Re: problem uncovered
That's a happy end to the day then - much better than if you hadn't worked it out!
AndrewT said:Hi all,
I am the idiot you already knew I was. 😱
I transposed the Drivers.
The worst ones to get into to examine the types inserted.
Get around to unsoldering soon.
That's a happy end to the day then - much better than if you hadn't worked it out!
Re: problem uncovered
you're not the only one! it happned to me also when i tried using 92pu06/92pu56 on the VAS, i inserted them wrong way! i had these makeshift fuses made out of 100ohm 1/2 resistors and each time i powered up they burn, luckily the ouput trannies did not fry with them.
😀
AndrewT said:Hi all,
I am the idiot you already knew I was. 😱
I transposed the Drivers.
Get around to unsoldering soon.
you're not the only one! it happned to me also when i tried using 92pu06/92pu56 on the VAS, i inserted them wrong way! i had these makeshift fuses made out of 100ohm 1/2 resistors and each time i powered up they burn, luckily the ouput trannies did not fry with them.
😀
Re: problem uncovered
Ohhh well, that is a pain, but be happy you only reversed the drivers.... the first 10 transistor version I build had all the outputs reversed.
\Jens
AndrewT said:Hi all,
I am the idiot you already knew I was. 😱
I transposed the Drivers.
Get around to unsoldering soon.
Ohhh well, that is a pain, but be happy you only reversed the drivers.... the first 10 transistor version I build had all the outputs reversed.
\Jens
Hi Tony,
build your self a series light bulb tester.
It only needs a plug top, light bulb socket, outlet socket and a little bit of mains cable. It would be nice to fix these into the faces of a box to make it look professional (and safer).
Use it EVERY time you switch on after maintenance or first build.
I am expecting my reversed drivers to be undamaged (although the legs are too short to fit back into Leach).
build your self a series light bulb tester.
It only needs a plug top, light bulb socket, outlet socket and a little bit of mains cable. It would be nice to fix these into the faces of a box to make it look professional (and safer).
Use it EVERY time you switch on after maintenance or first build.
I am expecting my reversed drivers to be undamaged (although the legs are too short to fit back into Leach).
hi andrewT,
thanks, done that too, but the resistor-cum-fuse worked for me lots of times, i have yet to lose power trannies using that method.
as i recall this method was recommended by prof. leach himself, plus i set the amps idle current by monitoring voltage accross the resistors, when everything is ok, very convinient too...😀
thanks, done that too, but the resistor-cum-fuse worked for me lots of times, i have yet to lose power trannies using that method.
as i recall this method was recommended by prof. leach himself, plus i set the amps idle current by monitoring voltage accross the resistors, when everything is ok, very convinient too...😀
Hi,
both the Leaches are playing well following my mishap. No apparent damage.
Has anyone tried altering the input resistors (r17) or the Feedback resistor (r8,12 &13) to reduce the output offset? Mine are sitting at -20mV to -26mV. A little more at start up and falling slowly as they warm up to about 10Cdeg above ambient.
The Iq of the drivers is about 6mA, set by r51.
Are there any disadvantages in raising driver Iq by lowering r51's value to say 150R? It would hold the drivers a little longer in ClassA before switching off on alternate peaks.
both the Leaches are playing well following my mishap. No apparent damage.
Has anyone tried altering the input resistors (r17) or the Feedback resistor (r8,12 &13) to reduce the output offset? Mine are sitting at -20mV to -26mV. A little more at start up and falling slowly as they warm up to about 10Cdeg above ambient.
The Iq of the drivers is about 6mA, set by r51.
Are there any disadvantages in raising driver Iq by lowering r51's value to say 150R? It would hold the drivers a little longer in ClassA before switching off on alternate peaks.
AndrewT said:
Great news, well done!
Changing R8,12,13 will change the closed loop gain, alter the loop stability and overall not do you or the amp any good.
20mV DC on the output is pretty good, did you use matched devises and 1% resistors?
