This is a very general question which depends on many factors of course but in general which one should sound better in a single ended class-A stage, BJT, MOSFET or JFET?
Try the new SemiSouth power JFETs. Aside from a stiff gate current, they are reputed to be very good by Nelson Pass.
Hugh
Hugh
I've seen more MOSFETS in single-ended class A amps than BJTs.
Must be because they require less circuitry as far as pre-drivers, temperature compensation e.a. go, therefore they fit in better with the minimalist philosophy of single-ended class A.
Must be because they require less circuitry as far as pre-drivers, temperature compensation e.a. go, therefore they fit in better with the minimalist philosophy of single-ended class A.
That would be the SJEP120R100A and SJEP120R063ASiC JFETs for audio
SiC Transistors | Semisouth Laboratories, Inc.
It would be nice if they published SPICE models...
I can see a tension towards audio morfets of new audio jfets. What about bjt.
constructions like these have remarkable sound from my listening impression.
Le monstre
constructions like these have remarkable sound from my listening impression.
Le monstre
MosFETs in ClassAB do not seem to respond to optimal bias voltage/current in reducing crossover distortion. This makes ClassAB for BJT & FETs somewhat different.
Most knowledgeable commentators draw attention to the FACT that mosFETs seem to get better the more the output bias is increased. I recall Borbely mentioned 500mA of output bias as minimum recommendation for every mosFET amplifier.
It is only a small leap of faith to move from >=500mA of bias to full ClassA and then there can be no crossover distortion to attenuate.
To me, that seems to explain why mosFETs predominate in small ClassA amplifiers.
Most knowledgeable commentators draw attention to the FACT that mosFETs seem to get better the more the output bias is increased. I recall Borbely mentioned 500mA of output bias as minimum recommendation for every mosFET amplifier.
It is only a small leap of faith to move from >=500mA of bias to full ClassA and then there can be no crossover distortion to attenuate.
To me, that seems to explain why mosFETs predominate in small ClassA amplifiers.
Mosfet switching transitions in Class AB
Andrew,
You may not need that much. Here is the (simulated) crossover event depicted by the source resistor currents on a mosfet AB output stage. The pmos (R18) is rather more linear than the nmos (R17), but you can see the transition is well controlled with no sign of ringing or instability. This is into 8R//100nF, a difficult load, at 1KHz.
Quiescent current is just over 150mA.
I feel that modern mosfets do not lose out too much to bipolars these days.
Hugh
Andrew,
You may not need that much. Here is the (simulated) crossover event depicted by the source resistor currents on a mosfet AB output stage. The pmos (R18) is rather more linear than the nmos (R17), but you can see the transition is well controlled with no sign of ringing or instability. This is into 8R//100nF, a difficult load, at 1KHz.
Quiescent current is just over 150mA.
I feel that modern mosfets do not lose out too much to bipolars these days.
Hugh
Attachments
Last edited:
Hugh,
I do not think that your example is a good guide to what modern FETs can achieve in terms of reducing crossover distortion and doing so at lower bias currents.
I do not think that your example is a good guide to what modern FETs can achieve in terms of reducing crossover distortion and doing so at lower bias currents.
oh, they will, but you won't notice, unless you have both 😀
but yeah, the different designs needed to get 'optimal result' may make the biggest difference 😉
I'm no expert, but from reading here I got the clear impression that BJT are better used fore classAB, rather than classA
and fore ClassA, better use mosfet
not saying that mosfets can't do classAB
I leave that debate to Andrew and Hugh
but sure, from reading only, I also got the impression that mosfets are better performers with classA bias
ehh, and Laterals like Renesas, different ?
hmm, I got the impression that they might end up somewhere in between, doing their best at higher biased classAB ?
hmm, I got the impression that they might end up somewhere in between, doing their best at higher biased classAB ?
I get the impression Neazoi is rather asking "What's wrong with BJT class A?" than how good are the usual Mosfet affairs, single ended or p-p.
Certainly Top, if by properly designed, you mean commercial style, optimum low distortion design, you could be right about similarity of sound but if instead, we designed for optimum perceived sound, it is more usual to hear that any different device, topology, wiring layout, supplies or grounding can produce significant differences. That's an opportunity and a great strength of DIY.
