Author Topic: Testing RF connectors and cables  (Read 16630 times)

0 Members and 1 Guest are viewing this topic.

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #50 on: September 20, 2019, 02:31:16 am »
I think I should note that most of the problem is a matter of not being able to grip the SMA male.   A knurled round wrench the diameter of an N male with a slot cut in it would get them tight enough.

I have tightened a *lot* of different connections while watching the 11801 display.  It's given me a very good sense of how much force is needed to get a good connection.  Lacking a lathe you could cut a small square of metal, drill a hole in the center and then saw to the hole to create a slot and clean up with a file.  While not elegant, it *is* good enough.

However, if you're playing above 6 GHz, using SMA you definitely need a wrench.  But the price tag for playing up there is so high I have to ask "Why?".  It's interesting to be sure, but do you *really* want to spend $50-100 for a *single* connector?

I first encountered the SMA in my HP 8601A sweeper.  I thought it was the cat's meow, but also assumed it was only intended for infrequent connections.  The life cycle connection limit reinforces that.

I consider it a great way to connect shielded modules.  I'm not real keen on them beyond that.  But they are cheap and good enough if you stay below 6 GHz and much better than BNC.  I've got some BNC test examples coming next.

Have Fun!
Reg
 

Offline syau

  • Frequent Contributor
  • **
  • Posts: 368
  • Country: hk
Re: Testing RF connectors and cables
« Reply #51 on: September 20, 2019, 02:46:14 am »
Planning to buy one, which torque value should I use for these brass based connectors ?
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #52 on: September 20, 2019, 04:19:41 am »
I don't know the standards from memory.  Look it up on the Internet.

I don't even know what mine is.  I need to check it and may need to buy another if it is set for stainless steel connector torque.

Have Fun!
Reg
 
The following users thanked this post: syau

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1208
  • Country: 00
  • mmwave RFIC/antenna designer
Re: Testing RF connectors and cables
« Reply #53 on: September 20, 2019, 05:39:59 am »
Planning to buy one, which torque value should I use for these brass based connectors ?

In my experience, it doesn't matter too much what torque exactly you use. The most important thing is that you use a torque wrench - a fraction of a turn with these connectors can be the difference between tight fit and damaged connector face.

From memory, all our 3.5/2.92 mm wrenches are 1 Nm, and our 2.4/1.85mm wrenches are 0.9 Nm, 1mm connectors uses 0.7 Nm? I am not 100% sure. I've seen 1.1 Nm as well for 3.5/SMA.

The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 
The following users thanked this post: syau

Offline _Wim_

  • Super Contributor
  • ***
  • Posts: 1514
  • Country: be
Re: Testing RF connectors and cables
« Reply #54 on: September 21, 2019, 08:21:30 am »
« Last Edit: September 21, 2019, 08:23:13 am by _Wim_ »
 
The following users thanked this post: bitseeker, FransW, syau

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #55 on: September 21, 2019, 05:48:47 pm »
FWIW the 2nd APC-7/N-F adapter from besttrade360 arrived today.  The first time I connected it, I saw a large reflection at the N-F interface which raised concerns about QC.  However, when I connected it a 2nd time after hooking up the 1st one and saving the TDR trace I got comparable results to the 1st one.

Measured rise time was 33-34 ps on the thru connection and the largest reflections were from the Astrolab SMA-F/N-F adapters.

I'll report on the $16.50 each pair when they arrive from China.  But certainly these are entirely satisfactory for use on an 85046A or 85047A S parameter test set.

With all the RF connectors one needs and the high prices of the name brands, getting a pair for under $50 vs a pair for $270+  makes a huge difference.  It makes buying Chinese connectors and simply returning those which fail testing worth the trouble.  But testing *is* mandatory.

Have Fun!
Reg
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #56 on: September 26, 2019, 06:19:31 am »
Well, too of the APC7-N Female arrived here today.

The first thing I noticed was the APC7 nut was really loose.  Then I noticed the lack of the threaded sleeve that enables them to mate with one-another, probably explaining the looseness.  And then the nail in the coffin was the surface finish.  The mating surface is not smooth - like it wasn't machined at all.  Unfortunately, the focus leaves a lot to be desired, but it's quite visible on the first attached picture.  The surrounding surface isn't smooth either, but it's not critical.

The N barrels and ends in contrast are nicely machined, but both are blemished.

I'm not going to use them.  Given the APC7 end, I'm not too keen to be mating my N standards with the N ends.

rhb: Do yours have the threaded sleeve that's clearly visible on the second attachment?

