Hi, Ian, thanx. I'm familiar with the later Mosfet version but not this one. Its nice to know that he had tried this topology with bipolar too, so I will try to design one too, with bipolar.
But when I saw and built my amps I didn't notice/realize that the FB and the input RC was connected to the mid-supply point, which I think is not possible. But I see from the list of JLH publications on the site you attached, there is the 4th series titled: "A Simple 30 Watt Integrated Amplifier - Errors". May be this article deals with that error? Or another error I noticed with transistor labeling in the mosfet version, I think.
Hmm... may be it is possible afterall.... Oh, it is single supply... 🙂
But when I saw and built my amps I didn't notice/realize that the FB and the input RC was connected to the mid-supply point, which I think is not possible. But I see from the list of JLH publications on the site you attached, there is the 4th series titled: "A Simple 30 Watt Integrated Amplifier - Errors". May be this article deals with that error? Or another error I noticed with transistor labeling in the mosfet version, I think.
Hmm... may be it is possible afterall.... Oh, it is single supply... 🙂
Last edited:
Hart Electronics sold a kit version of this amplifier and the later MOSFET one.
I have attached a couple of pages from an earlier article for a high power direct-coupled version designed for high quality industrial purposes but suited for high quality audio and studio applications.
The rest can be found along with a lot of historical material (with patience trawling through all the magazine advertising) at American Radio History: Documenting the History of Radio TV and FM broadcasting
The title of the piece for my two pages of the Australian edition June 1976 issue of Electronics Today International is "Amplifier Technology".
Two famous amplifiers mentioned on page 18 are the Bailey Amplifier published in Wireless World in May 1968 and the JBL T - Circuit triple output stage designed by Locanthi - published in the Journal of the Audio Engineering Society in July 1967 and Electronics World (USA) in January 1967.
The following passages on page 18 relate to problems in setting generator bias among other things - an advantage of using triple output stages for which Nigel P has been a strong advocate - see also figs 5-7 for symmetry considerations.
This with the Baxendale diode this looks good. I suspect the latter JLH class A would be worth finding as it was single input transistor with no output cap. I built a similar idea and had no problems. A comparator could be used to check for DC output. If a relay protector have a 10 uF 100V polyester bypass cap for the contacts.
Last edited:
Reading through Mjona No 458 the PDF of JLH is very well written and sums up all the things I found in my life of repairing. He is the only one who says things like despite measuring worse at 20 kHz it sounds better if you will forgive the paraphrasing. JLH is the only one I ever saw sum up the bootstrap current source so well. Strikes me a bootstrap CCS for the Naim would be worth a thought as the Naim could do with a bit more second harmonic. I have often speculated that a bootstrap CCS should inject the crossover distortion into the mix. As it is in phase and perhaps non additive I doubt it is a concern. I never found it in measurements.
I 100% aggree with JLH that an output capacitor is not a major concern. Far better than the best valve output transformers. It is the prime form of DC protection. I have never tried a 10 000 uF one. I can't see why is wouldn't work. As a rule of thumb 22 000 for the PSU if you did.
I like his other point. If the transistor SOA is not exceeded even zero ohms is a permissible load as the output fuse is enough protection, Rotel proved this. The thing I admire so much is despite the simplicity all technical areas are covered including reactive loads and TID. So you other designers this guy puts you to shame. This is the amplifer version of a Lotus 7.
I 100% aggree with JLH that an output capacitor is not a major concern. Far better than the best valve output transformers. It is the prime form of DC protection. I have never tried a 10 000 uF one. I can't see why is wouldn't work. As a rule of thumb 22 000 for the PSU if you did.
I like his other point. If the transistor SOA is not exceeded even zero ohms is a permissible load as the output fuse is enough protection, Rotel proved this. The thing I admire so much is despite the simplicity all technical areas are covered including reactive loads and TID. So you other designers this guy puts you to shame. This is the amplifer version of a Lotus 7.
I have tried a pair of 10mF back to back into an 8ohms load. Not for long, it was only as an experiment to hear for myself, but performance seemed to be pretty good.
Note in the link that the 100nF into ~ 150k input impedance that this sets the LF passband to ~10.5Hz.
This when applying my usual sqrt(2) factor for determining subsequent roll-offs would require the 8ohms output capacitor to be ~2650uF. He shows 2200uF, pretty close. If the speaker is 15 or 16 ohms then the cap just needs to be >1330uF
And it's worth noting that C5 also meets the no significant AC signal in being >22uF
i.e. the input filter sets the passband and all the subsequent DC blocking capacitors do not have any significant audio signal across them.
Note in the link that the 100nF into ~ 150k input impedance that this sets the LF passband to ~10.5Hz.
