up as a complete assy.

I did this on the TX side and swung the adjustments widely to be sure I was not on an image.
I had the exact same results, 1 Db per can of insertion loss on the TX frequency, showing 62.5 Db rejection per can.
Combined I have 2 Db insertion loss and 110Db rejection which is not 2 X 62.5 but I always see less than the math would suggest on that end.
I do have the high side on TX according to the original frequencies and the fact all the stems are at the same basic height I am confident that is not the issue.
I plugged in all the info I could into AI and its response was a well tuned duplexer would have between 1.5 and 2.0 Db insertion loss on 900 MHz depending on design, so I am going to conclude that is just what this set will and will not do.
2dB is not the end of the world, at a 35% loss, heck even if you could get it down to 1 DB that would still be 20% loss
He has a second duplexer, may as well give that one a shot.

On 900 it's all about terrain for the most part.
Funny story, I worked for 37 years as a com tech for a major pipeline company, the last 25 years it was mostly 900 MHz SCADA.
We had to move a MDS master from a tower that was condemned, about 5 miles down the road to another tower.
I did the paper work for the move and the original license was in feet, the FCC had moved to metric, we ended up with the antenna at 750' HAAT, it was fed with LDF5-50A.
There was not much of that 5W getting to the antenna but it worked like a hose all over that part of S. TX

 

[prcguy]

I did this on the TX side and swung the adjustments widely to be sure I was not on an image.
I had the exact same results, 1 Db per can of insertion loss on the TX frequency, showing 62.5 Db rejection per can.
Combined I have 2 Db insertion loss and 110Db rejection which is not 2 X 62.5 but I always see less than the math would suggest on that end.
I do have the high side on TX according to the original frequencies and the fact all the stems are at the same basic height I am confident that is not the issue.
I plugged in all the info I could into AI and its response was a well tuned duplexer would have between 1.5 and 2.0 Db insertion loss on 900 MHz depending on design, so I am going to conclude that is just what this set will and will not do.
2dB is not the end of the world, at a 35% loss, heck even if you could get it down to 1 DB that would still be 20% loss
He has a second duplexer, may as well give that one a shot.

On 900 it's all about terrain for the most part.
Funny story, I worked for 37 years as a com tech for a major pipeline company, the last 25 years it was mostly 900 MHz SCADA.
We had to move a MDS master from a tower that was condemned, about 5 miles down the road to another tower.
I did the paper work for the move and the original license was in feet, the FCC had moved to metric, we ended up with the antenna at 750' HAAT, it was fed with LDF5-50A.
There was not much of that 5W getting to the antenna but it worked like a hose all over that part of S. TX

A typical 1/4 wave cavity used in a duplexer is set for about .5dB loss and with two cans, cables and a Tee it should be about 1.2 to 1.5dB max per side. There will of course be a little more coax loss at 900 compared to VHF/UHF. Some of the six cavity UHF duplexers I have can do about 1.2dB insertion loss per side meaning they are set for less than .5dB loss per can.

 

[Mogman]

So I ended up tuning all the cavities separately and had the same results about 1db loss and around 60db rejection, when I put it all together I expected around 63 watts at the antenna port but actually only had a little under 45 watts and one of the TX cans would get warm after about 5 min TX So I am thinking they need more attention as in there is something wrong with them.
This is the Cellwave duplexer

I found this setup in my junk pile, it used to be a bandpass filter on a SCADA site that was decommissioned.

I was able to tune these to where I had around 1.5Db insertion loss and being bandpass only the rejection was -65DB at 25 MHz

I can probably improve the insertion loss as I did not spend allot of time on it, but I plugged the two high side cans between the Quanter and a watt meter/dummy load and got 66W which is much closer to what I would expect, the other 4W was probably lost in the 9913 cables I was using.
I am not sure what these smaller adjustments do, the rods are too short to go into the cavity, I am assuming they were the reject adjustments that were "disabled" when building these into bandpass filters.

But he ain't getting these cans, I am keeping them for myself, I am curious that if I am correct about the short rods if they only used longer rods or was there something attached to the rod inside, if it is the former I should be able to convert these to bandstop/pass.

The again is 65Db enough for a Quantar at 927 MHz????

 

[Mogman]

The RG142 jumper between the TX cans is getting pretty warm, that is a mystery as RG142 should be able handle allot more than the 75-80W that it is passing.

 

[prcguy]

The RG142 jumper between the TX cans is getting pretty warm, that is a mystery as RG142 should be able handle allot more than the 75-80W that it is passing.

Over an extended time RG-142 will get warm at 100w, that's normal. It will handle a good 500w at 1GHz but that's at a vey high temp.