73, Orin KJ7JQ.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #57 on: September 26, 2019, 04:34:09 pm »
The male threaded sleeve  can get stuck down in the female threaded nut.

You can take a pin spanner and unscrew the APC-7 end.  You'll find 4 pieces.

843154-0

The problem is that the pin that is supposed to keep the male threaded sleeve from rotating is too short and doesn't engage the slot properly.  If you turn it too far, the pin gets pushed out of the slot and the male sleeve gets stuck in the bottom of the female nut.

843158-1

So here they are put back together as they should be.

843162-2

If you look closely, you can see that the brass pin is not flush.  If you try to drive it in you run the risk of knocking it all the way through.  then it's a *real* nuisance.

Then use the pin spanner (I used some heavy tweezers) to screw the central bushing back on the body.

If you're constantly remaking the APC-7 connection it's worth fixing.  But if you're going to only attach it to an 85046A or 85047A and then leave it there it's not a huge problem.

NB:  After knocking the too short brass pin out, I was able to properly  fix it with a piece of 1/16" gas welding rod.  It took a good bit of time, so I suggest *not* doing that.  However, the male sleeve now operates reliably.
« Last Edit: September 26, 2019, 05:42:53 pm by rhb »
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #58 on: September 27, 2019, 06:57:53 am »
So, I disassembled the adapters and indeed, the sleeve had escaped.

A Park SPA-2 pin wrench will work, though you do have to squeeze it rather tight.  I wouldn't recommend it.  I suspect the cheap snap-ring pliers the auto parts stores sell would work better.

I'll see how they do back to back up to 3GHz on the 8753C at work...

 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #59 on: September 27, 2019, 02:55:21 pm »
Please post the results.  Once I get my 2nd pair I should be able to do the same with my 8753B.  Meanwhile I'll sweep the current pair back to back on my 8560A.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #60 on: October 10, 2019, 12:01:14 am »
I just got the 2nd pair of APC-7/N-F adapters which I ordered on ebay from hualun1988 for $16.49 each.  I made them up with Wiltron APC-7/N-M adapters, did a thru cal on my 8560A using an N-M/N-M and an N-M/N-F cable and then inserted them. The ripple is approximately trace width on all of them at 2 dB/div. I judge it as no more than 0.15 dB from 1 MHz to 2.9 GHz.  The cables alone had ~1 dB of ripple around 40 MHz and 2.5 GHz before normalization and are of good quality.

Both of the new adapters had to be taken apart and the pin put back in the slot,  but the pins are long enough to stay engaged.  At least if you don't get ham handed.

I don't have a way to test these above 2.9 GHz except by TDR with the 11801 & SD-24.  I'll add that and the 8560A photos later.

I also got a Chinese N-M SOL cal kit for $17.99, US stock.  I'll be testing those against my Anritsu SOL with the Wiltron APC-7/N-M & Chinese APC-7/N-F and also a probably Chinese APC-7/SMA-M.  The TDR will allow seeing which interfaces have the most mismatch.

I can see no reason these won't serve on an 85046A just fine.  With the ebay asking prices for an 85047A, I don't see myself getting one any time soon.  And no real need for one anyway.  Going above 3 GHz just makes *everything* cost more.

Have Fun!
Reg
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 388
  • Country: 00
Re: Testing RF connectors and cables
« Reply #61 on: October 10, 2019, 05:54:13 pm »
Has anyone used an indicator/gauge on any of these bargan connectors?

The ACP and N connectors are probably a little more robust/forgiving. 
I'd be particularly concerned with SMA adapters since they can be connected to 3.5mm connectors and cause potential damage.

rastro
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #62 on: October 10, 2019, 06:59:41 pm »
Sadly, I missed a Maury A-027A set on ebay because I set the clock wrong.  I'm planning to make a set of GO/NO-GO gauges myself.  But the 7" x 14" Chinese mini-lathe I bought to do the work is just a semi-finished  kit of parts assembled at the factory to make sure they didn't leave anything out.  Serious work required to finish it.

A big problem with the 3.5 mm compatible connectors is the tolerances required.
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #63 on: October 10, 2019, 11:49:35 pm »
Please post the results.  Once I get my 2nd pair I should be able to do the same with my 8753B.  Meanwhile I'll sweep the current pair back to back on my 8560A.


I finally got around to playing a little.

