Quick update... thinking some more on this.
The -8.8 volts on the IC being low will probably be compromising its performance so we do need to do something. It's running below its internal regulator drop out voltage.
What do the MJ outputs say exactly. Are they MJ21193G and 21194G. I am wondering if they are not the best examples... I'm not saying they are faulty or fake or anything... just the absolute minimum spec. CPC list MJ21193/4G's with a quoted gain of 75. I'm wondering if these are better specced than the ones fitted. You could fit them and find they were no different however so no absolute guarantee. I am surprised it seems so critical on this but it does...
It's a problem coming up with alternatives.
You could try 3.9k resistors in place of the 4.7k's which would give a couple of milliamps more drive... fractionally more than with the 27k's bridging them. 1 watt components would be advised although if you had 0.6 watt ones they would be OK medium term to see what it was like.
Or to give the same result having tried the 27k's now go a bit lower to 22k (which makes around 3.87k) and see if the bias adjusts smoothly to its correct value.
Reason for going a bit lower to 3.9k is to see if it brings the adjustment within the more linear part of the presets travel.
Remember to start with the bias on minimum whenever you change anything.
The -8.8 volts on the IC being low will probably be compromising its performance so we do need to do something. It's running below its internal regulator drop out voltage.
What do the MJ outputs say exactly. Are they MJ21193G and 21194G. I am wondering if they are not the best examples... I'm not saying they are faulty or fake or anything... just the absolute minimum spec. CPC list MJ21193/4G's with a quoted gain of 75. I'm wondering if these are better specced than the ones fitted. You could fit them and find they were no different however so no absolute guarantee. I am surprised it seems so critical on this but it does...
It's a problem coming up with alternatives.
You could try 3.9k resistors in place of the 4.7k's which would give a couple of milliamps more drive... fractionally more than with the 27k's bridging them. 1 watt components would be advised although if you had 0.6 watt ones they would be OK medium term to see what it was like.
Or to give the same result having tried the 27k's now go a bit lower to 22k (which makes around 3.87k) and see if the bias adjusts smoothly to its correct value.
Reason for going a bit lower to 3.9k is to see if it brings the adjustment within the more linear part of the presets travel.
Remember to start with the bias on minimum whenever you change anything.
Hi Mooly - they're just standard MJ21193/4s I think, as there's no letter on the end. They're definitely 'the real mccoy' too (been reading up quite a bit on fakes), but from what you say might not be ideal. Sound level does seem to get loud with very little volume turned, but perhaps that's because it was designed before digital sources with their 2v output were the norm.
Checking them from pics I took today, underneath the MJ21193/4 model numbers, they do say BM0233 and BM0231 (or could be 8M0233/1 - pics not too clear). I think that's just the date/manufacture code though?
Just before you posted I actually just finished setting the mV (21mV per channel after warm up) and nulling the DC offset.
Working fine at present, though running is at a gentle volume level for a few hours incase any nasties show up.
I'll definitely get some 22k resistors as you suggest, and will make sure they're 1 watters.
Checking them from pics I took today, underneath the MJ21193/4 model numbers, they do say BM0233 and BM0231 (or could be 8M0233/1 - pics not too clear). I think that's just the date/manufacture code though?
Just before you posted I actually just finished setting the mV (21mV per channel after warm up) and nulling the DC offset.
Working fine at present, though running is at a gentle volume level for a few hours incase any nasties show up.
I'll definitely get some 22k resistors as you suggest, and will make sure they're 1 watters.
No not late at all - and MANY thanks for checking out the mV reading. I take it yours has the original Hitachi output transistors? (Mine doesn't hence a few measurement anomalies).
If it's not too much trouble, you could take a voltage reading from the side of R340 and R390 (the sides connecting to the 100K adjustment pots). I imagine if you're running original output transistors you'll get a -8.8V reading...
Mucho thanks
If it's not too much trouble, you could take a voltage reading from the side of R340 and R390 (the sides connecting to the 100K adjustment pots). I imagine if you're running original output transistors you'll get a -8.8V reading...