 

[prcguy]

So I ended up tuning all the cavities separately and had the same results about 1db loss and around 60db rejection, when I put it all together I expected around 63 watts at the antenna port but actually only had a little under 45 watts and one of the TX cans would get warm after about 5 min TX So I am thinking they need more attention as in there is something wrong with them.
This is the Cellwave duplexer
View attachment 204107
I found this setup in my junk pile, it used to be a bandpass filter on a SCADA site that was decommissioned.
View attachment 204108
I was able to tune these to where I had around 1.5Db insertion loss and being bandpass only the rejection was -65DB at 25 MHz
View attachment 204110View attachment 204111
I can probably improve the insertion loss as I did not spend allot of time on it, but I plugged the two high side cans between the Quanter and a watt meter/dummy load and got 66W which is much closer to what I would expect, the other 4W was probably lost in the 9913 cables I was using.
I am not sure what these smaller adjustments do, the rods are too short to go into the cavity, I am assuming they were the reject adjustments that were "disabled" when building these into bandpass filters.
View attachment 204112
But he ain't getting these cans, I am keeping them for myself, I am curious that if I am correct about the short rods if they only used longer rods or was there something attached to the rod inside, if it is the former I should be able to convert these to bandstop/pass.

The again is 65Db enough for a Quantar at 927 MHz????

60dB is not enough tx/rx isolation for any repeater and for a four cavity Bp/Br I think 85dB would be a goal. How are you measuring isolation? Is it just the two cavaties are down 60dB at 25MHz or are you measuring from the tx port to the rx port with the ant port terminated?

 

[Mogman]

60dB is not enough tx/rx isolation for any repeater and for a four cavity Bp/Br I think 85dB would be a goal. How are you measuring isolation? Is it just the two cavaties are down 60dB at 25MHz or are you measuring from the tx port to the rx port with the ant port terminated?

I am measuring across the high side cans, just the two cans with the RG142 jumper, I get the same numbers on the low side cans.
I put the tee between the RX/TX cans, stuck a little 50 ohm load on the RX port and I lost about 3 watts at the ant port, the Quantar is reporting 4 watts reflected, the Bird is showing about 1-2 watts reflected from the dummy load and its feed line.
There are three adjustments on the big cans (plus the one that does nothing) there is allot of interaction between the input/output/pass adjustments I can make the skirts steeper but at the cost of insertion loss, I need to play with them some more.
I have not measured from TX to RX port, I would expect to get the RX cans rejection of the TX Freq, plus the insertion loss of the TX cans, but will try.
I found a piece of 1/4" aluminum rod, when I go back to the shack tomorrow I will insert it in the "do noting" hole and see if it magically turns into a BP/BR can
Of course that does not help the guy I am putting the Quantar together for....

 

[prcguy]

I think I have some of the same cans which are 3/4 wave 900MHz band pass only. They have adjustable coupling that should be able to do .5 dB insertion loss. If they are pass only and no notch that's not going to work out well for a duplexer. If the pass band is changing a lot when you combine them with the harness it could be the wrong length cables and that is time consuming to diagnose. You basically have to look at return loss through a pair of cans and a cable and if its not good as in 8-12dB return loss then you would tune up the cans higher or lower in freq to see which way improves the return loss.

I think about 17-20dB minimum for return loss of two cans and an interconnect cable is a good goal and it can get much better like approaching 25dB when everything is perfect lengths. Working with wrong length cables can pull freqs around on the cavity as its not presenting a good 50 ohm match and the notches may not reach desired depth. But if you have pass only cavities then I would not spend any more time with them and get a proper Bp/Br duplexer designed to cover the 900 amateur band.

 

[Mogman]

Yep, looks pretty much the same but for the one terminal that has dual N connectors and maybe taller, I will find out tomorrow whether my hunch is correct.
Have you ever tuned this can? If it is a pass only I am wondering if that shorter rod is too short to actually go into the cavity like mine are.
I got real close to .5Db per can, when combined it tossed in another .5Db for a total of 1.5DB due to interaction.
It's not like I will be installing a 900 MHz Quantar any time soon, although I have a couple getting on towers is not near as easy as it used to be.
But I would like to get this guy up if he can come up with a better duplexer, I have radios to build for some volunteer first responders so I am likely going to push this aside for a week or so, at least after tomorrow when I play with the TX/RX cans.

 

[Mogman]

Eureka!!!
Inserted the 1/4" aluminum rod in one of the TX RX cans and got this!


