Mixed Semiconductor Output Stage

[BKr0n]

Like it says on the tin.

Been looking at all kinds of transistors from bog standard BJT to RF HEMT. From my understanding, configurations like darlington, sziklai (can never remember the spelling of that), and composit amplifiers allow you to mix and match properties of different semiconductors in order to get better overall functionality. Is it worth looking into this further, or should I just stick with one set of transistors? If it is worth it, I would really appreciate some part recommendations. While I understand there's a few gold standard parts (Toshiba transistors for example) I'm sure there are a lot of others out there that may not be as prevalent because they are not used typically for audio, but can still perform as well if not better.

 

[stoo23]

I would 'Suggest' or even argue that you would Gain (sorry couldn't help myself lol) more Performance and reliability, by having 'Matched' Output transistors more so than merely 'Brand' but that said, perhaps a more Top End major brand offering may prove to be closer in individual tolerances and performance than a far cheaper brand ??

I am unaware of any amplifier output stage that would actually 'Benefit' from Mixed Devices, unless Specifically Designed for a specific mixture.

 

[BKr0n]

sorry couldn't help myself lol

It's OK I would have done the same thing lol

Top End major brand offering may prove to be closer in individual tolerances and performance than a far cheaper brand ??

Just from browsing it seems that Toshiba is usually a popular choice so there's probably something to it.

I am unaware of any amplifier output stage that would actually 'Benefit' from Mixed Devices, unless Specifically Designed for a specific mixture.

Yes that was kind of what I was getting at. When you composite a set of amplifiers properly, you can obtain a far higher net gain (see I did it too lol) in performance than with one alone.

 

[DVDdoug]

To me an amplifier is a "black box" and I don't care too much about what's inside. It's the performance that counts.

Most modern amps are class-D with MOSFETs for the output stage. Class-D has a microchip (or two or more) and there may be bipolar transistors or analog op-amps in the "earlier" gain stages.

Most solid state amplifiers are better than human hearing, except some have more noise than others. Whether the noise is audible or not depends on the sensitivity of your speakers, how close you are to the speakers, and ambient noise in the room.

P.S.
Of course, manufacturers usually tout their design decisions and trade-offs as a feature or advantage.

 

[BKr0n]

Most modern amps are class-D with MOSFETs for the output stage. Class-D has a microchip (or two or more) and there may be bipolar transistors or analog op-amps in the "earlier" gain stages.

For DIY though that's a bit more of a daunting prospect. Going back and forth between linear and nonlinear takes far more design consideration than say a class A or AB.

Most solid state amplifiers are better than human hearing, except some have more noise than others. Whether the noise is audible or not depends on the sensitivity of your speakers, how close you are to the speakers, and ambient noise in the room.

All true, but certain parts are better at different things than others. Good example is just the inherent difference between BJTs and MOSFETs that one is a current controlled device while the other is voltage controlled. In different situations, one can be better than the other.

 

[stoo23]

I remember some years ago, we were running these Huge "Australian Monitor" Mosfet amps to drive the EV MT-4 speakers we were running for 'Raves' and Similar High Energy 'gigs' at the time and they actually produced More Heat when Idle than they did once being driven with Signal

I guess, if you are designing / building your own design, it may actually be Worth the Extra expense of purchasing more 'devices', so as to test and 'match' them, so as to (theoretically) get their ultimate performance.
You could always sell the one's that didn't 'make the grade', on ebay

 

[DonH56]

I am not up on the latest audio devices. For a power amp, RF transistors likely have too low breakdown voltage, and far too high a noise corner for audio. A HEMT is likely to have a 1/f noise corner of 100 MHz or more, for example.

Darlington devices may be harder to compensate in a feedback loop due to their high gain though are easier on the driving stage.

 

[DVDdoug]

For DIY though that's a bit more of a daunting prospect. Going back and forth between linear and nonlinear takes far more design consideration than say a class A or AB.

That's a good point. I wouldn't build a class-D amp unless I was using a chip. In fact if I was building any amp, I'd build it around a chip even if there were MOSFET or transistor "boosters" on the output.

