21 - Diving Deeper Into The Diplexer

Part 2 of the Mixer Mini-Series.
Nick M0NTV explores the design and construction of a Diplexer for the IF port of a Double Balanced Diode Ring Mixer.
Part 1 is here: • 20 - Double Balanced D...
In terms of DESIGNING the diplexer I actually started with the Diplexer Calculator (Bridged Tee) at: www.changpuak....
I then took these values and also modelled the design in LTSpice. When it came to physically building it I tested the series resonant filter and the parallel resonant filters separately first of all with an S21 sweep on the NanoVNA. This is how I usually test all my filters these days.

[@M0XYM]

This is the best explanation on this subject I have seen. Thanks.

[@M0NTVHomebrewing]

Thanks very much Kevin.

[@alanmacdonald65]

In the grand scheme of things, building the diplexer really isn't that much work, the results of the tests are speaking for themselves! You, my friend, are doing an excellent and inspiring job with the testing and explanation (I wonder who taught you about rejectors and acceptors ;-). Great work as always Nick, look forward to the next one.

[@M0NTVHomebrewing]

It was a wise man ... who clearly taught me well! Thanks so much mate. See you soon. 73

[@G0FUW]

Genuinely looking forward to the 'so what?' video Nick. Does it make any real world difference? Thanks for sharing.

[@M0NTVHomebrewing]

Cheers Steve! Spoiler alert: I think it does ... but we'll see. Thanks again. 73

[@Steve-GM0HUU]

👍Thanks for posting, very interesting and useful.

[@M0NTVHomebrewing]

Thanks Steve - I'm pleased it was helpful. 73 Nick

[@johnwest7993]

Excellent presentation. Clear and concise, with what matters most to the way I learn - WHY use a diplexer. Why it's better than the other circuits. Until your presentation I never understood exactly what a diplexer did, or how. Thank you. My only other comment, (for those who work with much higher frequencies,) is that resistors don't stay resistors at higher frequencies. One needs to take their inductance (mostly) and their capacitance (possibly) into consideration depending on the frequency of operation. That's where the NanoVNA becomes the most valuable design tool you can have, since ALL components, including resistors have some inductance and capacitance, and it can be dealt with using the NanoVNA's. They have opened up microwave frequency design and construction to the amateur radio hobbyist.

[@M0NTVHomebrewing]

Thanks very much John and thanks for the correction about resistors too. I sometimes show my HF bias by forgetting there are frequencies where things don't quite happen the same as they do below 30 MHz!!! Cheers. 73

[@edbeckerich3737]

Nick, back again to say how great your videos are, man I learned a ton from this one, and can't wait for the next one!!

[@M0NTVHomebrewing]

Thanks so much Ed. I learn a lot myself from making them. 73 Nick

[@edbeckerich3737]

At what frequency does the SWR peak on the duplexer plot, and that seems like a bad thing, correct?

[@M0NTVHomebrewing]

Well actually Ed I think it may be a consequence of feeding it into the IF port with the other two ports unterminated. The final test (28:24 in the video) of the diplexer alone shows a flat SWR across the passband and a tiny blip in the Z. With some tweaking it might be possible to get it better but I can certainly live with it as it is!

[@Luappy13]

One thing you could do with explaining is how you arrived at the series LC values (23.9pF and 5983nH). At 13.3MHz these show a Z of 50.005102963016Ω and w of -0.81855029821357 degrees. Might be interesting to plot (using step params in LTSpice) how changes in L and C affect the w and how this affects the SWR.

[@M0NTVHomebrewing]

Hi Paul, good point! Apologies - I should have covered this in the video. I actually started with the Diplexer Calculator (Bridged Tee) at: www.changpuak.ch/electronics/calc_16a.php I then took these values and also modelled the design in LTSpice. When it came to actually building it I tested the series resonant filter and the parallel resonant filters separately first of all with an S21 sweep on the NanoVNA. Thanks very much for mentioning this. I'll reference it in the video description now. Cheers. 73

[@Luappy13]

@@M0NTVHomebrewing It got me thinking about the series LC and how they appear in terms of ZLC. I used your calculator mentioned and was surprised to see that when I calculated the | ZC - ZL | It came out with 500ohms... this made me think I was going doolally until I realised that the Q is multiplying. If you go to that calculator and give a Q of 1 then each the L and the C come out at 50ohm. Using your spice model you can play around with the impact of the q of the components using spice directive .param q=10 xc1=239p/q xl1=598.33n*q xc2=239.3p*q xl2=598n/q xzl=50 where xc1 and xl1 form the series LC and xc2 and xl2 form the parallel LC. q is obviously q and xzl is output 50ohm load Very interesting.

[@Luappy13]

Nicely explained video. i've been meaning to build a diplexer for my hb DBM for a while now.

