Offbeat microwave question

So one day soon I'm going to have these two SMA-mounted photodiodes that work up to about 10 GHz. Conveniently, they have built-in bias tees but don't have built-in load resistors, so they need a decent-quality termination.

They're part of a differential Doppler interferometer producing tone bursts of very roughly 200 cycles, immersed in wideband white noise, to the point of being completely invisible on a scope.

I need to build/buy/cobble together a splitter/combiner that works from DC to 8 GHz. I can buy resistive ones, but they have a 6- or 7-dB insertion loss, which means that by the time I'm done attaching an amp with a 3-dB noise figure, I'll wind up with about a 3000K noise temperature. Those photons are expensive, so this is clearly a non-starter.

I'd also like the two paths to be phase-matched to within 5 degrees or so, and the gains to be equal to within, say, 0.3 dB. That way I have some hope of getting 15 or 20 dB of laser noise suppression. On the plus side, because of the signal characteristics, I don't care much about the time domain response, and a few amplitude whoopdedoos aren't worrisome as long as they aren't too deep or too broad.

So, I'm thinking of just siamesing the two detectors together via matched pieces of RG-402 semirigid coax (0.141 inch) and either a T connector or some sort of artistically sculpted microstrip thing, e.g. a vanilla Wilkinson coupler tuned to some higher frequency.

I'm sort of leaning towards the tee connector to start with, because the actual input resistance of the average connectorized amplifier is around

25 ohms, iirc, so the match should be reasonable, and the degradation in the noise figure due to going a bit off the noise match shouldn't be too bad.

True? False? Smart? Stupid? Other suggestions?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs
Loading thread data ...

A Wilkinson is resonant when it comes to isolation but I guess you don't need isolation here.

If you use coax make sure both end right at the amplifier input. If you can't do that go from the point where the two cables joint to the amp input with two parallel coaxes. So that the impedance is half. Otherwise you'd see a big reflection from that point.

Regarding input impedance, are you sure the noise-optimal input impedance is 25ohms? If not, maybe it can be tweaked to be around there.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Why not dump each pd into a 20+ GHz MMIC first? That would get your s/n up before you start mixing and such.

If you're already SMA, the MMICS could be bought in cans, with connectors. Tweak the MMIC power supplies to trim differential gain, and stick an SMA trombone in one or both legs of the outputs to tweak phase.

The input impedances of MMICS does tend to be low, but Vcc trims that, too. I've seen some that cross right through 50 ohms. It also depends on the output load impedance, because of the feedback resistor in the usual Darlington gain stage. I have low-level-TDRd MMICS to measure their wideband input impedance.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

  • DIY termination? Perhaps use two 100 ohm 0603 in a physical "Y"to ground plane (am presuming 50 ohm system)?
** snipped to keep aioe happy **
  • Am assuming a balun transformer is not much better..

** snipped to keep aioe happy **

  • Did not know about the Wilkinson coupler and was going to suggest tapered transmission line "Y". But hoo nose? maybe the darn resistor can be tossed (when properly matched) for better noise and maybe lower loss.
** snipped to keep aioe happy **
Reply to
Robert Baer
** snipped to keep aioe happy **

Tapered transmission lines are one conventional way for impedance "conversion".

Reply to
Robert Baer

Hard to do over multi-decade bandwidths, though.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I'll have to get a few and do that. The difficulty with separate paths is that it'll be really hard to get the gain and phase to track well enough over frequency to have any decent subtraction performance.

Ideally I'd like to use a split cell, but they're not that easy to get. Discovery Semi supposedly makes them, but they give you the third degree before they'll even give you a datasheet, and that's pretty off-putting.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I thought about wiring them in series instead, with some Type 61 ferrite sleeves over RG-405 coax for broadbanding, and a pHEMT or something as the front end. I'd do that at 1 GHz in a heartbeat, and I've done the copper tape and XActo thing up to about 3 GHz, but 8 GHz is a little more difficult.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Yes, and you optical guys often need "DC to visible light" bandwidth.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Would you have to trim the gains, optically maybe?

Why do so many companies defend themselves against customers? It's not as if their competitors have any real trouble getting data sheets.

I've been trying to get ballpark quotes on deep-drawn aluminum cans, and some of the makers do everything possible to avoid selling them. One told me that they will have to go out and get pricing on aluminum before they can quote.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

Needing something bigger than cookie tins? ;-)

Tim

--
Deep Friar: a very philosophical monk. 
Website: http://seventransistorlabs.com
Reply to
Tim Williams

Well, I think I'm the local record holder, anyway--I've built 200 THz crystal radios. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Yup. The "wall in the sky" is around 100 ps. 8G is about 45 ps, and things are getting serious.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

The two beams come from a linearly-polarized beam going through a

45-degree tilted Wollaston prism (a type of crystal polarizer), so I can adjust their relative strength using a half-wave plate ahead of the Wollaston. That will adjust the overall gain very accurately.

At low frequency, parallelling the photodiodes guarantees that all their whoopdedoos are the same, because there's only one set of strays instead of two. At 8 GHz, the wavelength in cable is only an inch, which makes it a bit tougher to do well.

Weird. You obviously have to start drinking more Budweiser. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

You need a long, gradual taper, and then the losses getcha. It can be useful for stuff like increasing the peak voltages of pulses, where signal quality is secondary.

Oh, remember my pulse inverter experiment?

formatting link

formatting link

formatting link

At DC, of course, it's a dead short. Adding a ferrite or two extends the LF response.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

AM. Talk radio.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

I've used coax Christmas trees to make pulse sequences from an SD-24. Working out where to place the stubs and how long they should be is fun.

In the photodiode case it isn't as bad, because I can float the cases. Thus I'd only need the balun to work down to a megahertz or so, which is probably doable with some combination of 43 and 61 ferrite beads on a reasonable length of RG 402 or 405. I ordered a few dozen from Amidon yesterday, so they should be in by the time I get back from the wilds of Orange County. (How do people stand living in Irvine? Inquiring minds, and all that.)

BTW when did RG-402 get so stinking expensive? I used to pay more like

50 cents a foot, and now I see it's more like six bucks a foot. Pretty steep for just tinning the shield of some RG188!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

For the Nova laser, the predecessor to NIF, they needed an arbitrary waveform generator to drive the beam modulators. What they did was to machine a christmas-tree sort of pattern into the top layer of a shorted microstrip line, and use its TDR response as the arb. They developed software to turn a waveform into a machining program.

(We did the modulators for NIF, all electronic.)

There's tons of cool cables and connectors on ebay. I buy a batch now and then, to cut up and such.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

Me too. I just got a lifetime supply of MMCX jumpers (500 pcs for $50).

Any experience with the cheeep'n'cheerful Chinese semi-rigid?

Unfortunately I have to spend Saturday night down here in sunny-but-depressing Orange County--the system integration is going a bit on the slow side. :(

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Go to Disneyland. Or drive out to the PCH. Visit the Getty museum in Malibu.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.