Sansui AU-D907 needs some recaping...

[JoseHH]

I will replace those black flags eventually and do a full recap when I have time. I am thinking if mine can survive 39 years, it must have better black flags. Those with bad black flags are already dead or half dead. So, when they are replaced, I'd like to have them checked to confirm my theory. However, I don't have the equipment to do the testing.

If you are to replace the outputs with TO-3P modifications, will you get both channels done or just the bad one?

[/QUOTE]
Hi Robin,

There are no "better" black flag capacitors. To me, after reading about them, they appear to me as a bad capacitor technology. I am no expert in capacitors but this site by ConradH shows a lot of details and measurements:

http://conradhoffman.com/BF_caps.htm

The larger values seemed to be more stable than the low 3 and 6 pf. In my opinion, since polystyrene is a low melting temperature linear polymer which is known to exoerience cold flow, it will nit maintain its thickness tolerance. This is known to plastics in the machining community. It may be wise to replace them all, with C0G or NPO ceramics or mica capacitors.

You ask:

If you are to replace the outputs with TO-3P modifications, will you get both channels done or just the bad one?

In my case both channels had bad outputs. Only one par NPN-PNP were working on each channel. So I have a single genuine pair NMA/NMC1012, the other working pair are fakes that were installed by a "technician" who did a repair on this unit. The rest (4 transistors are open, or shorted but with cut pins so they are not connected. I removed them all already and installed the MJ21195/6 pairs and the amplifier works now.
But I am ambitious and would like to get the AU-D907 closer to the original specifications. There are only two ways to do it. For the time being I will experiment with the TO-3P devices. I have pins to adapt them to the sockets. The problem for me are the driver transistors. I have an idea which I exposed above on post #16. But this may be a terrible idea. I hope some of the big shots here have the time to comment on this.

Good luck with your capacitor renovation.

Jose

 

[JoseHH]

Hi everyone,
For the time being, I replaced the capacitors in the power supply. Here is the "before" picture:

Note the mylar capacitors mounted directly on the metal brackets when there is provision for soldering them on the PCB. This was the initial, original state. I noted that the unit I am working on, differs from the ones Hyperion, Scott Thompson and ilimzn showed in their photograps. Theirs appear to have thr foil capacitor installed in the printed circuit board. In this unit, the foil capacitors where taller and narrower and were not originally installed in the printed circuit.

I went on cleaning the glue, and replacing the small electrolytic capacitors in the F2845 power supply and made the modification that Echowars suggested. I think it is a logical thing to do. I used Panasonic FC 220 uF rated at 100V and in the four central capacitors I used Nichicon KZ 220uF at 100V. This is because I did not have 12 of the Panasonic type and I judjed that the Nichicon capacitors were just as good for this. After doing all that, the main power supply board looks like this:

With regards to the ground connections, in order to be able to dissasemble easier, I decided to use 4 terminals for branching from the M5 stainless steel screw:
Branch 1-Ground to star ground in chasis.
Branch 2-Speaker return and protection board ground
Branch 3-driver board (wires 21 and 22)
Branch 4 Preamplifier toroidal transformer ground "z" wires grey and black.

However, to remove the main power supply board F2845 one stil needs to desolder ground wires 19 and 20 from it which go to the star ground at the chassis, and the 6 pre-driver power supply wires (13, 15, 17, 14, 16 18).

Next, I will move on to the preamplifier power supply.

 

[Overundr1]

Had to put my own 919 on hold while other pieces demanded my time, will admit to glaring at that power supply board and what was involved in getting it out of the chassis, nicely done

 

[JoseHH]

Thanks to all the information kindly available in this forum, it is a real pleasure to clean up my amplifier.
I measured the ripple at the +/- 56 volts rails and all four rails were between 80 and 91 mV. Comparing this to the ca. 50mV values that John (Hyperion) posted when he repaired the main oval capacitors in :

http://audiokarma.org/forums/index....-capacitor-upgrade.561441/page-2#post-7388213

it seems that it is not urgent to renovate mine. I may do this in a second stage of renovation.

I move on to the preamplifier power supply now.

