Hi John,
as the J76 and K213 are recommended as complementary parts
I took a closer look to the datasheet. Maybe the the IRF610/9510
as driver are better because of the output characteristics,
but the J76 and K213 have a nice reverse transfer chapacitance.
And a lower Vgs.
On the other side using a MOSFET as source follower gives a lot
of local feedback which minimises the effects of different output
characteristics. I think, I shoud do some measurements about
this output configuration.
The advantage of combining a MOSFET as driver to a power bipolar
transistor is that there are no reverse effects from the output
load, transfer admittance change in class AB/B (crossover distortion)
or hfe changes of the output transistor to the VAS stages.
Marc-Oliver
as the J76 and K213 are recommended as complementary parts
I took a closer look to the datasheet. Maybe the the IRF610/9510
as driver are better because of the output characteristics,
but the J76 and K213 have a nice reverse transfer chapacitance.
And a lower Vgs.
On the other side using a MOSFET as source follower gives a lot
of local feedback which minimises the effects of different output
characteristics. I think, I shoud do some measurements about
this output configuration.
The advantage of combining a MOSFET as driver to a power bipolar
transistor is that there are no reverse effects from the output
load, transfer admittance change in class AB/B (crossover distortion)
or hfe changes of the output transistor to the VAS stages.
Marc-Oliver
I was holding out for more input, but this is what I can say at this time: The amp in the schematic is class A, because it has high standing current (without signal). Many suggestions have been to compromise the original amp with inferior output devices and LOWER standing current. This is bad! Why bother with an inferior design? There are newer active devices that can replace every device, that are superior to what was put in originally, about 30 years ago. I should hope so! Fets could replace virtually all of the bipolar devices, and even the driver stage could be removed, because it would not be necessary. Let's see if we can get some more responses that make sense, in this example.
Mr. Curl.
How about
a) removeing the output darlington and sub it with a couple of pairs of mosfets( IRF'S or Toshibas)
b) running more current thru the second stage
or even more radical
c) convert the second stage to a folded cascode with mosfet outputs, the whole thing would model as a two stage amplifier,
(The problem would be to get enough current thru the folded cascode to drive the outputs if we dispensed with the drivers.)
Regards,
Jam
How about
a) removeing the output darlington and sub it with a couple of pairs of mosfets( IRF'S or Toshibas)
b) running more current thru the second stage
or even more radical
c) convert the second stage to a folded cascode with mosfet outputs, the whole thing would model as a two stage amplifier,
(The problem would be to get enough current thru the folded cascode to drive the outputs if we dispensed with the drivers.)
Regards,
Jam
Can't see the schematic
MikeW,
Could you please e-mail me a slightly larger copy of the schematic you posted (or repost it in a larger format). I cannot get enough resolution out of the image to see the schematic's details to follow along with Mr. Curl's and everyone else's suggestions or analysis.
Thanks,
John Z.
jzimmerman@-remove_this_and_the_dashes-hgservices.com
MikeW,
Could you please e-mail me a slightly larger copy of the schematic you posted (or repost it in a larger format). I cannot get enough resolution out of the image to see the schematic's details to follow along with Mr. Curl's and everyone else's suggestions or analysis.
Thanks,
John Z.
jzimmerman@-remove_this_and_the_dashes-hgservices.com
jam said:
Hello Jam -
Personally, I would build it "as-is". That ciruit is wonderfully elegant in its simplicity. Besides, many consider the commercial version of that amp (called the Mark Levinson JC-3 or ML-3, depending on the vintage) to still be one of the best sounding solid-state amps ever. I believe they still command a respectable price on the used market.
There is also the aspect of historical context. When that schematic was published over 20 years ago, it was a rare glimpse into the ideas and concepts of one of the most brilliant and innovative audio designers ever. Building it "as-is" in kind of an homage would be very cool, in my opinion.
But if you really want to make a new design, I wouldn't mess with it all that much. I don't think you need to.
First of all, the driver stage acts as a folded cascode on the input FETs. You are already getting the extra bandwidth that a cascode would provide, by eliminating the Milller effect. Plus you get a bunch of current gain, so you don't have to run a bunch of current in the input stage.
About the only things I would *consider* doing are:
1) Adding more current gain in the output stage to reduce the loading on the drivers. This could either be done by adding another emitter follower to make a triple (read Leach's web site), or replace the first followers with FETs. This latter is much trickier than it seems at first glance, largely because all of the commercially available devices suck.
2) If you wanted to get really ambitious, you could get rid of the feedback loops. This is not a project for the faint of heart. You would probably spend a lot of time getting it to work right after that.
Just my opinion,
Charles Hansen
Mr. Curl,
This could be one great project.
How about some hints on biasing the diffs and folded cascode? I assume that for lowest distortion equal amounts of current has to flow thru the diff as the cascode, which means the tails of the diff amps would have to be be biased with current sources or are there some comp. fets that have a high enough Idss. I may be wrong here but would like to see your solution.
Another thought was to use a second stage much like Mr.Pass did in the A75 project.
http://passdiy.com/pdf/A75p1.pdf
Regards,
Jam
This could be one great project.
How about some hints on biasing the diffs and folded cascode? I assume that for lowest distortion equal amounts of current has to flow thru the diff as the cascode, which means the tails of the diff amps would have to be be biased with current sources or are there some comp. fets that have a high enough Idss. I may be wrong here but would like to see your solution.
Another thought was to use a second stage much like Mr.Pass did in the A75 project.
http://passdiy.com/pdf/A75p1.pdf
Regards,
Jam
jam said:
Hello Jam,
If you look at Curl's original schematic, you'll see that he has nearly the exact same "second stage" as Pass used in the A75. The only significant difference is that the two amps have swapped the inputs. Curl's amp inverts, Pass's doesn't.
