Repair of Tektronix FG502 11 MHz Function Generator

[xrunner]

Back in the Amiga days, drawing with a mouse was compared to drawing with a bar of soap.

It's funny, it seems effortless to use the mouse to move and point on the screen accurately, but when it comes to drawing a line with one it's total FAIL.  Just for a test I'm going to re-draw that notated schematic properly!

*****

The only thing I did today (no mail today) is replace R250 pot. I had ordered the replacement which I determined from the parts listing was 1k, but when I unsoldered it today the pot had "500" marked on it (this marking was hidden from view also). I went back to the parts list and it did have two listings for R250 - 1k for certain serial numbers below "X", and 500 for serial numbers above "X". I wasn't paying close enough attention when I ordered the 1K pot. But ... not to be deterred, I just substituted an on-hand 500 ohm, 25 turn pot for the 500 ohm, 1 turn pot because I don't want to order another one. It'll be OK, since it just has to be set one time. 

[Rerouter]

The golden trick to drawing with a mouse is use the line tool and hold shift, makes perfectly straight lines.

[Johnny10]

xrunner

Wondering if you could use your oscilloscope to show the square waveform at the collector of Q290.

My unit was molested before I started my repair.

[xrunner]

The golden trick to drawing with a mouse is use the line tool and hold shift, makes perfectly straight lines.

Yea I know in a drawing program you can do that but I'm using a screen annotation app called Epic Pen -

http://epic-pen.com/index.html#features

It's just a quick and dirty pen/highlighter for the screen. It's pretty convenient and with the simple tablet I ordered I can draw a lot more accurately on these old scanned-in schematics.

Wondering if you could use your oscilloscope to show the square waveform at the collector of Q290.

Sure thing Johnny but at the moment it's out of commission till I get some parts, but looks like USPS tracking says I will get some today. Yay.  ALso after it's working and calibrated I'm going to post a lot more scope shots like I did for the PG501, so in the future this thread can be a resource for others trying to fix this unit.

My unit was molested before I started my repair.

I hope someone reported that to the police! 

[David Hess]

Wondering if you could use your oscilloscope to show the square waveform at the collector of Q290.

My unit was molested before I started my repair.

What exactly is it that you want to see?  Levels?  Edge speed?  Transient response?

[Johnny10]

I realize if I can't define it I am not understanding the circuit.

I was thinking I should see a square wave at this point.

[xrunner]

Got parts today - should have something cooking - I mean working - by tomorrow (bad choice of words!) 

I have a bin full of brand new BNC jacks and was going to replace the originals since the center pin female are tarnished quite a bit. Problem is I don't have & can't locate the one specialized BNC, the one that has an insulated ground (see pics). I can find ones on Ebay but they are not made the same and go into a "normal" size hole (this hole is much bigger). I suppose I could use washers but does anyone know where to find the original part?

[tautech]

That is one unusual BNC socket, never seen the likes of one before. 
Washers and a bushing seems the simplest remedy to me.

[xrunner]

Well the trigger buffer is attached. I have two other plug-ins - PG501, FG501A, both have trigger output BNC on the front and neither is an insulated jack. As you can see the trigger out is routed to the backplane but that one has a direct ground reference.

I think when the transistor conducts there would be about 17V at the emitter. But I have no idea what the reasoning is for the insulated ground design. It goes to ground through a 1.5 ohm resistor. Maybe David can shed some light? 

[David Hess]

I think the unusual trigger output circuit and isolated BNC keep the fast edge from the trigger from getting into the function generator output which is a problem on my B&K and Beckman function generators.  This would be important when the trigger output goes directly to an oscilloscope used for measurements causing a ground loop.

There are other subtle design features in the TM500 series instruments for preventing ground loop problems.

[MarkL]

I think when the transistor conducts there would be about 17V at the emitter.
...

The emitter should be about 5.4V when on.  The base is driven to a high of 6.1V as determined by R290, CR264, and R262.

[xrunner]

The emitter should be about 5.4V when on.  The base is driven to a high of 6.1V as determined by R290, CR264, and R262.

Yea I forgot about a 50 ohm load.

Check my calculations -

If you connect a 50 ohm terminated load to the trigger output BNC you have 50 + 24 = 74 ohms in parallel with the 1 k. That gives 69 ohms. I calculate when the transistor is on it will pass ~72 mA given the above resistance plus the other resistors. 72 mA through 69 ohms gives 4.97V. if you look at the screen shot I posted at the beginning of this repair that's about what I measured at the trigger output with a 50 ohm termination at the scope. The 1.5 ohm has ~ 0.11 V across it.

