No, it will not work..Wrong connection in VAS current mirror with 2SB716, input dif. amp emitors blocked to GND (short at AC, so it is no more dif. amp) , no emitor degeneration resistors in current mirror in input, so it will not be able to drive push-pull VAS. Uce for BCV61 with shorted C,B will be about 0,6V. How could this drive VAS to 13mA at 100R common emitor resistor?? And here are more mistakes..
Thanks for the welcome Bob, glad to have you help with my dilema. 
Also thanks Nigel and RN for the info.
I went through posts by John Curl, Bob Cordell and Nelson Pass and lots more by you guys and managed to process some of it (total door knob here).
My current integrated amp (Roksan Kandy K2 BT) has this chain in the power section: TI LF353 > Toshiba 2SC4793/2SA1837 > Toshiba 2SJ200/2SK1529. So JFET buffer > BJT driver > MOSFET output. I'm curious to understand why the designer of the circuit has chosen this combo, as in what are the pro's and con's. Would BJT > FET > BJT be a better solution for driving low impedance speakers compared to this one?
Also thanks Nigel and RN for the info.
I went through posts by John Curl, Bob Cordell and Nelson Pass and lots more by you guys and managed to process some of it (total door knob here). My current integrated amp (Roksan Kandy K2 BT) has this chain in the power section: TI LF353 > Toshiba 2SC4793/2SA1837 > Toshiba 2SJ200/2SK1529. So JFET buffer > BJT driver > MOSFET output. I'm curious to understand why the designer of the circuit has chosen this combo, as in what are the pro's and con's. Would BJT > FET > BJT be a better solution for driving low impedance speakers compared to this one?
You would have to wait for Linear Audio Volume 8, "Cube Law class A" by Ian Hegglun
.jan
Thanks for the welcome Bob, glad to have you help with my dilema.
Also thanks Nigel and RN for the info.
My current integrated amp (Roksan Kandy K2 BT) has this chain in the power section: TI LF353 > Toshiba 2SC4793/2SA1837 > Toshiba 2SJ200/2SK1529. So JFET buffer > BJT driver > MOSFET output. I'm curious to understand why the designer of the circuit has chosen this combo, as in what are the pro's and con's. Would BJT > FET > BJT be a better solution for driving low impedance speakers compared to this one?
I think there's no real answer to that. Everybody has their own favorite devices, be they tube, BJTs, FETs, MOSFETs, or whatever. All too often, this dictates their use in various products.
Obviously, the man ahd an idea and went about implementing it. It is very probable that he was well acquainted with those particular devices, which of course helped him in his work. But other factors also come to play, such as device availability, price, etc.
For very nasty loads, I believe nothing beats a good BJT, some of them are capable if incredible output currents. However, for input stages, I think FETs are a better choice. How it goes in the middle will also depend on the power required, on the current required and what you have in the output stage.
It is about 6.5 ma per side in the Hitachi original . The Hitachi was either 2 x 3K9 or 5K1 in place of the BCV61. As I said road map for output side. I spotted the the error 8 years ago and had forgotten to correct all files.I didn't correct this version. The 13 mA is notes to myself. The dia was 10 minutes before going to bed so not bad. At work we pass ideas like that all the time. The detail comes later. If you never did the same errors I need you to come work for me as we never met your type before. Seriously. It is a road map and will be very very different when working. You can bet your life and your family I won't publish it here. We keep somethings to ourselves. Ask Hitachi the rest, it worked for them. Funny things is when building things and they don't work it is usually not a mystery.
Why no comment on Fetlington ? It is the part of the map we were going to?
Last edited:
I see no one advantage for "fetlington" compared to simple EF output stage with bipolars. Nor linearity, nor usable voltage swing, nor thermal stability...And I suppose also stability issues.
And current mirror in input stage will not work without emitor degeneration resistors. You need get at least 1.5V-2V (as in Hitachi schematic) at BCV colectors referenced to -Ub, to get push-pull VAS stage working with reasonable currents.