This when applying my usual sqrt(2) factor for determining subsequent roll-offs would require the 8ohms output capacitor to be ~2650uF. He shows 2200uF, pretty close. If the speaker is 15 or 16 ohms then the cap just needs to be >1330uF
And it's worth noting that C5 also meets the no significant AC signal in being >22uF
i.e. the input filter sets the passband and all the subsequent DC blocking capacitors do not have any significant audio signal across them.
Last edited:
Thanks Andrew. Good info. I heard I suspect this JLH when someone asked me to show them something better. I gave up and wanted one.
Sorry to get so wowed by this JLH. Do you have the distortion graphs? An output triple this simple is hard to believe and is so logical. I would imagine where the current source sits is critical. I never knew this exact version existed.
Also the VAS like a Quad 303 almost has nothing to do. I bet the IM distortion is very OK.
Also the VAS like a Quad 303 almost has nothing to do. I bet the IM distortion is very OK.
Last edited:
Hi Martin. You can blame Gareth for my reappearance. I quit the forum in January 2010 after they banned Andy. The last thing I was expecting was for someone to message me out of the blue after seven years asking for help and having the nerve to quote one of my own, self-flattering posts where I declared myself indispensable! Before I knew what was happening my ego had propelled me back in here.
FYI.
The accompanying pre- amplifier high pass filter has a hump in the response near the turnover frequency. C1 in the power amplifier is to flatten this. The value should be increased to say 470n.
Thank you Mjona. Wow that's some story. You know I have been down this path when testing things and word for word the same thoughts came to me as JLH more or less. I have looked at VAS current verses stability and square wave outcomes and usually found that less is more. No one at DIY Audio I can remember writes like he does. Some tell us what to think and others get told off for thinking the wrong thing. This man opens my eyes. It's years since I read this and to be honest some would have gone past me, hopefully less so now. A gentleman also.
I really like this JLH version in the PDF below. Remember this was written years before Douglas Self started to dislike MOS FET's ( Sorry Douglas if not true, as it seems to me ). This different use mimics the output triple of the loved design. The Douglas Self's blanket statement to pararphrase, use of FET's would offer less linearity and are more expensive. That's before he says dodgy stability plus gain droop. JLH who outshines any I know of not least when ears come into it simply shows they can excell.
I have always said an audio type MOSFET works much like a Darlington that can conduct tollerably well with a germanium bias level ( 0.2 V, 0.4 V per pair ). To get the Exicon devices to pass >50 mA usually takes about 1.2 V per pair which is marginally below a BD139/140 pair at circa 20 mA (which wouldn't make more than a headphone if amp not using Darlingtons ). BTW Self implies 5V gate drive required I seem to remember for his choice of MOS FET ( 10 V per pair ? ). That's industrial FET's which although very interesting are not ideal. T0220 FET's find it hard to give the power levels they seem on paper to offer, just too small. Better than bipolar T0220 I don't doubt. Audio FET's are I suppose TO247 ( TO3P ) and better still T03.
Audio FET's are very interesting. If the VAS current is made super low ( 2 mA, as a Triple ) they will work. If a fast signal is sent through it will show trouble. Realistically no musical signal is like this. As the device is super fast ( >10 MHz and often much more ) it doesn't need the driving stage to do the same as a bipolar stage and IMD or more TID is not automatically the result. The gate will discharge almost regardless as prime reason, bipolar devices would stuggle if Darlingtons. People state FET Cgs being high as the problem with audio FET's. What they think they see would mean the loop gain falling in the audio band which cleary it doesn't. This is because the source output when a source follower is bootstrapped to the speaker and not to ground as their beliefs suggest. The bigger problem is Cgd which is smaller than the typical Cgs of circa 700 pF.
If ever there was a proof of negative feedback it is this JLH design, it is not the conventional use of anything. It is good to see industrial Darlingtons replaced by audio FET's. Super slow verses super fast. Class B amps do need speed. Ironically the 10 watt class A JLH seems to favour many slow devices. Why not, it isn't switching and 100 kHz is plenty. JLH let a letter writter research that. Neither really came to a solid reason why slow was good. How honest. Some fast devices were fine. If you read JLH oscillation would be checked as would all other suspects.
In the JLH design the bipolar input device TR5/6 of the output pairs will set the linearity of the pair and should make the pair matching better. Also as these are the better type of bioplar devices we have at the drop of a hat very high speed. I have never really asked myself how fast a typical Darlington is even if 50 MHz devices. Not very is my guess.
