SMA phase adjuster construction?

[rhb]

Does anyone know the construction details of these things?

https://coaxicom.com/product/sma-male-for-rg402-semi-rigid-phase-adjuster-3/

https://storefront.brackemfg.com/storefrontCommerce/itemDetail.do?item-id=8940&order-quantity=1&item-index=0&customer-item=BM12000&order-uom=EA&warehouse-id=1&item-number=BM12000

They're rather pricey, so I'm wondering if a simplified version is possible.  I would expect that they are air dielectric,   tube in tube construction with a solid inner conductor sliding inside a tube for the center conductor and tapered walls.

I can get a ~1.5 ps variation by moving a loose SMA  to 3.5 mm connection, but I want about 10-20x that.

I'd potentially need enough of them to pay for a Sherline lathe to make them with money left over.

I found the attached photo on eBay. 

[KrudyZ]

The principle is a sliding fit of a trombone mechanism.
Colby Instruments makes these with much larger delay ranges.
https://www.colbyinstruments.com/
You can probably find their patent to see how it is done.
The difficult part is to maintain the impedance in the two sections and minimize any step effect between the two sections.
I would imagine this requires very tight machining tolerances to get low VSWR.

[rhb]

I would expect that the outer wall of the inner conductor is tapered and that the inner wall of the outer conductor is tapered.  But that might not be necessary.

If you have two 50 ohm air lines that slide together then the only discontinuity is the fringing effects at the edges.

I've been trying to determine what is used to support the center conductor in a 3.5 mm connector  I'd prefer not to buy one to cut in half.  And even if I did, I'd still have the issue of identifying the material.

The Tek 11801 is a pretty amazing instrument.  I can see the ringing from the calibrator step reflecting off the sampling head and calibrator ports and measure  the ringing in the calibrator and head at 25 GHz. No way, though, to tell if it's the head or the calibrator without getting an SD-32.

I'd love to get a dead SD-24  to take apart and map out which segments of the circuit are producing which reflections.  Reflection seismology at the speed of light :-)

All of this is in pursuit of my FOSS DSO FW project.  With two sources and two receivers I can solve for the ringing in each and correct for component variations in the AFE.  I'll need to add that to the "To Do" list.

[coppercone2]

I suspect these things might be made with a CNC jig grinder. I have been thinking about how someone might make these sorts of parts quite a bit.

https://www.mmsonline.com/articles/z-axis-opens-new-doors-for-jig-grinding

on precision cones:
https://www.practicalmachinist.com/vb/abrasive-machining/jig-grinding-tapered-lead-similar-countersink-265772/

[rhb]

A Sherline lathe is more than adequate, takes less space and is much cheaper.

[coppercone2]

well if their made to extreme precision you need some kind of grinder. the surface finish from the lathe is going to be poor in comparison. Thats how they get those tolerances in the millionths and very high quality surface finishes that won't degrade from repeated insertions etc.

I think the inside cone is harder to. here is someone doing a exterior cone face

[rhb]

In the late 40's, Dave Broadhead made screws for the ruling engine being built by John Strong.  The initial trial screw, done in 1947 as a warm up exercise, was within 0.000 002".  It was made on a Southbend lathe that Broadhead had trued up by removing all periodic errors an rescraping the ways to 0.0001".

You can find more in the June & July 1952 Amateur Scientist columns.

However, for phase adjusters, 0.001" is good enough.  That's not easy, but not all that hard either.  It's not like trying to meet 0.0001" tolerances.

[PaulAm]

Although you can cut threads on a Sherline (if you buy the attachment), it's not my idea of fun.  I would definitely not want to cut threads in steel.  In general, you may run into rigidity and chatter problems if you try to use something other than aluminum or brass on a Sherline.

I'd use a slitting saw and spin fixture to cut the collet grooves.  You could do that with a hand saw but repeatability might be an issue.  I wonder how burrs would affect the performance.  Seems to me when you get over 1GHz and look cross eyed at a component you can affect performance (usually adversely).

I'm not saying you can't do it, but making decent repeatable phase adjusters in a decently equipped machine shop would still be a challenge.  There's a reason they're expensive.

Personally, I would look for some surplus (unless you really want to get into precision machining - that is not an afternoon's task).  Trying to make your own is a much bigger detour from your project goal than you seem to think.

For the record, I do have a decent machine shop and I'd still think long and hard before making my own.  Then I'd buy them and get on with my project 

[rhb]

I also have a good machine shop. But i certainly would not claim to be a good machinist.   I would not considering duplicating the designs pictured unless I had a CNC machine.

I have in mind modifying some Chinese SMA connectors to fit suitable tubing, soldering tubing to them, sliding the tubes  together, adjusting to length and soldering in place to create a fixed delay.

I'd much rather do that on a Sherline than on my 10" x 20" Clausing.  However, I might go with a Chinese 7" x 12-14" machine.  The better ones have become quite good and I can make a hand crank for fiddly threading jobs.

Any  lathe can turn steel.  The chip size simply gets smaller for steel than for brass and aluminum.  If it's got good spindle bearings, you can make watch parts on a wood lathe. It would be clumsy is all.

[rfeecs]

I've been trying to determine what is used to support the center conductor in a 3.5 mm connector  I'd prefer not to buy one to cut in half.  And even if I did, I'd still have the issue of identifying the material.

