HEX-FET 60W/4ohm Amplifier

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
What do you think about this amplifier schematics ?
( AN-948 aplication from International Rectifier )

Q5 and Q6 (Hex- Fet power devices) are IRF532 and IRF9532.

1. It can be replaced with IRFP240/IRFP9240 and supply this amplifier with more voltage ( +/- 40v ) ?

2. Did anybody build this amp and can tell me something about the sound quality ?

3. Did it worth to build this amplifier ?

Thank you !


  • schema.gif
    8.6 KB · Views: 3,990
-> Eliminate R14 (short it)
-> Eliminate R8,R9,C4. Connect a constant current source (5mA to 10mA) between Q3 collector and +Vdd.
-> Connect a 50uF capacitor between Q3 emitter and collector
-> A la Nelson, add a gate series resistor to Q6, a similar value to R10 (both should be a couple of hundred ohms).
-> Optional depending upon the open and closed loop gains: add a capacitor between Q4 base and collector of between 50pF and 150pF (depends on the stability of the circuit).
-> For short-circuit protection of the FETs or output current limiting you can put a zener and diode in series between the gate and source of each FET. This will limit Vgs max.
Would you elaborate on the reasons the bootstrap can sometime improve the sound quality? I would like to understand this.

I could see that if R8 were a CCS then R9 would effectively be connected between the gate and source of the FETs. Perhaps this would provide some phase improvements at high frequencies or help to reduce the effect of Zgs mismatch.

I wish I knew the reasons. Perhaps a reduction in the number of active devices? Or the fact that a resistor is more linear than a transistor? Or a reduction in high frequency spuriae?

Nelson has indicated on many occasions his preference for a resistor, as opposed to a css, on the input differential of his amps because of the sonic benefits. Over the years, many highly regarded amps have used the bootstrap arrangement for the voltage amplification stage, including designs by John Linsley Hood and, more recently, the AKSA.

I will admit that my preference is to keep a circuit as simple as possible, with as few active devices as one can get away with whilst maintaining an acceptable level of measured and, more importantly, audible performance.

Perhaps if Hugh drops by he will be able to throw some more light on this subject.


Bootstraps - relic of the past, or glimpse of the future?

The most vexing problem of SS amplifier design, in my view, is stability. Because any amp with a global feedback loop has the potential for its NFB to turn positive and nasty at HF, we must find ways of pulling back the gain at HF so that at the upper pole point of the amp, the gain is below unity and it can never oscillate.

The big challenge is to do this without affecting the sonics of the amplifier. In short, pulling back gain at HF almost invariably affects the quality of the sound.

There are many ways of doing this, and usually a combination is used. Often there is a cap across the output feedback resistor, and always there is a small cap between collector and base of the voltage amplifier. There may even be a small cap between the collector of the voltage amplifier and the feedback node; this approach is suggested by Linsley Hood and is useful. These caps all conspire to pull back gain to just below unity at the point in the spectrum where global negative feedback turns positive and can send the amp into destructive paroxysms. This requirement is called the Bode/Nyquist criteria, after the mathematicians who originally postulated and researched it.

The voltage amplifier is crucial to the sonics of an amplifier, and must be thrown in manacles to meet the Bode/Nyquist criteria. These manacles, usually capacitive but sometimes inductive, can wreck the sonics, so they must be minimised whilst holding good stability into reactive - read difficult - loads.

A constant current source has DC to light frequency response. That is, it will hold very high output impedance at its collector well into the MHz range. This means that any voltage amplifier supported by a CCS needs a substantial manacle to keep it stable at HF, and this in turn mandates use of a large compensation cap between its collector and base.

A bootstrap using a conventional electrolytic capacitor starts to rapidly loose effectiveness at higher frequencies. This means that as the frequency rises above 50Khz - well into the ultrasonic range - the collector of the voltage amplifier sees a steadily increasing load - reducing impedance - and this pulls back its gain very nicely without recourse to large lag compensation.

Everything old is new again.........


Hugh R. Dean
Research/Technical Director
Thanks for the detailed explanation. I think I understand it - now that C4 is assumed lossy at high f.