Calibrated the 8753C for S11 with an old 85032B N calibration kit using an existing APC7-NM adapter on Port 1.

I put the two APC7-NF adapters from the 85032B back to back and added a 909F load.  This is the lower trace below, then did the same with the Chinese adapters.  The resulting upper trace speaks for itself:



To prove to myself nothing horrible was going on, I removed the adapters and attached the calibration load - (909F.gif).  The lower trace is now the calibration load.  It looks reasonable and is about what I'd expect measuring the load I'd used for calibration:



So I then tried one Chinese adapter with a 909C APC7 load - (SingleAPC7-909C.gif):



And finally, a 909C N-F load (909C-N-F.gif) - to see what we can expect out of a 909C compared to a 909F.  The 909C isn't as good as a 909F.  It's the flatter of the two traces.  (Unfortunately, the plot capture didn't draw the traces in different colors.)



I didn't do any S12 measurements...  I have two semi-reasonable Mini Circuits N-M to N-M cables.  One is intermittently bad.  I could only find one at work today and I think it's the bad one.  Let me know if there are any particular measurements you'd like.

Edit: Embedded the images and added final comment.
« Last Edit: October 12, 2019, 04:54:20 pm by orin »
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #64 on: October 11, 2019, 02:12:22 am »
Would you please edit the post to describe the plots in more detail?  I'm a bit unsure which trace is which.

Here are my results using the 8560A.  I forgot to take pictures of the logmagnitude response of the cables before normalization. 

The thru cal using  N-M/N-M & N-M/N-F cables:

852132-0

852136-1

The $16.49 Chinese APC-7/N-F & Wiltron APC-7/N-M inserted:

852140-2

852144-3

All of the 4 Chinese APC-7/N-F adapters I've bought produce comparable results.

As this is an insertion test normalized to the ends of the adapter stack, it seems to me a pretty good test.  I need to buy a name brand  N-M/N-F adapter for a comparison. But this looks pretty good to me.  Please bear in mind, I'm a complete novice at this stuff.  While there is a lot I do know, what I don't know is vastly larger.  It's infinite.

I spent the day trying to impose something resembling order around here.  A long way to go though.  I ordered 72 latch top bins today to help.  It will probably be a few days before I set up the 11801 & SD-24.  I suspect that will show that the poor finish on the APC-7 side of the Chinese adapter is the culprit for the ringing.  In which case I'll try smoothing that face with some 1200 grit wet or dry sandpaper one one adapter  and see what happens.

Have Fun!
Reg
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #65 on: October 11, 2019, 06:27:38 am »
I linked to the images inline.

Basically, I put a good load at the end of the adapter stack and measured S11.  It was fine for the adapters that came with the HP cal kit, but pretty poor IMO for the Chinese adapters.  I also tried a single Chinese adapter with a not quite as good APC7 load.  It wasn't great, but not as bad as two of the adapters stacked.

Orin.
 

Offline rastro

  • Frequent Contributor
  • **
  • Posts: 388
  • Country: 00
Re: Testing RF connectors and cables
« Reply #66 on: October 11, 2019, 01:52:21 pm »
Sadly, I missed a Maury A-027A set on ebay because I set the clock wrong.  I'm planning to make a set of GO/NO-GO gauges myself.  But the 7" x 14" Chinese mini-lathe I bought to do the work is just a semi-finished  kit of parts assembled at the factory to make sure they didn't leave anything out.  Serious work required to finish it.

A big problem with the 3.5 mm compatible connectors is the tolerances required.
That sounds like a good project.  I have an HP 3.5mm indicator/gauges but I would like to make something also for "N" and SMA connectors.

I've used my 3.5mm gauges to measure some SMA adapters/cables purchased from eBay and the results where concerning.  Quite a few male center pins extended too far out beyond specification.  This can clearly damage a quality 3.5mm connector and maybe even another mating SMA connector.  Until you have an indicator/gauge you're kinda flying blind.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #67 on: October 11, 2019, 05:37:29 pm »
@orin

There's something wrong with the plots.  20 dB is a large reflection change.  Other than theory, all I know about VNAs is a couple of hours playing with a nanoVNA.  But none of the curves look good to me.

S21 is simply incident minus S11 and I'm not seeing the level of difference you are.  Even without normalization I'm only seeing ~1 dB variations in S21 through the N-M/N-M & N-F/N-M cable connection.  And for the 4 cases I tested the mismatch at the cable connectors was 6-8x the mismatch of the Wiltron & Chinese N-M/N-F stack when I inserted them.

