Again, the blackest of the black arts in analog electronics. Nice PCB design - so much engineering effort went into getting all the impedances right, and having touched the bare basics of RF and high speed design, I don't grasp the rules of microwave electronics design. So, I'm looking at it like some alien technology.
The topology you see being used with the "parallel" amplifier arrangements is called a "balanced amplifier". Those screwed-down metal slabs I assume are 90-degree hybrids. The big benefit of doing this system is that the reflections of one due to poor matching get canceled out by the other. An arrangement like this therefore allows you to fairly easily make a wideband amplifier with a flat response and due to all the magical trickery you can account for a fairly consistent impedance at the in and output for a "wide" range of frequencies. Grab Pozar from your bookshelf, I remember seeing this topology in his book for the first time. I assume in the final stage, maybe it would've been cheaper to use 4 separate amplifiers rather than 2 big ones. It allows for reusing the same design rather than having to work hard designing something new a different semiconductor. Or they might've gotten a great bulk deal of course :-). Another reason could be redundancy? There could be many use cases for an amplifier in this range...
@1:30 The Blue resistor ~1" from the input SMA that connects the input trace to ground is so that the upstream equipment can sense if the input signal path is connected to _this_ amplifier Model, or detect (with another value of sense resistor) if it's connected to a different Amplifier model, or path is an short or open circuit. Expect to see such resistors in high quality RF electronics, and in ALL RF avionics.
That's a Wilkinson power splitter at 4:23. It has very high isolation between the three ports at the frequency where each of the legs is a quarter wavelength. The 100 Ohm resister at the output was Wilkinson's contribution...
hope we get to see oscilloscope waveforms coming out of each stage. there are some limitation in probing this high frequency on a scope, but fingers crossed. I would love to see those splitters in action.
Conventional scopes barely have the sample rate good enough to measure such high frequencies unless you have incredibly deep pockets :-). Oftentimes with RF design, you test every stage on a seperate PCB with seperate connectors so you have good coupling. Then you use a network analyser to tune it, do impedance matching, biasing etc. and characterize it. It sweeps an RF signal at the input, then it looks at the amplitude (if your have a fancy analyzer also phase) at both the output (as transmission) and input (as reflection). You can even do this in reverse putting a signal into the output and look what comes through at the input. He did a great video about scattering matrices not too long ago that explains this in great detail. I'm sure he'll demonstrate this in the coming videos! But for RF, you're not interested in the shape of the waveform, mostly the harmonics and the power. When everything works like expected, you jam everything together on a large circuitboard. I do have to say, the more seperate RF parts on such a board, the horrificness of doing all the matching grows exponentially, but I guess the charme of becoming an RF engineer is the hands-on-ness of the design process and large variation within the same series of components hahahaha.
@@BeesKneesBenjamin re: "Oftentimes with RF design, you test every stage on a seperate PCB with seperate connectors" With the P4 (Project 4 WiMAX) at Cisco (was Navini) the entire unit was simply proto-typed; debug/inspection of various stages for design troubleshooting employed in-circuit use (splice into the RF path) of Spec AN and Vector Net Ana but we did not, b/c of the density of the 8-ch phased array first do individual stages eval ... now, evaluation boards for various stages/components/system chips are available from the manufacturer so evaluation is made at the device level. Using quality FR4 or Nelco RF-capable board material takes out the guesswork or removes the need to evaluate every new production model since repeatability is achieved with pre-preg multi-layer board stackups.
@@uploadJ I think in this case then our differences design processes are more likely branche-related. We exclusively do measurement equipment for scientific research. The selection of chips and modules has gotten quite large but we often still need to do a lot of discrete designs where we need really high stability or a very specific function at the cost of repeatability. Hence we still do a lot of tuning by hand and extensive calibration processes in order to keep variation between units incredibly low. Ofcourse we also use computer aided design every now and then and in this branche manufacturing cost is only a fraction of the price of the entire unit so we use excellent PCB materials. But at the end of the design process we often end up with PCBs we can get somewhat reliably manufactured, but we will always need a couple adjustments here and there. For quite elaborate prototypes, we also tend to incorporate extra connectors which can be connected by replacing a capacitor or a 0R resistor somewhere. Once it's characterised the connectors are taken off and it's mounted in a case. Most measurements are taken with a VNA or spectrum analyser or we specifically design testsets for specific calibration steps. This is obviously unsuitable for mass-manufacture, but in our field it doesn't happen often we sell two identical devices :-)
@@BeesKneesBenjamin re: "The selection of chips and modules " For our project, we used a TI made demod chip that took I and Q LO and provided I and Q baseband for digitizing ... anymore, in the telecom/wireless biz it is incorporating a chip vendor's block' function device and less discrete design with transistors, diodes, Rs and Cs ... although power amp design still calls for discrete design skills. Modelling may be the first step, but it takes skill to realize when vendor parts aren't conforming or deviate from nominal specs; RF parts/components being the worst once into production.