Could be Rexolite:
http://www.rexolite.com/

This gives an idea of some of the APC-3.5 tolerances:
https://www.maurymw.com/pdf/datasheets/5E-062.pdf

[rhb]

Thanks.  I'd looked at several datasheets, but they were not as detailed as Maurey's are.  Small wonder they're expensive.  Holding .0002" in production is not easy. That requires temperature controlled flood coolant and probably a good bit of experimentation with feeds and speeds.  I doubt that a computer program could calculate the exact amount of heat carried off by the coolant accurately.

I've been using the calibrator output of my 11801 fed to a tee at the input of a 20 GHz SD-26 to look at reflections from various connectors and cables.  The firmware is awful, but what an instrument!  Because of my reflection seismology background, interpreting time domain data is very comfortable.  I can translate the trace directly into an impedance profile.

[mrflibble]

I can get a ~1.5 ps variation by moving a loose SMA  to 3.5 mm connection, but I want about 10-20x that.

Just give in to the dark side, and get that optical table. You know you want to... 

That would neatly solve several of your side-sidetrack projects.
1) 1.5 entire picoseconds equates to a huuuuuuuge distance traveled in photon country, and as such is trivial to adjust for.
2) Once you have all that laser goodness going, you might as well slap on a few gratings & prisms to do some of that pulse stretching & compression. That should easily take care of your desire for shorter-than-short pulses.

Personally I'd just buy a used phase adjuster at ~ $100 on ebay, and get the rest of the project done. You can probably sell it after you're done with it for a similar price if you treat it nicely.

[coppercone2]

i still think your surface finish will be poor on a lathe. I think you would need to grind those threads after you cut them optimally.

[rhb]

Err, uhmm

I need  7 phase adjusters, not just one.  If I only needed one I'd have ordered a new one from Bracke.

As for turned finishes, you've apparently not seen good work or don't know what to look for to distinguish turning and grinding.

[coppercone2]

are you going to get comparable surface roughness from turning vs a 300+ grit stone?

[coppercone2]

I go by this (as is practical to do with the black bars)

[rhb]

Both turning and grinding provide the same surface finish at the minimum roughness end of your chart your chart.

[coppercone2]

not typical results
needing 9 of them with cost sensitivity you need typical results.

I think with a engineer doing this your going to get some where to the left haha. maybe a experienced machinist can wing it.

And I honestly have trouble imagining that surface finish for a taper. I think it means for shaft.

[rhb]

The requirements for the best quality surface finish are a sharp tool ground at the correct angles, a very fine feed and a light cut.  It is much easier than you think.   Amateurs do it all the time.  The fact that it is not typical simply reflects the surface finish requirements of the enormous quantity of fasteners made by turning using automatic screw machines.  The sole reason for preferring grinding over turning in order to meet a surface finish requirement is speed.  The fine feeds required are much slower than grinding.

I can't think of any RF connector that *could* be made by grinding. You can't grind a 0.036" hole.

I've watched the machines used to make ball point pen tips to tolerances of millionths of an inch in operation.  They are very impressive.  Someone is almost certainly using the same machine I saw to make RF connectors.  The sales and marketing guy that gave the machinist's club to which I belonged  a tour simply mentioned BIC pen tips as an example of the sort of part the machine made.

[coppercone2]

video of that?

[rhb]

This was 25 years ago.  I don't even remember the company, though I think they were Swiss.

You can't see anything when they are running.  The coolant flood makes the loading door look like a front loading washing machine.  They are the modern version of a screw machine.  CNC instead of cams.  And the parts don't just drop into a hopper.

I'd bet they are the devil to get tuned to produce things to the specs of the Maury 3.5 mm connector.  And almost as much work to keep them in tune.  At 0.0002",  temperature control while turning is critical.  Just a little heating of the part and it will be under size.

[dcarr]

RHB,

Would it work to just make a set of microstrip or CPWG lines and then tap them with .085 hardline at the point you need to achieve your delay?  Yes, there will be a little impedance bump and you'll have to cut the line after the tapping point, but really cheap, easy, and decent performance.  I've heard that careful application of grease to the coax-microstrip transition can be used to tune out the mismatch, too.

David

[joeqsmith]

RHB,

Would it work to just make a set of microstrip or CPWG lines and then tap them with .085 hardline at the point you need to achieve your delay?  Yes, there will be a little impedance bump and you'll have to cut the line after the tapping point, but really cheap, easy, and decent performance.  I've heard that careful application of grease to the coax-microstrip transition can be used to tune out the mismatch, too.

David

I was wondering how two microstrips butted face to face would work.   Slot the PCB to make it adjustable.  Crude but I am really not sure what OPs requirements are outside the delay. 

[rhb]

David,

Thank you.  I think you just solved my problem.

Here's the situation:

The FPGA boards use a 2x6 0.100" IDC F header (PMOD) to bring out 8 signal lines.  It doesn't appear that anyone makes a male cable header, so I'll have to use an IDC F connector and M-M header pins to mate them if I use a cable connector as I had been planning.  However, by laying out a 2x6 M header on a board with microstrip lines I could adjust the electrical length before tacking the 8 coaxial lines in place with solder and then trim the excess stubs with a knife and peel them off.  That should be a lot less work. Then I can epoxy all the cables at the edge of the board to secure them without putting any strains on the solder joints.

I also need to make a 9 port resistive splitter to feed the calibrator signal to the head inputs to align the head skews.  So I actually have to solve the delay matching problem twice.  But matching lengths on a board should be a lot easier than matching cables.

I'll also need to feed a test signal in which is synchronized and the resistive splitter alone will not meet that requirement as the levels will be too low   I'm going to be learning a lot more about microwave construction than I had intended.

[joeqsmith]

If the signals are digital, does the FPGA you are using allow a fine enough trim to do what you need?