What do you have available to test with?  Can you assemble APC-7 loads from the various adapters and loads you have?  If so, normalizing on one of those and then substituting the others should provide a good picture of the variations among the APC-7 adapters.  Complicated of course, by using different loads.

What does S11 look like with each of the Chinese APC-7/N-F adapters and the same HPAK load?

If you normalize S11 from the HPAK APC-7 load and then substitute the Chinese APC-7 & an HPAK load that should give a reasonable good picture of how different they are.

Reg
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #68 on: October 12, 2019, 12:42:15 am »

S21 is simply incident minus S11 and I'm not seeing the level of difference you are.  Even without normalization I'm only seeing ~1 dB variations in S21 through the N-M/N-M & N-F/N-M cable connection.  And for the 4 cases I tested the mismatch at the cable connectors was 6-8x the mismatch of the Wiltron & Chinese N-M/N-F stack when I inserted them.


Actually, 1dB variation in S21 is quite large according to the rule of thumb here:  https://www.edn.com/electronics-blogs/all-aboard-/4430355/How-much-return-loss-is-too-much--Rule-of-Thumb--12

Quote

What do you have available to test with?  Can you assemble APC-7 loads from the various adapters and loads you have?  If so, normalizing on one of those and then substituting the others should provide a good picture of the variations among the APC-7 adapters.  Complicated of course, by using different loads.

What I don't have is an APC7 Open or Short, so I can't calibrate to an APC7.  I suppose I could measure the APC7 load with the HP adapter... but I ran out of time today.

Quote

What does S11 look like with each of the Chinese APC-7/N-F adapters and the same HPAK load?

If you normalize S11 from the HPAK APC-7 load and then substitute the Chinese APC-7 & an HPAK load that should give a reasonable good picture of how different they are.

Reg


I spent the time to measure S21...

Upper trace is with two back to back HP adapters, lower trace with two back to back Chinese adapters.  Note the scale is 0.1 dB/div.

Orin.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #69 on: October 12, 2019, 03:45:19 am »
That's a 1% difference in S21 which is quite impressive.  There is no reason I'm aware of to expect random connectors and cables to match more closely outside of a metrology lab which routinely tests and discards stuff when it gets worn.  Much of which now gets sold on ebay as "used". Most of my gear is all 30 years old.  I'm quite sure that there have been more than 500 N connections made to my 8560A ports in 30 years.  This is the reason for doing an  insertion test after normalization if at all possible.

When you get some time, please look into your S11 measurement results.  S21 should be the difference of incident and S11.  This is just conservation of energy.  If I understand the plots, that doesn't appear to be the case, so it may be the directional bridge is damaged or else something was set improperly.  The wonderful thing about embedding a computer in a measurement system is it can now do things on its own without informing the user.

Have Fun!
Reg
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #70 on: October 15, 2019, 06:05:29 am »

When you get some time, please look into your S11 measurement results.  S21 should be the difference of incident and S11.  This is just conservation of energy.  If I understand the plots, that doesn't appear to be the case, so it may be the directional bridge is damaged or else something was set improperly.  The wonderful thing about embedding a computer in a measurement system is it can now do things on its own without informing the user.

Have Fun!
Reg


I'll check the 85046 and redo the measurement, but I don't really see anything wrong... the 909C I measured came in at 45dB return loss and it's spec'd at 1.01 SWR to 2GHz which would be about 46dB return loss.  The HP adapters are spec'd at >= 30dB return loss, DC - 6GHz and my measurement is better than that.

The 18 or 19dB return loss of the Chinese adapters I measured is much better than the 0.3dB S21 measurement - if there is no insertion loss, I'd make that a return loss of 12dB: 10 * log(1 - 10**(-0.3/10))

...so I checked the bridges on the 85046A and they are better than 30dB directivity (I don't have an APC7 short so I have to measure it through an APC7-N adapter meaning the true directivity is better than what I see).  Calibration can easily handle 30db directivity as that is the spec at the 85046A's limits.

I did the same measurement using port 2/S22 and got similar results.  I also have a coupler from an 85047A where I can connect the coupled port to the 85046A port 2 and look at S21.  The results were similar, but noisier.

FWIW, I also tried a couple of N F-F adapters in place of the APC7-APC7 adapters.  One was even worse than the back to back Chinese adapters at 3GHz!  The other was 40-odd dB return loss - about the same as the HP APC7 adapters back to back.