By the way, the first time I saw this in print was in Borbeley's article in Audio Amateur around 1978 or so, called "A 60 Watt MOSFET Amp" or similar. Borbely credited a guy named Lender with that circuit, but Curl told me that Lender got the idea from him! Certainly Curl used that circuit before Borbely did.
Best regards,
Charles Hansen
I think Class A is for me as I am wanting to absolutely minimize distortion and I am willing to pay for the Electrical Power to do it (I am a BSEET working in the electrical power industry). 
-----------------------------------------------------------
I am a lover of Class A amps, having had several that sound much better than AB designs. In fact, I prefer them to tubes.
However, there is a problem. 100W/per channel gives off 300W as heat and they are just too hot for the summer or demand air conditioning which ruins the S/N. For many years I used them as winter amps, switching to heavily biased class AB in summer.
Perhaps you should consider a bias switch, allowing a standing 10-30W for hot conditions.
-----------------------------------------------------------
I am a lover of Class A amps, having had several that sound much better than AB designs. In fact, I prefer them to tubes.
However, there is a problem. 100W/per channel gives off 300W as heat and they are just too hot for the summer or demand air conditioning which ruins the S/N. For many years I used them as winter amps, switching to heavily biased class AB in summer.
Perhaps you should consider a bias switch, allowing a standing 10-30W for hot conditions.
Hi Charles,
You are probably correct on how the amp sounds. I have not heard the original. I have only listened to a ML 2 but the topology is much different.
You are right in that the second stage behaves like a folded cascode but does it and the diffs. behave as a single stage or is there another pole introduced that could cause problems?
I like your idea of no global feedback and I can vouch for the concept because of a certain preamp that I am familar with. I think this would show the full potential of the design. This would probably entail a larger output stage to keep the output impedence down.
Both you and Mr.Curl might have different ideas on the details of the changes to the amplifier but I see validity in both paths.
I have to admit that no global feedback idea has me thinking...........time to pull out the soldering iron.
Regards,
Jam
You are probably correct on how the amp sounds. I have not heard the original. I have only listened to a ML 2 but the topology is much different.
You are right in that the second stage behaves like a folded cascode but does it and the diffs. behave as a single stage or is there another pole introduced that could cause problems?
I like your idea of no global feedback and I can vouch for the concept because of a certain preamp that I am familar with. I think this would show the full potential of the design. This would probably entail a larger output stage to keep the output impedence down.
Both you and Mr.Curl might have different ideas on the details of the changes to the amplifier but I see validity in both paths.
I have to admit that no global feedback idea has me thinking...........time to pull out the soldering iron.
Regards,
Jam
jam said:
The open-loop output impedance of this amp is already pretty low. I think the reason for Charles' suggestion to beef up the output stage was to linearize the open-loop transfer function. If you don't like an output triple, you could omit the output stage NFB loop and leave the local loops from the driver stage back to the inverting input intact. I've heard good results from this setup in a roughly similar topology (a simplified A75).
BTW, I did hear the original amp (or whatever version of it was marketed by ML) in the late 1970s or early 80s driving DQ10s. Outstanding for the day, and a real motivator.
jam said:
Hello Jam,
The output impedance doesn't really depend on the number of devices. The output impedance of an emitter follower is simply 30 divided by the idle current in milliamps. So doubling the number of devices will halve the output impedance, but so will doubling the idle current in the existing devices.
Best regards,
Charles Hansen
Joe,
I had orginally thought of using mosfets for drivers but according to several people that was not a good idea, includeing Charles.
I suppose I could get enough current thru the second stage (20-40 mls.) I could drive a whole bank of mosfets.
I still think that getting rid off loop feedback is key. How about a Szkali pair for the outputs? It should be more stable if it does not have to work within a feedback loop.
Charles.
I ment with and without feedback.
Regards,
Jam
I had orginally thought of using mosfets for drivers but according to several people that was not a good idea, includeing Charles.
I suppose I could get enough current thru the second stage (20-40 mls.) I could drive a whole bank of mosfets.
I still think that getting rid off loop feedback is key. How about a Szkali pair for the outputs? It should be more stable if it does not have to work within a feedback loop.
Charles.
I ment with and without feedback.
Regards,
Jam
Hello Jam -
In my experience, the only MOSFETs that I would even consider using for drivers are the BUZ900P and BUZ905P from Magnatec. These are improved clones of the old Hitachi lateral MOSFETs designed for output devices. There are definitely some problems with using these, but it would be one option.
I don't like the Sziklai pair (also called CFP for Complementary Feedback Pair) because it has loop feedback around the two transistors. I'm personally not fond of the sound of feedback, but it may be worth a try.
Regarding the output impedance, my point was that (as long as thermal considerations were met) you could achieve the same results in an open-loop output stage from simply increasing the bias as you would by increasing the number of devices.
Let us know how your project turns out!
Good luck,
Charles Hansen
In my experience, the only MOSFETs that I would even consider using for drivers are the BUZ900P and BUZ905P from Magnatec. These are improved clones of the old Hitachi lateral MOSFETs designed for output devices. There are definitely some problems with using these, but it would be one option.
I don't like the Sziklai pair (also called CFP for Complementary Feedback Pair) because it has loop feedback around the two transistors. I'm personally not fond of the sound of feedback, but it may be worth a try.
Regarding the output impedance, my point was that (as long as thermal considerations were met) you could achieve the same results in an open-loop output stage from simply increasing the bias as you would by increasing the number of devices.
Let us know how your project turns out!
Good luck,
Charles Hansen
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.