A measured Sansui AU 7900 restoration

[Sci525]

I’ve recently picked up a Sansui AU 7900. I’d like to dedicate this thread to its restoration and measurement. There are plenty of threads on this and similar Sansui's restoration, so I will focus less on talking about parts and more on before and after measurements. I have an Analog Discovery 2 for an entirely different work purpose, but I’ve been playing with it for audio measurement. It should be just sensitive enough to measure any improvement to this amp.

The amp! It’s a little beat up on the outside, but all original inside, no cracked boards, and I don't see any bad solder joints.





For most measurements I'm just going to show one channel, whichever is worse, innless there are big differences.

Distortion. Showing 1K Hz THD across power out. I’m getting 48 watts from an amp spec’d for 75. If you push the amp beyond 48W it starts clipping, then oscillating! The load resistor heat up FAST! I’ve read oscillation is a common symptom of bad capacitors which fail to isolate and allow feedback loops.

The amp sounds good between around 1 and 10 watts although it is out of spec (0.1% THD) at 8K Hz as can been seen on the plot below, 20–20K THD at 5 Watts. The higher distortion in the highs seems to create a smooth sound that isn’t accurate but is still very pretty.

However, at 48 watts the highs are clearly distorted, and they wear on you quickly unlike the smooth highs at 5W.

Let’s look at noise. With a 1K Hz reference signal @ 10 Watts, noise is -82 dB with mains hum and its harmonics, and -84 without.

And now phono stage noise at 10 W out. About -52 dB with mains contribution, more like -56 dB without. Pretty bad.

Square waves look fine at 10W (10k Hz shown below). I'm curious what they look like at high output, but I don't want to push the amp into oscillation if I don't have to.


I've got a pretty basic recap and restoration planned and I've started soldering. More on that later. I'm excited to see how it changes things.

-Joe

 

[bolero]

Cool!!

 

[Monicker1]

I will follow with interest. Does it look to you like this amp is original from the factory since new? No replacements or mods? And has it lost its serial number sticker?

 

[Sci525]

Does it look to you like this amp is original from the factory since new? No replacements or mods? And has it lost its serial number sticker?

Yes, I've been through the main two boards. Looks all original to me. I don't see the serial number.

 

[Sci525]

The restoration:
I plan to replace 82 components including all ECs, a few pairs of mylar input/output caps, double diodes, zener diodes, bias pots, fusible resistors, and the transistors in the phono EQ. Normally I wouldn’t replace the phono stage transistors in less I was unhappy with it after the recap, but the noise on the phono stage is very bad and accessibility isn’t great, so I've just done it all in one go. All pretty standard, upgrading to MKP films where it makes sense and trying to select the right cap for each location.

What I've got so far:
The main regulator and amp board. I left some red glue, but it isn't touching anything. After I took this photo I went back and reoriented a couple films. Everything fit quite nice.

And the preamp power supply and phono EQ with the new filter caps hiding off to the side. I used a big pair 400v MKPs for the phono stage input. That isn't recommended, just parts I had on hand and wanted to use.


Cheers,
Joe

 

[Sci525]

Here is the recapped tone control board...

And the companion filter board.


Everything checks out and I'm letting it burn in for a few hours before I do a listening test and then take the post restoration measurements. TR13 which regulates the DC to 22V for the tone control gets very hot. I'm leaving it for now, but I would replace if doing it again or if I need to disassemble for another reason.

Cheers,
Joe

 

[bubhardy]

@Sci525, sounds like too much current is being pulled through TR13. Just my 2 cents.
I think I’d investigate the tone control board. It’s current draw should not be much.

 

[billyz]

Looking forward to seeing what you measure after the Cap job, Just sold my AU 7900 . It sounded great but, since I got ahold of the AU9900, I have not wanted to listen anything else.

 

[Sci525]

sounds like too much current is being pulled through TR13.

Hmm, I don't hear anything or see anything wrong with the tone board. Anything I should look/test for? I'll see if I can find any other hot components and check the transistor temp with a thermometer. TR13 drops the voltage by 12, would that be enough to produce a lot of heat without much current?

Here is the circuit in question.

Thanks!

 

[Hyperion]

It is true that series pass transistors can and do get hot especially when the voltage differential from input to output is high, and because we can't feel how hot it is, it could just be normal.

However, are you getting 22V out of the positive regulator? - something as simple as a bad or reversed electrolytic can cause overheating, but there are other possible causes.

 

[bubhardy]

At, say, 50mA, that 12V drop will produce 600mW. TR13 is a 2SD356, which is speced to be able to dissipate 1W into 25 deg. C free air. That 600mW will make that transistor feel hot to the touch, but the junction temp will still be a fair amount below the maximum junction temp of 150 deg C.

 

[Sci525]

However, are you getting 22V out of the positive regulator?

Yes, I've checked all 18 voltage points (9 per channel) which Sansui kindly denoted on the schematic. Everything looks good.

That 600mW will make that transistor feel hot to the touch, but the junction temp will still be a fair amount below the maximum junction temp of 150 deg C.

I appreciate the explanation. I can keep my finger on it a couple seconds so it's nowhere close to 150C. Not particularly scientific, but I think everything is probably running as specified.

Audio measurements coming soon.

 

[Sci525]

Listening at between 2 and 3W there is much less noise post restoration. Lows sound punchier and I find vocals more engaging, maybe more lifelike. Otherwise very similar sounding with those smooth highs I mentioned before.