 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #71 on: October 15, 2019, 08:44:50 pm »
What I don't yet understand is the magnitude of the changes in response with frequency.  That seems excessive to me.

When I started doing TDR on my RF connectors I discovered that I had several Pomona 6 dB BNC attenuators that were bad.  I've got all the bad parts I found segregated.  Eventually I'll apply red paint so I can safely use them for demonstrating various problems.  But I found enough bad name brand used stuff to not feel so confident about the difference between those and new Chinese, provided they are tested.

Thanks for contributing to the thread.  I'm in entropy reduction mode.  I bought 36x 15x11x3.5" and 36x 15x11x16" Sterilite containers with latching lids and am about to order a bunch of the half size containers in both depths.  I haven't gotten to putting labels on the tubs, but will eventually.  At the moment I'm emptying stuff out of the shipping boxes into the tubs.

Have Fun!
Reg
 

Offline orin

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Testing RF connectors and cables
« Reply #72 on: October 15, 2019, 10:12:37 pm »
What I don't yet understand is the magnitude of the changes in response with frequency.  That seems excessive to me.

Well, I looked at the F-F adapter that I found to be worse than the Chinese APC7s.  It's a Huber+Suhner.  I remembered Dr Kirkby had commented on them when searching for an inexpensive N F-F adapter: https://www.febo.com/pipermail/time-nuts/2013-March/074871.html

So I went digging for a datasheet.  This appears to be the one:

https://ecatalog.hubersuhner.com/product/E-Catalog/Radio-frequency/Adapters/31_N-50-0-2-133_NE

Mouser want nearly $28 for them!

Return loss spec'd at 15dB 2 to 8GHz, 30dB 0 to 2GHz.  So such variations in S11 aren't unheard of!

Orin.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #73 on: October 15, 2019, 10:58:33 pm »
Wow!   I'd love to put one on my 11801/SD-24 for a few minutes.  It would *not* look good.

I'm increasingly convinced that stopping at 3 GHz was a good idea.  Connectors and cables just get too expensive when you go above 3 GHz.  And propagation gets thoroughly weird.

I was going to put a bunch of time into the nanoVNA FW, but there are so many people doing so much there I decided to it's better to wait for the dust to settle.

I think I can contribute more by doing a comprehensive  tutorial on TDR and VNA testing of cables and connectors.  Seismic work involves switching between time and frequency so much I can generally look at one domain and draw a cartoon of the other domain.  But I was not taught that.  I just learned it. I *should* have been taught to do it.  It's not hard.

One can do much better than what Joel Dunsmore presents in his book and it does *not* require proprietary software.  Just a thorough knowledge of the Fourier transform and the 1D wave equation.

Have Fun!
Reg
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Testing RF connectors and cables
« Reply #74 on: October 24, 2019, 01:37:32 am »
More TDR porn.

This is the 11801/SD-24 with a good quality SMA-M/N-M cable.  This particular channel has a 23 ps rise time on the open APC-3.5 connector so it is able to test to about 15 GHz accounting for the pulser rise time and the input rise time.

From the bottom going up:

SMA-M/N-M cable only.  As can be seen the cable and the N-M connector are not a very good match above 12 GHz.  No real surprise there.

Cable and Astrolab N-F/SMA-F open.

Cable, Astrolab 16301 N-F/SMA-F adapter and a Radiall SMA-M terminator.

Chinese APC-7/N-F adapter attached to the cable with the APC-7 side open.  The reflection from the open APC-7 connector shows precisely the plane of contact.

A pair of the Chinese APC-7/N-F adapters back to back with the N-F open

The top trace is a pair of the Chinese APC-7/N-F adapters back to back with an Inmet 64671 Model 3070M terminator. 

The opens show the reference plane for each of the connections, though with a slight delay because of the 23 ps rise time.

The  capacitive discontinuity at the cable/connector interface is frequencies above 12 GHz.  So the  Chinese APC-7/N-F adapters I received are certainly acceptable up to 3 or 6  GHz.  Which is not to say the ones @orin received are.

I ordered an HP APC-7 short and open, a couple of HP 900A APC-7 loads and an APC-7 jumper. When those arrive I'll set up my 8753B/85046A and compare the frequency domain for these devices.

These plots show that the N interface is a bigger issue than anything else.

Have Fun!
Reg

Edit:  Added additional trace and rewrote the annotation.
« Last Edit: October 24, 2019, 03:05:12 am by rhb »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf