Im just going to give a shot out for the 18:10 - 18-11 changing of mixers and having them be so precisely aligned on camera that it fried my brain for a moment
Thanks for continuing to put this kind of "real" content on an otherwise mostly silly platform. I'm going to have to watch this two or three more times (while trying to hold back the typical lab equipment jealousy). Suffice to say I don't think I will be using a 10ghz scope to measure a 200hz encoder wheel anytime soon, let alone generate a 100ghz signal 😆
Great vid as always. I have made external mixers for my old Tek 492 using these diplexers. Just a tiny diode at the end of rigid coax and a flashlight reflector for a horn. Works to 61GHz using a radar module in test cw mode. There is a conversion loss list somewhere in Tek docs but 15th harmonic etc wont ever be strong.
I worked for Tek at one of their field offices back in the 80s, and worked on 492 series spectrum analyzers for several years, but never had the opportunity to play with the external mixers. Thanks!
Very impressive (or at least I THINK it is???)... It motivated me into building my own ultra high frequency oscillator... (around 454THz I believe) . . . . I grabbed a red LED and hooked it up to a battery thru a resistor.
Actually - you're spot on... LEDs are in fact oscillators. They obviously generate visible light, but in the process create ultra high frequency oscillations in the electrical side too at the same frequency as the electrons jump in the junction. Not many people realize this, so good for you!
One issue using an endmill to cut such a tapered slot by moving it up as you move is that you would get a concave bottom, as you are now taking an angled projection of a circle for your bottom surface. Let’s say the channel was 6mm wide and you had a rise over run of 1/10, the center of your endmill would cut 3mm ahead of the sides at the same level, so as you move forward these 3mm you will have a 0.3mm hollow. That’s a reason why you normally use a ball nose endmill as the projection of a sphere stays constant, at which point your surface dishing is a function of stepover and cutter radius. So this part was probably made on a fancy 5 axis machine, or depending on age and production volume at least an angled fixture
Thank you for explaining this so clearly. Would be nice to also see episode about Gunn diodes for generation of these high frequencies. Perhaps using EIP 578B locking counter could be great illustration of using mixer + Gunn oscillator to make programmable many-GHz clock source. Also could you use different harmonics in a mixer like this, for example 20th instead of 23th?
@@Thesignalpath The RF absorber looks like it suffered a brittle snap, maybe from a drop. Then, like you suspected, it would have resulted in damage at the bottom of diode, as the broken gasket basically wedged the diode. If someone didn't realize that it had been damaged and then tried to use it, that could have easily blown the thin oxide layer. A standard CNC probably wouldn't get you a surface finish R_a good enough for RF purposes. A surface grinder, well-designed jigs that you can precision pin to align against so you the right angle tapers, and a few thousand dollars in metrology equipment would probably get you on-par (or better) than what they were producing. But the diode is where all the magic really happens. What you should try to do is surgically remove/repair where the IC has localized damage and repeat the experiment Either way, as usual great stuff. Every episode that features failure analysis with the microscope earns you an A+ in my book
Excellent Video and Experiments again. I own some of those mixers up to 220 Ghz (got it from a ham radio OV working with THz). An succeded to repair the 90-140 GHz Version. The inside looked a litte bit different frm this video and the diode was in series. I needed around half dozen similar diodes and H20 glue to glue it on the tip reaching the filter with the other side. I have a spectrum analyser with integrated diplexer and signal id which helps (also biasing), -- Great to use a SMU for i-V curve, I will try to get this app for my 2650. Makes it safer to test the diodes. But at the time of my repair I didnt had one - will try it for more precise characteristics... For test I used a 120Ghz Radar Chip from silicon radar.
Wow, thanks for that. I was lying on my back outside here in Belfast watching the Perseid meteors last night, a fantastic natural phenomenon. Now a fantastic egineering/physics experience. I don't understand everything in your video, but there were quite a few "now i get it!" moments on first watch. I'll just have to watch it again.
Thanks for the wondeful video. I was wondering: is it really such a delicate diode that cannot be tested with a multimeter? When in operation, it is biased at 10 mA, which is a lot more than was used for characterization and is also more than the test current used by multimeters, at least in my belief.
The main issue is the initial discharge into the diode. A lot of multimeters produce an open voltage of several volts. The moment you connect the diode the inrush current can be high and damage the diode.
