Anritsu MG3633A Repair #3

Who's blown the fuse?

Slow start

Nothing much done yet, just a short diagnosis:

Just a bit of an update:
Remember the signal gen with the blasted +15V supply? I did a teardown of this unit (some pictures will be available soon), appears the synth unit (one of the inner boxes we didn't remove) is blown, there's quite a few ICs supplied from this rail, I found some of them released the smoke, and for sure, there will be more defective ones. The Processor board appears also defective, since it shows only garbage on the displays and does not start. So I'll try to scavenge parts from this one.

About the logic board, I was actually scratching my head and wondering what has +15V to do with the digital logic, the device is old, but not THAT old. It could be that there are some loads overloading the power supply, IMHO if they've used OPTICAL CONTROL of the modules, would have been very strange to mix the power supplies.

Have you now fully repaired the power supply, is it in a stable condition ? I was hoping that some "kamikaze" chips decided to shorten themselves and protecting the others, usually this is the case, and once removing them some modules will spring to life.

Here's the solution for your head scratching:
This 15V supply is used as the reference voltage most of the other output voltages. So if this voltage is too high, all other voltages will be too high also, go figure what might have happened to all the boards. A simple 15V TVS (or even a zener) might have saved the unit a lot of damage, but there is no such thing inside. Looks like we have a lot of blasted 15 operated OpAmps instead spread around all boards, as this particular +15V is quite powerful (it has a 3.15 Amp fuse), they had no chance to short out and protect others. Within the circuits (especially functional safety) that I'm developing today, this would have been considered a major design fail. And there's also some analog (+15V powered) stuff on the logic board. The optical control is used for the synthesizer part only, it's basically a 384 bit shift register with an SPI alike control.

No, I didn't repair anything yet with that particular unit, I've just used a lab supply for the missing +15V. It works only if the inner modules (Synth and Amp) are disconnected.

One good thing: this unit is still complete, no parts scavenged to repair the other two (except for a borrowed and returned oscillator hybrid).

Due to the vast damage, and some of the blast IC's are unobtanium (or at least obsolete) now, this unit is set aside for now.
magic smoke escaped magic smoke escaped too

More diagnostics

So this unit was waiting for some purpose. A defective signal generator doesn't have much of that, and I didn't need to salvage parts, so it's complete. Couldn't see that anymore, I've started some more attempts to find out what is damaged and if it'd be possible to repair.

Power Supply

Quite obviously, I found a shorted power transistor and a handful of blown fuses within the power supply.
power supply - regulator board transistor

Replacing them brought the power supply back to life. I disconnected the output connectors just to be sure, and then checked all the testpoints for the correct voltages: Everything turned out fine, all voltages within their expected range.

For the records, this is the block diagram of the power supply:
power supply block diagram

Look at the leftmost pass transistor, with its emitter tied to +15V-A: This one had collector to emitter shorted, applying the full raw voltage to this +15V supply rail. No wonder I've found some IC's releasing the magic smoke, since there's no means of protecting them from overvoltage. With this overvoltage on the +15V rail, current consumption raises significantly, leading finally to blow the related fuse. But blowing a fuse takes its time, so the IC's had enough time available to die. Even worse, this +15V rail is used as the reference voltage to most other regulators, so all the other supply voltages also were on overvoltage until finally quite a few of the fuses were blown.

This is the shorted transistor Q1, located on the power board:
power supply

Testing the modules

Expecting the worst due to this overvoltage event, I slowly started restoring power to all the boards. All of the "outer" modules (CPU, Modulator, Front Panel, Reference Oscillator) came back to life. I was able to operate the front panel, and observed a reasonable signal at the modulation output. So I supposed these modules to be OK for now.

Synthesizer module

Connecting the Synthesizer Module blew a handful of fuses again, most of the power supply rails were dead now. So I disconnected this module, replaced the fuses and checked again: No obvious damage.

