combining signal from two microwave power amplifiers? Diplexer Discussion.

[CopperCone]

if I had 2 generators being buffed by amplifiers, what would be the best way to combine the signal into an antenna?

signal gen -> PA -> isolator(s)
signal gen -> PA -> isolator(s)

Would combining them in a regular T-junction work (N-type) or do I need a special microwave circuit element here?

signal level ~ 10 W from each amp.

What if I also wanted to connect a spectrum analyzer to tihs circuit, can I just put an attenuator and another T junction? Is there a 'proper' method? (all to 1 antenna to measure reflections/NLJD).

Circulators are NOT an option because I see that the minimum order quantity for multi octave circulators is 5, at a price of 400-800$ each, I won't be catching the 3rd harmonic with the 'cheap' mass produced options.

Surprisingly high frequency (mid C band +) multi octave power circulators are available for ~+600$ in single quantity... but some cursory experiments showed that measurements above 4 GHz in a NLJD seemed to be unstable (then again, I just had a PA (with 80db isolation) and SA (with 20 attenuator to protect maximum power input of SA) connected to a T junction with a SA. This requires a very broad band ridged  horn

I think for this experiment to work with higher frequencies maybe its enough to use zip tie holders on the microwave cable (i am using high grade microwave cable... good to 26ghz)... but I found during my experiment that just standing next to the spectrum analyzer (a few feet away from the horn being aimed across the room) it was enough to just turn my body to significantly effect the harmonic readings being taken on a test subject (in this case a organic material with a blue-tooth radio tire pressure gauge inside of it)... any kind of motion on the cable had a great effect too. seem to be able to detect it somewhat even with 2 inches of wet organic material (cantaloupe) infront of it.... but only small difference. I assume the fruit absorbs ALOT of energy. Consistent readings with 2, 3 and 4 GHZ stimulus signals (but the 12Ghz of the 4ghz stimulus is difficult to measure.. very wonky).

it seems that commercial NLJD have a full scale of around 3db.

also, a trick, if you want to do it yourself, to equalize the 2nd and 3rd harmonic energy in the 'null' reading is to use a band pass filter and play with the frequency until the readings are equal. they will equalize in amplitude because there will be modeing (seen as a 'comb like' transfer function on the 2nd and 3rd harmonic of the pass band. if you get lucky with the filter you can operate on the edge of its pass band and have the modeing actually do you a favor.... (i found this on a 2 ghz microwave waveguide filter). I had two of the, one was disassembled (due to being 1900 mhz and the other one 2044), and it had little disks/rods in it (kind of like the tower in the first men in black movie where the alien bug hides his space ship in new york city).

I don't see my horn going above 18GHz, meaning 6GHz stimulus signal maximum. And I have no idea if this is even true, I need to do an experiment with 2 horns ( i have two identical ones) to determine 3db points. Perhaps I will try with zip tie holders first, then hardline.

I also have a very high frequency horns (18 ghz and 24 Ghz) set (lens horn, round conical shape (not rectangular prism pyramid)... I am interested in finding mixing products of semiconductor devices, as mentioned by another EEVBLOG user in my thread about NLJD. However, these horns only have a bandwidth of like 6GHz, as they are meant for communications purposes, but do have a high gain, I would need a smaller ridged horn to deal with higher frequencies in broad band...

I kind of see using two different microwave frequencies in a NLJD similar to doing geneology on a circuit (like mendel squares)... perhaps it will have a higher specificity.

[4CX35000]

Out of curiosity - what are you trying to do ?

Sampling the output using done with a coupler (Probably 30dB for example) were the forward and reverse power can be measured using a spectrum analyser. This allows measurement of the output and see if the transmitter meets regulations or not. Harmonics, IP's and so on.

Whilst combining stages; amplifiers are usually combined using 4 port 3dB hybrid couplers. Two ports for the amplifiers, output port and the remainder port will have a test load to absorb any wasted power from the small difference between the stages and the output.

There are plenty of books in the libraries and internet research which cover this type of thing. For example.
https://www.microwaves101.com/encyclopedias/hybrid-couplers

Comb generators are popular on channeled equipment were a couple of oscillators are used to generator a multitude of frequencies so many MHz apart, most cases I have seen them on was to cover output channels on TV transmitters post final mixer and filter off the required frequency. Used on all sorts of equipment at SHF and I think starting to become popular in amateur radio circles.

Around here in my workshop, I have a 800 MHz comb generator originally used for a military satellite uplink, this has a oven crystal oscillator and a voltage controlled oscillator which will generate a output around 800 MHz depending on the voltage. The output has tens of side frequencies which are 10 MHz apart from each other. The whole thing sits inside a small stainless steel case with a SMA connector at the top. Apply 12volt to the oven crystal oscillator and whatever the VCO voltage is required to gain the ideal output frequency and filter off. VCO doesn't change the output by much, probably a few tens of 10 KHz over a 6 volt range from 6volt to 12volt. Might use it on a project one day.