Mucho thanks
No not late at all - and MANY thanks for checking out the mV reading. I take it yours has the original Hitachi output transistors? (Mine doesn't hence a few measurement anomalies).
If it's not too much trouble, you could take a voltage reading from the side of R340 and R390 (the sides connecting to the 100K adjustment pots). I imagine if you're running original output transistors you'll get a -8.8V reading...
Mucho thanks
Yup, all four devices are Hitachi branded! You want those readings relative to chassis ground?
*edit* am I reading this right? Sony used the ICs heatsink as a jumper?! Hilarious!
Last edited:
8.75V off of pin 14 on each channel (should be 8.8V)
-8.58V off of pin 4 on the right channel (should be -8.8V)
-8.62V off of pin 4 on the left channel (should be -8.8V)
Should I care that they're just a smidgen off?
*edit* one thing to note: the voltage hit 8.8V at the initial application of power, however once the voltage settled down to 8.75V, then the protection relay clicks-in.
-8.58V off of pin 4 on the right channel (should be -8.8V)
-8.62V off of pin 4 on the left channel (should be -8.8V)
Should I care that they're just a smidgen off?
*edit* one thing to note: the voltage hit 8.8V at the initial application of power, however once the voltage settled down to 8.75V, then the protection relay clicks-in.
Last edited:
That'd be great - mucho thanks
Looks like! Haha - good old fashioned form follows function engineering!
It's 1am in the morning over here so going to hit the sack - will check in again tomorrow (I mean today!). Cheers for your help!
Ah you've done them already haha!
They look correct to me - some of those carbon film resistors set the voltages too so they may well be a bit of variation (I intend to replace R306/307/356/357 with more accurate/modern resistors next week).
Anyway many thanks for those - confirmed that the original transistors are very much tied in with the voltage readings one gets (as Mooly stated a few posts back). It's an intriguing design, this amp!
Thanks a mil for your help, Sir!
Morning folks... some more thoughts on this.
The more I look at this design the stranger it becomes... everything is in the detail.
If you want to try the 22k's then any small wattage will be fine. It was only if you fitted 3.9k's that 1 watt were advisable. Its easy to work out... assume worst case and say there is 50 volts dropped it. Watts = V squared/R so thats 50*50 / 3900 which is 0.64 watt. For a 22k in parallel its 50*50 / 22000 which is 0.11 watt.
I think there could be another better way though.
If you look at the circuit around the output transistors there are small differences between the resistor values for the NPN and PNP devices. R326/327 and R317/318. That's very unusual to see that and the only reason would be to try and compensate for the difference in characteristics between the NPN and PNP output devices. We always call them complementary but they aren't really.
So this could be another possible route to improving this. I wondering whether increasing slightly R317 and 318 might have the desired effect. And that means leaving the original 4.7k's in place feeding the IC. Perhaps initially leave R318 at 22ohms and increase R317 to 10 or 12 ohm and again see what the effect on the current setting is. Always start with the pot on minimum current though. The MJ devices are more complementary than most... this was a big feature of them. R317/318 are effectively across the B-E of the driver devices. If we increase the value of these resistors then the drivers will be able to conduct slightly more for the same level of drive from the IC.
The more I look at this design the stranger it becomes... everything is in the detail.
If you want to try the 22k's then any small wattage will be fine. It was only if you fitted 3.9k's that 1 watt were advisable. Its easy to work out... assume worst case and say there is 50 volts dropped it. Watts = V squared/R so thats 50*50 / 3900 which is 0.64 watt. For a 22k in parallel its 50*50 / 22000 which is 0.11 watt.
I think there could be another better way though.
If you look at the circuit around the output transistors there are small differences between the resistor values for the NPN and PNP devices. R326/327 and R317/318. That's very unusual to see that and the only reason would be to try and compensate for the difference in characteristics between the NPN and PNP output devices. We always call them complementary but they aren't really.
