G W Instek SFG-1013 Repair attempt

[waveform]

Hi All,

Bit of a long shot request. I am attempting the repair of a Instek SFG-1013 Function Generator that I acquired, second hand, cheap.  I can get a good TTL output but the Main output is completely shot. You will see from the attached pic that apart from the 2 resistors being burnt, G W Instek also felt that the components highlighted in red would likely be blown. They believe that the previous owner must have applied voltage to the output.  Repair by Instek was un-economical (more than the cost of a new unit!) so I've replaced all the suggested components but no joy. The replacement Transistor 2N2905 and replacement 47R (R917) resistor get exceedingly hot, just as the old ones did and there is still no output at the Main BNC.

I notice however that there are points on the board (highlighted in green) which are not populated. These are marked Q904 and C907 and I have no idea whether there should actually be components there or whether they are there for a different model (seems a bit odd)

I have been unable to obtain a schematic or service manual, online or from GW Instek and, although Instek originally offered to supply components they did not do so (I had to buy my own) and they have since stopped communicating or helping in any way. 

I was wondering whether, if anyone has this unit, they would be prepared to take a photo of that part of the board so that I can tell whether there should be components in the highlighted positions?  Any other thoughts/suggestions would obviously also be appreciated as, even if there should be components in these unpopulated spots, I will have no idea of the appropriate values.  Thanks for reading.

[Armadillo]

Is that a pcb discoloration effect or burnt surface? Or camera effect.
I have never seen such burnt spread pattern.
Would suggest;
All the vias, check continuity from top to bottom. Check the traces anyway. Eliminate got-you surprises.
Next, all the MLCC, check for short across. [don't underestimate even the 0603].

[waveform]

Is that a pcb discoloration effect or burnt surface? Or camera effect.
I have never seen such burnt spread pattern.
Would suggest;
All the vias, check continuity from top to bottom. Check the traces anyway. Eliminate got-you surprises.

Hi, It is a deliberate PCB effect (not a burnt surface) which just looks a bit darker/more prominent on the image.  I have done a considerable amount of continuity testing (which I find quite difficult on these multi-layer boards) and will continue looking. Traces all seem good. Cheers

[alm]

I am not familiar with this particular function gen, but the fact that the 47 Ohm resistor gets hot is interesting to me. I imagine, based on its value and the fact that it was blown, that it is a series resistor at the output of the output stage. There may be some sort of leveling circuit after it, I am not sure, but I would expect one side of the resistor to be pretty much directly connected to the output cable. So that suggests that there is a short very close to the output connector. I imagine this is with nothing connected to the output connector? I would trace the circuit around the output connector and look for shorted semiconductors, resistors or even shorts in the PCB.

[Armadillo]

Did you measure ohm the emitter of the 2219 to output common.

[waveform]

Did you measure ohm the emitter of the 2219 to output common.

Forgive my 'newbie ignorance' but I'm not sure what you mean by this or how I do it 

[waveform]

I am not familiar with this particular function gen, but the fact that the 47 Ohm resistor gets hot is interesting to me. I imagine, based on its value and the fact that it was blown, that it is a series resistor at the output of the output stage. There may be some sort of leveling circuit after it, I am not sure, but I would expect one side of the resistor to be pretty much directly connected to the output cable. So that suggests that there is a short very close to the output connector. I imagine this is with nothing connected to the output connector? I would trace the circuit around the output connector and look for shorted semiconductors, resistors or even shorts in the PCB.

There is quite a lot of circuitry between this resistor (including an opamp, standard & zener diodes and a relay) which may be what you describe as levelling circuit? Please see the attached image which gives a better idea of the layout (but ignore annotations as I have since replaced these components and now it is just the 2N2905A and it's connected 47R resistor that get hot).  Guess, I'm a bit out of my depth with these multi layer boards as far as continuity/ testing for shorts is concerned. SMD components which have no markings don't help much either 

[Armadillo]

I think;
The 2 transistors form the complementary output. The NPN transistor emitter is connected to the PNP transistor emitter thru some low value resistor. Depending on the switch selection, the emitter of the NPN transistor 2N2219 should not be lower than 49.5 ohms to output common.
Measure the ohm from the emitters to the output common [ground].