Consider the transimpedance of Q4 Ic to Vo. Ignore Q5 and Q6 input capacitance. At frequencies where C4 can be approximated by a short circuit the small signal transimpedance will be about R9xGmxRo (Gm is the combined transconductance of the two FETs and Ro is the load, say 8-ohms). When C4 can be approximated by an open-circuit the transimpedance is about R8+R9. Therefore R8+R9 << R9xGmxRo to cause a reduction in open loop gain at high frequencies.

But I'm not sure about the choice of circuit. I am concerned that C4 isn't a good short-circuit at low frequencies, that it may add distortion being an low performance electrolytic and that +Vdd noise is coupled through R8. Would it be better to have a CCS from Q3 collector to +Vdd and a series cap/res from the same point to ground? You could then use a low value, high f cap and a resistor of value R8+R9. Loop gain at low frequencies would be very high thus minimizing output offset at dc.


Hi Traderbam,

Thank you for your response. You wrote:

"Consider the transimpedance of Q4 Ic to Vo. Ignore Q5 and Q6 input capacitance. At frequencies where C4 can be approximated by a short circuit the small signal transimpedance will be about R9xGmxRo (Gm is the combined transconductance of the two FETs and Ro is the load, say 8-ohms). When C4 can be approximated by an open-circuit the transimpedance is about R8+R9. Therefore R8+R9 << R9xGmxRo to cause a reduction in open loop gain at high frequencies."

Yes, your comments on the transimpedance (current to voltage transformation at the voltage amplifier) is indeed right on the math, and though simplistic, it describes the principle well. However, the gradual deterioration of the AC short at the bootstrap cap is progressive, and it must be said that the various stabilising techniques used in practical amps all contribute to the one primary aim of meeting the Bode/Nyquist criteria. The full math for all acting together makes for very difficult math analysis, even by PSpice, which never seems able, predictably, to describe precisely the stability margins of an audio amplifier. One must build it and see............

"But I'm not sure about the choice of circuit. I am concerned that C4 isn't a good short-circuit at low frequencies, that it may add distortion being an low performance electrolytic and that +Vdd noise is coupled through R8."

I believe your concerns are groundless. First, we do have a global negative feedback mechanism which can correct for distortions in the output, albeit with lessening efficacy as the open loop gain is reduced, for whatever reason. Second the existence of an AC shunt from supply rail to the output is not a cause of hum in practice; the gain of the output stage is one, after all, and 100uF (a typical bootstrap value) is not good at coupling 120Hz hum to an 8R load since it is working into the emitters of the output devices, a low impedance, as well as the speakers. The Xc of a typical boostrap cap at this frequency is 13R, quite a large voltage division at the speaker, reducing say 10mV of supply rail ripple to 3.8mV at the speaker output. In any event, to improve PSRR still further I use a decoupling diode and CR network, and this reduces any hum at the rail to immeasureable levels at the top of the voltage amplifier supply.

"Would it be better to have a CCS from Q3 collector to +Vdd and a series cap/res from the same point to ground? You could then use a low value, high f cap and a resistor of value R8+R9. Loop gain at low frequencies would be very high thus minimizing output offset at dc."

Good, intuitive suggestions. However, I have tried this, and sonically there appears to be little benefit, although the low frequencies are tighter and more powerful sounding. BUT, they are not more musical, since the distortion introduced by the loss of bootstrap efficiency at very low frequencies is partly offset by the appearance of H2 which is introduced by a hard worked voltage amplifier and reduced OLG. This gives the overall impression of very musical bass, just like a tube amp, which is one of the endearing features of the AKSA amplifiers. Low damping factor is not as powerful sounding, granted, but it is musical, and it seems to confer a satisfaction to the listener which the market demands and I cannot ignore.