JLH seems not to want to use a long tail pair input TR1. By 1980 that would be a choice rather than saving money. My amplifier uses output capacitors. I have a different one for each speaker unit and use it as a first order filter. I can drive my setup to insane levels with open baffles and have standard RIAA with no 20Hz filter ( 115 dB average measured and not really obvious ). Unknown to many Celestion put a capacitor in the bass circuit of their speakers. When asking Adam their technical rep why he said dramatically better power handling. To my ears their bass for a ported design was better than usual and they seldom blew up. My caps are 4700 uF and 5R7 to the 15 inch bass ( Qts 1.2 measured ), 1000 uF 7R for the 12 inch bass mid ( Qts 0.49 ). And 3u3 for the high crossing 8R tweeter. I have a 5.85 mH coke to the 15 inch job which should be 4 mH. I often hit the current limiters without trouble. These drive units are like 1950's types and can do the job without excessive cone flap. I inferred recently I should be dreaming to have this work. Some say my RIAA needs a 3rd order bass filter.... NO. True if I insisted on off the peg speakers.
I hope people read this link. I am printing those of Mjona as I find paper ones better.
It might just be a Naim like design could use a bit of JLH.
John Linsley Hood
I really like this JLH version in the PDF below. Remember this was written years before Douglas Self started to dislike MOS FET's ( Sorry Douglas if not true, as it seems to me ). This different use mimics the output triple of the loved design. The Douglas Self's blanket statement to pararphrase, use of FET's would offer less linearity and are more expensive. That's before he says dodgy stability plus gain droop. JLH who outshines any I know of not least when ears come into it simply shows they can excell.
I have always said an audio type MOSFET works much like a Darlington that can conduct tollerably well with a germanium bias level ( 0.2 V, 0.4 V per pair ). To get the Exicon devices to pass >50 mA usually takes about 1.2 V per pair which is marginally below a BD139/140 pair at circa 20 mA (which wouldn't make more than a headphone if amp not using Darlingtons ). BTW Self implies 5V gate drive required I seem to remember for his choice of MOS FET ( 10 V per pair ? ). That's industrial FET's which although very interesting are not ideal. T0220 FET's find it hard to give the power levels they seem on paper to offer, just too small. Better than bipolar T0220 I don't doubt. Audio FET's are I suppose TO247 ( TO3P ) and better still T03.
Audio FET's are very interesting. If the VAS current is made super low ( 2 mA, as a Triple ) they will work. If a fast signal is sent through it will show trouble. Realistically no musical signal is like this. As the device is super fast ( >10 MHz and often much more ) it doesn't need the driving stage to do the same as a bipolar stage and IMD or more TID is not automatically the result. The gate will discharge almost regardless as prime reason, bipolar devices would stuggle if Darlingtons. People state FET Cgs being high as the problem with audio FET's. What they think they see would mean the loop gain falling in the audio band which cleary it doesn't. This is because the source output when a source follower is bootstrapped to the speaker and not to ground as their beliefs suggest. The bigger problem is Cgd which is smaller than the typical Cgs of circa 700 pF.
If ever there was a proof of negative feedback it is this JLH design, it is not the conventional use of anything. It is good to see industrial Darlingtons replaced by audio FET's. Super slow verses super fast. Class B amps do need speed. Ironically the 10 watt class A JLH seems to favour many slow devices. Why not, it isn't switching and 100 kHz is plenty. JLH let a letter writter research that. Neither really came to a solid reason why slow was good. How honest. Some fast devices were fine. If you read JLH oscillation would be checked as would all other suspects.
In the JLH design the bipolar input device TR5/6 of the output pairs will set the linearity of the pair and should make the pair matching better. Also as these are the better type of bioplar devices we have at the drop of a hat very high speed. I have never really asked myself how fast a typical Darlington is even if 50 MHz devices. Not very is my guess.
JLH seems not to want to use a long tail pair input TR1. By 1980 that would be a choice rather than saving money. My amplifier uses output capacitors. I have a different one for each speaker unit and use it as a first order filter. I can drive my setup to insane levels with open baffles and have standard RIAA with no 20Hz filter ( 115 dB average measured and not really obvious ). Unknown to many Celestion put a capacitor in the bass circuit of their speakers. When asking Adam their technical rep why he said dramatically better power handling. To my ears their bass for a ported design was better than usual and they seldom blew up. My caps are 4700 uF and 5R7 to the 15 inch bass ( Qts 1.2 measured ), 1000 uF 7R for the 12 inch bass mid ( Qts 0.49 ). And 3u3 for the high crossing 8R tweeter. I have a 5.85 mH coke to the 15 inch job which should be 4 mH. I often hit the current limiters without trouble. These drive units are like 1950's types and can do the job without excessive cone flap. I inferred recently I should be dreaming to have this work. Some say my RIAA needs a 3rd order bass filter.... NO. True if I insisted on off the peg speakers.
I hope people read this link. I am printing those of Mjona as I find paper ones better.
It might just be a Naim like design could use a bit of JLH.