[alm]

There is quite a lot of circuitry between this resistor (including an opamp, standard & zener diodes and a relay) which may be what you describe as levelling circuit? [...]

No, that sounds more like an output attenuator that can be switched off by a relay. If the 47 Ohm resistor is in the signal path, then it is further from the output then I thought, which also increases the area where the short might be. My suggestion would be to either try to find a service manual of a very similar instrument, reverse-engineer a partial schematic for the output stage starting from the overheating components, or find someone that has worked on this unit before.

[mikerj]

The 47R resistors look like a simple supply decoupling arrangement for the output stage.  The ends of the resistors that connect to the collector of each transistor also have a small electrolytics decoupling them to ground, one for the positive rail and one for the negative rail.  There are a pair of 5.6R resistors in the emitter circuit to help with bias stability (classic Class AB output stage) and the output impedance is increased to ~50R by the array of four SMD resistors, 3x200R and 1x150R all in parallel.

I would start by measuring the voltage on either side of the four 4 parallel resistors w.r.t ground.  If the output stage is properly biased you should see very close to zero volts on the output stage side of these resistors (right hand side in pic) with no signal output.  If there is a short somewhere towards the output socket then you would expect the voltage on the left hand side of the resistors to be lower to turn on the upper transistor i.e. the fault is forcing the output stage to source current, which is handled by the top transistor connected to the positive supply rail.

[ModemHead]

I was wondering whether, if anyone has this unit, they would be prepared to take a photo of that part of the board so that I can tell whether there should be components in the highlighted positions?  Any other thoughts/suggestions would obviously also be appreciated as, even if there should be components in these unpopulated spots, I will have no idea of the appropriate values.  Thanks for reading.

I don't have any specific knowledge about this unit, but I do have a working one.  I don't think I've ever opened it before, the case was a bit of a puzzle at first.

To answer your first question, the Q904 and C907 locations are unpopulated on this board also.

The controls were set as follows: default 1KHz sine, output on and unloaded, minimum output level, and I pulled out the offset control and adjusted it for minimum DC offset on the output.

The voltages on the finals are annotated on the attached photo. Looks like the 47 ohm resistors are collector resistors leading to the power supply rails.  The SMD 5R6 resistors are the emitter resistors.  Both transistors are biased on at about 40mA.  After about 30 minutes of on time, the transistor heat sinks both rise to about 72C.  Too hot to touch for very long.

[waveform]

I don't have any specific knowledge about this unit, but I do have a working one.  I don't think I've ever opened it before, the case was a bit of a puzzle at first.

To answer your first question, the Q904 and C907 locations are unpopulated on this board also.

The controls were set as follows: default 1KHz sine, output on and unloaded, minimum output level, and I pulled out the offset control and adjusted it for minimum DC offset on the output.

The voltages on the finals are annotated on the attached photo. Looks like the 47 ohm resistors are collector resistors leading to the power supply rails.  The SMD 5R6 resistors are the emitter resistors.  Both transistors are biased on at about 40mA.  After about 30 minutes of on time, the transistor heat sinks both rise to about 72C.  Too hot to touch for very long.

Many thanks ModemHead! This has given me a lot more to work with. Looking at the photo of your unit, it becomes clear to me that someone has tried working on my sick puppy before. There are no heatsinks on the output transistors of my board (I thought there should be judging as they are shown on the silkscreen). I'll now see how close, or far away, the voltages are on my unit.

[waveform]

Well, that's a shock...please see image with voltages There's clearly much more wrong with this device than I thought; the voltages with respect to ground (taken with the settings used by Modemhead) are out of this world. For example, the voltages across the 47 Ohm resistors drop by about 15V (5.08V on one side and 20.7V on the other) !  the 47 Ohm on the negative side is similar. Collector, Base and Emitter voltages are similarly way off.

I would start by measuring the voltage on either side of the four 4 parallel resistors w.r.t ground.  If the output stage is properly biased you should see very close to zero volts on the output stage side of these resistors (right hand side in pic) with no signal output.  If there is a short somewhere towards the output socket then you would expect the voltage on the left hand side of the resistors to be lower to turn on the upper transistor i.e. the fault is forcing the output stage to source current, which is handled by the top transistor connected to the positive supply rail.