So this could be another possible route to improving this. I wondering whether increasing slightly R317 and 318 might have the desired effect. And that means leaving the original 4.7k's in place feeding the IC. Perhaps initially leave R318 at 22ohms and increase R317 to 10 or 12 ohm and again see what the effect on the current setting is. Always start with the pot on minimum current though. The MJ devices are more complementary than most... this was a big feature of them. R317/318 are effectively across the B-E of the driver devices. If we increase the value of these resistors then the drivers will be able to conduct slightly more for the same level of drive from the IC.
Morning - hmmm very interesting. Sony really pulled out all the stops in designing this amp to get the best out of the devices available at the time. I know in more modern amps people often change to a more modern transistor without needing to worry about the rest of the circuit, but as you pointed out it really does seem to have been designed to fit these two like a glove, as it were.
I've got to do some work today so won't have a chance to try those changes for a day or so, but I'll give them a whirl sometime next week. It's all working fine today still and sounding OK, if a little cold and brittle sounding. Always seems to be that way after fitting new caps, which seems to disappear after a few hours use - the infamous 'burning in' thing perhaps? Noticed it with several DIY projects over the years.
Thanks for all your help/suggestions again Mooly - I really would be stuck with a nice looking (and rather huge!) paper weight if it wasn't for your posts/help!
I've got to do some work today so won't have a chance to try those changes for a day or so, but I'll give them a whirl sometime next week. It's all working fine today still and sounding OK, if a little cold and brittle sounding. Always seems to be that way after fitting new caps, which seems to disappear after a few hours use - the infamous 'burning in' thing perhaps? Noticed it with several DIY projects over the years.
Thanks for all your help/suggestions again Mooly - I really would be stuck with a nice looking (and rather huge!) paper weight if it wasn't for your posts/help!
Last edited:
P.S. I'm thinking is it worth coming up with all these fixes to get the existing output transistors working comfortably, or would it be better to take a more ground up approach and find a transistor which is better suited to the amp? (you mentioned the low gain of the MJs vs the higher gain of the Hitachi's a few pages back).
Ideally I'd love to get two pairs of the original Hitachi outputs, but I guess these are pretty much unobtainium now (aside from a few possible fakes on Ebay!). Even if I did find four I'd have no guarantee they were matched either as from what you said they had a wide gain window.
So, perhaps I should push the boat out (or at least blow up the dingy!) and get a better pair of outputs better suited to the characteristics of this design?
Having said all that it's sounding VERY good again this morning after spinning some vinyl for a couple hours (no longer 'frigid' sounding), but if I can get any extra performance out of it I might as well do it.
Ideally I'd love to get two pairs of the original Hitachi outputs, but I guess these are pretty much unobtainium now (aside from a few possible fakes on Ebay!). Even if I did find four I'd have no guarantee they were matched either as from what you said they had a wide gain window.
So, perhaps I should push the boat out (or at least blow up the dingy!) and get a better pair of outputs better suited to the characteristics of this design?
Having said all that it's sounding VERY good again this morning after spinning some vinyl for a couple hours (no longer 'frigid' sounding), but if I can get any extra performance out of it I might as well do it.
I still think I would try slightly raising the resistor values first perhaps... I really think that could be the key to this.
You may have to experiment somewhat and in small steps increasing both resistors and perhaps upward to nearer 33 or 47 ohm, but in small steps. Its a case of getting a feel for whats going on.
You may have to experiment somewhat and in small steps increasing both resistors and perhaps upward to nearer 33 or 47 ohm, but in small steps. Its a case of getting a feel for whats going on.
Hi Jaycee. Its certainly a bit of a puzzle and a bit more difficult not having the unit to actually work on and test. Its very odd for an output stage to be so dependant on device type, particularly DC characteristics.
Other than MJ15003/4 I can't really come up with much that is readily available. I was thinking MJ4502 and MJ802 but these have dropped off the radar from many suppliers.
Turn on volts of silicon junctions is pretty much an absolute but are these MJ's as fitted doped differently to try and make them more linear and complementary.
Other than MJ15003/4 I can't really come up with much that is readily available. I was thinking MJ4502 and MJ802 but these have dropped off the radar from many suppliers.
Turn on volts of silicon junctions is pretty much an absolute but are these MJ's as fitted doped differently to try and make them more linear and complementary.