\Jens
Hi,
I tried to follow Leach's matching advice but the difference between the NPN and PNP gains made this completely impossible, so I settled on NPN matching and PNP matching, but this leaves base current and the resulting offset current out by a small mile, hence the offset in my AC coupled amp. It could and should be a lot nearer zero output offset, particularly at the low gain (r16=750R +30.4db) I chose for the first pair of PCBs. I plan to raise the gain to about +35db on my next set and this will probably double the output offset. My resistors are generally 1% but usually <0.5% without matching. I did match the Re resistors (about 0.2%) to enable investigation of different Iq in the output sets.
That's why I am asking if anyone has tried to reduce the offset by deliberate mismatch of input/feedback resistors. I think -mV offset requires reducing r17 to too low a value and the r8,12 &13 combination is a lot of experimentation to find the optimum.
I too did not want to upset the two poles (or are they zeroes?) in the alternative feedback paths until I know more about how the amp operates and responds to tuning.
I tried to follow Leach's matching advice but the difference between the NPN and PNP gains made this completely impossible, so I settled on NPN matching and PNP matching, but this leaves base current and the resulting offset current out by a small mile, hence the offset in my AC coupled amp. It could and should be a lot nearer zero output offset, particularly at the low gain (r16=750R +30.4db) I chose for the first pair of PCBs. I plan to raise the gain to about +35db on my next set and this will probably double the output offset. My resistors are generally 1% but usually <0.5% without matching. I did match the Re resistors (about 0.2%) to enable investigation of different Iq in the output sets.
That's why I am asking if anyone has tried to reduce the offset by deliberate mismatch of input/feedback resistors. I think -mV offset requires reducing r17 to too low a value and the r8,12 &13 combination is a lot of experimentation to find the optimum.
I too did not want to upset the two poles (or are they zeroes?) in the alternative feedback paths until I know more about how the amp operates and responds to tuning.
Andrew,
On the versions I have build I have about the same DC offset (20mV) with no parts matched and 1% resistors everywhere but the emitter-resistors in the output.
Why do you need less than 20mV? I have heard people saying that anything below 50mV is great... or did I miss the point?
Why do you need more gain? Can this not be done in the preamp stage rather than the output stage?
\Jens
On the versions I have build I have about the same DC offset (20mV) with no parts matched and 1% resistors everywhere but the emitter-resistors in the output.
Why do you need less than 20mV? I have heard people saying that anything below 50mV is great... or did I miss the point?
Why do you need more gain? Can this not be done in the preamp stage rather than the output stage?
\Jens
way i understood it, the drivers themselves are operated classA, that's the way leach designed the output stage. only the ouput power trannie is AB, i guess operating those drivers at slightly higher collector current is good but you better have good heatsinks on them too.
having output tranies with higher betas will also be a good idea imho.
Hi,
the drivers run at about 5mA to 6mA depending on voltage but mostly on the r51 value.
They work as a mini output stage supplying quiescent current through each other and none to a quiet load. As the load demands current the drivers send some to the output. On alternate half peaks each of the drivers takes a turn to supply the current and the partner driver gets less quiescent current. As the load current increses the drivers continue supplying the extra current until one or other driver has almost no quiescent current flowing through it. Just before output current reaches this stage the drivers will be running at nearly double Iq and nearly zero Iq. This is the end of the ClassA portion. As the driver stops supplying current the driver pair start to work in a sort of ClassAB supplying all their current to the output stage.
For my output stage with hFE of about 80 to 100, the drivers move into ClassAB at about 400mA to 500mA of load current. The output stage went into ClassAB way back at about 200mA to 300mA depending on what output Iq one set up.
If I have this right then the amp transitions into ClassAB at two levels:- when output current reaches 2*Output IQ and again when output current reaches hFE*2*Driver IQ. I am wondering if there could be any or even a slight difference in sound quality/technical performance if the driver Iq was increased.
BTW. I have already substituted ALL the BOM transistors for what I think are better versions, with particular attention to hFE and fT, even at the expense of more fragility. I do not use 4ohm speakers.
Have I got the circuit operation correctly described?
reference to fig13 shows Leach stating "Q14 and Q16 Do Not Cut Off" but equally it also shows NO MODULATION of driver current into the opposite current peak. I believe this implies that the driver supplies no current to the output stage after it uses up it's reserve of Driver Iq.