[MarkL]

The emitter should be about 5.4V when on.  The base is driven to a high of 6.1V as determined by R290, CR264, and R262.

Yea I forgot about a 50 ohm load.

Check my calculations -

If you connect a 50 ohm terminated load to the trigger output BNC you have 50 + 24 = 74 ohms in parallel with the 1 k. That gives 69 ohms. I calculate when the transistor is on it will pass ~72 mA given the above resistance plus the other resistors. 72 mA through 69 ohms gives 4.97V. if you look at the screen shot I posted at the beginning of this repair that's about what I measured at the trigger output with a 50 ohm termination at the scope. The 1.5 ohm has ~ 0.11 V across it.

Which other resistors?

I calculated 5.4V without a 50ohm load.

I didn't notice the amplitude on your screenshot before.  I would expect the trigger output to be less than that with a 50ohm termination given R280 and the output impedance of emitter follower Q268.

The service manual for the FG502 says the trigger out level should be >=5V for 1Mohm, and >=2.4V for 50ohm.  I can't argue with something that's working, but what amplitude do you get if the trigger output is unterminated?


EDIT: Added a little more detail to schematic snippet.

[xrunner]

Thanks for your analysis. You're probably right but how this started was why they used an insulated BNC for the trigger output. I have two other Tektronix plug-ins that have trigger outputs and they do not have an insulated ground. Why did they design this one differently? I'd like to know your thoughts on that.

But today I have enough parts to get it going, assuming it doesn't have any more problems, then we can make some more measurements for the record. That's what I've been doing for the other units I've fixed, for the record, so that someone in the coming years can find correct waveforms for these units and fix them. Measurements and waveforms from internal test points that just aren't available anywhere else, nit even in the troubleshooting section of these manuals. That's my legacy to leave here for whatever it's worth. 

[xrunner]

More problems ...

First I replaced the 4 22 uF caps and tested without the two output transistors - it worked OK got a good clean output trigger.

I then replaced the two output transistors, the two 10 ohm resistors (R536 & R546), and the 10 uF caps (C536 & C546) - after a few seconds one of the 10 ohm res. started to smoke. So it should appear obvious to the most casual observer that there is more to this picture. No, I didn't put the tantalums in backwards (beat you to it  )

I checked the 10 ohm resistors and they still read 10 ohms, so after pulling the new final transistors Q540 & Q546 (see that's why I installed sockets ...) it does put out a very nice amplified signal at all switch positions indicating Q530 & Q532 operating well. So, there's an issue in the final circuit still which I will dive into today.

[Kleinstein]

Are the "new" transistors still ok or are they broken.
One possible cause could be trouble with the 1 µF caps C540,C542 and corresponding.
It is also possible that the replacement transistors were too slow (e.g. other source for the 2N2905) and this way made the amplifier oscillate at high frequency.

There might also be a problem with load sharing: if Q530 and Q540 are too different in U_BE one of them would get most of the load current.

[MarkL]

Ch1 (yellow) looks like you have some distortion going high.  It's flattening out.

Plus you have Ch1 AC coupled.  What does it look like DC coupled?  The output could be way out of balance.

I would scope E, B, and C of Q530 and Q532 (without Q540 and Q542 installed).

Also note that you have FOUR output transistors: Q530 and Q540 share the load equally from the +20V rail and Q532 and Q542 share the load equally from the -20V rail.  It's interesting it works with only Q530 and Q532 installed.  Another way to narrow down the issue is to see if it works with only Q530 and Q542 installed, or only Q532 and Q540.

I'm still wondering why the 33ohm resistors aren't cooking.  Are they even hot?

[xrunner]

Are the "new" transistors still ok or are they broken.

Still OK.

Ch1 (yellow) looks like you have some distortion going high.  It's flattening out.

Yea that due to the fact it needs a complete re-alignment

...

I'm still wondering why the 33ohm resistors aren't cooking.  Are they even hot?

Nope.

But good news - I believe it to be fixed and working good as new (but it needs to have the cal procedure completely done). 

What was the final issue? Let me eat lunch and I'll type it up - but it should prove to be interesting and will go to show that you shouldn't take anything for granted in this hobby. There was nothing wrong with the transistors or resistors, and I made no installation mistakes like reversing leads and such ... if you want to guess go ahead. 

[David Hess]

There might also be a problem with load sharing: if Q530 and Q540 are too different in U_BE one of them would get most of the load current.