Thank you for your insightful comments,

Hugh R. Dean

Just a note in passing...I've run across a number of places here and there in the last year or two where people have been reconsidering high damping factors vs. the perceived sound quality. Although a high damping factor is attractive in a logical sense, meaning that we're trying to control a willful driver, there seem to be a growing number of people who feel that the sound suffers. I've seen some vague possibilities batted about, but nothing concrete enough (other than high levels of NFB) to be worth debating.
I have only had one accidental test of this idea, which came about when I got my cables jumbled up and had my tube amps on my woofer panels when they should have been on the midranges. The midrange sounded poor, but the woofer panels (circa 70Hz to 250Hz) sounded absolutely amazing. So much so that I've spent the last year and a half trying to get enough amplifiers going at one time that I can try it again as a formal experiment rather than an "oops."
A damping factor of 1000 isn't 100 times better than a damping factor of 10. Some people have expressed the idea that it's more of a log function, leading to the idea that it's only, perhaps, two or three times as good. This might lead to the idea that there's an optimum range where the tradeoffs balance against one another.

Hey Hugh,
Interesting comments about the 'dual' use of components (actually they're quite depressing, while I can usually follow the math, the electronics give me trouble...).:confused: While the use of an old AB carbon resistor might be appropriate for its frequency consistency in a grid stopper, their use as a ccs or within a bootstrap position might well be self defeating??? Similarly, your own comments about the use of an elecrolytic in that boot position as opposed to a more linear film, follow that reasoning. As I'm about to recieve your AKSA and TLP and will ultimately tweak, its good to be warned off of ridicuclously expensive 'upgrades' that would actually be detrimental and understand the rational behind them.

Mr. Rollins -
I've recently heard a Cary V-12 and believe I follow your comments on sound quality. Is there a linkage between your thoughts and those of Lynn Olsen on his comments on 'hard' amplification?

Geoff -
Lacking the requisite expertise, its difficult to discern your reasons for your dislike of the ltp. Are you simply a retro sort of a guy advocating - 'JLH all the way!' with its implicit CFB? or is your dislike based upon an abhorence of non-inverting input? Would you post some links to some previous comments that might elaborate?

Apologies for these dumbass comments and questions in this group who actually know what they're talking about.

Hi Hugh,
Just a couple of things.

I didn't really explain myself well enough. I am averse to creating any currents in the signal path that are related to psu noise voltages. I don't only mean mains hum but all the wideband distortion products that appear on the psu rails as the various feedback systems divert it from the output. By having R8/R9 couple Q4 collector to +Vdd, Q4 must conduct a counteracting noise current to keep the collector voltage from changing. This current is then part of the overall loop error signal. Thus a small amount of psu noise pollutes the whole signal chain. In my experience this is usually a big no-no for high-end audio both from a distortion point of view and from a dynamic stability point of view. I acknowledge in your design that you have taken action against this by filtering the +Vdd before the bootstrap. Having said this, the Cdg of the output FETs may create noise currents that swamp this effect anyhow: depends on the component values. I hope this makes sense.

I also acknowledge your finding that the bootstrap circuit increases H2 and creates more of a tube sound and that your customers like this - fair enough. I don't have customers so I only have to please myself! My goal is to to make my amps inaudible - totally transparent without adding any character of their own. So the increased H2 solution is anathema to me.

Although somewhat counter-intuitive, my experience is that the musicality/tightness/dryness/boominess of the bass response is not as highly correlated to damping factor at low frequencies as it is to distortion mechanisms at high frequencies, often well above the audible range and exacerbated by feedback and push-pull topologies. I haven't tried this but I would put money on the fact that an amp that has high damping and sounds over-damped would probably still sound overdamped if a resistor was placed in series with its output to give it the same damping as a more musical amp.

There's no need to apologise for seeking further insight into the black art of audio amplifier design. You asked about my comment relating to the LTP and this is a little difficult to answer as it is mainly subjective, but I will try to put it in context.

During the past 30 years, I have been steadily improving my system (so I thought), ending up with a pair of highly regarded Class-AB commercial amps in a bi-amp arrangement. However, a few years ago I realised that I was not enjoying the music as much as I used to. It had lost its emotional impact. The many small, and sometimes larger, 'improvements' over the years had added up to a system that was no doubt highly accurate but at the same time seemed sterile and uninvolving. I therefore spent some time thinking about where I had gone wrong.

Looking back over the years, so far as amplifiers were concerned, I realised that the designs that had given me the most listening pleasure were all CFB, not an LTP in sight, whether they be Class-A or Class-AB. I therefore took several steps backwards and returned to a design that I first built 30 years ago, the JLH Class-A. My system may now be less accurate in measurable terms (THD etc), but the involvement and satisfaction are back.