[@M0NTVHomebrewing]

Thanks Paul. It's certainly worth considering I think. 73

[@smonair2806]

Hi, very interesting video and made me consider the diplexer I made for the Eamon Skelton EI9GQ transceiver. I followed Eamon's circuit but moved the IF frequency up from 10.7MHz to 11.2MHz to match the IF filter. So the capacitance and inductance values will require tweaking a bit. LTSpice simulation shows the mismatch I have to be quite small but I will put my VNA on it to check. Looking forward to your next video, Nick. Best 73's de Steve M0SHM.

[@M0NTVHomebrewing]

Thanks Steve. Yes, it was Eamon Skelton who first introduced me to the diplexer! 73

[@cthoadmin7458]

There's something here I don't understand. I get the part where you say the output from the mixers IF port must be 50 ohms to provide a good match and stop reflections back into the mixer. Couldn't you just feed the IF output into an impedance matching inductor, designed to match the input impedance of the next stage? Filtering would then be done in this next crystal filter (?) section... What have I missed here? Learning LTSpice at the moment, maybe I'll try to model it.

[@M0NTVHomebrewing]

Hi there, by 'impedance matching inductor' I presume you mean a transformer. The problem with this is that you need to know the impedance you are starting with and the impedance you want to transform to. The last part is easy if it's 50Ω: the first part is not. The impedance at the IF port could be anything and what is worse will probably be a different value at different frequencies. What you need is to ensure that the impedance is 50Ω regardless of the frequency. This is precisely what a diplexer does. 73 Nick

[@cthoadmin7458]

@@M0NTVHomebrewing Ah that makes sense Nick, and that's why in that 2 stage 2N3906 Rf or IF amplifier you designed in an earlier video (I think taken from experimental methods in RF design), there is an attenuation pad between the two stages, because even though there's an impedance matching transformer between the 2 stages, the output impedance of the first stage won't be constant. Have I got that right? I've often wondered why amplifiers have attenuation pads on inputs and outputs...

[@M0NTVHomebrewing]

Yes, Pi attenuators perform two important functions - a bit of impedance matching as well as attenuation. But the attenuation part is more significant than we often give credit to. So many of my issues with amplifiers have been fixed thanks to a well-placed bit of attenuation. Even just a dB or two can make all the difference between overdriving the next stage or everything running fine. The impedance matching is equally useful and gives a fixed amount of loss that you can design and account for - rather than just strapping a 50Ω resistor to ground and hoping for the best. This is one reason the 15/10m rig will be using predominantly TIAs (Termination Insensitive Amplifiers). 73, Nick.

[@davidbrodie5044]

Hi, thanks for the video, I found it really interesting and it moved me on a lot with a project I'm playing with. Suffice to say that my mixer IF is currently terminated in a band pass filter and I think I'll be inserting a duplexer now!

[@M0NTVHomebrewing]

Thanks very much David. 73

[@gradje74]

‘RF is lazy’ what a great one.

[@IanMcCrum]

Hi Nick, I ran the simulation (with a .NET command) and changed the 23.9pF to 22pF - the values for ZIn and Zout changed a lot, the peak is now 150 Ohm at. You may need to "tune" the diplexer to get it to work or pick a different Q value. Thank goodness for NanoVNAs!

[@IanMcCrum]

Oops, sorry, I mistyped some values. stepping C1 from 22p to 27p give ZINs of 80 to 90Ohms, You'll still need to get C1 within one picofarad of 23.9pF to get the diplexer to work well...

[@M0NTVHomebrewing]

Thanks for your work on this Ian. I appreciate your time and efforts. I've got to knuckle down and get a better grasp of LTSpice. It's such an amazing tool. 73

[@mfwwhelan7561]

Hi Nick, Excellent clear presentation. I've built several QRP transceivers but none with a diplexer. Next time. Quick question: I also use Me Squares. What glue do you use to attach them to the ground plane? Mike, EI6AU

[@M0NTVHomebrewing]

Hi Mike, thanks very much. I just tend to use superglue. It's not the nicest stuff to work with as it can dry with a messy white appearance sometimes; and I also end up gluing the squares to my fingers! But it is cheap and easy. What do you use yourself? 73 Nick

[@mfwwhelan7561]

@@M0NTVHomebrewing Hi Nick, I use a small blob of Loctite super glue but when I drop the Me Square in place it tends to float a bit and no longer sits parallel to the sides of the board. To make matters worse, when I apply pressure to the Me Square to set the glue it skews it more! My last xcvr used a series of small Veroboard modules and small Manhatten Me Square boards all sitting on a PCB ground plane. I often add PCB screens between stages on account of a traumatic childhood experience with an oscillating IF amp! Mike , EI6AU

[@johnwest7993]

I was re-watching this excellent video, because that's what it is, and I'm a slow learner. :) I do have a question for you, although a bit peripheral to the point of diplexers. I see that you mount your toroids directly on the copper-clad PCB material. I've watched some other videos that show how the inductances of them change with proximity to metals and dielectrics, so I've been trying to 'suspend' mine off the PCB's as much as reasonable, either by using heavy leads attached at the coils, or with small pieces of Styrofoam for the toroids to sit on, but your circuits appear to be quite accurate as built and affixed directly to the PCB, with little or no tweaking. Is this because the relatively low frequencies of operation are more tolerant to minor changes in inductance? BTW, I'm going to get one of those UV glue pens. That's an excellent tip. It allows one the control to make sure everything is properly positioned at the moment the glue sets, and I have numerous improperly positioned parts to demonstrate the need for one.

