SA-9800 left channel negative output clipping

[Fabio92]

Hi all, this is my first help request in this forum and I hope in somewone kind advice. About two weeks ago my friend gave me a SA-9800 saying that the left channel sounds distorted. I'm not familiar with this particular model but i wanted to give it a try to fix.

Before powering the unit I readed the service manual, to have a general idea of what might be the problem. After reading several posts here on AK about the PSU heating problem and component failure I started form there checking all the voltages.

The amply was untouched, all original. Moreover it seems it was not used so much, the PSU pcb is in a good state, no burnings, no brown areas. I already now that a complete recap is quite mandatory, even if not stressed, electrolitic capacitors have almost 42 years and leakage current and ESR might be a problem.

PSU assembly is the awr-196 and voltages at the outputs are quite good, mabe +18V to display a litte out specs:
pins 1&2 -52V -52.71V
pins 3&4 +52V +52.46V
pins 5&6 +32V +32.72V
pin 7 +18V +19.34V
pins 10&11 -32V -32.72V

input voltages
pin 17 55.4Vrms
pin 18 55.5Vrms
+68V is actually 73.7V
-68V is -75.6V

Outputs from the rectifier board awr-191 are a little out:
pin 13 -55V -59.9V
pin 14 55V 58.88V
pin 16 55V 59.7V
pin 17 -55V -59.9V

Then I moved to the pa-L assembly gwh-119. The first check I've made was hooking an oscilloscope probe to the outputs in parallel with a small speaker. Given a sine 1kHz input, just starting from very low volume adjustment, the left output is clipping to about -740mV. Attached a comparison betwheen left (yellow) and right (cyan) outputs.

It's not possible to regulate idle current, I always measure 8.8mV at pins 37-39 and 0v beetween 38-40. Dc offset is instead adjustable and at 5.3mV.

Note: all the measurement noted before desoldering components and reported on the image below are still the same even after re-soldering original components.

I started looking for damaged componets or cold joints on the pa-L assembly. I've disconnected the four output bjt (2SA1076 & 2SC2526) and tested BE and BC junctions with a multimeter giving about the same readings of Veb = 0.558V, Vcb = 0.548 for the pnp and Vbe = 0.579, Vce = 0.577 of the npn. Then I moved to the drivers 2SA913 & 2SAC1913 and they still seems good. Of course I've cheched that there weren't CE shorts. I've desoldered and checked in the same manner almost every transistor, except the monolithic current mirror and input differential pair. I've also checked every resistor with "one leg out" rule and they are on specs. Honestly I'm starting loosing ideas! The output wave is still clipping at about -730mV. I also disconneted pin 26 form the protection assempby witout any changes.

I measured most of the interesting currents/voltages in the circuit as reported in the image attached and below (currents are derived from voltages and previous resistance measurements):

emitter current of the differential pair: 570uA (500uA)
emitter currents of Q5 mirror, equally spread on both bjt is 283uA
Ir23 & Ir29 (top left current source) respectively 983uA (1mA) and 7.25mA (7mA)
Ir47 & Ir55 (top current source) respectively 1mA (1mA) and 1mA (1mA)
Ir35 & Ir59 (bottom current source) respectively 1mA (1mA) and 1mA
pin 26 (no load) 0v
Current throught balancing emitter resistor R79,83,81,85 0A.
D5 cathod 1.23V (1.5V)
Q7's collector 4.38V (4.5V)
Q33's base 1.17V (1.17V)
Q25,Q27 protection bjt both off

here I'm starting having throubles:
D5 anode -4.77V (-1.5V)
Q11's collector -15.37V (-4.5V)
Q11'emitter -47.5V (-50.5V)
Q33's base -0.91V (-1.4)
Q29,Q31 protection bjt both fully on
current throught Q11's collector resistance R33 24.6mA
Q13 Vbe 635mV
Q15 Vbe -490mV
Q17 Vbe -358mV (too low conducting)
Q19 Vbe 630mV (too conducting)

The negative side of the circuit is pulling too much current. I think It's not the driver Q35 because is barely conducting with a Vbe of -540mV.
My guess is that the culprits are the protection bjt Q31 & Q29 but I can't figure way they are conductiong.. Q29 Vbe is given by the the Wheatstone bridge at the output but load current is 0A.

Finally I did somethig that phreaps someone can judje crazy, but I turned the power on without Q29 & Q31 but in this case the situation was revesed putting into conduction upper Q27 & Q25 protection transistor turning off Q33 driver.

I very appreciate if someone followed me till the end, I spend several hours trying to figure out what is going and I wanted to give most details as possible hoping for your kind advice.

 

[markthefixer]

The feedback system may be pushing all the voltages to absurd (and hard to analyze) levels.