The lowermost (yellow) trace at 47:28 appears asymmetric, although the chopper-wheel's apertures look symmetric. What could be the source of this asymmetry? Would it reverse if the chopper-wheel's direction were to reverse?
Hi Shahriar , I have 2450 and 2461 KEITHELY , but I got KickStart software which you can do all curve tracer function for transistor and diode with all 4 quad. better then App.
1. Would there be any benefit in having a diode with a P/N junction width equal to a multiple or integer factor of the wavelength to be measured? 2. "How" tiny is the diode in the parts you have? How much current can it carry before it gets destroyed? Is it on the order of single digit mA or will it only carry a few uA before being damaged?
Wonderfully detailed video. Highly appreciated. It's just a pity that your microscope camera and your recording setup didn't agree on the interlacing field order.
At time 47:00, right after you start chopping the RF input, a series of feeble tones around the IF appear on the oscilloscope FFT. Later on, most die out, but one survives and is even picked up by the instrument @2.072727 GHz (726 kHz above the IF), -49.5 dBm. I suspect these are intermod products caused by the chopping. @47:00, if you watch closely, the amplitudes of these tones are not the same on either side of the IF (the amplitudes at below the IF are lower), which indicates both phase and amplitude noise introduced by the chopping.
This is super interesting! Thanks for posting this. I bought some absolutely minuscule SRD's (100G & 300G) years ago and they've been in a drawer waiting for this video! I always pondered how they'd get connected into a circuit and seeing that it is something so simple and just mechanical is awesome! I need to hone my milling skills to try to build a structure that can retain these things if only to get an IV characteristic from them. Is that IV characteristic app available for the 2450 ?
These days, the so-called mm-wave bands that 5G wireless uses. In the era when these were made, the basic research that has made these frequencies usable today.
Shariar I wonder about the construction of this diode. I would expect a whisker tip type like the old ceramic coaxial germanium mixer diodes. This would ensure a very small junction capacity. But I could be wrong. Do you know more about this?
This type of diode is called "beam-lead" (just google images and you will see a lot of them). These are planar diodes where a metal bridge connects to a very small epitaxial junction with capacitance about 50 femto farads or so. I believe the diodes used in these mixers were either made by Alpha(formerly TRG) (later SkyWorks) or Metelics [but I am not 100% sure]. The whisker type of diodes you mention were used by tektronix in the previous generation of the harmonic mixers employing the 1N26B and the 1N53B.
Is the high loss seen when introducing a gap in the waveguide due to free-space propagation loss or is it also due to the mismatch? If you have a pair of WR10 horns, I wonder if that would reduce the amount of loss. My lab had an old Tek 492 with a couple of mixer modules and the diplexer. We surplussed it a while ago, our mixers only went up to ~40 GHz as I recall.
Great video as usual! Just one question, when you show the transfer function of the step diode, shouldnt it be a square wave rather than a super narrow pulse? I understand the srd actually has a time dependancy which maybe helps create yhe pulse, but at first sight the vertical.respinse curve would create a square, wouldnt it?
The response goes to infinity vertically, there is no horizontal increase once it starts going up. As a result you can’t have a secure wave. If order to get a square wave, the I/V curve would have to look like a step function.
en.wikipedia.org/wiki/Step_recovery_diode shows very sharp, narrow pulses as well. I think the pulse is due to inductance in the circuit. The step-recovery diode turns off extremely quickly, so any inductance at all in the circuit produces a pulse much higher in voltage than the forward voltage drop of the junction, like on the order of several volts. But this dissipates quickly, depending on that same inductance. This is the main application of step recovery diodes, since those very fast rise times transfer to very rich harmonic content.
You could consider designing some unique pieces of lower-cost equipment using your knowledge and creativity - Similar to the NanoVNA, using existing tech for innovative purposes to bring gear to groups who otherwise wouldn't have access - That would be amazing.
Apparently the diode (a beam lead diode) was bent in the wrong direction. There is usually a minuscule spacer that sits on top of the rod and the diode lead is bent to sit on top of it. Then it would be possible to apply a small pressure to promote the contact of the LPF hat to the diode without smashing the diode, by the way beam lead diodes are extremely fragile. Maybe the diode is still good... even if not a new diode is very easy to apply and restore the mixer to the former glory.
