cheap THz experiments?

[coppercone2]

So I was trying to think of something fun and new for the next year, and I was wondering about diode detectors.

I bought castable 3d-printer resin recently, which would allow me to pour any shape I want, within reason. I see there are no problems making fancy geometry rings and such on youtube.

This makes me believe I can cast waveguide or antenna parts out of silver. Yes, somewhat expensive, but not as expensive as THz electronics. Combined with the silver-on-ceramic PCB methods that I discovered recently, it seems plausible to make some kind of circuit, i.e. a detector coupled to a horn in order to try to pick up THz emissions from a silicon carbide glo-bar source.

I also noticed that most diodes (i.e. conventional diode), is actually very small, for instance ceramic zener diodes, if you crack them open, they have something like a grain of sand sandwiched between two copper cylinders. If I was careful on some lapping media I could make that diode shorter. I believe I read that even normal diodes have a rather wide bandwidth, thats not including 'shortened' diodes, or really small diodes (i.e. 0402 or smaller).

Does anyone have any inspiration for this line of projects? If I cast the right size cavity and horn could I correctly position a modified diode to make a reasonable THz detector?

I am tired of being bared from this electromagnetic region because of financial reasons. I think I can make a glo-bar and I think I can make a cavity/antenna, I just need to know about the detector element in the system 

Thermal Imager/Optics forum version of thread
https://www.eevblog.com/forum/thermal-imaging/mmwaves-optical-methods/msg3929741/#msg3929741

[cdev]

I am also interested in this. I have access to a back yard. I think that outdoor environments have an advantage of being fairly free of reflections. One possibility is building a robotic platform to rotate an antenna in 6DOF. and then receive the signals a bit of a distance away and detect and record the RF signal strength, and perhaps a video of the view along the antenna's primary signal axis. (showing what direction it was pointing in)

This would be extremely useful in developing antennas.

You can also use satellites as signal sources with known locations.

[coppercone2]

Yeah, I am pretty confident it can be made. I think someones post somewhere on this site said common diode had like response to 90GHz or something. But I can't find it.

If you can cast it you don't need a multi-axis cnc machine to make the horn feed or anything like that, and its small enough so you can make it out of solid silver for not too much cost. Vacuum casting would be the best. The geometry of horn feeds is obnoxious.

Any candidates for misused diodes or transistors as detector elements? A simple power detector would be the most logical first step I think.

[Joel_Dunsmore]

not to rain on anyone's parade (PS, THz don't go through rain much either), but THz is really a challenge.  I've been doing a lot of work in the D-band (110-170 GHz) and some up to 330 GHZ, and everything is hard, and fraught with trouble.  You don't need to be outdoors for antenna testing as far-field is something like 10 cm, even for a large phased array.  And you don't have to worry about re-reflections as free path loss is something like 73 dB per meter at 110 GHz.  The thing that kills anything operating at these frequencies is shunt capacitance.  The capacitance for a corner freq of 110 GHz is on the order of 30 femto-farads (0.03 pF) so smallness is definitely your friend. 

[0culus]

not to rain on anyone's parade (PS, THz don't go through rain much either), but THz is really a challenge.  I've been doing a lot of work in the D-band (110-170 GHz) and some up to 330 GHZ, and everything is hard, and fraught with trouble.  You don't need to be outdoors for antenna testing as far-field is something like 10 cm, even for a large phased array.  And you don't have to worry about re-reflections as free path loss is something like 73 dB per meter at 110 GHz.  The thing that kills anything operating at these frequencies is shunt capacitance.  The capacitance for a corner freq of 110 GHz is on the order of 30 femto-farads (0.03 pF) so smallness is definitely your friend.

What sort of equipment are you using for your work up there? If you can't share exact setups that's fine, but I'm curious what's on your bench, so to speak.

[coppercone2]

I bought a bunch of diodes that claim to go into the high regions (fractional THz). I think you can just shrink a conventional waveguide diode detector down.

What makes a good detector diode anyway?

