AU-999 driver board mods

[stereofun]

A few month ago I restored a very nice condition AU-999 including the pre-amp and bass mod. I have since tweaked around a bit and wanted to share. My main focus has been the driver board - the feedback loop and the differential pair. This would not have been possible without direct or indirect help from Leestereo, Kevzep and Trnsfrmr - Thanks guys!

During the original restoration The only deviations from original besides the pre-amp and bass eq mod, was a new rectifier and 2 x 100v/12K filter caps. Then a polypropylene input coupling cap as well as a higher voltage feedback cap. The output transistors were replaced with MJ21194's as only two originals remained - the two substitutes were significant slower than the MJ's, which are almost on par with the originals. The amp sounded very good.

Mod#1

A 2.2 uf Polypropylene at the C801 driver board position instead of the 1.0 uf stock mylar.

Result: No appreciable difference until the amp gets cranked up to moderate or high output. Here the 2.2 shows better performance in the highest notes, but only gives little to no benefit in the low end.

Mod#2

Replacing the C807 bipolar feedback cap to Nichicon Muse 100uf/50v. Replacing feedback resistor R819 with a .5w metal film type for better temperature coefficient and replacing it's paralleled ceramic cap on the back with a CoG type as well as the low pass ceramic C803. (thanks Leestereo)

Mouser part number for the ceramic low noise caps: 581-SR211A560JAR, 581-SR211A150JAR Here a link to a post about the benefits of C0G ceramic caps.

Result: Noticeable - cleaner sound.

Mod#3
The goal here is to get the current through the differential input pair within 1%. The mod works best if you upfront replace differential input pair, TR801/803 with a low-noise, high-gain, high band-with set of transistors matched to within 1% hFE - a suitable candidate could be ZTX795A's, but other would work too (check specs). If you want to keep the original transistors then that is ok, but you will likely spend time swapping resistors until you get even current. We will get to that.

As an option you can begin by replacing the current source resistor R845 with a 1% .5w metal film - again the goal is to have the most stable and low noise resistors in these critical positions - but the main components here are the diff. pairs emitter resistors, R811/813. Assuming you have already gain matched TR801/803, you now replace R811/813 with 1% .5W metal film types that you have matched even closer out of a batch.That's it - as a check you can measure how close you got the currents (see below)

Now, If you kept the original transistors you need to probe mV across the resistors and convert to mA (Amp on, no signal) to see if these two currents are within 1% - If it is not you can calculate which resistance you need to replace in one side to match the other. For this operation it will be a huge help to measure your batch beforehand and document them so you know the resistance of those that went in as well as the candidates for swap if currents needs adjustment (see picture below).

Result: Cleaner and more spatial sound. It is like the amp has gone audiophile. Prepare to hear more heavily breathing jazz musicians on old recordings

The difference in current through the resistors, converted from their voltage drop and their actual measured value, caused a 10% variation on the left side and 3% on the right. After the new resistors were in, the change in current was within 1% on each differential pair and also within 1% channel to channel.

For instance - Left channel, before mod.:

R811: Voltage drop 65mV / value 56.3 ohm = current 1.15 mA

R813: Voltage drop 64mV / value 56.8 ohm = current 1.27 mA

After new resistors in place the numbers were: 1.116 /1.121 for the currents

R845 is connected to the + rail, then pulls the current (EF) up through the emitter resistors R811/13 of the diff. pair Tr801/3
Here you can also follow the feedback going through R819 and the bipolar C807 to the base of diff. pair Tr803.

Here an interesting post where AK member trnsfrmr discusses the idea of balancing the diff. input pair. He suggests a more sophisticated method by using a constant current source in place of the current source resistor R845.

If you give it a shot, let me know what you think.

Cheers,


P.S for MOD#3 You really need to have these probes with a tiny hook at the end (here shown on a different board of course)

 

[pete_mac]

Nice work! I've got another 999 resto coming up in 3 months. I might just incorporate these tweaks too.

 

[Karl vd Berg]

Nice work indeed! :yes:

 

[BLAH BLAH]

Nice work! I've got another 999 resto coming up in 3 months. I might just incorporate these tweaks too.

I hope you do pete!...:thmbsp:

 

[stereofun]

I hope you do pete!...:thmbsp:

...hmm I think I got it :thmbsp: - Lets hope Pete does. I've updated the post with the part numbers for the low noise ceramic caps where two are replaced on each driver board.