The "Harmonic Mixing Method" slide @6:30 had me a little bit confused. You used commas on the left side (7,328 GHz + 1,875 GHz) and on the right side you used a decimal for the IF (1.875 GHz). You also used a decimal point for your LO frequency. That is how rocket systems abort prior to launch or experience a "Rapid Unscheduled Disassembly" during ascent! 😁🧡
I really need a video about how analog signals mix and how transmitters use those mixed signals, what the if is for. There’s no good videos on radio that i have found for anyone attempting to make a transmitter and receiver
You’ve made a mistake at 5:30, referring to individual pulses as different harmonics (on an X-t graph), when in reality every harmonic should be localized at each pulse, and will show up in a similar looking graph when plotted as X-f.
Many quantum voltage standards used (and still use) 70-90 GHz still , since Josephson Junction device is in a nutshell discrete quantum-accurate AC current to DC voltage converter.
Lots of applications, I add 2 more: - Plasma physics measurements (interferometry, and slightly later reflectometry) - Supply pump for parametric amplifiers (in the tens of GHz)
Another very interesting video, thanks! Reminds me of an experiment to measure the speed of light... th-cam.com/video/bcKdh7cIC0U/w-d-xo.html it would be interesting to see how you would do that with your bench? Shorter distance, more accurate, etc?
Very cool!! Just a couple of notes: you're pronouncing "ridged" incorrectly, it's not the same as "rigid". And the low-mag microscope footage seems unusually poor quality (massive interlacing artefacts maybe??), but I assume you already noticed that during the edit. But those are just some minor constructive feedback notes, absolutely amazing content as always!
@@Thesignalpath Yes, listening at 1.5x. I loved this video BTW. Also why do we care at what voltage the diode turns on as long as the "knee" is sharp enough 18:30? Is a higher DC bias voltage bad somehow?
@@SDX2000To get as high power in the high harmonics as possible, LO signal should swing as close to the maximum nonlinearity of the diode as possible. The maximum nonlinearity happens to be at the knee. The diode rectifies LO and that generates positive DC component moving operation point away from the knee. Negative external DC bias counters that bringing operation point close to maximum nonlinearity: the knee. And that is why adjusting external DC voltage affects mixer conversion loss and thus IF level.
@@adamturowski3765 thanks for the explanation. I get that but it still does not explain why the actual value of the turn on voltage is significant. Can't you apply a negative DC offset of lower magnitude in the case of the diode with the higher turn on voltage?
@@SDX2000 Turn on voltage is significant because it determines how much LO voltage swing gets rectified and turned into DC that needs to be compensated for with external bias voltage. Is that more clear now?
The most impressive pile of test equipment on TH-cam! And one of the best guys to explain their operation
Honey wake up, new Signal Path video just dropped
I wish this kind of material existed when I was studying EE in university. Thanks for making yet another awesome video.
WOW I am staggered by how interesting this whole video is.
Im just going to give a shot out for the 18:10 - 18-11 changing of mixers and having them be so precisely aligned on camera that it fried my brain for a moment
Thanks for continuing to put this kind of "real" content on an otherwise mostly silly platform. I'm going to have to watch this two or three more times (while trying to hold back the typical lab equipment jealousy). Suffice to say I don't think I will be using a 10ghz scope to measure a 200hz encoder wheel anytime soon, let alone generate a 100ghz signal 😆
Lots of "real" content on this platform. TH-cam will recomend you stuff, if it has got idea you like. You give it idea by searching for things.
Aw come on! Cat videos are really fun! :)
Great vid as always. I have made external mixers for my old Tek 492 using these diplexers. Just a tiny diode at the end of rigid coax and a flashlight reflector for a horn. Works to 61GHz using a radar module in test cw mode. There is a conversion loss list somewhere in Tek docs but 15th harmonic etc wont ever be strong.
I've finally found a place to learn some RF magic in a practical and engaging manner. Many thanks!
I worked for Tek at one of their field offices back in the 80s, and worked on 492 series spectrum analyzers for several years, but never had the opportunity to play with the external mixers. Thanks!
Thanks for the great explanation regarding how a nonlinear devices actually result in mixing behavior.
Very impressive (or at least I THINK it is???)...
It motivated me into building my own ultra high frequency oscillator... (around 454THz I believe)
.
.
.
.
I grabbed a red LED and hooked it up to a battery thru a resistor.
Actually - you're spot on... LEDs are in fact oscillators. They obviously generate visible light, but in the process create ultra high frequency oscillations in the electrical side too at the same frequency as the electrons jump in the junction. Not many people realize this, so good for you!