What I was thinking if you liberate the actual diode that is inside of a conventional cerMELF package, it is literarly the size of a grain of sand. You can wedge this into a cavity of some kind to make a electrical contact I think (i.e. grind a tiny jacobs taper out of one side of the diode copper contact that is 1000x times bigger then the diode). What if you press fit the copper contact  of the diode so the diode is flush with a cavity (cold fit). That should make the package transparent I think?


I know its gonna be shitty but I would need to turn into a real big wig to try to even get that gear on a clean financial conscious.


Is that not an amusing image? BTW Props on building a enclosure for a test system! inconceivable. Now the secret is they need to be lego man sized.

[KE5FX]

The people doing the most interesting work at  this end of the spectrum are probably the NRAO folks and their international collaborators.  Some great papers and tech studies came out of the ALMA project in Chile, and are well worth tracking down.   

Darrel Emerson's essay on J. C. Bose and his early mm-wave work is also a must-read, if it's still online. 

[jjoonathan]

What sort of equipment are you using for your work up there? If you can't share exact setups that's fine, but I'm curious what's on your bench, so to speak.

Since Joel probably can't give us details very far beyond what is publicly available at the Keysight mm-wave VNA site, allow me to speculate 

They are working on a PNA-GTX with turbocharger. It does to 0-150 in under a second (GHz, of course), it has flame decals on the side, and the fans make "vrooom" noises that scare small animals. It finally does away with the three-box design (base VNA + mm controller + sample heads) in favor of just having small probe-size sampling heads that plug in to the main body. Different probes are available for different connectors / frequency ranges. To be announced in 2022, but none of our employers will be able to afford it until 2025.

------------

Regarding janky home setups, though, I have to warn about expectations on SLA printers: accuracy and repeatability offer rude surprises coming from the world of machining, even janky home machining, where the usual rule is that you can do .010" without thinking, .001" with care and good technique, and .0001" by really dialing things in and sweating the details. Controlling for time, effort, and experience, cheap equipment sets you back half an accuracy class. 3D printing sets you back 1.5 accuracy classes.

You can get to within a few thou, but you'll have to bust your butt to make it happen. The exposure process grows/shrinks your part as a nonlinear function of geometry, temperature, humidity, resin type, batch of resin, how well mixed the resin is, etc. It's not a problem for printing D&D figurines, but it's a big problem for engineering... which explains the relative adoption of 3D printing to the two applications. For a while I used a chart to map hole sizes modeled vs actual (test part + pin gauges + patience) but after controlling so many variables and still seeing the plot hop around, I basically gave up. Every time I need accuracy on a print I resign myself to the necessity of iterating a few times or post-processing. It's like going back in time to before the interchangeable parts era. I've had 10x less trouble and better results compensating the deflection in a cheap cnc aliexpress router, and that's really saying something, because those aliexpress routers are themselves floppy garbage.

Oh, umm, and be sure to get a printer with linear rails. They are cheap these days. Everything bad I said applies to the ones with linear rails, but the ones without rails are even worse, because the build platform deflects due to stiction and makes the parts wavy in addition to the usual problems of over/underexposure, potato-chipping when drying, potato-chipping over time, elephant foots, and chicken skin.

Of course, if you're willing to relax expectations it's fine. If you just want to see a signal, what are a few dB, or even tens of dB? If you just want one working sample, a few iterations will absolutely get you there. Keep the rat-tail files handy, stay patient, and sneak up on those dimensions 

One thing is for sure, it will be an adventure! 

[0culus]

What sort of equipment are you using for your work up there? If you can't share exact setups that's fine, but I'm curious what's on your bench, so to speak.

Since Joel probably can't give us details very far beyond what is publicly available at the Keysight mm-wave VNA site, allow me to speculate 

They are working on a PNA-GTX with turbocharger. It does to 0-150 in under a second (GHz, of course), it has flame decals on the side, and the fans make "vrooom" noises that scare small animals. It finally does away with the three-box design (base VNA + mm controller + sample heads) in favor of just having small probe-size sampling heads that plug in to the main body. Different probes are available for different connectors / frequency ranges. To be announced in 2022, but none of our employers will be able to afford it until 2025.

