Rohde & Schwarz SMIQ06B Repair - 6GHz board Level Preset Error

[smgvbest]

After following DaJMasta's Repair log for his SMIQ06B I'm finally getting started on repairing mine.
for his repair log see here -> https://www.eevblog.com/forum/repair/rohde-schwarz-smiq06b-repair-log-6ghz-board-alc-error/50/

Like DaJMasta I got mine of eBay with a stated status of a Level Preset error but output is stable and with-in .5dBm of requested value.
I do not have a SA or Scope that can go to 6.4Ghz but I can go to 3 Ghz on my frequecy counter and if I put a divide by 4 between the SMIQ and my SA I can at least see the signal but can not use it for amplitude checks.   
I have a E4418A and can check power and in deed it's is in the 0.5dBm range and the Level Preset is the only error reported.

I produced two items to help troubleshoot this
I cleaned up and stitched the E6GHZ Service Manual and it is available here
https://drive.google.com/open?id=0B0n5ltUlE8YHZ1laejFma245V00 I hope

I also created a excel spreadsheet to log Test data which you can get here
https://drive.google.com/open?id=0B0n5ltUlE8YHbjlQUG4yNlJuNlU i hope

here's a screen shot of the spreadsheet.  the TP's are not in numeric order but in order of the signal path from left to right.
this helps make it clearer where the problem is located. (at least I think so)



My initial thought based on the test point data available in the diagnostic menu indicated a problem in the IF900 path somewhere around or before N4.

I connected up some test points as seen here and measured bias voltages


TP                    Value          Actual
N39                  4.9 ± 0.1     4.67
V91                  4.7 ± 0.1   6.28
N3                   4.8 ± 0.1     4.95
N4                   4.8 ± 0.1     4.92

V91 is suspect and I found one of the SHF-0186-K  (datasheet here https://drive.google.com/open?id=0B0n5ltUlE8YHZlZrWkRvQXV6NEE
I have the SHF-0186 on order and should be here this week

Also N39 is a bit low  4.8 to 5.0 is specified value in the service manual.   the service manual though is for a 01 version and I have a 02


So as I am writing this up I decided with all the test points wired up I'd run a full calibration the bloody thing calibrated 
Doing some quick tests in the 0-3.3Ghz range its within -0.16dBm on my un-calibrated E4418A
in the 3.3-6.4 range it is now off 2dBm.   

So thinking about this I am still going to replace the V91 as it's voltage is very high. and at that point I will also re-run Level Preset Cal + ALC Table Cal.

and see what changes if anything.

I do plan once this part is fixed to also tune the various voltages on the unit.

here's some various images I took while working on it

This is the area i'm focusing on.  the amp marked H1 is the V91



This shows N3/N4 which is after the mixer in the IF900 section.



View of top of the unit



View of bottom of the unit



Power reading at 3.301Ghz.  Soon as i go above this it goes about 2dBm off



close up of the hybrid output section



close up of the hybrid input amp



Pic to remember how the connections go to the E6GHZ board



Here's N4 - This is the last amp before the Test Point that was the first one out of tolerance



Just because it looks neat



This has all the inputs/output on the back, I want to move them to the front




Question?
The rigid coax for the main output.   appears long enough to be moved from the rear to the front panel.  but I would have to re-bend the coax to make it work.   will doing so mess anything up?


edit:  made a few updates and hopefully you can see the images now.
my next planned update is after i get the v91 replaced hopefully this week (waiting on part)

[DaJMasta]

Good to see!  Though for whatever reason, I'm not seeing the pictures.

It's my understanding that bending rigid coax is not necessarily the end of the world, but it should be avoided.  The idea is that bending it introduces small changes in its performance (though I imagine these are most significant at much higher frequencies), and while they had to bend it at the factory, they then characterized the response of the unit with those initial bends compensated for.  The risk being run is that you won't be able to calibrate out these later bends in the instrument - and in the case of the final output line, the SMIQ's self tests won't compensate for it either because it doesn't have a detector right on the output.