RF output module

So I left the Synthesizer Module disconnected and tried the RF Output module. OK, this one doesn't blow the fuses, and no magic smoke escapes. To achieve more diagnostic, I used a signal generator to feed a 640MHz signal into the RF modules input: No output signal. I've tried a variety of frequency settings and found the RF module working and outputting a reasonable singnal in the 1280 MHz to 2.7GHz range, but no output in the 10kHz to 1279.9999MHz range. So this isn't a complete fail here.

Time to examine the block diagram of the RF module:
RF module

The amplifier is splitted into two paths, one less than 1280MHz (lower range) and one greater than 1280MHz (upper range). These paths use different output stages, but otherwise common components like the ALC loop.

To troubleshoot the module, I've set up a working signal generator and connected this module using an extension cable to it instead of its own RF module. test rig for RF module extension cable

I started checking the input, diverting the signal to the alternative paths and found everything fine here.
input diode switches

Next stop was the output amplifiers. Here's a view into a working RF module, I've taken this photo during the teardown. In the lower part of the picture is the upper range amplifier, the upper part shows the lower range amp. Note the rather large black inductors used to extend the frequency range to near DC here. Obviously the upper range amplifier was repaired before, one can see the bodged resistors and the FETs replaced. I forgot to take a picture of the damaged unit before I started to repair, but it looked quite similar with nothing been repaired before.
RF amplifier


low band amplifier

Measuring some DC levels within the low band amplifier, I found the transistors Q72 and Q73 were damaged. Desoldering base and collector, then diode testing the junctions revealed one transistor completely shorted and the other one failed open. So no more gain here, and even worse, these transistors are obsolete long time ago.

The manual provides me with a nice level diagram of the amplifiers:
RF amp levels
Showing each transistor has a gain of 10dB, and a final output power of +13dBm at a nominal 7dBm output. Since the signal generator allows for higher output levels, the amplifier must be able to provide more than 20dBm output, and operate from near DC (10kHz) to 1.28GHz.

To be able to check the RF module further, I bodged a pair of MMIC amplifiers into that output stage, bypassing the broken transistors. Looks ugly and is a rather low performer:
RF amplifier

This is the second attempt, for the first I've used one of the MMICs. The first one apparently didn't have enough gain and output power to replace the two-transistor amplifier stage.

At least it provides me with an output signal showing reasonable behaviour within the lower frequency range. Since the low band is split into some more paths (see above diagram), this is good news. All the frequency dividers and mixers for the low band appear to work, and the ALC also works within the limits of my bodged amplifier stage.

I've got now a RF module with a known defect, waiting for further repair, and a fuse-blowing synthesizer module. That's a partial success at least.

Synthesizer Module again

As I couldn't connect the broken synthesizer module to the signal generators own power supply (I ran out of fuses), I hooked up a bunch of wires to supply this module from lab power supplies:

wires hooked up to synth module lab power supplies

That was an easy job, since the necessary digital control signals from the CPU board are fed through plastic optical fibers. Now the module is isolated from the signal generator now and runs off the lab power supplies.

LF synth

For whatever reason, the designers of this module provided me with three independent +15V supply rails. The other power supply rails (+5V, -15V, +32V) are connected. So I could start the investigation by applying power to the module using only one of the +15V rails, I decided to start with the LF synth section.

Turning on the lab supply revealed 3 Amps on the +5V rail, I was somewhat concerned about this rather high current consumption, but this turned out to be normal for this module. +15V was at about 0.2A, and -15V was at about 0.3A, this looked pretty OK and no magic smoke appeared.

I started to search for faults at the 100MHz reference frequency input. The signal from the reference oscillator unit gets amplified and distributed to a bunch of frequency multipliers, band pass filters and diode switches. As these frequencies are essential to the operation of all the synthesizer loops, I considered this a good starting point.
Reference frequency distribution
I quickly found out the RF level all over the place beeing to low to operate the multipliers. None of the multipliers / filters worked. So I went back to the amplifier staged located directly at the 100MHz input.
Reference frequency amplifier

Reference frequency amplifier

The MMIC amplifiers Q120 and Q356 had some output level, but too low to operate the multiplier diode Q123. Measuring at C165 / R29 junction revealed a too low level signal here too. The level at this junction was quite the same as at the input connector (J1 / C160).

