Advantest R3361A spec-an - TG output 10dB low - help!

[philpem]

Hi there,

If you've seen my other thread about transceiver testing, you might have gotten the impression I wanted to avoid blowing out my spectrum analyser. It would seem my fears were somewhat unfounded, as the previous owner has already beaten me to it!      

I'm intending this to be something between a repair log and a Q&A - so if you can help me, or think I've made a howler of a mistake in my logic, please shout out!

First problem I noticed was that the tracking generator output was 10dB low - at 0dBm out, I was seeing -12dBm (connection was a 1m TE Connectivity BNC cable and two BNC-to-N adapters). The 1dB steps were gone, all I had were the 5dB steps. The analyser's been like that since I got it, and until I looked at the service manual I thought it was yet another software quirk (if I remember correctly, there are a few other things you can set in increments the analyser hardware can't actually handle).

A quick check of the AGC level on the TG revealed it was pinned at -13V; the AGC is negative-acting (it feeds four PIN series diodes), so 0V is "turn the bugger off!" and -6V-ish is "Och, ah' cannae give ye any more, Cap'n!". Negative thirteen is definitely a sign of brokenness. At this point the service manual pointed me at the AGC Control board (BLC-015680). Well, it turns out this board was fine. All the opamps were behaving sanely, the L-inputs (1dB-increment output power control) were switching properly, and there was a signal on the DET input which changed when I shorted the AGC pin to ground at the feedthrough (shorting the AGC pin is in the service manual test procedure, worry not!). Of course, the service manual doggedly insists that the AGC control board has to be at fault, the rest of the TG is infallible... 

So it's probably fair to assume the detector, 4GHz oscillator, mixer and AGC control are fine - the middle two being an absolute twat to test, requiring a 4GHz-capable power meter and/or spectrum analyser!     The rest of the circuit can be (partially) tested with an oscilloscope -- simply set the analyser to a centre frequency of 30MHz, zero span, and turn the TG on. Past the mixer, you won't see anything faster than 30MHz. 
The 4GHz oscillator output is the thing that the "fine gain" control (AGC Control board) adjusts via the AGC signal. This is fed into a mixer (the other input is the local oscillator), the output of which is fed into the output amplifier. Which is where we continue the story!

The troubleshooting flowchart indicates that there are five amplification stages between the mixer and the low-pass filter (a funky gold-on-ceramic thing which looks like an interestingly shaped PCB trace and positively exclaims "I am black magic you will never understand"  ).

If you'd like to follow along at home, I've attached the schematic for the power amplifier.

First step, "Thou shalt check the rails". 15V, -15V and 12V are right where they're supposed to be.

The first stage is U1, a MiniCircuits MAR-7. This provides 4dB of gain. Its cagemate U2 is the same part, and provides 18dB of gain.
This feeds into Q1/Q2, a combination of a 2SK878 AlGaAs/GaAs HEMT FET and a 2SA1162 PNP which should give 9dB of gain.
Q3/Q4 is more of the same, but with a 2SK571 GaAs MES-type FET and the same 2SA1162 PNP tuned for the same 9dB of gain.
Finally we have Q5 and Q6, a "900075" GaAs power FET and a 2SA1162. This has minus 7dB of gain (an attenuator?) and finally feeds into the funky ceramic microwave low-pass filter.

That gives us a total gain of 33dB leading up to the LPF. So let's trace the signal through.

30MHz centre, zero span, TG on, 0dBm. Go.

At "IN", we have 127mV of signal according to the DSO, in the form of a somewhat distorted sine wave (01_IN.PNG). After passing through U2 and C1, this becomes a respectable 219mV (02_U2in.PNG). A quick calculation pins that down at 4.7dB gain. Looks like that's working.
U1 is more of the same. Our 219mV becomes 1.8V at U1's output (03_U1out.PNG). There's our 18dB! By the time it's been through R3, C3 and C4, it's down at 530mV.

At the Q1/L3 junction (Q1/Q2 amplifier "output"), we see 2.91V (04_Q1out.PNG). Against our 1.8V that's only 4.1dB gain (we're expecting 9dB). Against 530mV, we're looking at 14dB, which is a bit much. This is starting to get interesting...