I got a small amount of bend in the two that directly attach to that front end board, but I think this is largely unavoidable.  I accidentally bent the long one with the ferrites going to the output by not being careful of what my fingertips were doing when lifting the instrument - but I gently bent it back to position and while the output was probably effected, I have not noticed any differences.  I'd imagine that big bends (like a 180) are more likely to kink and change the dimensions of the conductor more, being more likely to cause a problem.  While it's not a guarantee... maybe put the signal through a similar length low loss flexible or semirigid cable and see if the output level differs from what's coming out the back.  If there's no big difference, then maybe the bandwidth isn't so sensitive and maybe rebending the rigid (avoiding hard angles) is viable.

[smgvbest]

i tried from another computer and google is saying no permission so I'll try to figure that out
edit: hopefully images are fixed, sorry about that

What you posted is pretty much what I understood also on the rigid cables.

[smgvbest]

Replaced V91 and now the bias voltage is 4.65 which is good.   
recal'd it and 300Khz to 3.3Ghz is still in the <0.052 range but the >3.3Ghz range is off as much as 2.4dBm now.
Basically the IF900 section is still my focus.   going to go back and look at the other bias volts for everything upto N4 in the LO900 path

will update again once I have the change to wire up the test points.

V91 biad now looking better


6.4Ghz @ 0dBm not quite right


Dooh,  I only have a 8481D power probe,  good for 4.4Ghz as I recall.  so not sure how accurate that reading really was.

[technogeeky]

< pictures and stuff >
Question?
The rigid coax for the main output.   appears long enough to be moved from the rear to the front panel.  but I would have to re-bend the coax to make it work.   will doing so mess anything up?


edit:  made a few updates and hopefully you can see the images now.
my next planned update is after i get the v91 replaced hopefully this week (waiting on part)

You absolutely should not bend (and especially not re-bend) a rigid coax. Always substitute flexible coax (if necessary, tape the coax down in many places to keep it from bending and rotating) while troubleshooting.

Also, bends do not need to be extreme (like 90 degrees or near 180 degrees) to change the properties of the rigix coax. Even small bends (above 10 degrees or so) can introduce small ridges on the internal circumference of the cable which will change the properties significantly.

[smgvbest]

Just ran some checks now after replacing V91,  seen below.   
with this change N4 bias is now high @ 6.276Vdc so my next item will be to pull N4 and replace it.
hopefully tomorrow evening I can get to that point.  Once replaced I think I will run all the diags again and check them to reset my baseline

here's all the points I measured

N38   SNA486     4.9 ± 0.1   4.950 N39   SNA486        4.9 ± 0.1   4.994 V20   BFP450        4.4 ± 0.1   4.417 V87   BFP450        4.4 ± 0.1   4.416 V91   SHF-0186-K     4.7 ± 0.1   4.650 N3   SCA-4             4.8 ± 0.1   4.960 N4   SCA-4     4.8 ± 0.1   6.276

[DaJMasta]

Wow, with readings like that we may just have twin failures between the two units.  Wonder how common that is.


The pics work now and the macro shots are nice!

[smgvbest]

Maybe/Maybe not
So I replaced N3 and N4 and they are both slightly high now.
I replaced them with the MMG3012NT1

N38   4.9 ± 0.1   4.950 N39   4.9 ± 0.1   4.977 V20   4.4 ± 0.1   4.411 V87   4.4 ± 0.1   4.403 V91   4.7 ± 0.1   4.670 N3   4.8 ± 0.1   5.003 N4   4.8 ± 0.1   5.104

it's pretty level from 300Khz to 3.3Ghz on the output but is going from 0.7 - 2.3dBm in the 3.301 to 6.4 Ghz range so thinking something is still not right in it. 