2am in Toronto watching meteors and mixers. awesome 👍😎
That last experiment was amazing!
Another great video, from beginning to end.
One issue using an endmill to cut such a tapered slot by moving it up as you move is that you would get a concave bottom, as you are now taking an angled projection of a circle for your bottom surface.
Let’s say the channel was 6mm wide and you had a rise over run of 1/10, the center of your endmill would cut 3mm ahead of the sides at the same level, so as you move forward these 3mm you will have a 0.3mm hollow.
That’s a reason why you normally use a ball nose endmill as the projection of a sphere stays constant, at which point your surface dishing is a function of stepover and cutter radius.
So this part was probably made on a fancy 5 axis machine, or depending on age and production volume at least an angled fixture
I bet it was an angled fixture. CNC machines , especially fancy ones, were not really available back then.
So fucking awesome, especially the light/darkfield teardown of the diodes. Also lol @ "just a simple multimeter" re: that Keithley dmm
Great video!! Appreciate all the effort on getting that hardware and showing them up.
Dat is snel 🤩 morgen ochtend even kijken!
Thank you for explaining this so clearly. Would be nice to also see episode about Gunn diodes for generation of these high frequencies. Perhaps using EIP 578B locking counter could be great illustration of using mixer + Gunn oscillator to make programmable many-GHz clock source. Also could you use different harmonics in a mixer like this, for example 20th instead of 23th?
Really cool video. I wonder if the “damage” on the top of the diode was actually marks from a wafer prober during test?
I’d imagine that would be more in the center…
@@Thesignalpath The RF absorber looks like it suffered a brittle snap, maybe from a drop. Then, like you suspected, it would have resulted in damage at the bottom of diode, as the broken gasket basically wedged the diode. If someone didn't realize that it had been damaged and then tried to use it, that could have easily blown the thin oxide layer.
A standard CNC probably wouldn't get you a surface finish R_a good enough for RF purposes. A surface grinder, well-designed jigs that you can precision pin to align against so you the right angle tapers, and a few thousand dollars in metrology equipment would probably get you on-par (or better) than what they were producing. But the diode is where all the magic really happens. What you should try to do is surgically remove/repair where the IC has localized damage and repeat the experiment
Either way, as usual great stuff. Every episode that features failure analysis with the microscope earns you an A+ in my book
Excellent Video and Experiments again. I own some of those mixers up to 220 Ghz (got it from a ham radio OV working with THz). An succeded to repair the 90-140 GHz Version. The inside looked a litte bit different frm this video and the diode was in series. I needed around half dozen similar diodes and H20 glue to glue it on the tip reaching the filter with the other side. I have a spectrum analyser with integrated diplexer and signal id which helps (also biasing), -- Great to use a SMU for i-V curve, I will try to get this app for my 2650. Makes it safer to test the diodes. But at the time of my repair I didnt had one - will try it for more precise characteristics... For test I used a 120Ghz Radar Chip from silicon radar.
Wow, thanks for that. I was lying on my back outside here in Belfast watching the Perseid meteors last night, a fantastic natural phenomenon. Now a fantastic egineering/physics experience. I don't understand everything in your video, but there were quite a few "now i get it!" moments on first watch. I'll just have to watch it again.
Of course over here in my corner of the world it was the worst thunder and lightning storm with insanely heavy rains all last night. SMH.
Is there a reason why the bias-t was connected to the diplexer IF port, rather than directly inline with the mixer?
Thanks for the wondeful video.
I was wondering: is it really such a delicate diode that cannot be tested with a multimeter? When in operation, it is biased at 10 mA, which is a lot more than was used for characterization and is also more than the test current used by multimeters, at least in my belief.
The main issue is the initial discharge into the diode. A lot of multimeters produce an open voltage of several volts. The moment you connect the diode the inrush current can be high and damage the diode.
21:00 Great instructive video! Did you see the graffiti inside the end cover - ‘DB’ was here!
The lowermost (yellow) trace at 47:28 appears asymmetric, although the chopper-wheel's apertures look symmetric. What could be the source of this asymmetry? Would it reverse if the chopper-wheel's direction were to reverse?
Direction shouldn't affect this. It looks to me like the chopper pulses are saturating the detector, and it's taking some time for it to recover.
Great demonstration, thanks 73 de KT1R Lou
Hi Shahriar , I have 2450 and 2461 KEITHELY , but I got KickStart software which you can do all curve tracer function for transistor and diode with all 4 quad. better then App.