That's why I reminded myself to add them with my little drawing below. In truth I only did that to say to other people I was thinking it. Myself I knew it. As you say not workable as shown. Just my notes to myself. I don't reject how Hitachi did it as it works very well. Note no component has a value in that section so is not a real design. Much as a car designer will sketch. The car fails as cd awful, back to the drawing board. The idea of BCV 61 is to say , can we usefully get better slewing with same stability and not use the 3K9 resistors. Some speculation as to ideal would be welcome, although it was just a doodle. As said I was off to bed and thought out of respect for the gentleman who asked the question I should make clear what the words mean without saying Fetlington. Fetlington is old UK TV term and not the one used by many.
We have a chance with this to have an IGBT device of our own making. Doubtless a real IGBT is a different animal. The advantage is the IGBT here has large feed forward current if we choose. Choice is nice. SOA is very good and makes the Bipolar happier to work as a switch. The horrible notch in the output when the Bipolar switches on is another matter and might be awful. Bob Carver said his amp sounded much better than he dared hope so did very little to correct it. Quad also.
The linearity of the output stage is secondary to the ability off the amp to correct it, I was tempted to say irrelevant. Many people seemed to be locked into an argument centered around 1967 devices and their problems. These problems changed over time but thinking didn't. The class D people have no choice as that will not help them to live the old world ideas. For all that much is like AB when D.
Anyway I hope Vlad ( ? ) was happy although I think he meant the other way around?
We have a chance with this to have an IGBT device of our own making. Doubtless a real IGBT is a different animal. The advantage is the IGBT here has large feed forward current if we choose. Choice is nice. SOA is very good and makes the Bipolar happier to work as a switch. The horrible notch in the output when the Bipolar switches on is another matter and might be awful. Bob Carver said his amp sounded much better than he dared hope so did very little to correct it. Quad also.
The linearity of the output stage is secondary to the ability off the amp to correct it, I was tempted to say irrelevant. Many people seemed to be locked into an argument centered around 1967 devices and their problems. These problems changed over time but thinking didn't. The class D people have no choice as that will not help them to live the old world ideas. For all that much is like AB when D.
Anyway I hope Vlad ( ? ) was happy although I think he meant the other way around?
The Fetlington and the IGBT would both be nice to see in SIM and optimized for audio apps. Compare SIM results with 2EF OPS and Sziklai. However ---
It is not just static distortion numbers to compare but the benefits of both in one OPS (#13379). Low drive requirements and high current output for one. [yes might need to drive it with a EF for lowest distortion] And, the listening results compared with all FET or all BJT OPS. Will it have the characteristic performance of both?
THx-RNMarsh
Last edited:
What I think I might find is the feed forward resistor might be quite large. My reasoning is this. Up to a point negative feedback will fight to maintain a low output impedance. If not the hideous Ron of MOS FET's would be an insurmountable problem. My feeling was 0R33. who is to say it might not be much higher? This would also reduce the heat in the FET depending on choice. If so the Ron would be lower and possibly we win more than we loose. Setting the feed forward high might mean the dumpers only come on when silly requirements. That might be twice a week for me and twice an hour for Mr Wayne. That might make it far more like class G than class AB+C. Not in exact working principle, just switching spikes. I could even switch off the dumpers if it suited me.
My picture was exactly win win win. The Bipolar might be very cheap very fast devices working well inside SOA. I think people over estimate the cg of MOSFET's as a problem. If a Bipolar it might be the killer blow. It isn't. I warn people not to assume it is a bad thing. Try increasing drive current above 5.6mA. See if you like it. Marginally I didn't. 2SB716 can run 12 mA without distress at 55 V. Some special Sanyo To126 MOS FET drivers prehaps 18 mA and more with a clip on heat sink. BTW 1N4007 is good enough to make that mirror.
I really do suspect we can run the feedback we need as these devices will oblige us. Also the sky is the limit when driving the dumpers. A 2N3055 has a gain of 5 into 1 R. My guess is that is ball park for all bipolar devices. Should a speaker present that load this amp will say no problem even with the drawing design. Give it four bipolar and it should be OK for anything. Dumper base series resistor needs careful choice. It might be zero as the Ron might protect it.
http://www.onsemi.com/pub_link/Collateral/NJW0281-D.PDF
My picture was exactly win win win. The Bipolar might be very cheap very fast devices working well inside SOA. I think people over estimate the cg of MOSFET's as a problem. If a Bipolar it might be the killer blow. It isn't. I warn people not to assume it is a bad thing. Try increasing drive current above 5.6mA. See if you like it. Marginally I didn't. 2SB716 can run 12 mA without distress at 55 V. Some special Sanyo To126 MOS FET drivers prehaps 18 mA and more with a clip on heat sink. BTW 1N4007 is good enough to make that mirror.