John Linsley Hood
Member
Joined 2009
Paid Member
I'm pleased to know you were also a fan of Andy_c as I found him to be a very knowledgeable person who could explain things clearly and with a good temperament. There were other knowledgeable people around at that time whose inputs I valued - those were the golden years for the Solid State forum. My newer thread on TGM11 covers a design that Ancy_c helped me to understand.
If people read the JLH text maybe you will focus on his belief that the usual dominant pole capacitor to the VAS causes him to dislike the sound despite very good > 10 kHz distortion measurements. He says if I get him right that the input stage driving the base of the VAS and Cdom is too much. His method although taking the amplifer into the " only "-74 dB area at realistic levels sounds better to him. Theory has it we don't hear >10 kHz distortion as it's first nasty is at >30 kHz. Who do you believe? That is if that distortion isn't part of a nastier picture which JLH relates to in the text.
He says something if I said it I would be shot down in flames. The reduced need to sweep out the holes or whatever from the gate ( base ) allows the low level detail sound to be more realistic as the sound still resides in that mechanism ( hope I got it right John ). Self always seems to argue the opposite. That is only compound pairs can be properly biased and all other things we might " imgine " are wrong. He goes on to reinforce his arguement in other areas. Thus it could be everyone ignored the mild mannered JLH for the forceful Self. Did we have the wrong guru? Truth is a bit of both doubtless. The JLH is a compound pair except bipolar-FET.
Is that 1975 JLH the Naim ? It is very close. Note JLH went back to a more primitive design in 1980.
MJR7-Mk5 Mosfet Power Amplifier
If you listen to good class D amps they have almost the JLH sound. Hypex comes to mind. I measured the low level detail and it is there ( really hard to do if not usings ears, like JLH say + 400 kHz in the picture, I have to switch my router off also ). I had to build a 12 pole passive filter to do that ( thanks AP for saying how ). It could be somehow they don't kill the FET advantage when the Hypex. Hypex sounds bright yet nice. It rolls off at 47 kHz if I remember !
Maybe someone would like to think how the JLH Nyquist roll off could be made to work in the Naim.
He says something if I said it I would be shot down in flames. The reduced need to sweep out the holes or whatever from the gate ( base ) allows the low level detail sound to be more realistic as the sound still resides in that mechanism ( hope I got it right John ). Self always seems to argue the opposite. That is only compound pairs can be properly biased and all other things we might " imgine " are wrong. He goes on to reinforce his arguement in other areas. Thus it could be everyone ignored the mild mannered JLH for the forceful Self. Did we have the wrong guru? Truth is a bit of both doubtless. The JLH is a compound pair except bipolar-FET.
Is that 1975 JLH the Naim ? It is very close. Note JLH went back to a more primitive design in 1980.
MJR7-Mk5 Mosfet Power Amplifier
If you listen to good class D amps they have almost the JLH sound. Hypex comes to mind. I measured the low level detail and it is there ( really hard to do if not usings ears, like JLH say + 400 kHz in the picture, I have to switch my router off also ). I had to build a 12 pole passive filter to do that ( thanks AP for saying how ). It could be somehow they don't kill the FET advantage when the Hypex. Hypex sounds bright yet nice. It rolls off at 47 kHz if I remember !
Maybe someone would like to think how the JLH Nyquist roll off could be made to work in the Naim.
Thirty years ago, two of my friends exchanged amplifiers for a few weeks period. One of them had Musical Fidelity B200 lateral MOSFET, and the other had Musical Fidelity A100 classA bipolar amp. The one that had B200 complained that with A100 he was constantly going to the turntable to see if the rotation speed was right! He felt that A100 was slowing down the flow of music. I think that JLH was right about the subjective qualities of MOSFETs. They are not only faster in the specs, they sound faster too.
Member
Joined 2009
Paid Member
A darker sound is possible when bipolar class A. Some do not realise that with speakers as they now are class AB can sound better than A. If you like a class AB allows the mains transformer a secondary function. It becomes a storage choke which AB can exploit. Class A keeps the power supply at a constant current and constant ripple.
I have looked at a couple of Musical Fidelity circuits and seen the low frequency roll offs are not all the same. These depend on the RC relationships in the input coupling and the nfb decoupling circuits.
In the case of the B200 these are 1uF into 33k for the input and 390R into 220uF in the nfb decoupling arm of the feedback circuit.
The respective values for the Class A circuit were 1uF into 47k and 2k7 into 47uF.
I may have not got the right amplifier exactly right with this last one however this is the right area to be looking for reasons for your friends perceptions - low frequency roll off and phase differences.
Member
Joined 2009
Paid Member
It depends on the topology. I suspect most Class A is not constant current draw. The classic JLH 1969 Class A isn't constant current from what I remember.
But you make an interesting comment that modern speakers are better with AB (I like AB). I don't understand that one, why do you say this ?
Last edited:
- Status
- Not open for further replies.