I was wondering, on the diagram at 1:30, wouldn’t there also be an „RF+LO“ tone between 2LO and 2RF?
the amount of money in the background at 0:30 is insane
1. Would there be any benefit in having a diode with a P/N junction width equal to a multiple or integer factor of the wavelength to be measured?
2. "How" tiny is the diode in the parts you have? How much current can it carry before it gets destroyed? Is it on the order of single digit mA or will it only carry a few uA before being damaged?
Wonderfully detailed video. Highly appreciated. It's just a pity that your microscope camera and your recording setup didn't agree on the interlacing field order.
Thanks for pointing that out. I’ll fix it for the next video.
the windows login screen on the tek scope at the end of the video :D
Genius more rare experiments
Instrument Flex 💪🏼
At time 47:00, right after you start chopping the RF input, a series of feeble tones around the IF appear on the oscilloscope FFT. Later on, most die out, but one survives and is even picked up by the instrument @2.072727 GHz (726 kHz above the IF), -49.5 dBm. I suspect these are intermod products caused by the chopping. @47:00, if you watch closely, the amplitudes of these tones are not the same on either side of the IF (the amplitudes at below the IF are lower), which indicates both phase and amplitude noise introduced by the chopping.
This is super interesting! Thanks for posting this. I bought some absolutely minuscule SRD's (100G & 300G) years ago and they've been in a drawer waiting for this video! I always pondered how they'd get connected into a circuit and seeing that it is something so simple and just mechanical is awesome! I need to hone my milling skills to try to build a structure that can retain these things if only to get an IV characteristic from them. Is that IV characteristic app available for the 2450 ?
I bought some too, from Russia! At first I thought there were just some particles of dirt in the baggie!
love your channel. Especially since I can't afford Agilent test equipment.
Fascinating stuff as always! Thank you so much for your work ❤
What are the use cases for those super-high frequencies? I can imagine radar, but what else?
These days, the so-called mm-wave bands that 5G wireless uses. In the era when these were made, the basic research that has made these frequencies usable today.
Shariar I wonder about the construction of this diode. I would expect a whisker tip type like the old ceramic coaxial germanium mixer diodes. This would ensure a very small junction capacity. But I could be wrong. Do you know more about this?
This type of diode is called "beam-lead" (just google images and you will see a lot of them). These are planar diodes where a metal bridge connects to a very small epitaxial junction with capacitance about 50 femto farads or so. I believe the diodes used in these mixers were either made by Alpha(formerly TRG) (later SkyWorks) or Metelics [but I am not 100% sure]. The whisker type of diodes you mention were used by tektronix in the previous generation of the harmonic mixers employing the 1N26B and the 1N53B.
Those Keithley SourceMeters are so awesome aren't they?
Is the high loss seen when introducing a gap in the waveguide due to free-space propagation loss or is it also due to the mismatch? If you have a pair of WR10 horns, I wonder if that would reduce the amount of loss. My lab had an old Tek 492 with a couple of mixer modules and the diplexer. We surplussed it a while ago, our mixers only went up to ~40 GHz as I recall.
Great video as usual! Just one question, when you show the transfer function of the step diode, shouldnt it be a square wave rather than a super narrow pulse? I understand the srd actually has a time dependancy which maybe helps create yhe pulse, but at first sight the vertical.respinse curve would create a square, wouldnt it?
The response goes to infinity vertically, there is no horizontal increase once it starts going up. As a result you can’t have a secure wave. If order to get a square wave, the I/V curve would have to look like a step function.
en.wikipedia.org/wiki/Step_recovery_diode shows very sharp, narrow pulses as well. I think the pulse is due to inductance in the circuit. The step-recovery diode turns off extremely quickly, so any inductance at all in the circuit produces a pulse much higher in voltage than the forward voltage drop of the junction, like on the order of several volts. But this dissipates quickly, depending on that same inductance. This is the main application of step recovery diodes, since those very fast rise times transfer to very rich harmonic content.
@@Thesignalpath Thanks! I guess I need to think it a bit more.
You could consider designing some unique pieces of lower-cost equipment using your knowledge and creativity - Similar to the NanoVNA, using existing tech for innovative purposes to bring gear to groups who otherwise wouldn't have access - That would be amazing.