------------

snip

OK, that's hilarious! 

[coppercone2]

hmm well for some things I can see using some kind of broaches etc to clean it up (waveguide) but yeah I expect high losses and its just supposed to be some kind of detector.

[jjoonathan]

Yeah, I can't wait to see the results!

A while back MACOM bought what was left of Metelics and their THz line is here: https://www.macom.com/products/diodes/thz-diodes . I vaguely recall seeing some of them on eBay for prices I considered in-reach. I don't think I've ever seen Virginia Diodes products "in reach," but I haven't been aggressively looking either.

I know that there are a ton of special fabrication techniques to optimize diodes in this frequency range, but I don't have intuition for just how much you would sacrifice by going without. Is it just a 10dB suboptimal response? 30dB? Do they not work at all? I think you're on the right trail chasing Bose and NRAO for the "scrappy startup" type techniques. Of course, I'll still tune in for any Keysight announcements  even if I can't be an early adopter.

At the moment, I'm content saying "my eyeballs are the only THz detector I need," even if that's cheating 

[coppercone2]

How do you feel about the high GHz russian diodes on ebay?

I.e. 200GHz pin diode ?

Can a Pin diode be used for this?

I am curious about the quality of the castings I can get, the best quality would be silicone rubber vacuum casting silver, but that is getting kinda fancy.

If you want to use broaches or other odd cutting tools smaller would be better I think since its easier to fabricate smallish tools, there is probobly a sweet spot for every machinery setup that would need to be found... but I did get my hands on a depth micrometer which might be kind of nice for this.


ALso, about propagation modes... can circular wave guide be used OK for a detector? I.e. just a tube. Do you need the square geometry for such a primitive device? I did get some arco laps too.. I would prefer to avoid angles as much as possible.  Please let me know if I can cut any corners, literally.

I was hoping to get between 0.1-0.4 THz. That MACOM thing looks promising but for 1THz I guess you would need to really tune the manufacturing process down... as usual its probobly better to work in steps then shoot for the gold on a wobbly table.


I also wonder if casting quality can be improved, when you pour the cast around the wax blank, if you put it in a ultrasonic cleaner for a while to act as a vibrational compacter for the plaster.

Also wonder what happens if you try to rumble a molten casting that is cooling down. cavitation would be bad I guess

[Joel_Dunsmore]

What sort of equipment are you using for your work up there? If you can't share exact setups that's fine, but I'm curious what's on your bench, so to speak.

  I've use the two American producers of mm-heads (VDI and OML), there are two others as well.  Lately mostly VDI with up/down converters as well as mm-wave extenders (they are actually 2 different kinds of modules). Doing things like measuring phase noise at 144 GHz, as well as modulation quality (on the order of 1 Gsymb/sec also at 144 GHz).   And with Suren Singh (a.k.a. Mr. Keysight Millimeter), been trying  noise figure measurements at 110-170 and struggling a bit at 320 GHz.   Y-factor and cold-source.  It takes the better part of a day to setup, calibrate, verify, test the DUT, and verify that the test results make sense.  And almost every test setup has something wrong with it, so tear down, figure out what you did wrong (Isolator in backward, calibrated a power so low that broadband kTB noise dominates, loose screw on the waveguide flange, power off on the NF pre-amp). All those tests were with PNA-X (yes, it has a baseband AWG in it now).  Doing a little work the the new UXA and also UXR as IF receivers for mm-wave down converters.  Might have a paper in the fall on the world's lowest EVM measured at 144 GHz if I get myself sufficiently motivated.