So Q61 doesn't provide gain, and the MMIC amplifiers were toast? These MMICs are long time obsolete, so it would be a hard time finding a suitable replacement.

So I checked Q61 in the first place: Dead as a Dodo, shorted base - collector. It's a quite beefy RF transistor, I found a possibly suitable replacement in my junk box, an BFW16A transistor. I bodged that one in, et voila: There is well sufficient RF level at its collector visible.

reference section replaced RF transistor

More testing showed the shorted transistor also loaded the input signal and reduced substantially the RF level to the MMIC amplifiers. Having their intended input level again, they did in turn provide sufficient output level. So no need to find and replace these, phew! There's quite a lot of them (and similar ones) in the synthesizer module.

Now that the reference frequency stages were working again, I checked one of the synthesizer loops (LOOP13 / 14 that I repaired in unit #2) for its output signal and found it OK. Looks like I've restored the operation of the LF synthesizer part.

HF synth

And now for something completely different: The HF synthesizer. This is where I found the blown LM6165 chips pictured above. These amplifiers were supplied from the +15V and -15V rail and blown due to the overvoltage event. As this amplifier was obsolete, I first started to find a suitable replacement. I've ended up with the AD829A that offered similar GBW and slew rate specifications.

LM6165 amplifier

As the HF synth has two independent +15V rails, I've started with one of these. Before desoldering the LM6165 chips, current consumption was too high and drove the lab supply into current limiting. Replacing the first three LM6165 with AD829 brought the current consumption back to a reasonable value, but the loops didn't operate yet. Checking the supply pins of the AD829's showed the +15V was missing at two of them, the third one was fine here. The Anritsu engineers did a good job in decoupling the supplies, each amplifier has its own filtered supply:

supply distribution

Easy job now, these small inductors (e.g. L312) were fried and failed open. Replacing them restored the first three HF synth loops to operation.

Rinse and repeat for the second +15V rail, replace two more LM6165 and inductors, and the whole HF synth came back to life again. Some more checking showed a clean output signal at the 640...1350 MHz output connector.

reference section

The synthesizer module is fully operational now.

Power supply again

After replacing the synth module into the signal generators belly, the output showed a not so nice behaviour as seen before. For some frequencies (near the lower end of the 640 ... 1350 MHz range), the signal had a quite impressive phase noise, looking just like the classic "dick and balls" - but really big balls on the spectrum analyzers display. For frequencies at the upper end, the signal was just smeared all across the place.

reference section

Temporarily attaching the lab power supplies to the synth module restored normal operation. So the fault is located somewhere within the signal generator itself, with the power beeing suspect in the first place.

Remember, I've checked all the supply rails after replacing the transistor? Yes, all the supplies that have a designated test point, and with no load attached. Everything just looked OK then.

So check again: Now there's some nasty ripple visible on all rails, with the largest amplitude on the +15V_A rail. Digging through the regulator circuitry revealed some interesting low frequency oscillations, but they were totally unrelated to the observed ripple. Another observation: The ripple completely disappeared when I disconnected all outputs, including the reference oscillator. Moreover, the ripple disappeared with the reference oscillator and fan disconnected while the rest was left connected. WTF?

Some closer looking revealed a second +15V rail (+15V_B), that is used to supply the fan and keep the crystal oven warm. This rail has its own regulator that isn't turned off when the unit is switched to standby. And it didn't have a test point, so I neither noticed nor checked it while repairing the supply. The oscilloscope quickly revealed this rail as the primary source of the ripple that was spread all over the place:

probing ripple on 15V_B rail ripple on 15V_B rail

Now that's what I'd call an impressive ripple amplitude.