Feedback (negative feedback = "corrective feedback") can help to linearize EVERYTHING, reducing distortion, flattening frequency response, lowering the output impedance, and making the gain dependent on resistors instead of relying on raw transistor/MOSFET characteristics.

at the time and they actually produced More Heat when Idle than they did once being driven with Signal

That's unusual!!! I assume they were class-D amplifiers. Class-D amplifiers are highly efficient, but all amplifiers are zero-percent efficient at zero-output and the class-D might be burning more energy than class A/B when idle. Still, I'd expect more heat/wasted energy at higher outputs than at idle with any class, except class-A which pretty-much burns full-power full-time...

 

[BKr0n]

I am not up on the latest audio devices. For a power amp, RF transistors likely have too low breakdown voltage, and far too high a noise corner for audio. A HEMT is likely to have a 1/f noise corner of 100 MHz or more, for example.

From my understanding a lot of the usual manufacturers of lateral mosfets and the like are more focused on switched amplification than traditional transistor biasing. Toshiba and Exicon seem to be the two names that come up the most (sanket as well, but not as often). Another integrator called linear systems makes a lot of older transistor models as well. Think of it as kind of a hall of fame of sorts where they reproduce old designs that were popular, or make direct drop ins for people who like their equipment as is. Other than that, not a whole lot in terms of specifically discreet audio transistors. I do however think that just because something isn't "intended" to be used as something, doesn't mean it flat out can't. I saw a while ago an article about using half bridges as output stages so I may look into that as well. Heck, I even saw one about class A GaN, but I think that's begging for the silicon genie to pop out of the chip lol

That's a good point. I wouldn't build a class-D amp unless I was using a chip. In fact if I was building any amp, I'd build it around a chip even if there were MOSFET or transistor "boosters" on the output.

I'm actually considering doing a class D for something else. A possible 90% efficiency is too hard to ignore. That and frankly that's where all the money is going so there's a lot of sexy chips out there for just that. Using a chip amp definitely simplifies things, but they will never match discrete. Especially for class D where the secret sauce is in the linearization of a nonlinear system. You're depending on whichever manufacturer to do that for you.

Feedback (negative feedback = "corrective feedback") can help to linearize EVERYTHING, reducing distortion, flattening frequency response, lowering the output impedance, and making the gain dependent on resistors instead of relying on raw transistor/MOSFET characteristics.

This is why I never understood strictly linear amps. I get that there are some really good sounding ones out there, but one of the core principles of amplifications is the feedback loop. So why not use it?

That's unusual!!! I assume they were class-D amplifiers. Class-D amplifiers are highly efficient, but all amplifiers are zero-percent efficient at zero-output and the class-D might be burning more energy than class A/B when idle. Still, I'd expect more heat/wasted energy at higher outputs than at idle with any class, except class-A which pretty-much burns full-power full-time...

I would suspect a bad supply then. Bad supplies can be space heaters if not made well.

 

[stoo23]

No 'Bad Supply' !! The things (AM 1600's), were Solid as a Rock and performed Flawlessly, in fact they were rather famous for their Power etc and Sonic performance.
http://docs.australianmonitor.com.au/website/Discontinued/Australian Monitor/AM1600-Brochure.pdf

They Made (and still do), make some great and from my experiences, very reliable products.

 

[DonH56]

From my understanding a lot of the usual manufacturers of lateral mosfets and the like are more focused on switched amplification than traditional transistor biasing. Toshiba and Exicon seem to be the two names that come up the most (sanket as well, but not as often). Another integrator called linear systems makes a lot of older transistor models as well. Think of it as kind of a hall of fame of sorts where they reproduce old designs that were popular, or make direct drop ins for people who like their equipment as is. Other than that, not a whole lot in terms of specifically discreet audio transistors. I do however think that just because something isn't "intended" to be used as something, doesn't mean it flat out can't. I saw a while ago an article about using half bridges as output stages so I may look into that as well. Heck, I even saw one about class A GaN, but I think that's begging for the silicon genie to pop out of the chip lol

My comment was specific to RF transistors since you mentioned them, which are optimized for much higher frequencies and thus may (frequently do) have compromised performance for high-power low-frequency applications. Noise and lower-voltage operation are two I mentioned; there are others. It is difficult to design a device for high voltage and high current at high RF frequencies. A GaAsFET or HEMT may have a noise corner of >100 MHz thus noise at audio frequencies is very high.

There are other cases where a particular device may not be optimum for specific audio applications. For example, flicker noise in MOSFETs makes them unsuitable for small-signal low-frequency audio applications, where BJTs and JFETs are commonly used. For phono preamps, particularly MM cartridges, BJTs area poor choice because their base bias current introduces high shot noise due to the high impedance required, so JFETs are typical.