25:34 bet could be made on manual mill
Apparently the diode (a beam lead diode) was bent in the wrong direction. There is usually a minuscule spacer that sits on top of the rod and the diode lead is bent to sit on top of it. Then it would be possible to apply a small pressure to promote the contact of the LPF hat to the diode without smashing the diode, by the way beam lead diodes are extremely fragile. Maybe the diode is still good... even if not a new diode is very easy to apply and restore the mixer to the former glory.
Do you hire fleas to do the work on the diode assembly? :)
The "Harmonic Mixing Method" slide @6:30 had me a little bit confused. You used commas on the left side (7,328 GHz + 1,875 GHz) and on the right side you used
a decimal for the IF (1.875 GHz). You also used a decimal point for your LO frequency. That is how rocket systems abort prior to launch or experience a "Rapid Unscheduled Disassembly" during ascent! 😁🧡
as long as you never use either comma nor dots to separate thousands, it's not confusing
I really need a video about how analog signals mix and how transmitters use those mixed signals, what the if is for. There’s no good videos on radio that i have found for anyone attempting to make a transmitter and receiver
Excelente
the signal phat plis láser module ando plis
You’ve made a mistake at 5:30, referring to individual pulses as different harmonics (on an X-t graph), when in reality every harmonic should be localized at each pulse, and will show up in a similar looking graph when plotted as X-f.
I am referring to the equivalent series of pulses in the frequency domain, not the time domain one of course.
plis module lasr
What did they do in 100 GHz 50 years ago?
Carcinotron and particle accelerators.
Many quantum voltage standards used (and still use) 70-90 GHz still , since Josephson Junction device is in a nutshell discrete quantum-accurate AC current to DC voltage converter.
Lots of applications, I add 2 more:
- Plasma physics measurements (interferometry, and slightly later reflectometry)
- Supply pump for parametric amplifiers (in the tens of GHz)
I started a GoFundMe for the poor diode....
Above my paygrade, but super enjoyable!
According to the dimensions, the mixer should have a cut-off frequency of 40 Ghz.
How may of the 8 billion people on this planet have an RF lab equal to or better than this one? I would like to know the answer to that question.
This hertz my brain 🧠 😂
At the very least let the ads roll boys! Free for you and $$ for him.
Long time no vids. Everything alright ?
Yes! Thanks for checking. Just lots of travel and family stuff.
That’s the most microwaves I’ve ever seen! Does he repair them?
I repair everything! ;)
@@Thesignalpathwe need an 1. April video in wich you change the high voltage fuse in a microwave.
where the hell you get all these instruments man
30:36 ... 30:42 hehe :)
Missed in the edits. :)
Another very interesting video, thanks! Reminds me of an experiment to measure the speed of light... th-cam.com/video/bcKdh7cIC0U/w-d-xo.html it would be interesting to see how you would do that with your bench? Shorter distance, more accurate, etc?
Very cool!! Just a couple of notes: you're pronouncing "ridged" incorrectly, it's not the same as "rigid". And the low-mag microscope footage seems unusually poor quality (massive interlacing artefacts maybe??), but I assume you already noticed that during the edit. But those are just some minor constructive feedback notes, absolutely amazing content as always!
Yes, it is more like Rijd. I don’t know what the issue with the smaller microscope is, I’ll look into it. Thanks!
Talking faster than I can listen.
Have you tried 0.75x speed? Some people have mentioned they prefer it. Although I have had people who told me they listen at 1.5x...
@@Thesignalpath Yes, listening at 1.5x. I loved this video BTW. Also why do we care at what voltage the diode turns on as long as the "knee" is sharp enough 18:30? Is a higher DC bias voltage bad somehow?
@@SDX2000To get as high power in the high harmonics as possible, LO signal should swing as close to the maximum nonlinearity of the diode as possible. The maximum nonlinearity happens to be at the knee. The diode rectifies LO and that generates positive DC component moving operation point away from the knee. Negative external DC bias counters that bringing operation point close to maximum nonlinearity: the knee. And that is why adjusting external DC voltage affects mixer conversion loss and thus IF level.
@@adamturowski3765 thanks for the explanation. I get that but it still does not explain why the actual value of the turn on voltage is significant. Can't you apply a negative DC offset of lower magnitude in the case of the diode with the higher turn on voltage?
@@SDX2000 Turn on voltage is significant because it determines how much LO voltage swing gets rectified and turned into DC that needs to be compensated for with external bias voltage. Is that more clear now?