Looking again at the right hand half of the graph shows the driver current (Ie16) rising to about 15mA to 20mA for an output current (Ie18+Ie20) of about 1.2A. It appears that output hFE=60 to 80. This requires about 9V output into an 8ohm load. This is well short of the maximum ability of the amp and the simulation graph shows behaviour upto about 5W output rather than all the way to 120W output (Ie18 +Ie20=5.4A).
If the graph had shown the full range of currents I think you would appreciate even more, that the drivers have effectively stopped supplying current into the other half peaks at a very early stage in the hand over process.
Again have I got this right?
Feed back please!
the drivers run at about 5mA to 6mA depending on voltage but mostly on the r51 value.
They work as a mini output stage supplying quiescent current through each other and none to a quiet load. As the load demands current the drivers send some to the output. On alternate half peaks each of the drivers takes a turn to supply the current and the partner driver gets less quiescent current. As the load current increses the drivers continue supplying the extra current until one or other driver has almost no quiescent current flowing through it. Just before output current reaches this stage the drivers will be running at nearly double Iq and nearly zero Iq. This is the end of the ClassA portion. As the driver stops supplying current the driver pair start to work in a sort of ClassAB supplying all their current to the output stage.
For my output stage with hFE of about 80 to 100, the drivers move into ClassAB at about 400mA to 500mA of load current. The output stage went into ClassAB way back at about 200mA to 300mA depending on what output Iq one set up.
If I have this right then the amp transitions into ClassAB at two levels:- when output current reaches 2*Output IQ and again when output current reaches hFE*2*Driver IQ. I am wondering if there could be any or even a slight difference in sound quality/technical performance if the driver Iq was increased.
BTW. I have already substituted ALL the BOM transistors for what I think are better versions, with particular attention to hFE and fT, even at the expense of more fragility. I do not use 4ohm speakers.
Have I got the circuit operation correctly described?
reference to fig13 shows Leach stating "Q14 and Q16 Do Not Cut Off" but equally it also shows NO MODULATION of driver current into the opposite current peak. I believe this implies that the driver supplies no current to the output stage after it uses up it's reserve of Driver Iq.
Looking again at the right hand half of the graph shows the driver current (Ie16) rising to about 15mA to 20mA for an output current (Ie18+Ie20) of about 1.2A. It appears that output hFE=60 to 80. This requires about 9V output into an 8ohm load. This is well short of the maximum ability of the amp and the simulation graph shows behaviour upto about 5W output rather than all the way to 120W output (Ie18 +Ie20=5.4A).
If the graph had shown the full range of currents I think you would appreciate even more, that the drivers have effectively stopped supplying current into the other half peaks at a very early stage in the hand over process.
Again have I got this right?
Feed back please!
Hi Tony,
I have used 21C/W sinks and they barely run at over blood heat.
I guess Tc~=35degC to 40degC. They can easily afford to run an extra 50% dissipation and possibly even more if there is a benefit.
Q8 & Q9 share a 21C/W sink and Q21 has it's own 21C/W sink.
what temperature are your To220s/To126s running at?
I have used 21C/W sinks and they barely run at over blood heat.
I guess Tc~=35degC to 40degC. They can easily afford to run an extra 50% dissipation and possibly even more if there is a benefit.
Q8 & Q9 share a 21C/W sink and Q21 has it's own 21C/W sink.
AndrewT said:
Andrew, in prior versions Leach had a higher current in the drivers. He reduced it in v4.2 to keep them cooler. The value of this resistor was 120 ohms in version 4.1. If you use higher beta output transistors than Leach's version, the driver load will be less anyway. He also lowered the bias in the predrivers over the years. R27/R31 have gone from 240 -> 270 -> 330.
AndrewT said:
I have been looking at reducing the resistors to reduce noise, and to better match my input-transformer ideal load requirement of 10K. It's not that easy, since you have to also recalculate the values in the split feedback portion. It would be easier to try if you remove the split feedback portion, at least for experimentation purposes.