At the Q3/L7 junction (Q3/Q4 amplifier "output"), we have 3.34V (05_Q3out.PNG), but our sine wave has cast off its spectral purity and become something quite distorted! The positive peak clips at about +1.5V and folds back to about +1V. Oh dear.

By the time we've got to the output of the power amplifier (Q5/Q6), we've got something that bears a striking resemblance to a cardiac sinus wave (06_Q5out.PNG), and there just isn't any point in measuring the amplitude (it's about 543mV).

What I can't tell is, are the transistors acting up because something further down the line is loading them down?

What should my next step be? Start desoldering capacitors to isolate sections and re-measure? That seems a little dicey - the PCB is an alumina-base (from what I've been told) thing with gold tracks. Heating and reheating the thing seems likely to cause damage, and tricky to repair damage at that.

Does anyone have any suggestions?

Thanks,
Phil.

[nctnico]

You could have a crack in the ceramic base. What happens of you try heating it a little?

[KJDS]

Is the TG 10dB low across the whole frequency range?

[philpem]

You could have a crack in the ceramic base. What happens of you try heating it a little?

You mean the ceramic that the circuit is built on? Heat doesn't have any effect on it.

Is the TG 10dB low across the whole frequency range?

More or less. It falls off a little more (maybe a few dB) at the high end, but otherwise it's fairly stable.

I've just done some more checks -- I've isolated Q5 (NEC NE900075 power amp MESFET) from the circuit by desoldering C12 and got these signal amplitudes, all peak-to-peak of the :

• U2 pin 1 - 133mV
• U1 pin 1 - 200mV (+3.5dB)
• C4/Q1 junction ("Q1 in") - 631mV (+10dB)
• C8/L3/Q1 junction ("Q1 out") - 2.41V (+21dB vs. U1 out, +11dB vs. "Q1 in")
• C8/R62/Q3 junction ("Q3 in") - 2.41V
• C12/L7/Q3 junction ("Q3 out") - 7.6V (!) with distortion on the negative half cycle. 250mV with no distortion into a 50R load.

DC bias points:

• Q1 at R61: -1.147V
• Q1 at L3: 2.75V (vs. schematic 3V)
• Q3 at R62: -1.485V
• Q3 at L7: 2.97V
• Q5 at R63: 0.40V
• Q5 at L10: 0.58V

The lack of bias on Q5 is extremely worrying... I'm not sure if Q1's bias being a quarter-volt down is anything to worry about, but 0.5V bias on Q5? That's silly.

The distortion on Q3's output looked exactly the same as that in 05_Q3OUT.png, except it was affecting the negative half-cycle instead of the positive.

I've noticed that the schematic has some incorrect component designations. What it calls Q9 is actually Q4 on the PCB...

[nctnico]

I've had cracks/open circuits in ceramic circuits before. It's a bitch to find though.

[KJDS]

I'd change Q5 and the PNP that biases it. It's a classic FET bias circuit, and it could be either device that's gone, or possible one has died and killed the other, hence it makes sense to change both. Pbviously spend a few minutes checking that the resistors and inductors round it as ok.

[G0HZU]

At the moment the circuit has failed in a safe way for Q5. (but of course Q5 itself could be the part that has failed). Normally the active bias circuit Q6 (drawn wrong in the circuit as an NPN. Should be PNP) will produce a bias voltage at Q5 input that causes about 75mA drain current in Q5. This will mean Vds will be about 3V normally.

I'd be tempted to probe around Q6 and I'd expect the emitter to be at about 2.6V and the base at around 1.9V if the circuit was healthy. The collector would be somewhere around 0V or just below at a guess but this depends on the bias point of your individual FET.

With your faulty circuit Vgs is 0.4V which would bias it hard on. This is what you are seeing. Normally, I'd expect to see about -1V at the gate of Q5.

So Q6 or one of its biasing components may have failed although it is in a circuit that should give it an easy life. I'd also check that the shunt capacitor at Q6 base hasn't gone leaky/short as this would cause your symptoms.

Also check the dc blocking cap in the negative feedback RC network from drain to gate of Q5 hasn't gone leaky. Or Q5 itself may have developed an internal leakage path.

[G0HZU]

Thinking about it a bit more... even if Q6 went short circuit collector-emitter you couldn't see +0.4V bias at the gate of Q5.