I really do suspect we can run the feedback we need as these devices will oblige us. Also the sky is the limit when driving the dumpers. A 2N3055 has a gain of 5 into 1 R. My guess is that is ball park for all bipolar devices. Should a speaker present that load this amp will say no problem even with the drawing design. Give it four bipolar and it should be OK for anything. Dumper base series resistor needs careful choice. It might be zero as the Ron might protect it.
http://www.onsemi.com/pub_link/Collateral/NJW0281-D.PDF
It's bed time again and I have been doing speakers again so just a road map guys and yes I do make mistakes.
This is the IGBT as far as I understand . Douglas Self book of spells says in all ways comparable to the better type of all NPN output stage used by Naim with Baxendale correction diode. Much better symmetry and looks very good into low resistance loads. Still favour Fetlington myself. the distortion of normal MOS FET is disgusting at about 0.5%. It's ability to work with high feedback seems to more than compensate. Do not take 27pF VAS comp as usable here.
IGBT is 0.05% and EF 0.03% CfbP 0.014%. Quasi 0.07% Baxendale 0.05%.MOS SF 0.5%. Fetlington ??
I somewhat optimistically assumed the Self example was this IGBT. No it is bipolar input. My instinct is to think it not so different. He would not like it as bias point is not a defined voltage. In my book 4 I don't think it gives this detail.
http://www.embedded.com/print/4015850
Last edited:
Nige, that seems to be ON Semi's version of Toshiba's 2SC5200/2SA1943.
Whuile I like Toshiba's version, and have not tried ON Semi's, "just about everything" may be right only if you have enough pairs lined up. And no, I don't give a hoot that some manufacturers (mis)use it for more cheap power.
When it heats up to 85 deg C, it will deliver only 80W of useful power. More only if you want to fry eggs on it.
On the other hand, they are very affordable, so using multiple pairs should not be too much of a problem. For example, I wouldn't dream of making anything at 100W/200W into 8/4 Ohms with less than 3 pairs, and that only in a pinch.
You will remeber than in another thread, Aniket used 4 pairs to obtain 250W/4 Ohms, and you may also remember that I told him that was a close call because, as I understood it, he was planning to really run that thing at very high power levels, and he never mentioned forced air cooling. To no avail, I'm afraid.
Whuile I like Toshiba's version, and have not tried ON Semi's, "just about everything" may be right only if you have enough pairs lined up. And no, I don't give a hoot that some manufacturers (mis)use it for more cheap power.
When it heats up to 85 deg C, it will deliver only 80W of useful power. More only if you want to fry eggs on it.
On the other hand, they are very affordable, so using multiple pairs should not be too much of a problem. For example, I wouldn't dream of making anything at 100W/200W into 8/4 Ohms with less than 3 pairs, and that only in a pinch.
You will remeber than in another thread, Aniket used 4 pairs to obtain 250W/4 Ohms, and you may also remember that I told him that was a close call because, as I understood it, he was planning to really run that thing at very high power levels, and he never mentioned forced air cooling. To no avail, I'm afraid.
It was just flavour of the day. I suspect they are not very critical. That's the dream here. Bang out the current and do the job. Can you imagine. Re = >0R and totally safe if a bit of care taken.
The Exicon FETs are about $8 and the Bipolar about $2 in 100's.
The LTP emitter re is 50 R when only 500 uA. 26/ic mA. I always use 47R in real life. I also use a tail CCS as it is no big deal. It measures no better and my test gear is not bad.
Question to the world. If you increase tail current did you find you lost the advantage in needing larger VAS caps. I seem to and also get some DC offset for my pains. The transistors I have in stock are very high gain so the values here might not be possible today. A friend bought them, then never did the project. I must have 1000 ? I suspect most of these you will buy are fakes. 2n5551/5401 in a big package. OK up to a point.