----------------------

Did some more listening test. Yes, this differential tweak really does change the amp for the better. While all freqs. stands cleaner, you especially hear it at the higher notes - an area where the 999, lacked a bit resolution. Things like high-hats and guitar plugs now have more detail and life. On some of my re-mastered test cd's I pick up small details like transient distortions or tiny pops from the original master. There are no mercy on bad recordings and nothing but heaven on the good ones.

The goal have been to zero in on a few key areas, where it was worth going the extra mile. In short, the best components for the feedback as well as the differential pair - here both transistors and the resistors should be matched to less than one percent. I have also toyed around with different input coupling caps - 1uf, 2.2, 3.0 - Dayton, Mundorf and Solen - that seems to have very little influence and not where the effort should be put...at least not until the resolving power may or may not warrant it :scratch2:

 

[mikstick]

would these mods etc be applicable to the au888 i wonder

 

[SoundOfSound]

 

[stereofun]

@ at Milstick - Yes I would believe so as the config looks the same. You have the same differential pair - one side fed by the scaled down feedback from the outputs via R819, C807 to the base of TR803, and then on the other side of the pair you have the base of TR801 receiving input from from the pre-amp via coupling cap C801. You have the same emitter resistors R811/13 connected via current source R845.

I think there are improvements here because these early ss amps must have been some of the first Sansui amps to employ differential pairs ? They came right after the single rail/cap coupled topology so perhaps their first attempts were great in design, but required tighter tolerances than were feasible, or deemed important at the time ? It seems that in later designs great effort was put into balancing via diamond differential and even double diamond differential circuitry. At any rate, all we are doing here is to swap in some better spec'd parts at key positions in order to realize more potential of the existing design.

 

[roger2]

great post stereofun :thmbsp:

I am interested in the diff pair, NFB circuit components, etc. Haven't had time to experiment with it yet, but I have got some quality resistors for NFB (and C0Gs for the bypass) and also the resistors you refer to in Mod#3, for a couple of my project amps. Just hadn't realized how important tight matching of current was, thank you! No doubt I will be checking voltage drop across a lot of resistors...

 

[roger2]

I am curious...

Does adjusting VR801 have an effect on the current in R811 or R813?

 

[stereofun]

I have been pondering the same - before changing the resistors I set DC-offset, here I would as usual get a little fluctuation from -2mv to +2mv, not bad!. After replacing the three resistors, there was no change in offset, even though the current measured had changed. My reasoning is that VR801 - connected to the base - must change the current in Tr801, this to compensate for differences in gain between the pair, or some other imbalances in the supply - case in point many kenwoods don't have an offset adjust, so if you are outside specs, you have to match the pair. My pair was already matched, and starting out by dialing in the offset, the last variable would be the ER's - keep in mind here that the differences in currents are minuscule, so maybe those tiny variances would not translate in my subsequent dc-offset measurement.

I'll be honest here - I don't truly understand the interactions :scratch2: But I do know the balance of the current through the differential pair is critical - so on my end it is a search on what parts of the amp really matters and what can we do - some trial and error so far indicates, that it is not about chasing boutique caps - it is the feedback loop and differential pair that really makes an audible difference, and probably more so in earlier designs. Meanwhile if there are any hardcore techs out there, please chime in! and for those fellow tweakers - give it a shot and let me know if you hear the same substantial improvement Meanwhile I'll truck on and for good measure also replace the R807/09 with a 1% .5w metal film. I'll update if it makes a difference. Cheers.

 

[roger2]

.....

I'll be honest here - I don't truly understand the interactions :scratch2: But I do know the balance of the current through the differential pair is critical - so on my end it is a search on what parts of the amp really matters and what can we do ...... it is the feedback loop and differential pair that really makes an audible difference, .......

This pretty much sums up where I am too. Except that I have not yet had time to actually experiment on amps as you have done here. I agree it is best to focus on that parts that really matter...as opposed to, say, shot-gunning every resistor in an amp. I have dug through lots of old threads and have found several references to the importance of NFB components including this one, and Bob Cordell's book references the importance of the diff input pair and how it is fed.

I was reading a thread at diyaudo (posts 14, 24, 25) where one mod (of three) on a Rotel amp was to lower ripple on the power supply to the diff pair and VAS sections by adding RC filtering.

May I ask what MF resistors you are using and do you know what their temperature coefficient is ?

 

[gslikker]

For me all AC behavior is voodoo, more for the guys able to design.

Nevertheless for DC, I think the null offset potentiometers are designed in for worst case situations, which do not exist anymore when you pair transistors and resistors.

