Okay, so I’ve already found problems I have to work out. I’ve only tested the output stage. I need to go further back, but on the left channel that’s fairly close to each other I’m seeing appropriate voltage, a tad high though. 49.08 Vdc on the NPN collector, 24v on the emitters and 25, on the bases. The right hand channel on the other hand that is way off, 49v on the NPN collector, 47v on the emitters and 48v on bases.
I thought oh crap, the coupling caps are only rated for 35v and the right channel is going to have 47v on it. I check it and it has 47v on both sides of it. So I check the speaker output posts and sure enough I have 47v on the right speaker out and 24v on the left speaker out. Somewhere a mistake was made. It was with the few changes in the post above when it started having issues so at least my search is limited.
Dan
I thought oh crap, the coupling caps are only rated for 35v and the right channel is going to have 47v on it. I check it and it has 47v on both sides of it. So I check the speaker output posts and sure enough I have 47v on the right speaker out and 24v on the left speaker out. Somewhere a mistake was made. It was with the few changes in the post above when it started having issues so at least my search is limited.
Dan
Well I figured out the high voltage, went over my work and I must have desoldered the emitter of Q18
With that soldered back up I have the normal voltage in the right channel outputs. Now to check bias again.
Dan
With that soldered back up I have the normal voltage in the right channel outputs. Now to check bias again.
Dan
Alright, now with that repair done we are much much better. The channels are very well balanced, but I need to increase the bias. Here are the NPNs
Here are the four transistors.
What is causing the difference in the NPN vs PNP if their internal resistors are fairly well matched?
wg_ski, when you said shoot for 15 mA total, you meant per device or per channel? I’m about half way there on the NPN side. To increase the current what should I adjust? Decrease resistance across the bias diode? By how much? Or do I need to adjust the the diodes? Hopefully not, those are a bit of a pain lol.
I am still having an issue with the outputs though. I have about 24 Vdc going into each output coupling cap and on one channel I have 24v on the output and the other it drops to 19.5v. So still got to figure out that.
Edit: nevermind, I applied a load and the voltage is gone. Didn’t realize it could go up to full voltage without a load.
Dan
Here are the four transistors.
What is causing the difference in the NPN vs PNP if their internal resistors are fairly well matched?
wg_ski, when you said shoot for 15 mA total, you meant per device or per channel? I’m about half way there on the NPN side. To increase the current what should I adjust? Decrease resistance across the bias diode? By how much? Or do I need to adjust the the diodes? Hopefully not, those are a bit of a pain lol.
I am still having an issue with the outputs though. I have about 24 Vdc going into each output coupling cap and on one channel I have 24v on the output and the other it drops to 19.5v. So still got to figure out that.
Edit: nevermind, I applied a load and the voltage is gone. Didn’t realize it could go up to full voltage without a load.
Dan
Last edited:
To increase the biaising current you ll have to increase, step by step, the resistance in parralel with the bias diode,
if it s 220R then increase it to 240R if you have this value on hand, if that s not enough to get the right current
go to 270R and then to 300R, 330R and so on, of course power off the amp during the change.
if it s 220R then increase it to 240R if you have this value on hand, if that s not enough to get the right current
go to 270R and then to 300R, 330R and so on, of course power off the amp during the change.
The difference in current is the current in R89. This is the emitter current of Q15, which is a handful of mA. It does provide a little bias in the NPN side, even if the PNP is completely cut off. I’d try to bump that up to 15 or so mA, with a bit higher in the NPN. If you’ve got the darlingtons in there that would make sure there is some actual bias current in the outputs, not just the drivers. Even one miserable mA in the outputs (versus zero) makes a huge difference in the sound. Increasing that parallel resistor (a little at a time) is the proper method.
So far I’ve been going up in small increments, but not a huge changed, need to make bigger jumps. I’m up to 330 ohms from the 220 I had in there a the NPNs have just reached 8 mA, so an increase of one mA. I’m going to go for 470 next, skip 390. I am powering up on dbt.
Okay great. Well the outputs definitely have more than 1 mA. So for at about 8 on the NPNs and a little over 5 on the PNP.
Dan
I just want to make sure this is looking okay. I’ve gone from 330 ohm which was giving me 8mA of current for the NPNs and I am now up to 1000 ohm going in small increments and I have not just hit 9mA. Should the jumps in current be bigger than this? I’m not going to, but I kind of what to go to 2.2k next to see where that puts me. Going to 15 mA seems like I have a way to go.
Dan
Dan
I’m going stop here until I hear back. I’m up to 3.3k and after about 5 minutes of warm up it settles at about 9.2 mA. So going from 1k to 3.3k had increased the bias on the NPN by 200 mA. It’s like it’s peaking out. The PNP are at about 6.5 mA.
Dan
Dan
May ask that you sketch out the revised circuit from Q18 forward to the outputs with latest values, parts types, etc?
Many thanks!
Many thanks!
Certainly, I don’t have the best drawing program and using my finger.
I changed C80 since it’s right behind Q18, but it went to 100 pF.
I didn’t draw in D5, but that’s the four diodes in series, two 1N4001 and two 1N5819.
Q18 - 2N5551C, I believe Onsemi
Q19 - BD139, Onsemi
Q21 - MJ11015 Mospec
Q22 - MJ11016 Motorola
If I missed anything let me know.