So it looks bad for Q5 (internal leak?) or the RC feedback network from its drain to gate has gone leaky. Or maybe C12 is leaky but this is unlikely.

[philpem]

I'd change Q5 and the PNP that biases it. It's a classic FET bias circuit, and it could be either device that's gone, or possible one has died and killed the other, hence it makes sense to change both. Pbviously spend a few minutes checking that the resistors and inductors round it as ok.

The catch is, the GaAs FET seems to be unobtainium:

Q5   NEC   NE900075   GaAs Power FET
Q6   2SA1162     TRANSISTOR PNP

Q6 I can see listed at a few places, Littlediode have them at £1.87 for two I'd likely get a few spares. It's expensive for a PNP, but still pocket change (anything less than £10 is "pocket change", I spend more than that on a night out!).
Q5 on the other hand... Only seen those listed at Littlediode, and they want £76.19 plus post for one! Mouser can apparently get them, but they're non-stock "call for quote". Beside that, nada, unless someone knows of another source or has some squirrelled away?

At the moment the circuit has failed in a safe way for Q5. (but of course Q5 itself could be the part that has failed). Normally the active bias circuit Q6 (drawn wrong in the circuit as an NPN. Should be PNP) will produce a bias voltage at Q5 input that causes about 75mA drain current in Q5. This will mean Vds will be about 3V normally.

Yeah, I noticed the PNP/NPN swappage...

I'd be tempted to probe around Q6 and I'd expect the emitter to be at about 2.6V and the base at around 1.9V if the circuit was healthy. The collector would be somewhere around 0V or just below at a guess but this depends on the bias point of your individual FET.

With your faulty circuit Vgs is 0.4V which would bias it hard on. This is what you are seeing. Normally, I'd expect to see about -1V at the gate of Q5.

So Q6 or one of its biasing components may have failed although it is in a circuit that should give it an easy life. I'd also check that the shunt capacitor at Q6 base hasn't gone leaky/short as this would cause your symptoms.

Also check the dc blocking cap in the negative feedback RC network from drain to gate of Q5 hasn't gone leaky. Or Q5 itself may have developed an internal leakage path.

I'll check those tomorrow afternoon - if I'm right, I should have equivalent parts for the passives on hand. I'm not so sure about the transistors... see above

Thinking about it a bit more... even if Q6 went short circuit collector-emitter you couldn't see +0.4V bias at the gate of Q5.

So it looks bad for Q5 (internal leak?) or the RC feedback network from its drain to gate has gone leaky. Or maybe C12 is leaky but this is unlikely.

It does, doesn't it? :-/

I'll have a poke around Q6 and its surrounding parts tomorrow afternoon, but I have no idea what voltages to expect (seeing as I don't have a working unit to check against).


Thanks,
Phil.

[G0HZU]

If Q5 has failed and you want to try a cheap and temporary bodge I can suggest this:

Swap Q5 for an Avago MGA-82563 if it can be made to fit the space.

Also, lift up R63 (not needed with the Avago part)

Also fit a 4.7V 500mW zener at the junction of R25 and L9 as a voltage limiter at power up.

This might work as a temporary solution and give you time to find the correct device.The Avago MGA-82563 costs about £2   (or £5 on ebay or £3 from Farnell).

There will be similar (better?) alternatives than this but that's what I found after a quick search. The Avago part is quite small but it is capable of 100mW at Psat and 50mW at P1dB when biased at 3V at 80mA . This is fairly typical for a 3V RF driver device

What package is the original Q5 in?

Also, I would be tempted to check the health of the 4 protection/limiter diodes D3-D6 that are just ahead of the attenuator section. These could be blown due to very high reverse power and may have died along with Q5 if someone zapped the TG port with many Watts of RF

[philpem]

What package is the original Q5 in?

It's the "Package Outline 75" package shown in the datasheet: http://radioshop.co.uk/surplus/NE9000.pdf

Also, I would be tempted to check the health of the 4 protection/limiter diodes D3-D6 that are just ahead of the attenuator section. These could be blown due to very high reverse power and may have died along with Q5 if someone zapped the TG port with many Watts of RF

I'll have a look at those - thanks. They're apparently HSM2838CTR parts (aka HSM2838C) dual fast recovery diodes.
Datasheet for that is here: http://www.datasheetarchive.com/dlmain/Datasheets-13/DSA-246771.pdf  and seems to be more unobtainium...