So myself, I just modify nulling circuits to lower the influence of the pot to have a nicer range.

As you can see on the schematic you just posted, the base resistance for the diff pair are 33k, leading to a situation that for DC the base resistance to the input transistor is higher because of the pot than the 33k from the feedback.

So lowering the pots resistance would make sense to me.

Anyway you having just a few millivolts offset is about the best obtainable.

I modified the circuit in my AU666 and even added a smoothing cap "just in case a pot taper can get noisy", the latter being a "guess" only. The 75 ohms resistor to ground severely lowers the pots sensitivity so there is a good tweaking range. Comments of people more educated than me are welcome since I want to learn from this also......

I also "guessed" base to emittor resistors on the output transistors could be useful as they appeared to have very high HFE, to force the drivers having comparable current flowing through them. Also on this one, comments welcome...

 

[EchoWars]

Fooling with the offset doesn't make any sense. If the diff pair transistors are decent enough, all you are sourcing from the offset circuit is a few microamps. And there's no detrimental effect on the amp by having it in there.

The best matched transistors will still leave you with some offset (since base current will never equal zero), a situation where the trimpot is nice to have. And a CCS in place of the resistor will not balance the diff pair, but does have other nice effects (on paper, with the math) on the front end.

 

[stereofun]

@ roger2 - thanks for the reading. In Bob Cordells "Designing Audio Power Amplifiers, page 266/67 he goes in depth about the feedback cap. The reason I bumped up the voltage in mod#2

As for the MF - I used Vishay/Dale for the R845 18K Ohm 100 ppm/c For the ER's: Xicon 56 Ohm 50 ppm/c. I reckon that better temperature coefficient is almost entirely due to metal film vs. carbon comb and not so much the bump from 1/4W -1/2W - in fact the ratings are identical in 1/4 package for the Xicon, yet in theory a higher rating of heat dissipating would indicate a better co-efficient all things being equal ?

@ gslikker - I am not there yet, as to modify the actual design. This makes me all Socratic, - the more I learn, the more I realize how much I don't understand :sigh:

@ EW - Any input from you is greatly appreciated :thmbsp:

 

[roger2]

Yes, you are right that the difference in TC is because they are metal film as opposed to carbon film.

I was surprised, when I went through and studied spec sheets, that some Vishay/BC metal films that I had bought had TC of 250ppm. So not all MF's are created equal in that regard. That number is almost in the neighborhood of carbon films, and using that particular MF would partially negate the main benefit of switching from CF to MF.

My favorite metal film resistors are Vishay/Dale CMF and CPF, which have non-ferrous end caps. And for a more budget conscious selection I have been buying TE Connectivity LR1 series which are .6w, 50ppm, 6.2mm long. Those are .15¢ and Mouser seems to have a wide range of values in stock.

 

[gslikker]

I am curious...

Does adjusting VR801 have an effect on the current in R811 or R813?

Yes it does, the two sides of the trimming resistor as seen from the wiper are more or less "parallelled" to some "virtual ground voltage", and have to be added to the 33k, is the way I see it.
So, if balancing is your goal so much that you are looking for exact resistors, exact VBE, HFE and currents, within one %, you would have to take this into account I think. At the moment any current deviation over the 33k+(something) would lead to a different voltage than current over 33k only, which is at the base of the second transistor?

 

[stereofun]

Lets keep it real simple:

1) Gain match your differential pair as a precursor. Adjust offset.

2) Replace R845 with a high quality 1% metal film
3) Replace R811/813 with a high quality 1% metal film, matched even closer (I bought 8 instead of 4, then picked two pairs that were near identical on my DMM)
4) Adjust offset
5) lean back and crank the amp. You should hear quiet a difference, even in my case where the amp was already completely rebuild, modded and sounded great by any measure.

That's it - give it a try.....

I just did a lot of measuring before and after to verify the changes - PITA, since the amp is on and space limited, and you need probes with tiny hooks (no difference between no signal vs a 1 khz test tone) The balancing is done in the selection of the components you replace with. Surely there are more elaborate schemes - but this worked really well for a few bucks, beating any boutique caps

 

[slimecity]

Great mod stereofun. Just completed it on my 999 and replaced C801 on the driver boards with a good poly cap while I was in there.

Better bass definition, better stereo imaging and definition overall IMO. The top end of well-mastered vocals also have extra "crispness" I think.

Wonderful, thanks!

 

[stereofun]

Glad it worked well for you - thanks for the feedback