GaN does not involve silicon...

Use whatever device you want, but with many decades of device research behind them, there are good reasons why certain devices are used for certain applications. History can be useful at times.

As an aside, a "linear amp" does use feedback. Those that do not exhibit higher distortion and higher sensitivity to environmental conditions and so forth (e.g. biasing changes with temperature, etc.) Even "zero-feedback" amps typically have some local feedback around the devices; it is virtually impossible to avoid having a little bit just from device parasitics adding local degeneration and so forth. Some (not all) of the zero-feedback amps I have seen do include feedback in the biasing circuits for operating-point stability.

 

[MaxwellsEq]

Welcome to the rabbit hole of gain-stage design! I've been reading debates about this for decades and I don't think I've ever seen a consensus. I've always had a soft spot for cascode arrangements.

 

[DonH56]

Welcome to the rabbit hole of gain-stage design! I've been reading debates about this for decades and I don't think I've ever seen a consensus. I've always had a soft spot for cascode arrangements.

Me too, especially for phono inputs, to reduce the Miller Effect.

 

[BKr0n]

Use whatever device you want, but with many decades of device research behind them, there are good reasons why certain devices are used for certain applications. History can be useful at times.

100% agree with you there. Whenever people ask me where to start in electronics, I always tell them to Google some of the history of it because it's much easier to understand the tech once you know where it comes from and how it evolves. Relating it back to audio, I can't find a specific technical reason as to why the switch to FETs became a thing. It seems more about either marketing or availability. I've seen some literature say that FETs do have less noisy frequency response, but I've also read that BJTs can do this far better, hence why I made the post. @MaxwellsEq has the right of it that there's no clear consensus. I figured since both technologies have merit, maybe running them composite would provide an overall benefit.

Welcome to the rabbit hole of gain-stage design! I've been reading debates about this for decades and I don't think I've ever seen a consensus. I've always had a soft spot for cascode arrangements.

I've been considering cascode, but I'm holding off on device selection until I figure out the output stage so I can better match the Idss.

 

[DonH56]

FETs are cheaper to make and generally easier to bias. In the early days their resistance to thermal runaway was touted, but in power devices that is not always true. An ideal FET exhibits lower distortion than a BJT, albeit with lower gain and higher output impedance.

 

[BKr0n]

FETs are cheaper to make and generally easier to bias. In the early days their resistance to thermal runaway was touted, but in power devices that is not always true. An ideal FET exhibits lower distortion than a BJT, albeit with lower gain and higher output impedance.

So really it's all on a case by case basis and not apples to oranges?

 

[DonH56]

So really it's all on a case by case basis and not apples to oranges?

I am not sure what "apples to oranges" means in this context. I am (was) an IC designer, not (usually) for audio, but device choice is one of many variables in the design process. Most everything is application-specific, and usually a trade (compromise) among conflicting features (or, pros and cons). There are good and bad amps using FETs, BJTs, etc. and there are often different devices sprinkled through the design to use the best device for the circuit.

 

[BKr0n]

I am not sure what "apples to oranges" means in this context.

As in it's more about how they're used as opposed to the fact they're two different devices.

There are good and bad amps using FETs, BJTs, etc. and there are often different devices sprinkled through the design to use the best device for the circuit.

So then in the case of an output stage, is there a specific transistor that is more optimal for that amplification stage, or does it really all depend on the circuit as a whole?

 

[DonH56]

As in it's more about how they're used as opposed to the fact they're two different devices.

They are different devices... Operation is very different, fabrication is different, etc.

So then in the case of an output stage, is there a specific transistor that is more optimal for that amplification stage, or does it really all depend on the circuit as a whole?

It's a choice to be made in conjunction with all the myriad other variables. BJTs may give lower output impedance but feedback can counter that, FETs may allow voltage swings closer to the rail but circuit techniques can counter that, it's complicated.

Note different folk may define "output stage" differently. There may be no voltage gain in the output stage, so may not be referred to as an amplification stage though current (and thus power) amplification does occur.

 

[BKr0n]

Note different folk may define "output stage" differently. There may be no voltage gain in the output stage, so may not be referred to as an amplification stage though current (and thus power) amplification does occur.

Being that they are devices for different purposes, is there a rule of thumb for which kind of transistor should go where?