Can you please (either in this series or with sometime else) zoom in a bit more on the design of the transmission line and more specific the track (volume/width) and impedance etc. and perhaps some light of board material and their coatings a bit. Just a bit of sharing your knowledge and pointing in the direction. Thanks!
There was a lot done with the circuit. Maybe some of the active components have died and were not replaced with matched ones so matching was tried by adjusting the coupling. Maybe even by fine tuning with heat compound. The flying blue pot is also an indicator for later adjustment as well as the first mounting screw in the final stage with the slit which does not affect the circuit but is really worn out. Considering the power and the supply voltage: 4 parallel elements provide 30W so 7.5W per element. With 50 Ohms 7.5W is 19.4V eff and 27.4V pp. With 15V supply voltage one will be able to reach an amplitude of up to 2x15V. This might also be the limit of the transistors (some of which gave up when testing the limit). (FLL120 goes up to 15V Uds) With 30W at 2GHz one might already warm up a party sausage, at least half way in case of a single side band signal.
If the cover didn't come with it (along with ~50 4-40 stainless screws), you should make one and screw it down before you try powering it up. Or use separate covers as you power up each section. That will keep it from turning into a high power oscillator. That much white stuff is a very good thermal insulator. I have never seen that much used, even by the dumbest techs that worked for me. If you can't see the metal through the white thermal paste, you have used too much.
With luck the transistors are still ok. Normally they seem to be bolted down with the board before they solder them to the board to remove mechanical stress between the tabs and the ceramic body. looks like the repair department decided that it was beyond the point of no return, and somebody caught it before it fell too far in to the WEEE bin. Hope it just takes a few new screws to get it up and producing power again, would be nice to have 30W's on 23cm's
Around 9:28 I was wondering when you'd get to those. At first glimpse when you showed the board I was like what are those? Then I started thinking about how small microwaves are and question if those act as capacitors because of the longer PCB traces/blocks making a T shaped type filter? Wouldn't at those frequencies every solid wire starts looking resonant? So could be some sort of way to reduce that? Maybe that was what you were getting at describing. I've only experimented with RF up into the UHF range so the microwave stuff requires almost a new way of thinking when dealing with such a small wavelength. Edit: sorry if you made it evident. The more I watched I realized you already understand the purpose. Would make for an interesting video describing how though. BTW to get really crazy I wonder what PCB font lettering may do when put around the fine tuning traces at higher RF power lol.
@@NickNorton OK Maybe I might need two of them. One might look a little odd strapped to my Thinkpad though, and somewhat limit the battery life. On the plus side however, I cold probably work from home anywhere within twenty miles of the router, but then again I would have to keep a watchful eye out for the WiFi police.
Why is 50 Ohms a standard for RF (and 75 Ohms seems to be for TV & Satellite too)? I picked up on the 200 Ohm section you mentioned, and that led me to remember to ask.
Those extra little solder pads are tuning stubs. And those looped traces are Wilkinson dividers/combiners.
Wilkinson.
@uploadJ yes, Wilkinson
Again, the blackest of the black arts in analog electronics. Nice PCB design - so much engineering effort went into getting all the impedances right, and having touched the bare basics of RF and high speed design, I don't grasp the rules of microwave electronics design. So, I'm looking at it like some alien technology.
Me too, this is serious witchdoctor stuff.
@@monteceitomoocher not there yet either, I still got to write my thesis, haha.
The topology you see being used with the "parallel" amplifier arrangements is called a "balanced amplifier". Those screwed-down metal slabs I assume are 90-degree hybrids. The big benefit of doing this system is that the reflections of one due to poor matching get canceled out by the other. An arrangement like this therefore allows you to fairly easily make a wideband amplifier with a flat response and due to all the magical trickery you can account for a fairly consistent impedance at the in and output for a "wide" range of frequencies. Grab Pozar from your bookshelf, I remember seeing this topology in his book for the first time.