EDIT --> And here's a bit of info on the FET biasing circuits seen in the R3361A TG: http://www.odyseus.nildram.co.uk/RFMicrowave_Theory_Files/Bias_Circuits.pdf

Thanks,
Phil.

[G0HZU]

I'd get a bit of shim metal eg an old tinned screen wall, drill the two holes in it and then cut it to the size of the flange of the 75 package.

Then preheat it and then solder the avago part to it at a slight angle so it's input and output pins are pointing out as required (after bending the legs). It would then hopefully fit the slot and screw down the same as the original part.
The NEC part has a higher power specs but this is when biased at 8V. At the 3V bias in the analyser circuit it will have similar power capability as the Avago part. 
 Worth a try?

[eb4fbz]

You could try an MGF1801 as substitute for Q5. You would need to adjust bias circuit for it.

NE900075 Vds=8v Ids=50mA
MGF1801 Vds=6v Ids=100mA

[philpem]

I'd get a bit of shim metal eg an old tinned screen wall, drill the two holes in it and then cut it to the size of the flange of the 75 package.

Then preheat it and then solder the avago part to it at a slight angle so it's input and output pins are pointing out as required (after bending the legs). It would then hopefully fit the slot and screw down the same as the original part.
The NEC part has a higher power specs but this is when biased at 8V. At the 3V bias in the analyser circuit it will have similar power capability as the Avago part. 
 Worth a try?

Sounds like a plan, except that the Avago part has a minimum frequency spec of 100MHz. The SA goes all the way down to 9kHz... losing TG below 100MHz would make it difficult to align, for instance, radio IF strips.

It seems the drive transistors are fairly standard 150mA parts, and a BC857B would probably work at a pinch (albeit with an Ic rating of 100mA, 150mA absolute max).
I'm crossing my fingers for a biasing issue...

Would, say, an ERA-2SM or MAR-6 be a possible option for a temporary substitute? (just thinking out loud here - I have more of those MMICs in stock than I care to count...)

You could try an MGF1801 as substitute for Q5. You would need to adjust bias circuit for it.

NE900075 Vds=8v Ids=50mA
MGF1801 Vds=6v Ids=100mA

That seems to be just as hard to find as the NE900075...

[Mr Simpleton]

Easy...
http://www.ebay.com/sch/i.html?_from=R40&_trksid=p2050601.m570.l1313.TR0.TRC0.H0.Xmgf1801&_nkw=mgf1801&_sacat=0

[G0HZU]

Sounds like a plan, except that the Avago part has a minimum frequency spec of 100MHz. The SA goes all the way down to 9kHz... losing TG below 100MHz would make it difficult to align, for instance, radio IF strips.

If you keep the existing Drain to Gate 320R  'RC' negative feedback network fitted to the Avago part then I would hope/expect it to regulate the gain response to make the device usable down to very low frequencies.

i.e. the reason the RC network is there for the NEC part is to control/stabilise its gain right down to the low frequencies the TG is designed for. It does this with negative feedback because the 320R will feed back the output to the input in antiphase and will regulate/flatten the gain response across a very wide frequency range.

I think the Avago part already has an RC feedback network fitted internally but the tiny internal dc blocking capacitor in series with the resistor only allows the feedback to be effective down to a ballpark 100MHz.

Obviously there is the risk it won't be stable and it might not give the same harmonic suppression but I still think it is a contender as a temporary fix and it will allow you to find the real NEC part in slow time

Note:
I don't know what the existing 320R  RC network looks like in terms of package or how it is fitted but I suspect that it is a unique part and you should be very careful when going near it with a soldering iron. The capacitor part of it could have quite high capacitance in a very small footprint making this another bit of unobtanium.

Is it fitted in piggy back on top of Q5?

[philpem]

Is it fitted in piggy back on top of Q5?

It's a black ceramic package with two wire leads -- Q5 is screwed into the metalwork and soldered into the circuit, then the R-C bypass network connects across it.

I've spent this evening checking the parts in the circuit, and everything looks fine. The only thing I can't rule out is.... Q5.
The bias transistor measured ~0.6V diode drop between B-C and B-E, and seems to be intact (and not shorted). Shorting B-C and B-E changes the bias voltage a little, but not enough to get to 3V. I could shotgun it, but I'm not convinced that's the issue. Given the relatively high resistances involved, I'm more inclined to think that Q5 has developed a short, possibly to ground.