Notice this is not calibrated.
So with a couple hours fiddling with one can (4 adjustments and they all interact)
I ended up here, undetectable insertion loss and 43Db of rejection at the RX frequincy.
Notice the "set reference" light is lit meaning the unit is in calibration, anyone know where I can get some silver plated rod?

 

[prcguy]

Interesting to say the least. If you take on a project to make them a pass/notch cavity then tune them up as a duplexer you are a braver man than me.

 

[Mogman]

I am starting to remember why I enjoyed working with our low band radios (49.xxx MHz) more than the SCADA stuff, 900 MHz can be very frustrating.
I connected the above setup but the return loss was so bad it would trip the PA into failure.
So I tuned three of the adjustment in/out and pass for max power transfer, I ended up at 70W, I also noticed if I simply put a barrel connector in place of the can I was seeing 80W so I was only loosing 10W in the can but 20W attenuation in 10 feet of 9913 seems awful high.
With this setting I also reduced the rejection of the RX frequency to 25Db and that will not get me where I want to go.
Time to set these aside for a while.
The guy is going to bring me a WACOM duplexer, why he did not start there I have no idea.
I went to trade school in Waco and when I was taking the two way course we toured the WACOM facility, hatted to hear of their demise, there were some great folks working there.
EDIT
I just went through the coax loss calculator, even with a perfect match I would lose 15W and that does not take into consideration the connectors/barrel connector so I guess there is no mystery there.

 

[prcguy]

I am starting to remember why I enjoyed working with our low band radios (49.xxx MHz) more than the SCADA stuff, 900 MHz can be very frustrating.
I connected the above setup but the return loss was so bad it would trip the PA into failure.
So I tuned three of the adjustment in/out and pass for max power transfer, I ended up at 70W, I also noticed if I simply put a barrel connector in place of the can I was seeing 80W so I was only loosing 10W in the can but 20W attenuation in 10 feet of 9913 seems awful high.
With this setting I also reduced the rejection of the RX frequency to 25Db and that will not get me where I want to go.
Time to set these aside for a while.
The guy is going to bring me a WACOM duplexer, why he did not start there I have no idea.
I went to trade school in Waco and when I was taking the two way course we toured the WACOM facility, hatted to hear of their demise, there were some great folks working there.
EDIT
I just went through the coax loss calculator, even with a perfect match I would lose 15W and that does not take into consideration the connectors/barrel connector so I guess there is no mystery there.

When coax is operated in a mismatch condition the advertised loss can be much higher.

 

[Mogman]

When coax is operated in a mismatch condition the advertised loss can be much higher.

The Wacom is a strange one, it has one BP/BR can and one BP can on each side, it was tuned to .1375 where I am tuning .0125 so I did not have to even touch the pass adjustment and only had to "touch" the reject adjustments, BUT because it only has one BP/BR can on each side it only had 65Db of TX freq. reject (from antenna port to RX port)
I am beginning to understand the measuring from TX to RX port thing as there is another -65 DB to the RX frequency in the TX cans

So if I have over 85 DB total RX freq. rejection from the TX port to the RX port (with 50 ohm load on antenna port) then I should be OK?


EDIT I have only about 12 watts total insertion loss through the duplexer (compared to a barrel connection between the transmission lines)
With some better cables I am getting just under 80 watts to the ant port when considering the loss in the cable between the tee and the bird 43 (72 measured at the bird) I am running a load on the RX port, not convinced yet to connect the receiver.

 

[prcguy]

The Wacom is a strange one, it has one BP/BR can and one BP can on each side, it was tuned to .1375 where I am tuning .0125 so I did not have to even touch the pass adjustment and only had to "touch" the reject adjustments, BUT because it only has one BP/BR can on each side it only had 65Db of TX freq. reject (from antenna port to RX port)
I am beginning to understand the measuring from TX to RX port thing as there is another -65 DB to the RX frequency in the TX cans

So if I have over 85 DB total RX freq. rejection from the TX port to the RX port (with 50 ohm load on antenna port) then I should be OK?


EDIT I have only about 12 watts total insertion loss through the duplexer (compared to a barrel connection between the transmission lines)
With some better cables I am getting just under 80 watts to the ant port when considering the loss in the cable between the tee and the bird 43 (72 measured at the bird) I am running a load on the RX port, not convinced yet to connect the receiver.

I would test for isolation going from the transmit port to the receive port with the antenna port terminated in 50 ohms and at both the tx and rx frequencies. That will be a combination of tx and rx passbands plus notch contribution. Your setup might reach 85dB with a 25MHz split. On UHF it usually takes at least two Bp/Br cans in each leg to get 85dB but the split there is 5MHz and your twice the freq but with more than four times the UHF split.