A little more specificity: I use a N5245B PNA-X 423 with the extra 13.5 GHz rear panel option XSB source that also supports now Arbitrary Waveform Generation 10-6000 MHz. I generate a 1 GHz QAM centered at about 5 GHz with that, and then use the VDI CCU WR6.5 mm-up converter to generate a 144 GHZ output. The PNA internal port 1 or port 3 is used for the VDI LO; then a VDI CCD is used to down convert. I put a waveguide attenuator in the path to make sure the CCD is in the linear range and use the down-converted output to the PNA port 2 ; where I measure and pre-distort the XSB source to get the lowest possible EVM. then I insert the amplifier under test into the path and I can record the EVM of the amplifier+CCD, most of which will be from the amplifier.  Oh, and you need a band pass filter on the CCU output to select the proper carrier sideband out of the CCU, if you use the same LO for the down converter; or you can use another one of the PNA sources to generate a different LO so the image doesn't down convert in-band.

I've done the same setup with a Marki MMIQ mixer up to about 100 GHz, but in that case I use N5291 system as the down converter.
N5291 is spec'd to 120 GHz but you can push it to 130 GHz if you hold your mouth right.   Then I can use it to cross-check my results that I got with the external D-band converter in the 110-130 GHz range.  Then I spend a day trying to figure out why they aren't the same. 

[0culus]

What sort of equipment are you using for your work up there? If you can't share exact setups that's fine, but I'm curious what's on your bench, so to speak.

  I've use the two American producers of mm-heads (VDI and OML), there are two others as well.  Lately mostly VDI with up/down converters as well as mm-wave extenders (they are actually 2 different kinds of modules). Doing things like measuring phase noise at 144 GHz, as well as modulation quality (on the order of 1 Gsymb/sec also at 144 GHz).   And with Suren Singh (a.k.a. Mr. Keysight Millimeter), been trying  noise figure measurements at 110-170 and struggling a bit at 320 GHz.   Y-factor and cold-source.  It takes the better part of a day to setup, calibrate, verify, test the DUT, and verify that the test results make sense.  And almost every test setup has something wrong with it, so tear down, figure out what you did wrong (Isolator in backward, calibrated a power so low that broadband kTB noise dominates, loose screw on the waveguide flange, power off on the NF pre-amp). All those tests were with PNA-X (yes, it has a baseband AWG in it now).  Doing a little work the the new UXA and also UXR as IF receivers for mm-wave down converters.  Might have a paper in the fall on the world's lowest EVM measured at 144 GHz if I get myself sufficiently motivated.

A little more specificity: I use a N5245B PNA-X 423 with the extra 13.5 GHz rear panel option XSB source that also supports now Arbitrary Waveform Generation 10-6000 MHz. I generate a 1 GHz QAM centered at about 5 GHz with that, and then use the VDI CCU WR6.5 mm-up converter to generate a 144 GHZ output. The PNA internal port 1 or port 3 is used for the VDI LO; then a VDI CCD is used to down convert. I put a waveguide attenuator in the path to make sure the CCD is in the linear range and use the down-converted output to the PNA port 2 ; where I measure and pre-distort the XSB source to get the lowest possible EVM. then I insert the amplifier under test into the path and I can record the EVM of the amplifier+CCD, most of which will be from the amplifier.  Oh, and you need a band pass filter on the CCU output to select the proper carrier sideband out of the CCU, if you use the same LO for the down converter; or you can use another one of the PNA sources to generate a different LO so the image doesn't down convert in-band.

I've done the same setup with a Marki MMIQ mixer up to about 100 GHz, but in that case I use N5291 system as the down converter.
N5291 is spec'd to 120 GHz but you can push it to 130 GHz if you hold your mouth right.   Then I can use it to cross-check my results that I got with the external D-band converter in the 110-130 GHz range.  Then I spend a day trying to figure out why they aren't the same.

Neat stuff. And one expensive lab bench. 

[Joel_Dunsmore]

yes, a bit dear.

[jjoonathan]

Wow! Nice! Thanks for sharing! That's some crazy awesome kit -- and research activities to match!