Thanks,
Phil.

[philpem]

Guys,

I haven't managed to find any NEC NE900075s, but I did find a supply of NE900175s. They're covered by the same datasheet:

http://www.radioshop.co.uk/surplus/NE9000.pdf

Could anyone comment on whether that'd work as a substitute, and what I might have to do to the bias circuit to make it work reliably? Seems like it'd involve less modification than the other parts mentioned upthread.

Thanks.
Phil.

[G0HZU]

Looks good!

You probably wouldn't have to do anything to the Q6 auto/active bias circuit because it is designed to automatically bias the device at around 3V with a 75mA drain current.

So with the 175 part fitted at Q5 the Q6 circuit will automatically adjust the gate bias voltage 'for you' to hit the above bias target of 3V with a 75mA drain current.

It's certainly the best option if you can get the 175 part at a reasonable cost. If works as above then I'd consider it a permanent repair. The Avago part was only suggested as a stopgap solution to buy you time

[G0HZU]

The only slight performance risk is if the 175 part can't achieve the same gain up at a few GHz. But hopefully this won't be significant and the gain of the two devices could be very similar with the 320R feedback RC network fitted.

Definitely worth trying if the original part is unobtanium or £££

[philpem]

Managed to score three '075s!

Next job: get some of the bias transistors too.

More updates when the transistors arrive and I've installed them

[philpem]

Hi guys,

She's fixed!

I installed the new NE900075 FET and swapped the 2SA1162 at the same time (just in case!). I measured 7.6V of bias -- contrary to the 3V the service manual suggested. After noting this I quickly powered down and checked everything. On closer inspection, I noticed that the new 2SA1162s from Littlediode were labelled "3E", which the Marsport SMD Codebook identifies as a BC557A. The parts in there were marked "M6", which Marsport IDs as a 2SA1179. I will be discussing this with Littlediode! 

I swapped the old transistor back in (bit of a gamble!), but the odd bias voltage persisted... Given that I had a stable output, I assume this is (yet another) service manual error. 

After checking all this, I noticed that connecting the TG Out to RF In resulted in no signal on the display. I set the SA to 50MHz Zero Span and -10dBm TG Out, and both the MSO-X scope and the Racal 1992 frequency counter reported a 55MHz output. Oh dear... I traced this back to the 4.066GHz dielectric oscillator being a bit off frequency. I had removed the cover to this while I was testing the TG block, and in doing so knocked the oscillator frequency out of adjustment! 

Thankfully this is an easy fix.

• There's a second cover under the TG top cover. This shields the 4.066GHz dielectric oscillator. Don't remove this cover! (though if you're doing this adjustment you've probably removed it already -- in which case, put it back and tighten up the screws!).
  
• Loosen the nut which holds the gold adjustment screw. Replace the TG top cover and hold it down with a few screws (you don't need to fit all of them). (If the top cover isn't in place, the TG will oscillate and otherwise misbehave)
• Set the analyser to: Zero Span, Centre Freq 50MHz, TG on at -10dBm
• Monitor the tracking generator output with a frequency counter
• Adjust the gold adjustment screw until you get 50MHz out. Required tongue angle may vary! 
• When you've finished, remove the TG top cover and tighten the nut with a spanner while holding the adjustment screw steady with a screwdriver.

 

A few tips if you find yourself working on one of these units:

• The TG module needs to be shielded -- if you have the TG top cover off, the RF modules will misbehave (self-oscillation and other fun). Covering the little submodules with your hand will make them "mostly" work, covering them with the metal cover will make them work fine. Tinfoil would probably also work if you need to access some parts of the circuit while shielding others.
• Under the TG top cover there's a second cover with a gold adjustment screw in the middle-left. Whatever you do, do NOT touch the gold screw or remove the cover. If you do, you'll have to re-adjust the 4.066GHz oscillator (see above).

There may be a better way to trim the TG output frequency (4.066GHz oscillator), but if there is, it's not in the service manual... and if it is in the service manual, it's missing from my copy! Suggestions on how to properly adjust this thing would be gratefully received!

Cheers,
Phil.