I assume in the final stage, maybe it would've been cheaper to use 4 separate amplifiers rather than 2 big ones. It allows for reusing the same design rather than having to work hard designing something new a different semiconductor. Or they might've gotten a great bulk deal of course :-). Another reason could be redundancy? There could be many use cases for an amplifier in this range...
@1:30 The Blue resistor ~1" from the input SMA that connects the input trace to ground is so that the upstream equipment can sense if the input signal path is connected to _this_ amplifier Model, or detect (with another value of sense resistor) if it's connected to a different Amplifier model, or path is an short or open circuit.
Expect to see such resistors in high quality RF electronics, and in ALL RF avionics.
The input is likely still 50 ohms. The difference in trace width is due to the output using microstrip, but the input using coplanar waveguide.
That's a Wilkinson power splitter at 4:23. It has very high isolation between the three ports at the frequency where each of the legs is a quarter wavelength. The 100 Ohm resister at the output was Wilkinson's contribution...
I think I see 90 degree hybrids in there; That's what gives the extra port labeled "ISO" for Isolation ...
Here we go!
hope we get to see oscilloscope waveforms coming out of each stage. there are some limitation in probing this high frequency on a scope, but fingers crossed. I would love to see those splitters in action.
The tool you want is Spectrum Analyzer; the 8-bit digitization/quantization of an O-scope is inadequate to see harmonics down 40 dB and more.
Conventional scopes barely have the sample rate good enough to measure such high frequencies unless you have incredibly deep pockets :-).
Oftentimes with RF design, you test every stage on a seperate PCB with seperate connectors so you have good coupling. Then you use a network analyser to tune it, do impedance matching, biasing etc. and characterize it. It sweeps an RF signal at the input, then it looks at the amplitude (if your have a fancy analyzer also phase) at both the output (as transmission) and input (as reflection). You can even do this in reverse putting a signal into the output and look what comes through at the input. He did a great video about scattering matrices not too long ago that explains this in great detail. I'm sure he'll demonstrate this in the coming videos! But for RF, you're not interested in the shape of the waveform, mostly the harmonics and the power.
When everything works like expected, you jam everything together on a large circuitboard. I do have to say, the more seperate RF parts on such a board, the horrificness of doing all the matching grows exponentially, but I guess the charme of becoming an RF engineer is the hands-on-ness of the design process and large variation within the same series of components hahahaha.
@@BeesKneesBenjamin re: "Oftentimes with RF design, you test every stage on a seperate PCB with seperate connectors"
With the P4 (Project 4 WiMAX) at Cisco (was Navini) the entire unit was simply proto-typed; debug/inspection of various stages for design troubleshooting employed in-circuit use (splice into the RF path) of Spec AN and Vector Net Ana but we did not, b/c of the density of the 8-ch phased array first do individual stages eval ... now, evaluation boards for various stages/components/system chips are available from the manufacturer so evaluation is made at the device level. Using quality FR4 or Nelco RF-capable board material takes out the guesswork or removes the need to evaluate every new production model since repeatability is achieved with pre-preg multi-layer board stackups.
@@uploadJ I think in this case then our differences design processes are more likely branche-related. We exclusively do measurement equipment for scientific research. The selection of chips and modules has gotten quite large but we often still need to do a lot of discrete designs where we need really high stability or a very specific function at the cost of repeatability. Hence we still do a lot of tuning by hand and extensive calibration processes in order to keep variation between units incredibly low. Ofcourse we also use computer aided design every now and then and in this branche manufacturing cost is only a fraction of the price of the entire unit so we use excellent PCB materials. But at the end of the design process we often end up with PCBs we can get somewhat reliably manufactured, but we will always need a couple adjustments here and there. For quite elaborate prototypes, we also tend to incorporate extra connectors which can be connected by replacing a capacitor or a 0R resistor somewhere. Once it's characterised the connectors are taken off and it's mounted in a case.
Most measurements are taken with a VNA or spectrum analyser or we specifically design testsets for specific calibration steps.
This is obviously unsuitable for mass-manufacture, but in our field it doesn't happen often we sell two identical devices :-)
@@BeesKneesBenjamin re: "The selection of chips and modules "
For our project, we used a TI made demod chip that took I and Q LO and provided I and Q baseband for digitizing ... anymore, in the telecom/wireless biz it is incorporating a chip vendor's block' function device and less discrete design with transistors, diodes, Rs and Cs ... although power amp design still calls for discrete design skills. Modelling may be the first step, but it takes skill to realize when vendor parts aren't conforming or deviate from nominal specs; RF parts/components being the worst once into production.