It's wild to think of the UXR as an IF receiver, and even more wild to consider that it has higher frequency range than the UXA, and not by a little    I see that ENOB becomes a challenge with mm-wave oscilloscopes -- how much can process gain typically claw back in this application? Close to the theoretical amount or does something else become limiting?

Re: Marki MMIQ, are we talking about harmonic mixing on the MMIQ-4067L? I'd be really interested in hearing tips and tricks on this, since I've been doing a bunch of 3rd harmonic mixing on MM1-1467L. I suppose nobody ever told me I was allowed to use odd LO harmonics on mixers designed for fundamental mixing and I'm still a bit caught up on that even though the results are in and it seems to work just fine -- I got a lot of awkward silences asking around about it. The story is straightforward (diodes clip, odd harmonics appear), the Henderson model says it should work for a very modest 10dB penalty, spur tables say it should work, and it does in fact seem to work. It would still be comforting to hear that it was a respectable thing done by respectable people in respectable settings, lol. Also, if you are doing it with a quadrature mixer, and on non-CW signals with distortion requirements, those are two levels of sophistication beyond what I've been doing and it would be nice to hear that it works in that regime as well.

[0culus]

I am just happy with narrowband microwave work! Regardless, it is very cool to see a glimpse of the state of the art. 

[coppercone2]

sounds about right, I had similar experience trying to learn microwave with non PCB components. Always a problem. At least its fun, until you have to unscrew n connectors too many times, or drop something. (much more simple characterization then yours)

At east with MMwave you have small threads, I swear I tweaked my hand before with too many n connectors.

[0culus]

sounds about right, I had similar experience trying to learn microwave with non PCB components. Always a problem. At least its fun, until you have to unscrew n connectors too many times, or drop something. (much more simple characterization then yours)

At east with MMwave you have small threads, I swear I tweaked my hand before with too many n connectors.

Yeah no kidding. I ordered a 20mm 8 in lb torque wrench. It'll mainly be used for when I need to tighten the test port NMD connectors on my test set, but it'll come in handly for N connectors too.

Also I can only imagine how careful the connector hygiene would have to be in Joel's lab. The tiniest micron of dust will probably upset the whole setup. 

[coppercone2]

If they had like a 5 inch knipex syphon pliers for knureled connectors I would buy it. I should get  knipex parallel jaw mini pliers wrench too.

I would recommend assembling that in a dust free hood actually.. if its that sensitive.

[0culus]

If they had like a 5 inch knipex syphon pliers for knureled connectors I would buy it. I should get  knipex parallel jaw mini pliers wrench too.

I would recommend assembling that in a dust free hood actually.. if its that sensitive.

Or you just work in a clean room.

[coppercone2]

unless its low level people would HATE that. No one wants to put on that suit.

But, can it really be worse then optics? I figure any of these technologies is more dust resistance then actual light.

[0culus]

unless its low level people would HATE that. No one wants to put on that suit.

But, can it really be worse then optics? I figure any of these technologies is more dust resistance then actual light.

I think it might be worse than optics...a fiber or something stuck in a center pin at these frequencies will cause madness, mayhem, and confusion for the impedance of that connector. This would be less of an issue with traditional optics experiments I would guess.

[coppercone2]

True they are so small the fiber is physically large compared to the contact, I guess light might still pass through but not a conducted signal.

[jjoonathan]

Speaking of soft wrenches, I've been playing around with the idea of 3D printing them. The major challenge is material strength -- even with special high-strength resin and three generations of chunkier design, they still barely work to 1Nm torque. Also, tolerances are a challenge. They need to be snug because sloppiness exacerbates the strength requirements, but "snug" fits just aren't something resin does well. You have to really dial it in, but then a resin change or temperature change or humidity change throws it right back off.

I've had a lot more success with "wrench stands," as the geometry is better suited to provide strength. They have a more unique value prop, too: they solve the three wrench problem and they chop a degree of freedom out of the alignment process. I've "standardized" on designing them to hold the center of the connector exactly 3cm off the table (or, more recently, chopping block, since my desk is wavy enough to frustrate the process without some kind of surface plate).