My experience has been that real, commercial and often not so well designed amplifiers can readily show their colours as Mosfet, Tube or BJT "sound". Assuming you are familiar with Musical Fidelity products, for instance, you may have noticed. Single ended class A amps go beyond this however, due to the intentionally strong and characteristic harmonic distortion patterns of the topology and semis used. A visit to Nelson's forum will give the idea of just how much Mosfet distortion is possible and actually desirable. It is not quite so pleasant at 1% THD with SE BJT's I think.
Unfortunately, we can easily descend into heated debate of the "rightness" or "wrongness" of these design goals, making pointless, nasty threads - not a good idea.
Certainly Top, if by properly designed, you mean commercial style, optimum low distortion design, you could be right about similarity of sound but if instead, we designed for optimum perceived sound, it is more usual to hear that any different device, topology, wiring layout, supplies or grounding can produce significant differences. That's an opportunity and a great strength of DIY.
My experience has been that real, commercial and often not so well designed amplifiers can readily show their colours as Mosfet, Tube or BJT "sound". Assuming you are familiar with Musical Fidelity products, for instance, you may have noticed. Single ended class A amps go beyond this however, due to the intentionally strong and characteristic harmonic distortion patterns of the topology and semis used. A visit to Nelson's forum will give the idea of just how much Mosfet distortion is possible and actually desirable. It is not quite so pleasant at 1% THD with SE BJT's I think.
Unfortunately, we can easily descend into heated debate of the "rightness" or "wrongness" of these design goals, making pointless, nasty threads - not a good idea.
Perhaps this diagram might further elucidate the point, n'est pas? You can see that the crossover 'dysjunction' is in fact unbelievably clean. The whole notion of crossover distortion is somewhat overhyped, I feel.
Hugh
Attachments
Last edited:
Spiced
It takes a bit of juggling with formulae, but LT Spice can be made to produce a wingspread diagram - a highly graphical representation of semi crossover behaviour at stepped power levels. I know AndrewT has read the old thread this came from.
The benefit of wingspreads is obvious, as can see at a glance how very wide the region is, at about 13 volts with the Toshiba VFETS simulated here.
As AKSA points out, the absolute discontinuity is quite small, in the mVs, but we love to fret about anything that doesn't look straight. 😉
It takes a bit of juggling with formulae, but LT Spice can be made to produce a wingspread diagram - a highly graphical representation of semi crossover behaviour at stepped power levels. I know AndrewT has read the old thread this came from.
The benefit of wingspreads is obvious, as can see at a glance how very wide the region is, at about 13 volts with the Toshiba VFETS simulated here.
As AKSA points out, the absolute discontinuity is quite small, in the mVs, but we love to fret about anything that doesn't look straight. 😉
Thread titl is a bit 'wider', but this is Neazoi's direct question
This is a very general question which depends on many factors of course but in general which one should sound better in a single ended class-A stage, BJT, MOSFET or JFET?
Tinitus,
Mosfets have two distinct advantages: tempco over about 150mA, and great thermal robustness.
They generally do not have high transconductance (although the FQA series from Fairchild I'm using are high, the FQA36P15 is rated to 294 watts and 29S, which is very good) but they do have wonderful speed and a square law transfer function. OTOH, they are NOT easy to drive, since Ciss is upwards of 2-8nF, depending on current rating, and this requires some high duty shaking.
It's swings and roundabouts, but I'm very impressed with the latest crop, even the switchers are quite linear enough for audio.
However, it has to be admitted that bipolars are getting better all the time too, with their gain linearity, speed and SOAR consistently improving.
Cheers,
Hugh
Mosfets have two distinct advantages: tempco over about 150mA, and great thermal robustness.
They generally do not have high transconductance (although the FQA series from Fairchild I'm using are high, the FQA36P15 is rated to 294 watts and 29S, which is very good) but they do have wonderful speed and a square law transfer function. OTOH, they are NOT easy to drive, since Ciss is upwards of 2-8nF, depending on current rating, and this requires some high duty shaking.
It's swings and roundabouts, but I'm very impressed with the latest crop, even the switchers are quite linear enough for audio.
However, it has to be admitted that bipolars are getting better all the time too, with their gain linearity, speed and SOAR consistently improving.
Cheers,
Hugh
Lateral fets zero temp co is about 150mA. Vertical types can be several amps depending a bit on Gm. Lateral fets generally do not have as much Gm and exhibit square law transfer. Vertical types have an exponential transfer, better suited for switching. The vertical types can have much higher Gm but are not any easier to drive, in fact more complicated.