[@M0NTVHomebrewing]

Hi John, the thing about toroid cores is that the magnetic field is pretty self-contained. As long as you pay attention to the winding and spacing of the turns I've found the actual mounting of them to be of little consequence. Having said that I always mount them upright (not flat across the board) and where there are two or three I try to ensure that they are all oriented at different angles to the nearest one. The UV 'glue' is very useful and I find I'm reaching for it more and more. 73 Nick

[@johnwest7993]

@@M0NTVHomebrewing, thanks for the reply, explanation, and suggestions Nick. 73

[@chadx8269]

Nice explaination. Check out Mini-Circuits Labs Reflectionless Filter ( XLF-151) series. These are a low pass versions instead of bandpass. Question: for wide band Data use what is the phase response in the passband, how bad does it distort the IF signal.

[@M0NTVHomebrewing]

Thanks for watching and for the tip. I've only ever built diplexers for narrow frequency ranges so I'm afraid I can't comment on wideband usage. You'd need a different design I would imagine. 73 Nick

[@nevmarr]

Hi Nick, another option to put on the IF might be a Wes Hayward termination insensitive amplifier, what'd think? I've seen/read in lots of places that DBM like to see 50R but I've never worked out why. What is it about the architecture of a DBM that makes the IF 50R? Perhaps others could comment too?

[@M0NTVHomebrewing]

Hi Neville, absolutely! In fact in the last transceiver I built ('The Optimiser') I actually have both i.e. a diplexer on the IF port followed by a TIA. I didn't try just using the TIA alone because I wanted the filtering that the diplexer offers. But Wes' TIA is a great way to get that 50Ω termination and I see he has now produced a modified design with variable gain to be used as an AGC. As to why 50Ω I confess I'm not certain and will bow to the greater knowledge of others. Clearly there is an inductive reactance in the transformers and surely some resistance in the diodes. I measured a bit less than 50Ω Z when I built the transformers. Anyway, thanks very much for watching and for your questions and comments. 73

[@foxbat888]

Perhaps 50 ohms is a reasonable approximation for the frequencies at which a typical DBM is used, just an idea

[@kerwincarreon7621]

See u nxt video.cheers

[@M0NTVHomebrewing]

Thanks Kerwin.

[@twobob]

Small Correction "The 'Entire Universe' is as lazy as possible." ;) Not just RF

[@LydellAaron]

Thank you so much. Diplexers are an important topic. Have you done a video on phase rotators?

[@M0NTVHomebrewing]

Thanks very much. No I haven't come across those yet. Glad you found it useful. 73 Nick

[@chronobot2001]

Why wouldn't you use the same value of capacitance and inductance for the bandpass and the bandstop portions of the diplexer?

[@M0NTVHomebrewing]

Good question! You should model it in LTSpice as I did and you'll see. SPOILER: If you make the values of both parallel and series filters the same you get a narrower filter (higher Q) but it shelves off to a plateau at about -12dB. Bear in mind there is a predicted insertion loss of -6dB too so you are only getting 6dB of attenuation out of the diplexer. Building it the way I suggested means a little bit less Q but a lot more attenuation in the stop band. It is a fascinating mathematical correlation between the values in the two respective filters. Thanks for your interest. 73 Nick

[@drstrangelove09]

Thank you, good video. What is the black wire on the right at 0:08? A DC bias supply for a diode? (but there's no diode, there is there?)

[@M0NTVHomebrewing]

Hi there. That's just a temporary coax connection for testing purposes. There was a BNC on the other end of it (out of shot) which went to port 1 of the NanoVNA. It enabled me to disconnect the IF port of the mixer and feed a signal directly into the diplexer. I also reconfigured it at one stage to be the IF port output and to disconnect the diplexer - just so I could compare the results through the mixer alone and then through mixer + diplexer. I'll be covering this in the next video. 73

[@drstrangelove09]

@@M0NTVHomebrewing Thank you so much! You are great!

[@frequencywatchers]

WHy Do We Went 50Ohms There? Can Someone Explain?

[@M0NTVHomebrewing]

Hi! Good question! If by "there" you mean the IF port of a double balanced diode ring mixer - that is the characteristic impedance which that port wants to "see". All RF devices have input/output impedances and if these are not matched then maximum signal transfer will not be achieved. Some signal will be reflected back towards the source. Sometimes you can get away with not perfectly matching impedances and you can live with a little loss. However, the whole point of a double BALANCED mixer is the balance. The attenuated LO and RF signals are dependant on balance in the matched diodes and transformers and the IF port particularly needs a good 50Ω termination. If not then the reflected signal will unbalance your balanced mixer! Why it is particularly 50Ω and not another value I guess is a question of mathematics but I will defer to others on this. Thanks for your question. 73 Nick