[CopperCone]

I want to combine the output of two signal generators connected to two power amplifiers connected to a single antenna which is also connected to a spectrum analyzer to look for mixing products generated by a nonlinear junction.

So the spectrum analyzer would be tuned to look in an area of an expected mixing product of the two frequencies irradiating the nonlinear junction.


Or, since this would not be an initial ratiometric test (which typically measures the difference in the 2nd and 3rd harmonics, to distinguish between a corrosive metal and a semiconductor), it might be possible to use another antenna connected to the spectrum analyzer aimed at the junction.. but in the end it would be preferred to use a single antenna (wide band dual ridged horn is what I am using).

This would occur in frequencies above 2 GHz.

I managed to get isolators that can isolate the power amplifiers from each other, but any kind of circulator is out of the question due to minimum order quantity and cost for a multi octave device.



So, if two power amplifiers are connected through appropriate directional isolators, and hooked into a standard T junction, which is connected to an antenna, is there any kind of problem?


Then, if that is extrapolated, so there is two T junctions connected, so two power amplifiers and a spectrum analyzer is connected to an antenna, is there a problem? It annoys me because its hard to get a wideband isolator so that the spectrum analyzer can measure things with standard gain, if I switch hardware before taking different measurements it means I need to readjust a sensitive system.

An option is to get a switch and two different isolators to cover each of the mixing products (kind of like a power preselector), i guess.  big problem is that the SA is not compatible with the power levels, so I have a power attenuator.

I kind of like to imagine it as a microwave microscope. with some samples it might be preferable to use two antennas, but both scenarios are interesting.

[4CX35000]

Problem with isolators (Circulator with test loads) is they are not designed for combining transmitters together for the same antenna. You still need a diplexer of some type to combine the two RF signals together, reduces losses and maintains the correct output impedance. You can use the isolators on the output of each amplifier and then combine the outputs of the isolators two together using a diplexer of some type. That combination of equipment would produce two pure signals and protect each amplifier from each other and maintain reasonable output impedance which each amplifier needs to see on its output port.

The easiest combiner is a type of diplexer which many amateur radio books by the ARRL, AMSAT and RSGB cover in reasonable detail. You could also look at using a hybrid ring (Some call it rats race ring) with a pair of stubs or a commutating line type of combiner with a pair of 3dB hybrids. Look around eBay for any former telecoms 3dB hybrids, just check the frequency matches your requirement, otherwise you can look at building your own.

Good video here which will get you started and give you a idea of what is involved.
https://youtu.be/wv79DSqVi9w

Some software which might help.
http://tonnesoftware.com

[CopperCone]

To clarify, without them I am just getting a waste and dissipating power in my isolators right?

Can an amplifier still be damaged so long the reflections from the impedance mismatch past the isolators is dissipated by them and not allowed to reach the amplifier (with enough DBs of isolation)?

[4CX35000]

It really comes down to how much power do you require at the antenna and how much power are you prepared to end up inside the isolator resistors. Yes the isolators will protect the transmitter and yes the isolators will protect transmitters from each other, but combining them with a T piece on the outputs of each isolator will have so much power being lost due to impedance mismatch and all that power will be wasted inside each isolator. If your happy with the loss of power then go with it, but remember to check the isolators can handle the power, check in operation that they are not getting too warm and keep an eye on a transmitter warning/fault indicators (If it has them).

Off top of my head, without working it out properly; you will be 6dB down in power which at 1 watt Tx output gives you 0.25 watt. Hell of a loss. It is not a test I have ever done as I would naturally avoid such mismatches and try to get everything equal to 50 ohms and properly filtered for maximum performance. I suppose if it allows you to do the experiments then what the hell, go with it.

[CopperCone]

Assuming

PA1 -> Isolator -> LPF
PA2 -> Isolator -> HPF

When making a diplexer, how many DB's of attenuation do I need with the filters? That is, how sharp do these filters need to be, or how apart the frequencies need to be, or how much should PA1's frequency be attenuated by the HPF (and vice versa for PA2's frequency be attenuated by the LPF) in order to maintain low loss (from the filters passband, for which I will assume low loss).

 I figure it would be best to get the attenuations provided by the filters as close to each other as possible, then to adjust the signal power to be equal (assuming I want equal powers). (at least in the case of the advanced NLJD described by Dmills and evb149)


I'm not sure how to calculate the loss in the signal power transferred to the antenna vs the filter signal (reflection? it's reflected right.. I never used to think about reflection, just that the impedance is high and it does not tend to go there for whatever reason).

If I am understanding this right, the energy from PA1 (with a low pass filter), will somewhat enter the antenna, and somewhat reflect so many dB's from the high pass filter of PA2, and go back into the antenna. Does the reflected signal also get absorbed by the isolator in this case? The whole concept of reflection and the isolator bugs me, I'm not sure what the hell happens at this point.

Do you know what the minimum theoretical loss is?