A well balanced LTP does have the benefit of cancelling 2nd harmonic distortion, which is therefore less than that in an amp with a single input transistor, but I believe that it generates more 'unpleasant' distortion products, particularly intermodulation. Another benefit of the LTP is the degree of control over dc offset at the output, but I am prepared to put up with a lower spec for this in return for my perceived benefits.

Am I a 'retro sort of guy'? Yes and no. Do I advocate 'JLH all the way'? The same answer. I do believe that some of the earier circuits (mid '60s to mid '70s) sound as good as or better than more recent designs, possibly due to their simplicity and avoidance of the ccs etc currently used to bring the figures down. I also think that these early designs can be improved by using more modern components. These beliefs are supported by the response shown to the AKSA, which is a simple (old-fashioned?) circuit. JLH designed some very good amps. He also designed some that I just couldn't live with - generally his later MOSFET offerings. This is not to say that his MOSFET amps were bad, they just didn't suit my listening tastes (back to the sterile and uninvolving).

As an engineer, I can appreciate the technical advances that have been made in circuit design, but I do think that a lot of the developments are more appropriate to, for example, an instrumentation amplifier rather than an audio one.

I usually try to keep my posts short and factual, avoiding personal preferences and subjective observations, but I've failed on this occasion. I hope other members will excuse this lapse.

Geoff wrote: "I must try to...avoiding personal preferences and subjective observations"
Hell no. How things sound is subjective. How circuits function is objective. It's superstition that we all need to fight off :p .

Please don't poo-poo the BJT LTP :( . This is one of the most linear ways of subtracting two voltages known to mankind and it works. It works just as well in audio amps as instrumentation amps. Absolutely no question in my mind :) . Many high-end audiophile products use them. However, your subjective observations of your amps is also correct. So what's going on? When you change from LTP to CFB designs what else changes? Were the LTPs being used correctly in the first place? I guarantee the lack of involvement you mention is nothing to do with the LTP topology as such - it is something else. Find that something else and you will have your involving sound back with greater accuracy.

I was not denegrating or dismissing the LTP, merely saying that my personal experiences have led me to a preference for amps that do not use it.

Every circuit module (LTP, CCS, CM or whatever) has its place and its optimum usage. Circuit design is always a compromise based on various conflicting requirements. The designer/constructor is free to chose which compromises he/she wishes to make based upon the final objectives.

I prefer simple amps with little or no added compensation required to ensure stability. The bootstrap arrangement previously discussed helps in this respect, as does the single transistor input stage. The LTP may well reduce the 2nd harmonic, but it usually needs additional high-frequency stabilisation which, particularly if it takes the form of the traditional Cdom across the voltage gain stage, can cause problems with slew rate limiting and intermodulation distortion.

Geoff -
Thanks for the informative post. Indeed, the JLH reference was to his earlier work. I owned a Audio Innovations single rail, CFB, bootstrapped, quasi-complementry amp years ago. And while it was certainly lacking in resolution (low biasing?), it did have that certain 'je ne sais pas'. Certainly, CFB presents problems as an ic opamp, but are adressable in the context of a audio component. but then again, visual symmentry is so appealing. Without strong partisan feelings, the art does not advance. While scientists may advance the technical (Self), it takes someone like Hugh [or you], who takes those findings, such as combining extended beta transistors and switch off caps (please hold me 'blameless' for any misrepresentations) and subjects them to the cauldron of empiricism to advance the subjective art and goes 'retro' (no ccs and/or ltp) where appropriate.

I recently had the pleasure of comparing my own amps, Llano A50s, ltp input, AB bias MosFet, with wonderful specs and in subjective comparisons, trouncing (IMO) many highly thought of amps, to Hugh's AKSA's. As a closet tree hugger, I don't do class A output. Suffice it say, as soon as the AKSAs arrive, the llanos are Audiogon. Damn Hugh's eyes!

While my own electronics experience is 25 years out of date, my academic and work experience has given me an unerring ability to detect when someone is blowing smoke up my ***, and you don't. I've gotten the 'measure of the mettle' of many on the Asylum, and am now faced with the tedious task of doing so and using the search feature on DIYAudio to uncover those gems from your and GRollins posts.