I have also shown, on the pic, voltages measured on either side of the parallel resistors which are negative (- 3.78 on one side and -3.73 on the other).

[ModemHead]

I assumed you had taken the heat sinks off for the photo.  You will definitely need to replace them.  Even in a quiescent state, there is a fair amount of heat being generated.

From the voltages, it looks likely that the bottom output transistor Q906 is fried, and maybe the top transistor Q903 is also, or it's being driven really hard.  That means pretty much everything in the area is suspect.  Check every base-emitter and base-collector junction with your multimeter diode range.  For reference, all the SMD transistors are NPN, and Q904A is PNP.

You will also need to verify that the D/A converter is actually producing the correct waveform which will require an oscilloscope.  I'm not an expert in amplifier topology, but it looks like Q908/Q909 are a differential input pair, which makes the base of Q909 the input to this stage.  The whole thing is direct-coupled, and has a gain of 10.  I see a 2Vpp signal on Q909 base for an output of 20Vpp (with output level control maxed out.)  This 2Vpp signal also appears on the right-hand side of the unpopulated J701 header.

Something to note, when the output enable switch is off it appears to stop the clock as well as de-energize the output relay, leaving the input waveform at an arbitrary DC level.  So keep the output enabled for predictable results.

[waveform]

Thanks ModemHead. I really appreciate your interest in helping with this.  I did start with replacing ALL the transistors in this section (the smd componenents Q905,907,908 & 909 are the worst as they are miniscule but that seemed to go OK with the aid of Chipquik). I also replaced Q904A (2N3906) and of course Q903 & Q906.

I'll now have to buy another Q903/906 pair before I continue as I suspect one or both are fried again (I have plenty of the SMD MMBT3904LT1G and 2N3096 left to swap out). In the meantime I can however, as you suggest, check the waveform on my scope (The TTL output is good but I guess that doesn't prove anything).

The components I replaced (i.e all the transistors) was on the advice of Instek (it was their annotation on the image I included with my opening post).  I do wonder though whether the 2 diodes D906 & D907 which seem to connect the Base and Emitter of Q903 & Q906, could be faulty?? What type of diodes do you think would be used here (eg a switching diode such as RLS4148) ? 

EDIT: All waveforms (sine, Square & Triangle at different frequencies) are good at the base of Q909 , so the DAC is good 

Once again thanks for your help.

Regards
Erik

[waveform]

Well, this unit just keeps on giving in terms of surprises!   I completely removed the two main output transistors, Q903/Q906 to run a few tests. Out of curiosity, I hooked the main output to my scope and hit the output button (not expecting anything but a bit of noise perhaps) and was shocked to see a perfect waveform at the default 1kHz generator output 

What was even more surprising was that:-

- I can adjust the output voltage up to 19.8V Pk-Pk (I've set it to around 6V Pk-Pk at the moment)
- I'm sure the manual suggests just 10V Pk-Pk
- I get a perfect Sine/Triangle & Square wave up to about 2 mHz (after that there a touch of distortion in the Waveform)
- I can adjust both the Duty Cycle and the offset without issue.

So, everything seems to be working reasonably well without the output amplification circuitry (obviously, as the transistors have been completely removed ). I'm now wondering therefore if the output amplifier merely boosts the output so that it can achieve more than 20V Pk-Pk or whether also prevents the slight distortion I see at higher frequencies? In any event, I'd still like to fix the issue but at worst case, if I cant, I do have a decent FG with some minor limitation.

[ModemHead]

That's really good that the waveform source is okay and the problem is definitely in the output stage.

Regarding the diodes, I'm not sure of their role.  They are reverse-biased.

The final pair of transistors is a class A/B push-pull amplifier as already pointed out.  Each Q is essentially an emitter follower, NPN for the positive half of the waveform, PNP for the negative half. They lower the output impedance of the amp, so it has enough current to drive loads that are much lower impedance than your scope. The entire output circuit is essentially a big discrete operational amplifier in non-inverting amp configuration.  But without the final stage, it has very little current available.  Connect too much of a load and the output will diminish and/or distort.  Probably not very useful.

