Rohde & Schwarz - IQ Modulator A240 (1084.9800.02) - Level Problem

[rg58]

Having investigated this issue for several days now, I am stuck with this problem. I only have the incomplete service manual for 1084.9800.08 here, containing part list with X/Y coordinates which helped me to identify parts. But it doesn't contain any schematics.

The problem is with the output level only doing the 5dB steps from the output attenuator of the unit, no levels in between. The output level is also around 6-10dB too low. The self calibration fails (getting aborted). The signal on output socket X249 is constant, not following the level setting of the unit. It should change in a 4 or 5dB range (measured +8.5...12.5dBm in a known-good unit).

Among a lot of other things, I investigated N35 - AD744 (level control loop). At first I thought it was the culprit. Inverting OpAmp, in CW mode -input (pin 2) -0.31...-0.69V (following level setting), output (pin6) sitting at +14V constantly. Strangely, power supply for this OpAmp is +14.8V and -5V which I know is correct.

Removed this OpAmp (N35/AD744) and put it in a SOIC-8 socket of my programmer, plugged it onto a breadboard and wired it up with a potentiometer. It worked fine! With N35 desoldered, the output pad (pin 6) was now sitting at 0.156V (IQ mode) and 0.58V (CW mode). So where the heck did those +14V come from on its output? It must be related to the feedback loop or other circuitry that I am not aware of (without schematics, likely a 4-layer PCB). My other working 1084.9800 has +0.64...1.01V on N35 output (pin 6) to control the output on X249, depending on the level setting - not constant 14V. But here I am stuck now 

Previously I looked at the AM modulator ('O') part but wasn't able to identify a faulty component, although the problem might be here in this section I believe. The 3 double PIN diodes work at least.

Today I looked at the output stage 'R' again. As this is a loop for the ALC, it might as well be a problem here with too low gain. The 3 FETs (SHF-0186K = 'H1') are self-conductive and 'might' be OK, so is a diode which in-circuit shows 0.3V bidirectional. The 3 power transistors BSP31 and all other diodes have been tested and didn't show any obvious signs of damage. Not quite sure if I should touch the small potentiometer DETECTOR LINEARITY... it should be part of the ALC loop, measuring output power with the detector/coupler.

Besides, it's quite a hassle working with this module which is inaccessible during operation, without the proper test jig. So I keep removing all cables, pulling it out, soldering wires to my test points, plugging it back in, reconnecting all cables. Let alone the 47 screws to open the "B" side...

One more weird thing to note is that date of first calibration failure (showing up on screen) is the SAME DATE that R&S did the last calibration (written on the label at the front). How strange! Looks as if R&S messed something up!?

I attached photos with labels, it might also help others in the future. Bottom right on B side is output SMA socket X249 (FIQOUT), left of it input SMA socket X247 (FIQFIL).

Does anyone know more about this and could possibly help me out here?

[Harry_22]

Hi! I can't see what this point on the body is.

[rg58]

This is just some resin from the soldering, just removed it. This double diode (BAR64-04 "PPs") is testing OK.

Attached the result from going over all 24 internal test points at address 21, accessible via the service menu. All values at 1 GHz / -30dBm in IQ mode. Only value 2112 (D_SWITCH) is out of range, sitting at 0.000V, so 5mV below the minimum voltage (0.005-0.400). Tried with higher level -6dBm but no change. The meaning of this test point is described as "Level subsequent to AM modulator".

Not sure what to do with this info and where to go from here.

[Harry_22]

I found some information regarding SMIQ modulator but I need some time to understand the scheme.

[rg58]

Thanks for your effort, Harry! Take your time.

I also just found the schematics, finally! Yes, will indeed take some time to get through it. The block diagram alone is scattered over 4 pages. Everything is huge, therefore scattered over many pages.

[Harry_22]

There are different schematics. I found broken modules .08 and .12. They are different from yours .02.
Step adjustment may be due to the fact that the level does not reach the true level. Therefore ALC reaches its maximum.

If you will need help you are welcome!

[rg58]

Thanks! Nice photo.

I'm having the complete signal path in front of me now, printed on 5 pages, side by side. Things start to make sense. I took notes and labeled components on the block diagram.