Attached is a screen cap of my spreadsheet for reference

The values in the IF900 section have cleared up but now the "Level ahead of filter bank" TP2409 and "Level ahead of AM Mod." TP2410 are showing off some

which brings N25 and V92 into play so I  pulled the board and add a few more test points to measure
N25   MGA82563           3.4 ± 0.2     3.038  LOW
V92   SHF-0186-12GHZ   7.1 ± 0.1     7.147  OK

I really don't cherish the thought of replacing N25,   such a tiny part but is out of tolerance.
plus have to find some and digikey has them and of course ebay

[smgvbest]

Before replacing that one component I decided to check the LO4200-6O00 path out
Those amps are in bold below

Component No.        Device Type   J:c/mA      0/V   Actual N38   SNA486   4.9 ± 0.1   4.950 N39   SNA486   4.9 ± 0.1   4.977 V20   BFP450   4.4 ± 0.1   4.411 V87   BFP450   4.4 ± 0.1   4.403 V91   SHF-0186-K   4.7 ± 0.1   4.670 N3   SCA-4   4.8 ± 0.1   5.003 N4   SCA-4   4.8 ± 0.1   5.104 N15   SNA486   5.25 ± 0.1   4.103  << N17   SNA486   5 ± 0.1   4.075  << N14   MGA82563   2.8 ± 0.2   2.950  V90   SHF-0186-12GHZ   7.1 ± 0.1   7.050 N25   MGA82563   3.4 ± 0.2   3.038 V92   SHF-0186-12GHZ   7.1 ± 0.1   7.147

N15/N17 Both the SNA486 are about 1v low and look to have been replaced before.

This path feeds in to U9 which mixes the IF @ 900Mhz and the LO @ 4200-6000Mhz (talking like I know what I'm talking about  )
The IF path seems much better after replacing the N3/N4 there.

I also noted that when checking the BIAS on V90 i saw a ~1dBm loss on the output

I'm going to replace N15/N17 first and see if there is an improvement base on that.

[smgvbest]

OK,  so here's a really dumb moment.   My previous tests where without the shield in place and the budge wire attached.  I decided to give it a full test and removed all the wires. re-assemabled the board shields including all those screws and put it in.

with the shield in place and the previous repair done it passed all tests and calibrates.  no more level present error.
Also the output levels are now <0.6dBm in the 3.3  - 6.4 Ghz range.   it's not very linear though.  with a low of ~.25dBm and high of 0.6dBm which is likely my power sensor which is only rated for 4Ghz.   I'm trying to find a good deal on a E4412A sensor but no luck so far.

I think before I really look at part replacement I should attempt to tune the voltage references, bias and other levels first that are with in my equipments ability.

[smgvbest]

Finally got ahold of a better power meter (Agilent U2001H) and took some reading with it.
It's pretty flat upto 6.4Ghz varying around .20-.35dBm across the entire range of -40dBm to 16dBm so I don't see that I actually have any other repairs to be done.   

If I didn't mention it already I've also moved all the connectors back the front.   I used a semi-rigid cable for the main RF for now.  I did find a piece of hardline and just need to solder the ends on it for a permanent cable.

time to button it up and call it repaired.

[DaJMasta]

Grats!  Nice to have the connectors in front when it's on the bench too - anything special to mount them in front, or is it the same board/just panel mount connectors?

[smgvbest]

nothing special really.  just allot of work to reroute the BNC's from the back.  the main RF connector hardline cable needs to be replaced.  That was my question on bending them.  I thought I might be long enough to make it but would require multiple bends to do it. decided to not bend it and replace with RG401 for now.  it does have some RF suppression on the existing cable that is not on my replacement.   

Though I guess this means my E4418/8451 is out of cal. 

[BFX]

Great repair 
You remind me that I have to fix my old HP8559A 22GHz spectrum analyzer

[rastro]

Wow, with readings like that we may just have twin failures between the two units.  Wonder how common that is.

I think the N3 and N4 on this PCB are a weak point.  I had the same repair on an SMIQ6 a few years ago.  The board in my system had different MMICs for the N3/N4 and it looked like they where original to that PCB.  This suggests that RS may have noted problems with this component. 

Also I gambled on an untested 6GHZEXT on ebay which also turned out to have an issue with the N3.  I also found an old repair log from a German blog which had focused on the the N3/N4 as a culprit. 