Regarding the 10Vpp output spec, that would be into a 50 ohm load.  The output impedance of the generator is set to 50 ohms with the parallel array of SMD resistors on the output line, so driving a 50-ohm load cuts the voltage in half.  Think voltage divider.  This is typical of most function generators.

When you get output transistor replacements, don't operate them without heat sinks.  That may have contributed to their early demise.  In fact I might be thinking about some kind of current-limiting for the next attempt.  Maybe temporarily replace the 47 ohm resistors with a higher value.  Or get lots of spares...

[waveform]

....... But without the final stage, it has very little current available.  Connect too much of a load and the output will diminish and/or distort.  Probably not very useful.

Regarding the 10Vpp output spec, that would be into a 50 ohm load.  The output impedance of the generator is set to 50 ohms with the parallel array of SMD resistors on the output line, so driving a 50-ohm load cuts the voltage in half.  Think voltage divider.  This is typical of most function generators......

Mmmm...interesting. I'll have to try feeding some of my audio circuits with a generated signal and see what impact that has on the signal . As I say, I still want to restore the output amp and will replace all the components again (although I find that the original 2N3906 that I replace was fine). There is clearly one component that is causing this 'domino effect' failure and blowing Q903/Q906...just a case of identifying what it is before I burn out a couple of dozen transistors  . I find the smd transistors the most difficult to test as, out of circuit it's virtually impossible to attach DMM probes (even with grab connectors) and in circuit the readings are very random. That's why I asked about the diodes but I did manage to detach one end of each of them and they do seem to check out OK. Bit baffled but I'll keep going because it's a fun project and I'm learning a bit . Thanks for your support and education

Regards
Erik

ADD: I have ordered TO39 heatsinks - not quite the same (I cant find originals) but hopefully will do the job (see pic)

[ModemHead]

... There is clearly one component that is causing this 'domino effect' failure and blowing Q903/Q906...just a case of identifying what it is before I burn out a couple of dozen transistors  .

Yeah, that's what happens with a DC-coupled feedback loop.  A fault anywhere can throw the whole thing out of whack.

I find the smd transistors the most difficult to test as, out of circuit it's virtually impossible to attach DMM probes (even with grab connectors) and in circuit the readings are very random.

You definitely need some needle-sharp probes for SMD parts in-circuit.  Most default multimeter probes are like shooting marbles with mittens on.  Just for grins I went through the output circuit on mine and got solid in-circuit diode-junction readings on all the transistors.  (Roughly 0.68V on all.)

[PA0PBZ]

If you want to measure them out of circuit put a piece of double sided tape on the bench to stop them from running away when you poke the probes at it.

[waveform]

If you want to measure them out of circuit put a piece of double sided tape on the bench to stop them from running away when you poke the probes at it.

Good idea that  :-

[waveform]

I find the smd transistors the most difficult to test as, out of circuit it's virtually impossible to attach DMM probes (even with grab connectors) and in circuit the readings are very random.

You definitely need some needle-sharp probes for SMD parts in-circuit.  Most default multimeter probes are like shooting marbles with mittens on.  Just for grins I went through the output circuit on mine and got solid in-circuit diode-junction readings on all the transistors.  (Roughly 0.68V on all.)

You do realise that you've just added to my ever growing shopping list? 

[alm]

I like these probes for small SMT work. They are fairly affordable, but you have to order the black and red probe separately.

[waveform]

Update - Success! Today, I replaced all 5 surface mount transistors (MMBT3904LT1G), 2 1N4148 and the two power transistors (2N2219A and 2N2905A) and fitted heatsinks (which were not in the unit when I got it). I originally bought the unit for £25 and have spent £16, on parts so,  £41 all in, not at all bad for a unit that costs circa £140. Everything is now working perfectly. Thanks to all of you for your time and contributions to my questions. I particularly want to thank Modemhead who not only took voltages, and posted photos from his working unit but offered a great deal of help & support. 

[ModemHead]

Good job on the repair!  Hope it works great and lasts a long time.