I checked 3 more voltages during operation, nothing suspicious. The 2112 value is depended on the mode. In CW mode it shows 0.188 which is in the range now. I can tell that N38 (MMIC) draws 30mA of current which could be right, too. I see 5V on it's output (Vcc). The 2112 value (D_SWITCH) I also checked with voltmeter and is close to the one shown in the service menu. Furthermore I checked (T)P55 which only seem to give me 3 different values: -6.2V, -5.9V, -0.78V when moving through the level setting in 1dB steps. I would have expected this to be linear as this signal also includes the V_AMOD signal from N35. Need to investigate further and compare with my working one.

Step adjustment may be due to the fact that the level does not reach the true level. Therefore ALC reaches its maximum.

Right, this is what I basically said earlier.

As this is a loop for the ALC, it might as well be a problem here with too low gain.

Will continue tomorrow.

[Harry_22]

The following adapter can greatly simplify the module measurements.

[rg58]

Yes, it probably would. But there are 6 more coax cables that connect to it which would need to be extended as well, not all of them might be required though. I am also not sure if the (open) module could be placed in another slot towards the back which would make it more accessible from the front. All slots have stickers with the module name/number that goes into them. In case this is a proper bus it would work, but not knowing for sure, I don't wanna risk it and rather stick to my cumbersome, troublesome troubleshooting

Had not a lot of time today, but did take more measurements and reading through the circuit descriptions. The "level preset" is controlled by a constant voltage coming from a DAC (level setting independent). At first I though it was a fault, but if I understand the manual correctly, this is only attenuating the AM input signal (if any). There's a lot of things going on in this masterpiece. It's also interesting to see that they use those pin diodes for a controlled attenuation (as it seems), but maybe I am wrong? I only knew they are used for RF switching.

Will need to investigate more, now that I have the schematics.

Attached photo of today's session of some serious troubleshooting...

[Harry_22]

Hi! What do you think about testing IQ Modulator separately? You will need at least 4 voltage sources, generator and power meter/spectrum analyzer.

[rg58]

Problem found. V103 (SHF-0186K) in the output stage is blown. Already ordered the part but will take about 2 weeks to arrive.

So our early assumption was correct that we have too low gain, hence ALC is turning N35 fully up to +14V on its output pin 6.

I did check this transistor several times, but as D-S already conducts if Vgs=0 V it's really hard to find when the board is sitting on the table. Unless you apply external voltage to it.

Rds at Vgs = 0V (in-circuit measurement)
V144 - 8.6 ohm
V103 - 6.8 ohm (faulty)
V104 - 12.9 ohm

Of course I found out when checking the voltage on the Drain which was only 0.89V instead of 7.0V, again by soldering wires to it (as well as to the other 2 FETs) and plugging it back in the machine.

I should have checked that Vds earlier, but that N35 thing brought me off track. Not to forget that this is a loop, making things harder. Plus, lots of other signals added along the path, summed and/or switched. There are many ways to get lost during troubleshooting.

Thanks Harry for your help!

[Harry_22]

rg58 my congratulations!
Finding a fault in a looped circuit is very difficult. Everything affects everything in the circle.
Great!

[rg58]

Exactly!

Guess what? Someone DID mess up! I wouldn't be surprised if it was Rohde & Schwarz themselves during their last calibration.

Remember I said that the date of calibration handwritten on the calibration sticker on the front plate is equal to the date when the error occurred for the first time, as stored in the machine?

Apparently they took off the heat sink from the output stage. There is some heat-conductive black rubber pad underneath, but only covering one third of it!! The 2 FETs were not touching the heat sink and couldn't dissipate their heat. In case of the blown V103 this was 840mW! In this tiny package! Seems these were 2 pads and the larger one fell off. Only V104 made contact to the heat sink through that pad, dissipating 1.2 W of heat.

[rg58]

I must correct myself. The pad probably shifted down because of the heat as it overlaps the heat sink. The length can cover the 2 last FETs with the highest power.

But that still tells me that the pad didn't make contact, otherwise it can never shift like that if it would have been pressing against the 2 transistors. I have tried to move, shift and remove it, but it broke off. Will replace it with something better.