Checking the schematics again:

power supply

Remember this diagram? Yes, that's the place where Q1 was replaced. Notice Q8 and Q6, forming the second +15V regulator, this is the origin of the ripple. Quite obvious now, I replaced C5 and everything was fine again. Still having my bodged MMIC amplifier in its guts, the signal generator now showed a nice and stable signal at its main output.

Back to the RF module

As my first and second attempt to replace the toasted RF transistors with common MMICs (some SCA-14 I had as leftovers from another repair) didn't give a satisfactory result, I've started to investigate how to rebuild a somewhat more suitable RF amplifier.

From the above mentioned level diagram I derived the requirement of 28dB gain in case I replace the whole amplifier cascade (one MAR4 MMIC and two NEL2301 RF transistors). The output power level should achieve at least +15dBm to ensure +7dBm output power (this is the nominal output level into the attenuator), somewhat more would give headroom for higher output level settings. The maximum power of the original amplifier isn't specified, my guess would be in the +27dBm ballpark.

Finding a MMIC provinding +27dBm is kind of difficult, I've found the Quorvo ECG003B that is specified +24dBm P1dB. Using this one, I've build a prototype amplifier using an SCA-14 and an ECG003B together with some inter-stage attenuation.

Amplifier prototype 2 Amplifier prototype 1 Amplifier final

The pictures show the prototype using the SCA-14 and ECG003B, another less successful prototype built using some random MMICs from the useless samples box, and the amplifier that is to be placed into the RF module. Testing the prototype (first picture) showed promising results, so I built the a second one (third picture) that I intend to use as a replacement for the toast RF transistors and the (still working) MAR4.

Simply putting this amplifier into place, just where the RF transistors were located (and still are, since one cannot remove them without removing the whole PCB, which would be a royal PITA), turned out to be a naive idea. Yes, the amplifier itself did its job, but the ALC went mad on me. Setting an output amplitude of 10dBm (in continuous level mode) worked, lowering this value showed quite a non-linear behaviour ending up in -40dBm output at -3dBm setting. WTF again?

There's the RF power detector (Q82, see schematic way above) located near the output connector. It isn't shielded from the amplifiers in the original design. This works obviously, since the strip line PCB design keeps the RF to where it belongs to. So my new amplifier was constructed in an open frame dead bug style way, and didn't do a good job in keeping the RF in its place, especially because there was no easy way to merge this amplifiers ground plane with the PCBs ground plane.

The radiated RF found its way into the peak detector circuit, so the ALC was fooled leading to a way too low output level.

So do I have to put the amplifier into its own shielded box? Yes, that'd be the commonplace solution. Placing the amplifier outside the amplifier area and replacing the shield confirms this. But wait, building a box requires some effort, and where should I put this box into the RF module? I found no suitable space in the original amplifiers vicinity.

So I experimented moving the unshielded amplifier around and changed its orientation. Found this place a good place within the available area:

Amplifier placement Amplifier placement

So I installed all the shielding and screws, put the RF module back into its place. Checked the maximum achieveable output level over the 10kHz to 1.28GHz range: +10dBm over the whole band, but with a maybe 2dB dip around 1.1GHz. This dip might be related to the new amplifiers mismatch to the further signal path. The MMICs are matched for 50 Ohm, while the RF transistor most probably had a lower impedance. The additional 50 Ohm resistor in series just before the signal exits the RF module supposes this. I can't remove this resistor since it is still required for the upper band amplifier.

Finally a working signal generator

The total damage was less than I expected from the first analysis. Most +15V powered devices survived, and components supplied from the other rails were not harmed though there clearly must have been some overvoltage on these rails.

  • I've repaired the power supply and found the obligatory dried capacitor
  • I've found a replacement for the blown LM6165 amplifiers that works flawlessly
  • The replacement RF transistor in the reference section does it's job as well as the original
  • The improvised replacement lower band amplifier works quite well, though the original amplifier is believed to have performed better
  • In total, this is a satisfying result for an unit that was believed a total failure. The signal generator is useable again and I found a way around all the broken obsolete components.

    Back ... und ein Zaehlpixel hab ich auch :-)