Remove Q25, Q27, Q29 & Q31, then disconnect the four output transistors.

Test all 8 transistors. Leave them disconnected. Eliminating their influence from the circuit.

R73 between the drivers Q33 and Q35 (IF connected to the feedback point) will establish a DC feedback target, Probably at -0.7 volts - if you want to be a stickler for symmetry.
IF NOT, and your BIG X there on the schematic is correct ( I am examining the patterns) establish a DC feedback point by splitting 470 ohms into 2 ( 220 + 220?) and the middle tap of the two halves
should go to the DC feedback point. (top of R21, 2.2k)

Then power up (NO signal) the Left Channel and check the voltages (like you did) again. Might spot a bad Vbe voltage.

Then run in a sine wave and SEE what the BASE drive waveforms for Q33 and Q35 look like,
then look for the same waveforms at Q33 and Q35 on their emitters (just one base - emitter drop lower) which is what would be sent to the outputs.

The NSA circuit (Q13, Q15, Q17, Q19) should be splitting the upper and lower halves of the waveform.

Once it is doing that correctly, then we can add outputs.I suspect a problem with (Q13, Q15, Q17, Q19) or their power feed Q21 and Q23.

After THAT then we can add the extra fiddling that Q25, Q27, Q29 & Q31 do.

This image, even if it's for a SA-7800 NSA amp, may help visualize things:

 

[Fabio92]

Thank you . Tommorow I'll do as you suggested. Meanwile, to give you an answer about if R73 is or not connected to the feedback path, I've taken some pictures of my board. I've to say I've done a little mistake with assembly name: actually my revision is the GWH-125 and not 119. Manual says that schematic is the same for both. I've looked carefully to both schematics and they are identical. But I noticed that my circuit have some differences:

1. betwheen C-E of Q27 and Q29 there are two 4.7nF ceramic capacitors. (when I see them connected in this way in power circuits I usually call them snubbers, am I right?)
2. protection diodes across bases of Q3 are zener in anti-series, instead of normal diodes in anti-parallel as shown in schematic.
3. I can confirm that R73 is not connected to the feedback path (blue line in the image below). (I also remember to have read of this error in the scematic in another forum's post)

 

[markthefixer]

Excellent response.

Remember, the amplifier needs to have the feedback loop closed, thus the needed connection.
Unless you spot a bad transistor in the (Q13, Q15, Q17, Q19) group without powering it up.

 

[Fabio92]

Hi, I'm here again! I've done like you suggested, splitted R73 into two 220Ω connected to the feedback path. Removed Q25,Q27,Q29,Q31 and all four output transistor. Then I rechecked voltages in the circuit.

• Current sources are in specs.
• Current through R33 is now 7.18mA (previouslly 24.6mA)
• Current through the two 220 splitted R73 is 10.5mA
  
• Q13 Vbe 577mV
• Q15 Vbe -630mV
• Q17 Vbe -594mV
  
• Q19 Vbe 600mV

I didn't tryed to re-solder neither the output nor the protection bjt,because I found another problem, this time affecting both channels (I didn't noticed this behaviour before on the right channel): without applying any input signal, the output is very noisy / oscillating.

In the picture below of the left PA channel, ch1 (yellow) is the input measured beetween pins 35 & 36, ch2 (blue) is the base of Q33 npn driver and ch3 (purple) is base of Q35 pnp driver. Even at the input, between pins 35 & 36, there is a noise of about 300mV pk-pk at 32.9MHz

here a zoom of the above picture


then I've re-checked the output, measured at the speaker connectors, of the right channel (untouched): without any input and no load, it is also oscillating at about the same frequency!



I think I've to to separete both channels from the pre-amplifyer, because I found stange that this behaviour is affecting both channels. In theory they are independant

 

[markthefixer]

Sigh, I had hoped that they wouldn't oscillate and that you could check the dc operation of the circuitry.

There would be a lot of fiddling around just to kill the oscillation.
Bigger neutralization capacitors (they are found connected collector to base on various transistors, LIKE but not limited to C25, C27)
Trial values of 0.01uf or so.


Just check out of circuit Q13, Q15, Q17, Q19 - or just replace them. This has been my advice in the past - to check EVERY transistor and diode out of circuit.

We needed to get the bad effects of Q25, Q27, Q29 & Q31 out of the circuit.

You could try putting R73 back to original, installing output transistors (but NOT Q25, Q27, Q29 & Q31) and testing to see if the amplifier does NOT oscillate like it did in the last post. The outputs would provide the feedback connection needed.
VR3 would be set at zero ohms, and VR5 would be carefully adjusted to set the turn on (emitter to base) voltages of Q15 and Q19
THAT is why all the Q21 & Q23 circuitry is in there - to provide that voltage - no matter what the output voltage of the amplifier is.