 

[JoseHH]

The preamplifier power supply board was removed without unsoldering the wires. This board had no glue in it. There were 5 100 uF capacitors rated at 25, 35 and 50 volts, and two 1000 uF. The old flux on the conductor side was quite abundant. Here are the pictures of the initial condition:

I replaced the 100 uF wit 5 identical panasonic FC rated at 100V. The 1000uF were replaced by a pair of Nichicon FW 1000uF rated at 100V. As it was well documented in this forum, the silk screen printing for the capacitors C07, and C08 is wrong, so I installed the new ones in the same orientation that the originals had. Here is how they look after doing this:


I left the rest of the components untouched since the voltages of this power supply were measured to be correct before I worked on it.
I finally cleaned the old flux, and applied a thin coat of fresh flux. I put the power supply back in the amplifier and checked the output voltages. At the driver board, I verified that the DC offset and bias were still the same as I had before.
The heat sink in the preamplifier power supply gets quite warm after a few minutes of having the unit powered. I am not sure this is the normal condition. There is a 17 volt drop at each of these transistors, so that even a few mA of idle current can easily get to over 100 mW of dissipation. I believe that once the bottom cover of the amplifier is bolted in place it will also help dissipating this heat, so this may not be an issue to worry about. Please correct me if I am wrong.

 

[RobinTW]

Jose,

I have an idea about TO-3P pins. See details in http://audiokarma.org/forums/index.php?threads/to-3p-subs-in-my-ba-f1.749242/page-2#post-11173082

Robin

 

[JoseHH]

Jose,

I have an idea about TO-3P pins. See details in http://audiokarma.org/forums/index.php?threads/to-3p-subs-in-my-ba-f1.749242/page-2#post-11173082

Robin

Hi Robin,
Thanks for the suggestion. I have seen that thread. It seems like a good idea, but tin is a bit too soft. TMy concern is that the kind of "claw" in the TO-3 socket is quite strong and may rip off some of the tin and may create an unreliable connection touching in some points rather than along a line like it happens with a smooth round pin. But honestly, i am thinking aloud. The idea has to be tested to be certain. Some other thread offered another idea about filing and forging the pins into a nearly cylindrical shape but I can not find where i read this now. I got some pins but are different from those that Echowars (Glen) suggested. I will post some pictures soon. But the TO-3p devices are one of the next things I am going to try. Right now I completed the flat amplifier cleanup and capacitor replacement. I am going to post pictures soon.

Jose

 

[JoseHH]

I had some time to continue removing the black flag capacitors in the flat amplifier (F-2841). I loosened the front face of the chassis, with the unit standing up on its side, with the power transformers down. Then removed the nuts of the balance and volume potenciomenters, and the 4 screws that secure the 4 switches S01-S04. Disconnected the power supply connector and then carefully desoldered the signal pins to separate this board from the control bard (F-2846). I needed to remove the spade of an orange jumper wire from the control board (wire labeled JP02) which was on the way to detach the pins from the control board. This are the pictures of the flat amplifier before I removed the black flag capacitors, and the old electrolytic capacitors:

Cleaning the glue in this board it was rather easy because it was softer, but required to remove the mylar capacitors that were also glued, cleaned them and soldered them back.

I used mica capacitors of the same capacitance in the locations where i found the black flag capacitors. In this unit there were two 5 pf and four 3 pF capacitors. The electrolytic capacitors were replaced by Panasonic FC and Nichicon KZ, for the output series capacitors I used a bipolar type 50uF 25V made by Elna placed at the C19 and C20 locations and a wire jumper at the locations C20, C22. Then I cleaned the flux on the conductor side and this is the end result for this board:


I checked the board to make sure there were no lifted tracks, solder bridges and dry joints. I reinstalled the flat amplifier onto the control board,and re-soldered jumper JP02, then mounted the joined boards and the volume potenciomenter to the front face of the chassis.
Just in case I powered the unit using a dim bulb test, no problems there. Then checked DC-offset upon full mains power up, followed by adjusting the DC-0V in the flat amplifier. I did not replace the single turn variable resistors at this stage, because i could not find a 470 or 500 ohms trimmer for side adjustment. But I could adjust the offsets to the allowed range in the manual.
The next task will be to remove the last Black flag capacitors from the head amplifier (F2835). I will post an update when I finish.

 

[Hyperion]

because i could not find a 470 or 500 ohms trimmer for side adjustment

https://www.mouser.co.uk/productdet...GAEpiMZZMvygUB3GLcD7n4W/LGLppVQSr8ovdhw%2bAA=

 

[JoseHH]

https://www.mouser.co.uk/productdetail/bourns/3296p-1-501lf?qs=sGAEpiMZZMvygUB3GLcD7n4W/LGLppVQSr8ovdhw%2bAA=

Thanks John, I will place an order soon, I am building a new list for digikey, but Mouser will probably carry these items too.
I was just reading two of your posts about checking for oscillation! Very clear and helpful. Thank you very much!!.
Those tests will be most important when i put the TO-3P output transistors.