[Harry_22]

When disassembling the module I also discovered an aluminum plate that hung in the air above HFET. The gap was 2.6mm. The radiator was coated with heat-conducting paste on top which clearly indicated its purpose. The other module had no radiators at all.
It should be noted that the heat was supposed to be removed through the Source terminals. To do this HFET was placed close to the bulkhead on which the radiator was attached.
R&S does not repair modules immediately but replaces them by new or refurbished. Maybe your module is one of these. It shows smudges from the flux being washed off.
What about your second working module?

[rg58]

This one doesn't have a heat sink. It looks like the one on the right in your photo.
Yes, you can tell that some rework/modifications have been done here. It might be a refurbished one, who knows...

Makes sense that the heat will also be dissipated through the source terminals.

[Harry_22]

It's also interesting to see that they use those pin diodes for a controlled attenuation (as it seems), but maybe I am wrong?

I found following:

"A PIN diode is a semiconductor device that operates as a variable resistor at RF and microwave frequencies. The resistance value of the PIN diode is determined only by the forward biased dc current. In switch and attenuator applications, the PIN diode should ideally control the RF signal level without introducing distortion which might change the shape of the RF signal. An important additional feature of the PIN diode is its ability to control large RF signals while using much smaller levels of dc excitation."

[Astralix]

Hi!

This seems to be a revision 6 board, as all earlier boards did not have any heat-sink screwed in.
The early boards all had the chamber shielding leading over the SHF output driver and massive ground areas with many through-holes into the other layers.

That made changing the output driver nearly impossible without damaging the PCB and surrounding parts.
In a later revision, that bridge from one side to the other of the casing was removed. There was also a change in the BIAS settings, probably lowering the BIAS voltage a bit.
And they used SHF-0286K instead of SHF-0186K sometimes.

But as we have not yet tested and repaired enough boards, we cannot say for sure, what to change and set to make the thing more reliable.
SHF-0186K is available from several sources as "used", sometimes "NOS". SHF-0286K is not available.

Wonder if there is a good hint how to remove the bridging shield on the old units and improve cooling after a repair without interfering with the electrical parameters. Bringing a metal plate in may influence the signal by pure existence... The Revision 6 PCBs have a clearly different layout compared to the older units.

[rg58]

Thanks Harry! Just learned something new, something I should have known actually. They are doing the AM (amplitude modulation) and additionally the output level attenuation in a range of 0...-5dB with them.

I ordered the SHF-0186K from China. Let's see how it performs (if it's original?).

About the thermal pad, it could probably introduce some attenuation at higher frequencies, sitting right on top of the tiny FET packages (=>capacitance). But then again, the unit does a self-calibration and should even this out I believe.

Will try the get an ordinary, sticky, white thermal pad with a thickness slightly bigger than the old one which is 1.1mm.

On the bottom of the board's A side, besides the socket, there's a circle with a "97" inside and on the other side "-04-". Is this the revision number you are talking about? It suggests 04/1997.

[rg58]

The other board is part number 1084.9800 and has "00" (as in 2000) and "-10.01-" on it - as if this was a "later" revision. Although the part number suggests this to be the FIRST version. Confused 

[rg58]

Could it be possible that this rubber pad is not even a thermal pad, but only used as a spacer?

You might be right that they ONLY use the source terminals for heat dissipation. From the FETs package, S-terminals, on to the bulkhead, then to the radiator (thermal paste), then to the upper lid (thermal paste again). There was no thermal paste between the pad and the FETs.

[rg58]

Alright. The other board (1084.9800 printed on the PCB) turns out to be 1084.9800.10 - a higher revision. So at some point R&S decided not to print the revision number on the board anymore, rather putting a sticker on it. I overlooked this on top of the metal. It also has another tiny sticker "MGA1" on it, whereas the older faulty one (.02) has "MGA5" written on it.

Also I forgot to mention that right after finding Vds is too low, of course next I checked was the supply voltages to the gate and drain - which were OK.

[wriely]

I was just looking at my SMIQ03B and wondering if it may be possible to get probes on the IQCON board without an extension card and extending the myriad of RF cables.

How?  Well, the front panel easily comes off of the unit.  Behind it is a metal bracket that is secured by what I count to be 8 screws.  With that out of the way, one could have direct access to the front side of the IQCON board with ease.

I've not gone any further than taking the front panel off of my unit so I can't say for certain whether this work, but it could be a major time saver if it does.

Walt