The Exicon FETs are about $8 and the Bipolar about $2 in 100's.
The LTP emitter re is 50 R when only 500 uA. 26/ic mA. I always use 47R in real life. I also use a tail CCS as it is no big deal. It measures no better and my test gear is not bad.
Question to the world. If you increase tail current did you find you lost the advantage in needing larger VAS caps. I seem to and also get some DC offset for my pains. The transistors I have in stock are very high gain so the values here might not be possible today. A friend bought them, then never did the project. I must have 1000 ? I suspect most of these you will buy are fakes. 2n5551/5401 in a big package. OK up to a point.
Last I checked the only audio amplifiers for which on resistance is a figure of merit were switching amps. Linear amps are clipping when you hit the on resistance range.
D Self has a lot of great insights but doesn't seem to step far out of his comfort zone to explore the other options enough to see a benefit from them. There are other solutions that give good results by his metrics and don't necessarily cost more.
Except for very high end work today I use switching/digital amps with DSP for commercial projects. There is no economic reason not to and the performance can be really good without much effort. Real concerted effort is needed to outperform what you can do with two TAS5711's direct driving good drivers. That's a BOM of about $40 for the electronics and drivers for a good stereo speaker system. Designing them as a system will always yield better results than designing pieces in isolation.
D Self has a lot of great insights but doesn't seem to step far out of his comfort zone to explore the other options enough to see a benefit from them. There are other solutions that give good results by his metrics and don't necessarily cost more.
Except for very high end work today I use switching/digital amps with DSP for commercial projects. There is no economic reason not to and the performance can be really good without much effort. Real concerted effort is needed to outperform what you can do with two TAS5711's direct driving good drivers. That's a BOM of about $40 for the electronics and drivers for a good stereo speaker system. Designing them as a system will always yield better results than designing pieces in isolation.
Dejan kicked it off by asking about the JLH work on IGBT. Looking up the common devices it seemed they never happened. Rather sad as JL Hood had made a good case for them. My instinct is the D Self Lingtonfet if you like and HBIGBT ( Home-brew IGBT) take us one step forwards and two steps back.
It might be a bit imaginative of me to think a home-brew IGBT using L FET can equal the Douglas Self version or Lingtonfet. If I do imagine that the bipolar must work hard to be dominant. I call it book of spells as there are great examples. I have pulled rabbits out of hats using that book.
I think the Fetlington might work well with 0R1 feed-forward. I will stick my neck out and say no bipolar input resistor and rely on Ron to prevent base punch trough. The feed-forward resistor will also speed the operation of the bipolar bass as it is almost a short circuit when switched off. That switching off might be at 50 kHz+.
If you have never built with special Audio FET's it is the world you always wanted. You set the bias by ear. Even zero bias works unusually well. As the dumpers do the hard work in a Fetlington it might be possible to go a bit mad with that. Usually if you do the sound improves at HF and looses at LF when conventional source follower. 0R1 might be ideal for FF and allow much greater range of bias options. If the nasties where together or the FET gain doubling higher all the rubbish would be when ears start to bleed and we should stop anyway. If doing PA at the local barn dance we might welcome the raw power. Sure as eggs are eggs the speakers will be public enemy no one anyway if that loud.
Commonsense says the VAS compensation will be higher than the Hitachi original. It might be less than guessed as the FET's will be controlling the Bipolars better than some nice driver transistor is my guess. The standing current at 100 mA might be helpful. My first thought of 0R33 is too high as we are not to 1 watt at 4R before kick in. 0R1 seems about right. It also makes a more compact resistor possible. If the load becomes ridiculous the FET can even assist a little. I wonder if the FF resistor at 0R1 will smooth the switch on a bit, I suspect it will? It is hard to imagine something that hard shunted being naughty.
One thing I suspect will happen is that the slightly overly friendly sound of FET's will vanish and that's no bad thing. People think it is cg ( gate capacitance). Not really, it is Ron going up like a sky rocket. If Bipolar it would have died so not so bad.