 

[JoseHH]

Start at 1vac peak to peak or less square wave at say 1000hz watch for overshoot at the start or end of the pattern. If none observed work your way on up the frequency curve to at least 50khz if your scope and signal generator will handle it. Since you do not have original outputs yet there should be no oscillation but once you secure the fast ouputs this becomes a very important step.
It also is good bench practice to have the device under test on a current limiting device such as a dim bulb tester until you are sure the amp is stable.
Note the 50k upper limit, this series of amps will amplify signals at that level and even higher so we want to catch any poor behavior on infrasonic frequencies.

Hello Lee,

I am getting ready to do the square wave tests of this amplifier following your recommendations. I have also seen the post by John (Hyperion) where he explains how he does this, and these other two posts which I think are really useful.
post 143 in
http://audiokarma.org/forums/index.php?threads/restoring-an-au-x1.709469/page-8
post 172 in
http://audiokarma.org/forums/index....istor-replacements.684619/page-9#post-9498921

At the beginning I thought that I understood how to do it, but now I have some doubts and I wish to have some advice to prevent damages.
The first issue I have is that my amplifier is connected to a 220:100 volt transformer because the mains here are 220Volts. The unit powers on normally with full mains which is what I used to adjust the DC offset and bias. The amplifier works with speakers no apparent problems.
When I use the dim bub tester (100W bulb), it does its job at telling me if there is a major fault like a short, but the unit does not come out of protection.
Since the DBT can not be used, and I do not have a variac to limit the current, to do these tests. Should I do then with full mains?

In your post you did not mention if a dummy load should be connected at the output of the amplifier, so I just probe with the oscilloscope?

I can at frequencies between 500Hz and perhaps 150KHz and the oscilloscope that I will use has a 60MHz bandwidth (it is a basic digital scope from agilent).
Sorry for my long post
Thank you,

Jose

 

[Hyperion]

When I use the dim bub tester (100W bulb), it does its job at telling me if there is a major fault like a short, but the unit does not come out of protection. Since the DBT can not be used, and I do not have a variac to limit the current, to do these tests. Should I do then with full mains?

The only time you use a DBT is, if you suspect a fault, and need to safely make fault finding measurements with power applied (through the DBT).

OR You want to check that bias, and/or DC offset adjusts before doing the 'real thing' on full mains.

OR You have just made some major changes to components or wiring, and just need a sanity check before proceeding.

You never play music, or test tones through the amplifier when on DBT.

 

[Overundr1]

Well stated, I also use a standard sig gen not a sweep generator so yes to having to adjust things a bit as the frequency climbs. Not being as brave as John I have been known to hang a 200w bulb on my dbt for initial volume run ups to a couple of volts just in case I do have an amp try to run away under initial square wave testing before putting the unit on mains. Yes, you need to use those dummy loads for these tests. Remember what you are looking for is overshoot either at the beginning or tailing off at the end of the square wave pattern indicating there is a phase shift anomaly somewhere in the amplifier stage.

 

[Overundr1]

Don't remember if this has been mentioned. As an amp begins to (run away) or oscillate it will start to really heat up and draw current i.e. get hot in a hurry so if the heat sinks start to get toasty shut things down even if you cannot see any obvious pattern errors.

 

[JoseHH]

Hello John and Lee,

I have done some tests using a square waves and I see some overshoot. I wish to have your opinion about this.
Besides the Information provided by John (Hyperion) and Lee (Overundr1), I found this thread also useful to do this tests:
http://audiokarma.org/forums/index.php?threads/amp-testing-w-scope-function-generator.279830/

The condition of the tests were:
1-Amplifier: Au-D907, all electrolytic capacitors renovated with same capacitance values but higher voltage rating. All black flag capacitors replaced by mica capacitors of same values. Ouptput transistors ON Semiconductor mj21195/96. Drivers: MJE15030/31 Power supply boards with new electrolytic capacitors but original main capacitors. All new components sourced from digikey.
2-Load: 8.2Ohm (10 resistors 82 Ohm, 10W in parallel)
3- Input signal 1V peak to peak signal at the "Power Amplifier IN" function generator 50Ohm output, GW model GFG-8016G
4-Oscilloscope, Agilent DSO1002A. 10X probe compensated with the reference signal at the scope.

Results RIGHT CHANNEL:


Results LEFT CHANNEL:


There is a clear overshoot which becomes more prominent at 50KHz and higher frequencies.The amplifier went into protection when I took the picture of the signal at 75KHz on the right channel so I shut it off and took a look at it. After I checked that everything was fine, I did the tests for the left channel but I limited the maximum input frequency to 60KHz for this channel.
Both channels are similar. The voltage gain is nearly 30V/V. I did these tests quickly to avoid any heating. I think at frequencies higher than 50 KHz the heating started to be more noticeable but the heat sink never reached 35 degree Celsius, but each time the signal was applied for a few seconds. When I tested the amplifier with speakers I did not notice any distortion but I only played music at low volume. I can not play loud music here.