And I would prefer not to have losses, but its good to know I can do stuff right now (though limited). And it helps me understand whats really going on. And if it helps at all, this will run in the multi GHz range.

I see that commercial diplexers/duplexers offer some kind of degree of isolation between the channels, and people say they just have some kind of small insertion loss.  I need to make a custom one. I assume its just from the edge of the pass band to the other signal peak (isolation value).

but it would still help to explain it with impedances

[4CX35000]

It comes down to what frequencies/bands you are using. I suppose your using the amateur radio bands ? or the ISM bands ?. It really comes down to that.

In your case it might worthwhile looking for some amateur radio and/or ex telecoms diplexers. They tend to appear on eBay (Search GHz Diplexer) and a little research will reveal if they are suitable or not. The ex-telecoms tend to be well made and remove the hassle of deciding what performance you want from the diplexer.

[CopperCone]

I'm looking at maybe 5 and 7 ghz, some non harmonically related frequencies.

Do you know how to calculate loss based on the diplexer isolation?

[4CX35000]

At those frequencies, you would better off looking for a diplexer which is ex telecoms on eBay or maybe directly source one from a manufacturer. Search for GHz Diplexer or Microwave Diplexer and you should see plenty on offer. Look for one which looks to cover frequencies similar to what you require and then do some further research to see if it can be retuned to match your frequencies. Only the manufacturer would know.

I don't think your using waveguide so that limits your search down to items which are using various types of connectors such as SMA and so on.

Designing one tends to be a complex issue with all the individual cavity design required and this can take weeks depending upon the method of design taken. I have seen radio amateurs develop there own using steel tubes with plates and threaded rods to adjust the plates inside, but that is after spending months experimenting and building the thing. Whilst buying a diplexer the most difficult option will be to tune the cavities using a Spectrum Analyser with a tracking generator and hybrid, again something which only comes with practice and not covered by any textbook.

[CopperCone]

Can't I just use low and high pass filter assemblies? like put a high pass filter on one, then the low pass filter on the other, then connect them with a T? Thats what the video showed. I have a bunch of filters 'rods' already (N, SMA) of various orders.

Since my frequencies are not set in stone, I could just adjust the frequency of one of the signal generators so the attenuation by the filters is equal, as it will track the slope.


i.e. if i did 5 and 7 ghz, lets say the filters come out that the LPF attenuates the 7ghz by 20 db, and the HPF attenuates the 5GHz by 23db, I could move the 5GHz to something like 4.5 GHz so their both 20db (I mean, I would have to experiment, but it seemed to work with a bandpass filter I have to equalize the harmonics (I used the modeing behavior present in the filter to my advantage).

[4CX35000]

I've no idea what you have for filters but give it a try and see what results you get.

[CopperCone]

I got some diplexers, but in the meantime I am wondering if I could use microstrip filters (the PCB filters) to make a diplexer.

I tried a few different filter simulations and it seems that a bandpass filter consisting of a microstrip shunt stub or open stub resonator followed by a microstrip spaced stub low or high pass filter will work for one side of the diplexer. It seems that both the shunt stub and open stub bandpass basically pass the 3rd harmonic through, so you need to follow up with another filter.

The PCB etch dimensions for a low insertion loss stub filters seem reasonable (other filter types seem to yield widths and lengths in the micrometers range, this is in the MM range)...

Hairpin filters look really nice, due to increased attenuation on the harmonics, but I am getting numbers in the 600uM range. I feel that underbite in PCB etching might make these difficult to realize. I'm not sure what a practial resolution is for home etching. Perhaps 0.1mm? Foam etching 0.05mm? 

Since the filters are terminated to 50ohm, there should not be a problem right?

[4CX35000]

Should be fine, so long as the micro strip design can handle the power and provides the isolation between the frequencies you require. Most of what your doing is probably going to be fine using a PCB design. You might be able to look at modelling the diplexer before building if you have access to ADS or similar, otherwise use the alternative amateur radio software packages around which tend to be fine for what most people need. A software package called PUFF comes to mind, there are others which might be more practical.
http://www.its.caltech.edu/~mmic/puff.html

Are you looking at producing a ring diplexer (Rats race) ?

[CopperCone]

How do you estimate power handling capability in a stripline filter?

And I am not sure, I did not know a single thing about stripline up until yesterday. Just fooling around really, since diplexers are expensive. I got several but their not wideband like I would like, the spacing is fairly close (though the mixing harmonics should be discernible on a spectrum analyzer).

And of course the waveguide ones are rather nonportable. One looks like it would be good for smashing pork and chicken cutlets.

Still have to do more research. I might just etch one tonight on some FR4 to see how good the Nuhertz default solution is and how it will work with bubble based home etching on MG chemicals FR4. Perhaps it will make me feel better about paying for 15$ connectors and 40$ microwave cables if I see poor performance lol.

Is it just the capacitor formed between the ground plane and the trace with the dielectric in between? So you calculate the capacitive dissipation of the circuit?