Those old MJ's would be even worse at any rate. The MJ21193/4 is a TO-3 version of the MJL21193/4 (TO264) transistor and these are excellent audio transistors. Onsemi have better still, but they are in modern packages like TO-3P or TO-247. Faster fT might also cause a big problem in such an old circuit.
Better driver transistors might help in this circuit... but then we are getting further away from the original design. I would say live with the small performance difference TBH!
Better driver transistors might help in this circuit... but then we are getting further away from the original design. I would say live with the small performance difference TBH!
Thanks for your comments Mooly/Jaycee.
After giving it some thought over the weekend, I think I'll stick with the output transistors that are in there now - the unit sounds very nice so I think I should quit whilst ahead there rather than make a sideways move. I will however restore the 4k7s to stock, and try increasing R317 and see what affect that has on the voltages.
Spent most of yesterday evening replacing every electrolytic on the EQ and the control boards. Very fiddly job - glad that's done now. There were at least 5 caps that looked like they'd leaked slightly so the unit was obviously on its last legs with some of the originals caps. I also desoldered that huge diecast 4-gang ALPS volume pot and sprayed some contact cleaner into that to get rid of the crackle when using it. All back together again now and working fine - definitely sounding clearer now and the volume pot is working as it should.
Whilst going through my notes I remembered that I'd replaced the following electrolytic polarised caps with MMK non polarised types:
C311/312/361/362
Was it OK to do this, or should I have replaced with polarised types like the original design?
Thanks.
- John
After giving it some thought over the weekend, I think I'll stick with the output transistors that are in there now - the unit sounds very nice so I think I should quit whilst ahead there rather than make a sideways move. I will however restore the 4k7s to stock, and try increasing R317 and see what affect that has on the voltages.
Spent most of yesterday evening replacing every electrolytic on the EQ and the control boards. Very fiddly job - glad that's done now. There were at least 5 caps that looked like they'd leaked slightly so the unit was obviously on its last legs with some of the originals caps. I also desoldered that huge diecast 4-gang ALPS volume pot and sprayed some contact cleaner into that to get rid of the crackle when using it. All back together again now and working fine - definitely sounding clearer now and the volume pot is working as it should.
Whilst going through my notes I remembered that I'd replaced the following electrolytic polarised caps with MMK non polarised types:
C311/312/361/362
Was it OK to do this, or should I have replaced with polarised types like the original design?
Thanks.
- John
Those four caps are probably better specced than the originals. A small value like across a rail is used to provide a low impedance path to HF noise and hash. Large electroylitics in theory are better... larger the cap the lower its reactance (resistance to AC)... in practice it doesn't quite work like that due to real world electroylitics having significant self inductance which negates the desired effect. So a small cap is actually better and a film type probably best of all.
Be very interesting to see if the resistors can be tweaked to bring the current up. I think that would have real benefts in bringing the voltage up on the IC too.
Be very interesting to see if the resistors can be tweaked to bring the current up. I think that would have real benefts in bringing the voltage up on the IC too.
Thanks Mooly - figured they'd be OK but thought I'd best check with an expert first
Definitely - so only start with R317 first, right?
Do you think I'd get any mileage in upping the value of the following PSU caps (or does it make less difference with an SMPS?):
C801/805 (1000uF / 200v) - immediately before the SMPS from what I can see.
C802/803 (2200uF / 50v) - immediately after the SMPS.
- John
Definitely - so only start with R317 first, right?
Do you think I'd get any mileage in upping the value of the following PSU caps (or does it make less difference with an SMPS?):
C801/805 (1000uF / 200v) - immediately before the SMPS from what I can see.
C802/803 (2200uF / 50v) - immediately after the SMPS.
- John
Yes start with R317, just working on the one channel. Be prepared to experiment a little increasing the value in small steps and always retesting with the bias on minimum after each change.
I would tend to leave the PSU caps as designed, changing the values changes time constants and with some SPSU's that can be critical for correct and safe start up.
I would tend to leave the PSU caps as designed, changing the values changes time constants and with some SPSU's that can be critical for correct and safe start up.