Yes, you have it right! There are moments in this black art when the poor quality of some components in measured terms - electrolytics in this case - is a boon. Of course, you could use a quality film cap as the bootstrap with a series inductor, but it's much easier just to use an electrolytic.

It comes down to this: for stability, the HF gain of any amp with global negative feedback - tubes included - must be degenerated to below unity at the pole point where NFB turns to PFB. This is text book Bode/Nyquist - absolutely nothing new at all, and pioneered back in the 1930s a few years after Black's analysis of negative feedback (and the beginning of the end for exquisitely linear triodes!).

This gain degeneration process is highly destructive of sonics because it introduces pervasive non-linearities into the amplifier, creating intermodulation and, extending down into the audio band, non-musical tones. In my opinion (and nothing is more dangerous than one man's opinion, particularly as it comes from a background in English literature, Maths teaching and IT consulting!!) these non-musical intermodulations are the primary culprit in 'bad' sounding amps. Of course it goes without saying that most distortion measuring techniques do not quantify this phenomenon at all, and this partly explains the fatuousness of modern distortion measurement techniques, at least for audio.

The most obvious way to cut gain with rising frequency is to nobble the voltage amplifier. Overdone, this sounds terrible. A second way is to configure the VAS load so that it is heavier at higher frequencies, neatly delivering less gain. A third way is to capacitively link the output of the voltage amplifier - or the output stage - directly to the feedback node. Same effect. A judicious mix of all these artifices, some more sonically damaging than others, will give us an amplifier which is unconditionally stable, or at least one which is acceptably stable into most loads. One must remember, we are not building a jumbo jet here; this is a fighter; a jumbo would do good service as a sub-woofer amplifier but sound impossibly slow at higher frequencies.

I see these compromises as little different to the rest of science and technology; the design of a truck v. a car engine, a sports versus luxury suspension on an auto, a turbo prop for a transport aircraft v. a jet fighter engine, and so on. Since we require a compromise purpose (viz. reproducing the entire audible range through one amplifier) we are stuck with design compromises which must cover quite a range of circuit blocks and which work best in synergy, with no one correction mode dominating.

Long Tailed Pairs are a special case. Like Geoff, I railed against them, and for some time built LH type circuits using the singleton. I tactically withdrew to the LTP for one reason; superior offset control, which is not a trivial matter in a SS direct coupled amplifier. While it is true that the H2 generated is minimal, this is only so if it is perfectly balanced, and even with a current mirror, this happy outcome rarely prevails. Otherwise, its low H2 is a function of good device matching and balanced currents, something which is actually easier to achieve with careful design and resistive loading than by forcing the issue with a current mirror. In my view, the current mirror does not make for better sound (any more than the current source supplying a LTP); however, its undisputed mathematical and engineering elegance has raised it to unjustified prominence in audio design for many decades now.

Geoff makes an apology of his subjectivity. Hell, no! Why do audiophiles listen to music? Is it because they like the lines of the amp enclosure, the grace and elegance of the circuit design, the familiar feel of those exquisite controls, the glimmer of a lovely blue LED? Certainly overhyped marketing departments would capitalise on these qualities; one is regaled with advertising copy of the patrician designer standing beside his life sized mock up of his latest amp, with a Testarossa lurking elegantly in the background....... These images appeal to power hungry technocrats, perhaps successful business executives, but to me they betoke the marriage of high consumerism and technology worship for its own sake. For some absurd reason, the sonics seem to be secondary in this priestly imagery. Not for all of us - and arguably none in this forum. I believe audiophiles listen to music because it talks to their soul. Who could fail to be moved by the second movement of Brahms violin concerto? Or Massenet's famous violin solo? Or Night on Bald Mountain? Or 12 Moons from Garbarek? Music is a language with all the subtlety of spoken language and more - with intonation and harmony a fundamental requirement, all of it necessarily sympatico with human hearing. We do not choose to gaze upon a superb painting through dirty spectacles; something very similar surely applies to music and its proper, dignified reproduction with technology.

I hope this stuff is not too pompous; I love this art, and have never tired of it in 40 years.......



This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.