Here is a paper Leach wrote that can help you with these calculations:
W. M. Leach, Jr., "Feedforward Compensation of the Amplifier Output Stage for Improved Stability with Capacitive Loads," IEEE Transactions on Consumer Electronics, vol. 34, no. 2, pp. 334-338, May 1988.
I had a link to this paper at the IEEE site at work, but it doesn't seem to work at home.
Anyhow, your offset isn't bad. Mine was around 10mv with beta-matched transistors and resistors. One way of reducing the offset is to use high beta transistors to reduce the base currents. I've been looking at some low-noise Toshibas (2SA970, etc.) to do just that, although they have a different pinout. Self states that it's better to match the base currents of the input devices rather than Vbe for best offset.
AndrewT said:
unfortunately i have no way of checking temps as i do not have access to a temp meter, but just like you the drivers are really not that hot.
my bad, i may have misunderstood you, can you link the scheme you are reffering to so i can be sure that we are talking bout the same thing? is it the scheme jens made or is it leach original v4.5?
😀
Hi,
I am still on topic, meaning Jens' 3pair PCB and component values (well, nearly).
But Pooge's posts1374 & 1375 are enlightening.
I wonder just what voltages Leach was running at ( I assume about +-58V to +-60V) and what size sinks he was using if he felt that reduced temperatures were necessary. Tc and Tj around 40degC hardly seem to be a problem at the VAS, predriver and driver semis. These typical temperatures certainly don't need fixing purely on temperature considerations.
Now, if he had said it sounded better at the lower Iq, then I would be listening.
No feedback on my post1372 !! Read and argue.
I am still on topic, meaning Jens' 3pair PCB and component values (well, nearly).
But Pooge's posts1374 & 1375 are enlightening.
I wonder just what voltages Leach was running at ( I assume about +-58V to +-60V) and what size sinks he was using if he felt that reduced temperatures were necessary. Tc and Tj around 40degC hardly seem to be a problem at the VAS, predriver and driver semis. These typical temperatures certainly don't need fixing purely on temperature considerations.
Now, if he had said it sounded better at the lower Iq, then I would be listening.
No feedback on my post1372 !! Read and argue.
If I recall correctly, the predrivers used to get a bit toasty. He may have changed this component over the years, thus changing the bias. But I don't have the info on that right now.
Hi,
a thought (a rare occurance), did Leach specify To5 or To39 for the hotter Qs? His original design's are over 30year old.
a thought (a rare occurance), did Leach specify To5 or To39 for the hotter Qs? His original design's are over 30year old.
Andy -
Out of curiosity, what devices did you choose for hfe and Ft? I plan on using BC546C/BC556C turned around (higher hfe and have lots) for the TO-92s and MJE15034/35 (availability, but also higher hfe) but otherwise the recommend devices.
I think you are correct on rail voltages that Leach used. His predrivers were TO-5s that you couldn't get much heat sink on. That may have been the reasoning behind lowering their bias.
The drivers had room for adequate heat sinking in Leach's layout. No idea why he thought they were too hot. Perhaps he just played with it and found that there was no sound quality impact by lowering their bias.
Post 1372 seems to make sense, but I don't have the knowledge/experience to fully analyze. I guess we can ask you to experiment and report back whether changing the driver Iq has an audible effect. 😉
Out of curiosity, what devices did you choose for hfe and Ft? I plan on using BC546C/BC556C turned around (higher hfe and have lots) for the TO-92s and MJE15034/35 (availability, but also higher hfe) but otherwise the recommend devices.
I think you are correct on rail voltages that Leach used. His predrivers were TO-5s that you couldn't get much heat sink on. That may have been the reasoning behind lowering their bias.
The drivers had room for adequate heat sinking in Leach's layout. No idea why he thought they were too hot. Perhaps he just played with it and found that there was no sound quality impact by lowering their bias.
Post 1372 seems to make sense, but I don't have the knowledge/experience to fully analyze. I guess we can ask you to experiment and report back whether changing the driver Iq has an audible effect. 😉
- Status
- Not open for further replies.