Can you please (either in this series or with sometime else) zoom in a bit more on the design of the transmission line and more specific the track (volume/width) and impedance etc. and perhaps some light of board material and their coatings a bit. Just a bit of sharing your knowledge and pointing in the direction. Thanks!
There was a lot done with the circuit. Maybe some of the active components have died and were not replaced with matched ones so matching was tried by adjusting the coupling. Maybe even by fine tuning with heat compound. The flying blue pot is also an indicator for later adjustment as well as the first mounting screw in the final stage with the slit which does not affect the circuit but is really worn out.
Considering the power and the supply voltage: 4 parallel elements provide 30W so 7.5W per element. With 50 Ohms 7.5W is 19.4V eff and 27.4V pp. With 15V supply voltage one will be able to reach an amplitude of up to 2x15V. This might also be the limit of the transistors (some of which gave up when testing the limit). (FLL120 goes up to 15V Uds)
With 30W at 2GHz one might already warm up a party sausage, at least half way in case of a single side band signal.
It's incredible the kind of stuff that ends up in junk stores, I wonder if it was sold by someone that didn't know what it was?
Because it's at least 20 years old obsolete
@@StbLexBoussole I guess I didn't consider that possibility lol
If the cover didn't come with it (along with ~50 4-40 stainless screws), you should make one and screw it down before you try powering it up. Or use separate covers as you power up each section. That will keep it from turning into a high power oscillator. That much white stuff is a very good thermal insulator. I have never seen that much used, even by the dumbest techs that worked for me. If you can't see the metal through the white thermal paste, you have used too much.
With luck the transistors are still ok. Normally they seem to be bolted down with the board before they solder them to the board to remove mechanical stress between the tabs and the ceramic body. looks like the repair department decided that it was beyond the point of no return, and somebody caught it before it fell too far in to the WEEE bin.
Hope it just takes a few new screws to get it up and producing power again, would be nice to have 30W's on 23cm's
Around 9:28 I was wondering when you'd get to those. At first glimpse when you showed the board I was like what are those? Then I started thinking about how small microwaves are and question if those act as capacitors because of the longer PCB traces/blocks making a T shaped type filter? Wouldn't at those frequencies every solid wire starts looking resonant? So could be some sort of way to reduce that? Maybe that was what you were getting at describing. I've only experimented with RF up into the UHF range so the microwave stuff requires almost a new way of thinking when dealing with such a small wavelength.
Edit: sorry if you made it evident. The more I watched I realized you already understand the purpose. Would make for an interesting video describing how though.
BTW to get really crazy I wonder what PCB font lettering may do when put around the fine tuning traces at higher RF power lol.
Just a silly question. Wouldn't grubby fingerprints dust and other airborne junk effect the output of a device like this???
120's are 37.5W devices (L band) ... nice find
OK I think that thing would just about pacify my WiFi range anxiety. ;~)
;~) and never be able to acknowledge a packet
@@NickNorton OK Maybe I might need two of them. One might look a little odd strapped to my Thinkpad though, and somewhat limit the battery life.
On the plus side however, I cold probably work from home anywhere within twenty miles of the router, but then again I would have to keep a watchful eye out for the WiFi police.
Where did you get that from what state do you live in, and was it sky craft in Florida
ua-cam.com/video/WSjood3v7TI/v-deo.htmlsi=PARN5nmB0P3f2OGS
Why is 50 Ohms a standard for RF (and 75 Ohms seems to be for TV & Satellite too)?
I picked up on the 200 Ohm section you mentioned, and that led me to remember to ask.
www.digikey.com/en/blog/the-reasons-for-50-ohm-and-75-ohm-transmission-lines#:~:text=The%2050%20%CE%A9%20value%20is,%CE%A9%20is%20a%20better%20choice.
@@IMSAIGuy Perfect! Thank you 👌
Looks like it was made for space, accept it's weight.
I'll give you $50 for it😝
1-2ghz is a shit band that is full of nothing
nice videos but your pooing is not really pleasant ;-)
Can't see much, I would have loved it, of you got a camera closer, to see how it's made..
watch parts 2 3 4