Dan
Thanks for your sketch--- it looks just like what I might generate. 😉
There seem to be some current measurements that that don't seem consistent. If you have a scope, take a look to ensure there aren't oscillations that could confuse DC measurements. My instinct would be to keep the C80 cap at 1000 pF until bias is sorted out. Then experiment with making it smaller for improved distortion.
The current flowing through Q21 and Q22 should be about equal, except for about 2.4 mA flowing through R96 to bias Q18, Q19. So Q22 current should be about 2.4mA higher than Q21, and difference in currents larger than about 2.4mA are curious/suspicious. You can measure drop across R96 to confirm my estimate of bias current.
I suspect the combo of 2 silicon + 2 Shottky diodes slightly under-compensates the output Darlingtons, and that you may need to add a few ohms in series with the diodes to get the desired bias current. The shunting R112 may be necessary. Bias current of about 20mA or even a bit more seems reasonable.
There seem to be some current measurements that that don't seem consistent. If you have a scope, take a look to ensure there aren't oscillations that could confuse DC measurements. My instinct would be to keep the C80 cap at 1000 pF until bias is sorted out. Then experiment with making it smaller for improved distortion.
The current flowing through Q21 and Q22 should be about equal, except for about 2.4 mA flowing through R96 to bias Q18, Q19. So Q22 current should be about 2.4mA higher than Q21, and difference in currents larger than about 2.4mA are curious/suspicious. You can measure drop across R96 to confirm my estimate of bias current.
I suspect the combo of 2 silicon + 2 Shottky diodes slightly under-compensates the output Darlingtons, and that you may need to add a few ohms in series with the diodes to get the desired bias current. The shunting R112 may be necessary. Bias current of about 20mA or even a bit more seems reasonable.
Last edited:
Like I was saying earlier, it might take 3 PN junction and one schottky to get it where in needs to be. 2 and 2 was suggested so it didn’t overshoot the first time it was powered up. Only 8 mA in each darlington probably isn’t enough. It may sound ok like that but it will be noticeable when it is high enough. The internal Rbe’s set the minimum current needed, and those are fixed. The ratios are about 200:1, regardless of type. So if there is 10k between base and emitter in the first stage of it, there is about 50 ohms across the second “output” stage. You need to develop half a volt across that BE junction just to get the 2nd internal stage to crack on. That takes about 10mA, all of which will flow through the internal 1st stage. Once you overcome that 2nd vbe, the 2nd stage will finally get some bias and run AB. It sounds much better that way. If you were using a vbe multiplier with a pot, there would be a point in the rotation where the rate of increase vs. position suddenly seems to increase. That’s the 2nd stage kicking in. If you set it to 15-20mA, you re far enough past the point where this happens, and the whole output stage runs in AB.
That 2.4 mA imbalance is par for the course with a singleton input stage where it’s DC (and possibly AC) feedback comes off the speaker output. With split supplies and differential inputs the DC offset is low and that does not happen.
That 2.4 mA imbalance is par for the course with a singleton input stage where it’s DC (and possibly AC) feedback comes off the speaker output. With split supplies and differential inputs the DC offset is low and that does not happen.
Okay, to first start off I’ll go ahead and remove one of the schottkys and add a 4001. Go from there.
Dan
Dan
Remember to go back and start with a lower parallel R when you do that, and sneak up on it again.
Absolutely! I certainly remembered. I’m very close. I just finished up the new strings of diodes, three 1N4001 and one 1N5819. I started with 470 in parallel. Boy did that make a difference, just powered it up and let it warm for about 5 minutes. Sitting at about 17 mA on the NPNs and about 14.6 mA on the PNP.
Here are the NPN
And the PNP
Seems to be exactly 2.4 mA lower. Should I keep it there? Or take it up slightly more until the PNP hits 15 or just over 15?
Dan
It's fine as is. The bias current will be governed by the diode string and current through it, i.e. Q18,Q19 current.
According to wg_ski datasheet the darlingtons internal resistance that goes from base to emitter
of the power transistors is 40R, so at 15mA most of the current is actualy supplied by the drivers,
it can be increased to 25-30mA with care as once the power transistors kicks in the current
increase at a quite higher rate when increasing by a given amount the resistance that shunt the diode.
of the power transistors is 40R, so at 15mA most of the current is actualy supplied by the drivers,
it can be increased to 25-30mA with care as once the power transistors kicks in the current
increase at a quite higher rate when increasing by a given amount the resistance that shunt the diode.
The B-E resistors were running 20% high on these, so 50 ohms not 40, so the 2nd stage will kick in around 10-11 mA. Running somewhat higher won’t hurt anything, but once you start to get some bias in the 2nd stage it hits diminishing returns quick. 15 to 17 (in the PNP) is probably a good number.
At 15-17mA most of the distorsion is eliminated, but things will keep improving quite well up to 30mA, those devices optimum is at 50-60mA but i wouldnt go as high given the amp and an eventual positive temp coefficient, so 15-17mA should be enough, moreover due to the EF2 that surely did reduce distorsion by a 10 factor compared to an EF1, that s now a incomparably better amp
than what it used to be.
than what it used to be.
It seems like going higher would only benefit the amplifier, with little downside. Would you take the PNP up a little higher closer to 17 mA and the NPN to 19.5 mA? If you were doing it?
Dan