Sansui AU919

[Michael Maurice]

I have been asked to repair this beast after its owner had replaced some capacitors by following instructions from this forum.

His work seems to be of a reasonable standard but when he brought it to me, it had two faults. The first fault was no left hand channel from the AUX input, this was simply a broken wire on the ribbon cable.

The second is no sound from the right hand channel on any of the PU inputs.

The fault is on the head amp MM, I've included a diagram below. The output instead of being at 0V is at +38V.

I've done the following:

Checked for dry joints
Checked for broken print.
Removed the feedback loop (temporarily)
None of these make any difference.

I then removed R59 and R61 or rather R60 and R62 I will from now on use the LH channel diagram for ease of understanding.

The output as expected was 0V. but the voltages on TR17/19 collectors are not what one would have expected.

They are at +38V. My next move will be to replace all 6 transistors Q9 - Q19.

The voltages around FET1 and Q1 - Q7 appear correct.

Any help will be gratefully received.

[G6ONEDave]

It might be a silly comment but you have got the negative supply to the actual panel at joint 35.

[Radio Wrangler]

It might be a moving mgnet cartridge RIAA stage, but it's built as a mini power amplifier. All of the complexity (and maybe more than most) and all of the voltage swing, but lighter currents.

Powered off of crazily high voltage rails for point-scoring on overload margin figures. For some crazy reason with a JFET input stage which is liable to be a quite suboptimal match for the source impedance from the cartridge.

Check the -ve rail is there.

Check the DC feedback path along R's 71. 73, 75, 77, 79, 81.

Measure the voltages on the two gates of the dual JFET. The left hand one should be zero and the ritht hand one should be a scale model of the output DC you've found...

47/(47+390+1.8k+6.8k+20k+47) * 38V = +61mV

If you have a big DC difference across the gates, it shoulld be driving the rest of the amplifier to reduce the output DC.

As you can see, for reasons best put down to ****** stupidity, this circuit is set to deliver not only huge AC gain, but the same and more at DC.

It is non-inverting, so not many millivolts DC from a preceding stage or leaky capacitor will also do the deed.

What the desperate will do for one more bullet point on a marketing epistle....

David

[Michael Maurice]

Dave: The -40V is present and correct.

David, the resistor chain is present and correct. I found R45 to be burnt and O/C but replacing it made little difference.

The voltage between the gates is near on 0v on the good channel and about 12mV on the bad one.

I agree its well over engineered, i doubt it would sound much difference that a couple of NE5532's!

[mikej56]

This may also seem like a silly comment;

If the amplifier was working before the owner changed the capacitors, wouldn't the most likely fault be a new capacitor fitted the wrong way around - reverse polarity. I'm looking at C47 which decouples the feedback line to the +40 DC power rail? Could this have become a short circuit?

[Michael Maurice]

C47 decouples the +40V to ground, if it went s/c you'd blow the fuses.

[mikej56]

Sorry, I see that now. It was a silly comment.

[Radio Wrangler]

OK, let's start with the externals.... The power rails you've said are OK.

So how about any DC coming in on the input... any DC with respect to ground on R03? we're looking for millivolts upwards because of the daft gain. This could be from a capacitor mistake on a previous board. I don't know what S01 does.

How about the bias voltage on the bases of the cascodes TR01, TR03

then the voltages on their collectors.

Because it's high gain, there can be small differential voltages between balanced signal lines, but the lines may both be a long way from ground. So measure each with respect to ground, and if they are close to equal, do a differential measurement so your meter can go on a more sensitive range.

It's more work, but it stops you shooting past and missing something.

David

[samjmann]

I worked on one these beasts a few weeks ago. It was basically working. But had bad hum on the phono stage shown. It was the two capacitors in series that are in the output of the preamp that had failed. On both channels. I can't see that giving your fault, but they might be what was first wrong.

I couldn't set the preamp mid point to zero for every MM. As the input selector was moved it was correct for one MM but then not the other.
The preset has a huge control range.

I put mine into AC coupling on the back switch to protect the output transistors in case of DC spikes in fault finding. If these fail it might be the end of it.

This amp did sound great all the same though.

Sorry, not the best of typing as done on my phone...

SJM.

[Michael Maurice]

Originally Posted by samjmann

I couldn't set the preamp mid point to zero for every MM. As the input selector was moved it was correct for one MM but then not the other.
The preset has a huge control range.


SJM.

I haven't touched the preset, but it might be worth looking into.

[Michael Maurice]

ts taken all this time to get some parts. I went through the amplifier again today, I found the owner had fitted C9 the wrong way round, so I replaced both C9 and C11 but thatmade no difference.