BTW I swapped EXT PCB's between an SMIQ4 and an SMIQ6 - they seem to be to be interchangeable.

-rastro

[smgvbest]

Well,  here's an update.
After working great for a while I am now getting the same ALC error DaJMasta had.
Freq is spot on and very stable.  but startying at the 3.3Ghz range i get from -2db to +2db depending on which band it's in.
i'm going to re-run my speadsheet to see what's changed if anything from the last run.   it's still usable but no longer repaired 

[rastro]

Well,  here's an update.
After working great for a while I am now getting the same ALC error DaJMasta had.
Freq is spot on and very stable.  but startying at the 3.3Ghz range i get from -2db to +2db depending on which band it's in.
i'm going to re-run my speadsheet to see what's changed if anything from the last run.   it's still usable but no longer repaired 

When you replaced the MMIC did you change the bias resistor to accommodate the substitute part?  The bias resistor(s) are located on the other side of the PCB from the MMIC's.  I changed the resistor values to accommodate the substituted replacement amp.  The original bias resistors (SMD) where double stacked on my 6GZEXT PCB - Parallel aided.

-rastro

[smgvbest]

Hi,

No I did not change the bias resistor.   how did you determine the correct value for the replacement bias resistor?

[DaJMasta]

Well, I guess I hadn't used it as much, but mine seems to be exhibiting the same error now - smaller variation from what is set than before, but still a failure on the ALC.

Looking into it a bit though... according to the schematic, the bias resistors shouldn't need to be changed.  The replacement MMG3012NT1 is rated at 82mA nominal supply current and has an absolute max of 300mA, while the schematic indicates an 80mA supply at both amps... I could be missing something, but voltage and current should be correct in that instance.

Clearly it's not working, but maybe the schematic is different from the built board (not the first time) and maybe it's something else.  On the schematic,  the resistors on the bias for both N3 and N4 total 30 ohms, which at 80mA should be a 2.4V drop from the 7.5V supply, plus whatever resistance there is from the inductors... which seems to be right in line with the indicator and our testing of those points.

I've got a new instrument open for a fix at the moment, but looking at the test points from the screen, TP2409 is only 10-40mV down from my original recorded results, but TP2010 is 40-270mV down.  While the prime suspects may still be N3 and N4, it may not be happening at those... hard to tell.  I'll be going back to give it a look eventually, but at my usual working speed, it could be a couple months.

[smgvbest]

[quote author=DaJMasta link=topic=90644.msg1294184#msg1294184 date=1504501507

I've got a new instrument open for a fix at the moment, but looking at the test points from the screen, TP2409 is only 10-40mV down from my original recorded results, but TP2010 is 40-270mV down.  While the prime suspects may still be N3 and N4, it may not be happening at those... hard to tell.  I'll be going back to give it a look eventually, but at my usual working speed, it could be a couple months.
[/quote]

Sorry to hear you're experiencing the same.
But I have to admit I got a laugh out of your 'usual working speed' comment.   That is so me also.
I finished the repair of the SRS PS350 and have been tracking down a undocumented circuit in it.  to no avail so far.   loose it under a resistor divider.

I'll get back to SMIQ hopefully soon.
do wish someone would out the service kit up on ebay. sure would be easier LOL to troubleshoot.

[rastro]

Hi,

No I did not change the bias resistor.   how did you determine the correct value for the replacement bias resistor?

DaJMasta  is probably correct.  This may not need a different bias resistor based on the MMG3012NT1 datasheet I downloaded.  Table-4 shows a typical current of 70mA with a maximum of 82mA.  Then by using figure-6 it seems you can derive the collector current based on the collector voltage.  So 5.0V and 5.1V collector voltage is in the 60-70mA range according to the graph.

The part I used was different and did require different bias resistors.  To determine the bias resistor I used the datasheet to find the typical bias current.  After replacing the MMIC I removed the original bias resistor(s) and hooked up a DMM (DC amp setting) in series with a 500-ohm multi-turn pot  (using only 1 leg and sweep arm).  I adjusted the pot to reach the appropriate current.  I took the test rig out off the PCB and measured the resistance of the pot to determine the correct bias resistor value. 