Any comments are welcome.

Jose

 

[Hyperion]

If I understand the tests you perform correctly, the test you performed is pretty severe. You shouldn't expect it to 'behave' at these frequencies (60/75 Khz) when delivering anything like full output. For this test 20Khz is really the maximum frequency, (at full power) and even that is too high for some amplifiers which are otherwise well behaved and sound good.

 

[JoseHH]

If I understand the tests you perform correctly, the test you performed is pretty severe. You shouldn't expect it to 'behave' at these frequencies (60/75 Khz) when delivering anything like full output. For this test 20Khz is really the maximum frequency, (at full power) and even that is too high for some amplifiers which are otherwise well behaved and sound good.

Thanks John, It is very nice to have your opinion.
Before trying 1 volt input, I also experimented with the unit using only 400 mV square wave input. Under these conditions, I can go over 100 KHz:



If I use 200 mV input square waves, the overshoot/undershoot seems a lot less severe. Before I did any of the square wave tests I checked with sine wave signals and i applied up to 2 volts input at some frequencies between 1kHz and 50kHz. Above that I restricted the input to one volt or less. With sine wave at 100kHz using 1.0 volt input I get nearly the same voltage gain that get at 1kHz:


I take this as an indication that the bandwidth is larger than 100 kHz. My function generator can go a bit higher but I am not sure it is safe to carry on a series of measurements at higher frequencies to make a plot of the gain versus frequency when using the set of transistor replacements I have. Any thoughts about this?

Thanks again,

Jose

 

[Overundr1]

Thought I mentioned that running square wave tests really needs to be limited in total voltage as it puts quite the load on the amp, the fact you were able to run the amp that far up the voltage scale before things started to tail indicates just how good Sansui got at building fast amplifiers. Impressive graphs, now I know what to look forward to when I get my own 919 back on the bench for the rest of the boards. As a side note, this may be the first time we have had full graphical representation of this series of amps, too cool
-Lee

 

[JoseHH]

Thought I mentioned that running square wave tests really needs to be limited in total voltage as it puts quite the load on the amp, the fact you were able to run the amp that far up the voltage scale before things started to tail indicates just how good Sansui got at building fast amplifiers. Impressive graphs, now I know what to look forward to when I get my own 919 back on the bench for the rest of the boards. As a side note, this may be the first time we have had full graphical representation of this series of amps, too cool
-Lee

Thanks Lee,
My interpretation was that i had to apply 1 volt peak to peak to the input. Did you mean that I should have applied a signal such that the output of the power amplifier is 1Volt or less?. If you read the post that John (Hyperion) quoted from @dr*audio, it really seems to be that the later is the case.
In any event, If you I input a 1 volt peak to peak signal, the output is circa 30 volts. My estimate was that this would correspond to 3/4 of full power for an amplifier rated at 100W RMS like the AU-919 or AU-D907 which is a sine wave output voltage of 40 volts (peak-to-peak) over an 8 Ohm load. I used:
Power_RMS=(V_RMS^2)/R
Looking the observations about my tests, it looks like the amplitude of 1 volt peak to peak for the test signal was too high.
Anyhow, like you say the bandwidth that the Sansui engineers achieved with their design is very impressive indeed.
But I am curious to find out what would be the bandwidth with the ON MJ21195/6 devices. I read lots of comments that they are vastly slower than the originals, but wouldn't It be nice to see an actual measurement?
What do you think about the overshoot/undershoot we see in the attached pictures in comparison to other units you have worked on?

Good luck with your AU-919. If you post your work, I look forward reading and looking at photographs if you upload some.

Thanks for your opinion.

Jose

 

[Overundr1]

1vac p-p was on output side during square wave test, input side typically is @1/2vac. My sig generator allows me to vary that input voltage and I find that at 3/4vac I am getting to the upper limit of most transistor pre-amps before they start to overload. Also worth a mention that most modern gear is more tolerant of higher input levels than the typical 70's stuff was as back then the high input sources were normally cassette or reel to reel format at 3/4 volt or so not including pro audio gear of course.
Your scope appears more accurate than my old analog one, however those pics are very favorable, the only time I have really run into overshoot issues was on one of the Pioneer's I was rebuilding a couple of years back that had to have modifications to the output drivers.