My view is the Fetlington is like a modern diesel engine. With proper care it is better than it should be. Toyota made a better Fetlington diesel . The take off point for diesel is about 14 to 1 compression, if a supercharger is used that can work. That's the bias if you like. Methanol was injected at low revs and fired with a conventional spark plug. Methanol resists detonation at 14 to 1. The result was a smooth engine with minimal consumption of the cheap methanol that could run without it. Then the killer blow. Diesel engines will never completely burn the fuel. It lubricates the upper cylinder and results in long life. Methanol tends to ruin engines. The result was an engine of slightly poorer wear characteristics. It was felt the pro diesel buyers would instantly reject it. Now you might say they should have realized that? That is a bit worse than the mistakes I make. And it is not a mistake to try ( read that as you best please, means dreaming is key to invention). Those dam bumble bees will never fly, maths has proved it. Sometimes things live on the edge and do work. Supersonic flight is full of that. The will to do it makes it work. One man I worked with and was my engineer said " I could have been rich if I had the ideas". Problem is he thought it was pure luck. My boss came up to me and said " did I just hear what I thought I did ". Some golf player said he noticed his luck improved if he practiced. Peter designed the now universally used shut off valves for the oil industry so wasn't such a bad engineer. That is resisting 150 bar now 450 bar. The seats are normal Stellite to high precision. That's the difficult part as the material is hard. The main problem is not to crack under that load. It was after the Piper Alpha disaster.
It might be a bit imaginative of me to think a home-brew IGBT using L FET can equal the Douglas Self version or Lingtonfet. If I do imagine that the bipolar must work hard to be dominant. I call it book of spells as there are great examples. I have pulled rabbits out of hats using that book.
I think the Fetlington might work well with 0R1 feed-forward. I will stick my neck out and say no bipolar input resistor and rely on Ron to prevent base punch trough. The feed-forward resistor will also speed the operation of the bipolar bass as it is almost a short circuit when switched off. That switching off might be at 50 kHz+.
If you have never built with special Audio FET's it is the world you always wanted. You set the bias by ear. Even zero bias works unusually well. As the dumpers do the hard work in a Fetlington it might be possible to go a bit mad with that. Usually if you do the sound improves at HF and looses at LF when conventional source follower. 0R1 might be ideal for FF and allow much greater range of bias options. If the nasties where together or the FET gain doubling higher all the rubbish would be when ears start to bleed and we should stop anyway. If doing PA at the local barn dance we might welcome the raw power. Sure as eggs are eggs the speakers will be public enemy no one anyway if that loud.
Commonsense says the VAS compensation will be higher than the Hitachi original. It might be less than guessed as the FET's will be controlling the Bipolars better than some nice driver transistor is my guess. The standing current at 100 mA might be helpful. My first thought of 0R33 is too high as we are not to 1 watt at 4R before kick in. 0R1 seems about right. It also makes a more compact resistor possible. If the load becomes ridiculous the FET can even assist a little. I wonder if the FF resistor at 0R1 will smooth the switch on a bit, I suspect it will? It is hard to imagine something that hard shunted being naughty.
One thing I suspect will happen is that the slightly overly friendly sound of FET's will vanish and that's no bad thing. People think it is cg ( gate capacitance). Not really, it is Ron going up like a sky rocket. If Bipolar it would have died so not so bad.
My view is the Fetlington is like a modern diesel engine. With proper care it is better than it should be. Toyota made a better Fetlington diesel . The take off point for diesel is about 14 to 1 compression, if a supercharger is used that can work. That's the bias if you like. Methanol was injected at low revs and fired with a conventional spark plug. Methanol resists detonation at 14 to 1. The result was a smooth engine with minimal consumption of the cheap methanol that could run without it. Then the killer blow. Diesel engines will never completely burn the fuel. It lubricates the upper cylinder and results in long life. Methanol tends to ruin engines. The result was an engine of slightly poorer wear characteristics. It was felt the pro diesel buyers would instantly reject it. Now you might say they should have realized that? That is a bit worse than the mistakes I make. And it is not a mistake to try ( read that as you best please, means dreaming is key to invention). Those dam bumble bees will never fly, maths has proved it. Sometimes things live on the edge and do work. Supersonic flight is full of that. The will to do it makes it work. One man I worked with and was my engineer said " I could have been rich if I had the ideas". Problem is he thought it was pure luck. My boss came up to me and said " did I just hear what I thought I did ". Some golf player said he noticed his luck improved if he practiced. Peter designed the now universally used shut off valves for the oil industry so wasn't such a bad engineer. That is resisting 150 bar now 450 bar. The seats are normal Stellite to high precision. That's the difficult part as the material is hard. The main problem is not to crack under that load. It was after the Piper Alpha disaster.