The Gm droop at small currents creating an inconsistency of output Z through crossover is the biggest issue IMO. Although the Ft of lateral fets is fast, it is hampered by an overall lower Gm, the effective Ft of vertical types is exceptionally high thus they usuallly require local HF compensation.
I can confirm the ruggedness of those FQA (or FQP for the TO-220 package) Q-fets. The planer stripe geometry is easier to manufacture and can provide a higher effective well density than cellular architecture like hexfets leading to lower Rds on and less charge per change in conductance. Of course Rds on is pointless in a linear application but they are tough. Perhaps the area of conduction on the die is more evenly spread out due to the construction, less likely to have 'hot spots'... a show stopper in linear audio! Derate FQA36P15 at 1.96W/C, for 150C, 294-245= 49W. Pretty impressive. Take FQP47P06, derate 1.06W/C, for 150C, 160-133= 27W. Really impressive for a TO-220 device. That part being wonderful, it is likely these figures are a bit high for linear use, also it is very tricky to get these to work properly in class AB with high BW and also have a completely transparent sound, but it can be done.🙂
BTW Here are a couple papers of interest about the construction comparison of planer stripe and cellular fets....
The Gm droop at small currents creating an inconsistency of output Z through crossover is the biggest issue IMO. Although the Ft of lateral fets is fast, it is hampered by an overall lower Gm, the effective Ft of vertical types is exceptionally high thus they usuallly require local HF compensation.I can confirm the ruggedness of those FQA (or FQP for the TO-220 package) Q-fets. The planer stripe geometry is easier to manufacture and can provide a higher effective well density than cellular architecture like hexfets leading to lower Rds on and less charge per change in conductance. Of course Rds on is pointless in a linear application but they are tough. Perhaps the area of conduction on the die is more evenly spread out due to the construction, less likely to have 'hot spots'... a show stopper in linear audio! Derate FQA36P15 at 1.96W/C, for 150C, 294-245= 49W. Pretty impressive. Take FQP47P06, derate 1.06W/C, for 150C, 160-133= 27W. Really impressive for a TO-220 device. That part being wonderful, it is likely these figures are a bit high for linear use, also it is very tricky to get these to work properly in class AB with high BW and also have a completely transparent sound, but it can be done.🙂
BTW Here are a couple papers of interest about the construction comparison of planer stripe and cellular fets....
Last edited:
Wingspread assumes a fixed bias spread that reacts slower than the lowest audible
note (usually just to temp) , or not at all. Fast corrections that watch and correct
both voltage and current errors in time with music like Hawksford, Circlophone, and
SRJLH do not leave wingspread to match by chance, but take control and shape it.
Since this feature can be implemented with as little as 3 added components, I think
its a waste of time to worry about a problem that need only exist in legacy design.
I can match MOSFET to BJT with near perfect AB crossings, not even a problem...
No preference for me other than whatever is cheap and convenient.
In linear class A, I don't even know why we would need to discuss a wingspread?
Not operating anywhere near there. We could describe Square law vs Square law
Class A as-if it were a really big wingspread, I suppose...
Or discuss legacy JLH class A as a non-linear wingspread with a huge hump in the
middle. That nobody ever seems to take notice till the amp overheats on the hump
at idle, or clips earlier than expected due to beta droop at the far edges...
But the wingspread discussed above looks like AB to me....
note (usually just to temp) , or not at all. Fast corrections that watch and correct
both voltage and current errors in time with music like Hawksford, Circlophone, and
SRJLH do not leave wingspread to match by chance, but take control and shape it.
Since this feature can be implemented with as little as 3 added components, I think
its a waste of time to worry about a problem that need only exist in legacy design.
I can match MOSFET to BJT with near perfect AB crossings, not even a problem...
No preference for me other than whatever is cheap and convenient.
In linear class A, I don't even know why we would need to discuss a wingspread?
Not operating anywhere near there. We could describe Square law vs Square law
Class A as-if it were a really big wingspread, I suppose...
Or discuss legacy JLH class A as a non-linear wingspread with a huge hump in the
middle. That nobody ever seems to take notice till the amp overheats on the hump
at idle, or clips earlier than expected due to beta droop at the far edges...
But the wingspread discussed above looks like AB to me....
Last edited:
- Status
- Not open for further replies.