OK, some details since some of these measurements now hit the sensitivity limits of my somewhat noisy equipment. When measuring across sweeps in power or frequency, THD vs power or THD vs frequency, the measurement floor is around 0.02% THD but around 0.04% THD between 20-30 Hz due to 60 Hz noise in my measurement equipment that I couldn’t resolve. For noise testing with a static reference or test tone I can consistently measure down ~-94 db excluding 60 Hz noise. Readings will go lower, but they are not consistent. As before I’m just showing one channel, the worst of the two (right), which has about 0.01-0.02% more THD above about 6kHz. All before and after testing is shown with tone control defeated. I did plenty of testing with tone control on and off, but the results are complicated, and the differences are very minor.

The clipping and oscillation issue that was limiting power is resolved. Power output looks clean @ 1k Hz to about 85 W! A big improvement from 48. The zigzagy lines are THD.


The largest improvement to THD is at around 1 W and may be better than shown since the values are at or near my measurement floor.


At 5 W we see only a minor improvement in THD from the restoration. I believe that high frequency distortion is partially controlled by the amps slew rate. It measures at 5.7 V/us on the right and 7.0 V/us on the left. That might also explain why distortion is worse on the right channel but only at high frequencies. Although only 3 V/us is needed to produce a 20k Hz signal at the amps full output voltage, the general rule of thumb is that the slew rate should be 5 times this value (15 V/us) to minimize distortion from ringing and feedback.


It's worth remembering that we can’t hear the harmonics of signals beyond 8K Hz (ish), so the plot below is of “audible” distortion for different frequencies across power output, including only harmonics in the 20-20K band. Notice that at 5K Hz the 2nd and 3rd harmonic are in the audible range and distortion is outside of Sansui’s spec. Maybe there is still something wrong with the amp, but it’s not extreme and may just be age on the transistors and mylar caps. Happy to hear any thoughts.


Here is a spectral analysis of a 5K Hz tone at 20W. The third harmonic is @ -55 db. Not good but...


...It occurs to me that I’m doing all this testing on an 8 ohms dummy load, but speaker impedance is often much higher than 8 ohms at higher frequencies and would load the amplifier significantly less. Since the distortion is so power dependent, it makes sense that it should also be load dependent. Here I compare the THD across frequency at 20W on a speaker (blue) and the dummy (pink). The real world (ish) scenario looks much better and is near enough to factory spec.


Noise is greatly improved for both line level inputs and phono inputs. The 60 Hz noise and fundamentals on the line level input measurement is from my measurement setup. On the phono stage, 60 Hz noise is real, audible, and dominant, but the measurement likely makes it look worse than it is. Sorry I can’t be more precise.
Line level input:

Phono:


In summary, a simple restoration greatly improved the power output, stability, and noise of my Sansui AU 7900, but made only minor improvements to distortion which allows the amp to retain its smooth highs and a touch of vintage character. Speaker selection to minimize load at high frequencies should help maintain the amps performance at high output (>10W) where otherwise audible distortion is possible.

Cheers,
Joe

 

[bolero]

I wonder if replacing those remaining few mylars would make a difference?

Would be interesting to see graphs comparing tone ctrls bypassed to non bypassed, too?

Thanks for posting these: it's always good to see quantitative info

 

[Sci525]

I wonder if replacing those remaining few mylars would make a difference?

I wonder this too, but as long as none of them have failed, I think the difference would be very small. Maybe one day I'll find out, but right now I have other Sansui's that need TLC.

Would be interesting to see graphs comparing tone ctrls bypassed to non bypassed, too?

Sure. If I just showed any one figure it could be misleading. But I've done the measurements, so I'll just need to find some time to organize them.

 

[Sci525]

Below is a quick measurement and analysis of THD and noise of the tone/filter section of the Sansui AU 7900 after a recap but retaining all transistors in this section.

The tone/filter section reduces gain, 17.7 (tone/filt on) vs 20.4 (tone defeat) @ 5W. So, we need to compare based on the same gain, not power. Below is distortion at 25 db of gain with the tone/filter section defeated and engaged. Pretty darn close.

But more harmonic distortion is 3rd order with the tone/filter section engaged. Many people prefer 2nd order distortion to 3rd.

That said, the change from 2nd order to 3rd order distortion can move high frequency distortion out of the audible range, so audible band distortion is reduced above 6.5k Hz.


The tone/filter section will also add a tiny bit of noise, but it is not measurable with my equipment which suggests it is less than -94 db or masked by the noise in the power amp section. Very good considering I did not replace any of the tone control transistors. In general, the SNR of the amp (restored) through the line level inputs is great.

In my opinion, good tonal balance is much more important than any of the minor factors I’ve shown/discussed in this post.

 

[Monicker1]

I had no idea there is a nearly 3db drop between tone/filter on and defeat. Usually I don't shift to compare when the tone controls are in the flat setting, but yes the difference is audible. It would seem like the drop is barely audible, so would put it in the 1 to 2 db range if I had to guess. Is the drop or attenuation pretty much constant across the audio spectrum? BTW Listening to the same amp.

 

[bolero]

This is good: I've always wondered about that, but have no tools/skills to investigate. Impressive, and Thanks!

 

[Sci525]

Is the drop or attenuation pretty much constant across the audio spectrum?

In short, yes. In long, not exactly. When you engage the tone control, the frequency response generally changes a bit. This is because the components aren't so precise that the 0db spot on the control is exactly neutral. You usually get +/- a couple tenths of a db on each control (in this case bass, mids, treble), but you won't notice because the frequency response of speakers has much more variation, not to mention of effects of the room.

 

[Overundr1]

There have been reported instances of those green mylar caps not aging gracefully. Kudu's sir for the great visual walk through pre and post rebuild. It does make me look at all the gear on my bench wondering if the modern computer age is going to put all the analog test equipment out to pasture.