I also replaced transistors TR9, 11,13, 15, 17 and 19 but this made no difference whatsoever. I tried adjusting the pot VR01 but again it made no difference.

One thing, the voltage on the output does not go straight to +38, it takes a few seconds.

I am totally stuck.

[Radio Wrangler]

OK, let's break the circuit down into stages.

The output is rammed hard against the positive supply rail.

The input amplifier has strong overall DC feedback applied to it. If the input amplifier is working, it should sense that the output is much more positive and should be trying to fight it.

You said there is 12mV difference between the gates of the dual FET. It is very very important which polarity this is. The gate of the FET with C03 on it should be the negative end.

The amplifier is really a chain of several amplifier stages. Somewhere along the chain is a fault. The fault is driving the output to full positive. All stages after the fault will be saying "Go full positive!"

The input stage (if working OK) will see the full positive output and be trying to oppose it. All stages before the fault will be trying to drive the output negative to correct it.

So, if we stab a stage with a voltmeter and we see it's trying to drive the output negative, we are before the fault, if we see it trying to drive the output positive, we are at or after the fault. This lets us home in on the fault and rapidly reduce the number of suspects.

There are three stages of voltage amplification, follower by two stages of current amplification. This amplifier stage has to have a lot of closed loop gain at the low frequency, so in order to have a reasonable amount of feedback there, it needs an even larger amount of open loop gain... hence the three stages of voltage amplification. Most power amps only have two. But then they don't need to do so much gain, closed loop.

The output is stuck at +38v

Follow that through R71 (47) R73 (20k) R75 (6.8k) R77 (1.8k) R79 (390) and we have 29037 Ohms. It looks like R81 to ground is 47 Ohms.

So the DC voltage on the right hand side gate should be 38x (47/(47+29037)) =61.4mV
The DC voltage on the left gate ought to be 0v so long as no DC is coming in from off left.

What's the gain of the first stage? Well production tolerances on FETs are ginormous so we won't get a terribly certain figure, but let's take a stab. Yfs (another name for gm!) is 40mS typical (= 1/25 Ohms) add in the 2.2 Ohm resistor gives 1/27.2 Ohms. TR01 TR03 are common base stages (current gain close to unity) so the current from a FET winds up in 1.91k

So the DC gain in the stage will be 1910/27.2 = 70.2

This will be reduced by loading from TR9, TR11, TR13, TR15 but they have 180 ohm emitter resistors and should have current gains over 100, so the loading is going to look like 18k/2 = 9k

9k parallel with 1910 Ohms is 1576 Ohms, so I'll revise that gain estimate to 58.

So with +35 on the output, the difference on the gates should be 61.4 mV then with this gain, that should be 3.5 volts between the collectors of TR01 and TR03.

With the right gate positive, TR03 should have the more negative collector.

THat's a lot of voltage swing

TR11 should be cut off (measure Vbe) and TR09 should be carrying all the current
TR13 should be cut off and TR15 should be carryiing all the current.

On to the third voltage amp stage TR17 should be cut off and TR19 should be pulling current. So the voltages on either end of C25 should be down near the negative rail.

Follow this through and see where reality diverges from theory. and we'll be on the trail.


David

[samjmann]

If it's any help, I've got one of these to look at again. This one does run, but the owner complained of hum on phono, which I've reduced by 50%, but not eliminated.

There is a switch on the back to AC couple the output stage, I moved this from the DC position it's set to DC from factory. Just in case of spike etc when servicing.

One thing I did notice is that when switching between MM1 MM2the mid point of the preamp section shifted DC level. I'm sure that this was wrong.

As soon as it's with me again I'll post again.

Regards, SJM.

[Michael Maurice]

I’ve now replaced the amplifier, the cause was the FET. They are not available but an IFN146 from Mouser proved a good replacement.

You have to swap the drain and source leads on one side of the device before fitting. After a setting the output to 0V and testing the amplifier was returned to the customer.

[Radio Wrangler]

That's good. There's a lot in that circuit to lead people astray, if not scare them off completely.

It's likely that the optimum Zsource for noise figure of that JFET will be a long way from the source impedance presented by moving magnet cartridges, but it is what it is, and you've kept it original, which must be the best option.

I wonder if the JFET died from just being physically disturbed, or maybe an ESD zap during the recap operation.

Good work!

David

[Michael Maurice]

I eventually found that he put a cap in the wrong way round! This was before I changed the FET's. I did both sides for balance and completeness.