Through 33 ohm emitter resistors?  Not a chance, the excess Vbe from the triple diodes produces more current than any mismatch.  The bypass capacitors in parallel with the 33 ohm resistors should be checked though.

Also note that you have FOUR output transistors: Q530 and Q540 share the load equally from the +20V rail and Q532 and Q542 share the load equally from the -20V rail.  It's interesting it works with only Q530 and Q532 installed.  Another way to narrow down the issue is to see if it works with only Q530 and Q542 installed, or only Q532 and Q540.

Exactly, there are 4 output transistors in two parallel pairs and having either pair or a complementary transistor from each pair installed will produce the output.

I'm still wondering why the 33ohm resistors aren't cooking.  Are they even hot?

I would look for oscillation and the bypass capacitors carrying the DC current.  I would also change all four transistors at least temporarily to 2N4401s and 2N4403s until the problem is figured out rather than risking more expensive TO-39 parts.

[xrunner]

David see my post above - 

[MarkL]

Thanks for your analysis. You're probably right but how this started was why they used an insulated BNC for the trigger output. I have two other Tektronix plug-ins that have trigger outputs and they do not have an insulated ground. Why did they design this one differently? I'd like to know your thoughts on that.

David's explanation of reducing ground-loop currents is what I what I would have assumed also.

The SG502 also has this type of connector on the trigger output, but there's no detail in the schematic of how it's connected.

If you're up for an experiment, you could try shorting R284 to see if it has any noticeable effect on the signal output waveform (such as increased glitches at the trigger transitions).

[MarkL]

What was the final issue? Let me eat lunch and I'll type it up - but it should prove to be interesting and will go to show that you shouldn't take anything for granted in this hobby. There was nothing wrong with the transistors or resistors, and I made no installation mistakes like reversing leads and such ... if you want to guess go ahead. 

Ok, I'll take a wild guess:  The problem was mechanical.  Somehow, the heatsinks for the output transistors Q540 and Q542 got shorted together.  That would short their collectors together and put quite a bit of current through both 10ohm resistors.

[xrunner]

So remember I had ordered new parts - the transistors, the 10 ohm resistors, and 22 and 10 uF tantalum caps. The original failure burned the two 1/4 W 10 ohm resistors and also I found the 10 uF tantalum caps had failed, but I didn't find any other failures even the two output transistors checked out good.

The first things I replaced were four 22 uF tantalum caps, and checked it at that point and it worked OK. I then replaced the rest of the components in the output stage I ordered and turned it on and the magic smoke started coming out of the 10 ohm resistors. I turned it off before they were damaged too much to continue troubleshooting. I then pulled the last two transistors from the sockets and turned it back on and it worked OK even without the last two transistors.

At this point I was befuddled. I studied the schematic again and couldn't see what else it could be since I had checked most everything else in that area. I then inspected the board again and I noticed that the two new 10 uF tantalums I installed were discolored. I pulled them and found they were now bad. They both measured something like 45 - 55 ohms and failed the capacitance test.

So did I install them backwards? Nope. So what happened? I turned my attention to the rest of them in the bag I bought - 10 total. About half of the "new" caps tested bad! Bad as in they had no capacitance only resistance. So this is very very odd, that the original failure was bad tantalums and the following failure was the same using what was supposed to be new tantalum caps.

I decided I wouldn't use any of the 10 uF tantalums I bought, even though half of them tested OK. I installed two 10 uF electrolytics and the thing is humming along now like a finely tuned engine for the last hour up to 10.9 MHz. 

I will now write a message to the seller of those 10 uF tantalums asking for a refund.  Also I'm now confident this part is done and will shorten up the transistors leads and re-re-install two new 10 ohm resistors.

Now I can move on to the calibration procedure and lastly the cosmetics/cleanup.

OK johnny10 - your turn.  You can use this thread if you need any help or measurements.

[xrunner]

xrunner

Wondering if you could use your oscilloscope to show the square waveform at the collector of Q290.

Here's your waveform. CH 1 is the collector of Q290, CH 2 is the trigger output for reference.

[MarkL]

But why, when you removed "Q540 & Q546" (I assume you meant "Q540 & Q542"), did it work fine without smoking the 10ohm resistors?

The newly installed bad capacitors would have still been in place, if I understand your sequence correctly.

Did you ever go back and test the original capacitors with a power supply to make sure they were bad?  A little post-repair failure analysis can never hurt for confirmation, especially if it's easy.