-rastro

Also thanks for posting the additional documentation excel and schematic cleanup.

[smgvbest]

Happy to give back to the community.

ok so it was the brute force technique.  i was wondering if there was a magic method 
I'm not sure when I'll get back to it.  The output power is stable once selected.  it's just off by varying amounts.  if I really need a exact power level I can hook it up the power meter and get it dialed in so not dead,  just walking with a limp

[charlyd]

Hi a lot of thumbs up for DaJMasta, Smgvbest & Rastro  for doing all the investigation.   

as i followed both topic(s) and took my 6Ghz apart i found out the different components are used in my SMIQ06B

i wanted to share these pics with you all figuring out where the problem lay in the output level

[smgvbest]

Hi a lot of thumbs up for DaJMasta, Smgvbest & Rastro  for doing all the investigation.   

as i followed both topic(s) and took my 6Ghz apart i found out the different components are used in my SMIQ06B

i wanted to share these pics with you all figuring out where the problem lay in the output level

Took me a minute to figure out you zoomed in on the photo to show N3/N4
those were replaced and not sure with the correct components

N3/N4 are DC-3Ghz MMIC Applifiers and if substituted they may need to BIAS network redone to work correctly
I replaced mine with MMG3012NT1 (thanks DaJMasta)

[smgvbest]

by the way,  the evil ebay has one seller with them
eBay auction: #https://www.ebay.com/itm/132863299750?_trkparms=ispr%3D1&hash=item1eef4510a6:g:PKUAAOSwXHxb9QZA&amdata=enc%3AAQAGAAACkPYe5NmHp%252B2JMhMi7yxGiTJkPrKr5t53CooMSQt2orsSHYXPhGXR5uguexJBeHwfgU%252FMAp%252BmZYsUr3k34eAw%252BVmA2J4dFbFW47fHDX2MRUlN52XzvUv9Ss9gy%252BCxXorxZGAzwOhnsK5xpcXT9QUB%252BPCQGYe4auJblKNGu8kgx3Y3Au4zo53hMlY2SAHFassn9b%252FOg8cOxgFc6ls%252Fu7TE288vpuYQMR4vBp7fK4zfArN1yGAw3BdHhk97Db4QAMyDK805HqAgXnyWjwVSOnfsl1JePkdq0SixHuOLtkSBVKhpOPKnRhtMRxIGHeC4aGKskUoUAnfE95vA6zNq16PyQKmUED47tvCAr%252BoYj0HZIEWnYi6jFsmfwsNaOb%252FEcWLZJjRcECFV2AiFx4v83hwJxsu416iHooxLQDvdsRBpBcJh8%252Fl4DUImawWl8ywf7WnRKyIoCRWDd2tn7nR0O3fICz%252FDDJuB16n%252B4wmek6N7MeQsME6fMlZVH9w0LpenapuSkSe1iaucIfGldpdJWtq%252FUHSosy02lAA02cXfp3gctrvHnTxdoiIeGmBClrtNHurTUFdvHqvftnv1wFa%252BQ48JStQIsTVMECBQtM6eX3wVReULuz0s2rFsOVixNYZRB6vsdtuuW8A8hX3Vsi%252F9WnQptByIGnAJK11esfVXJwwryvIV27IJaleWMZ5E86%252FpZzpWfbzVU2fuKk9xdrnty6g59%252BIrHOE%252BwjpBFyyn3zCA2Z%252FtHcZmIbBcP0wjcx%252FPZmg1XRGNFIwGHo5IZEsJxgnXfy7ZWSWiI2QzqVKeVnOcZ5s%252FtZGmmBSbrKATJ10lNLZB9IP2YxeWJDSJ7KAfucIZjSrwd9UAUFPE80SHnpdDclzs%7Cclp%3A2334524%7Ctkp%3ABFBMgu2FjuNf

but very high cost
you could make an offer and see what they'll take

Links seems to not be working
just search for the part number,  there's 1 hit for it