Last edited:
Nige, it's not just what drives the output stage, it's ALL of it.
It's not at all the same thing at what current is your VAS running, whether that's say 6 mA or 14 mA. There is a reason why the amps of old sounded more muscular than most of the moder weenies. Way back then, NOBODY, not even the Japanese, ran their VAS as anythuing less than 11 mA, as far as I have looked. Just take a peek at their highest rated models at no small amounts of money.
If memory serves, Otala/Lohstroh ran their VAS at something like 20 mA or so. When you have that much at the base of your driver transistor, you have no reason to expect it will not be able to drive the output stage into oblivion if necessary.
By such currents, the transistors in those circuits are at their best, delivering everything they are capable of. Then, using a FET driver will not make half as much of a difference as they do these days of reatively low current values preceeding the output stage. In other words, it has no incoming juice to work with. So when it meets a difficult load, it starts to choke and falls into deep current starvation catatonia.
By and large, the Amerians had that pinned down just as they should. That is, I believe, one of the key reasons most of their amps tended to sound big and meaty, almost palpable, most of the time. That, and the fact they were not frugal with the output stages.
I do not question the quality of the drive circuits, obviously the whole will work better with better driver circuits, but I also believe a well proportioned triple, or the Locanthi model, is still the best we have to this day, even if some other topologies also produced some memorable models. I don't think it's so much the gain as the reduced output impedance, always good for the "natural" (i.e. open loop) damping factor and current delivery.
I remain unconvinced by MOSFET output stages, even if they too have come a long way since way back then. For the sheer force of it, BIJT still rules supreme, in my book.
But on the input side, I would rather use FETs than BJTs. Rock steady and with small DC offset, mostly.
It's not at all the same thing at what current is your VAS running, whether that's say 6 mA or 14 mA. There is a reason why the amps of old sounded more muscular than most of the moder weenies. Way back then, NOBODY, not even the Japanese, ran their VAS as anythuing less than 11 mA, as far as I have looked. Just take a peek at their highest rated models at no small amounts of money.
If memory serves, Otala/Lohstroh ran their VAS at something like 20 mA or so. When you have that much at the base of your driver transistor, you have no reason to expect it will not be able to drive the output stage into oblivion if necessary.
By such currents, the transistors in those circuits are at their best, delivering everything they are capable of. Then, using a FET driver will not make half as much of a difference as they do these days of reatively low current values preceeding the output stage. In other words, it has no incoming juice to work with. So when it meets a difficult load, it starts to choke and falls into deep current starvation catatonia.
By and large, the Amerians had that pinned down just as they should. That is, I believe, one of the key reasons most of their amps tended to sound big and meaty, almost palpable, most of the time. That, and the fact they were not frugal with the output stages.
I do not question the quality of the drive circuits, obviously the whole will work better with better driver circuits, but I also believe a well proportioned triple, or the Locanthi model, is still the best we have to this day, even if some other topologies also produced some memorable models. I don't think it's so much the gain as the reduced output impedance, always good for the "natural" (i.e. open loop) damping factor and current delivery.
I remain unconvinced by MOSFET output stages, even if they too have come a long way since way back then. For the sheer force of it, BIJT still rules supreme, in my book.
But on the input side, I would rather use FETs than BJTs. Rock steady and with small DC offset, mostly.
Last edited:
Bateman capacitor articles
For those of you interested in 'Capacitor Sound' or lack thereof, I have posted Cyril Bateman's landmark article series on my personal website.
Thanks to Cyril and the Editor at Electronics World for making this possible!
Enjoy!
Jan
For those of you interested in 'Capacitor Sound' or lack thereof, I have posted Cyril Bateman's landmark article series on my personal website.
Thanks to Cyril and the Editor at Electronics World for making this possible!
Enjoy!
Jan
Last edited:
- Status
- Not open for further replies.