Hi, quick question re. filtering FETs when search on, for example, rs uk. In the past I've worked on switched motor drive projects that have utilised power FETs with sharp turn on characteristics ie. when the Vgate threshold is reached Rds falls sharply. I'm now interested in audio amp outputs and have noticed one spec'ed device has much slower turn on. I'm thinking that modern devices might all have sharp transfer characteristic and this can be softened with a drain resistor. Is this right? When I search for FETs on rs uk capable of dissipating 10W or so the Vgate thresholds seem to be ~2V rather than ~0.2V for the SK134 device below. Any help in my understanding would be appreciated.

Robin

2SK134 NMOS, graual turn on:


NVD5865 NMOS (fast turn on):

 

The original MOSFETs were lateral types, and they had much lower Gfs, much higher Rds(on) and lower Vgs(th). They were quickly superseded in most applications by vertical types with higher Gfs and Vgs(th) and much lower Rds(on), which gave them higher efficiency in switching applications, but by then, thanks to Sempei and Ohashi's application note they had made a name for themselves in audio, and instilled a belief in many people that they were the only MOSFETs suitable for audio.
Any MOSFET can be used for audio - why use one with low gain when higher gains are available - you wouldn't do that with a bipolar transistor! (You can, as you said, add a source resistor to reduce the Gfs, but it is not necessary.)
The lateral MOSFET's only neat trick is the position of the zero-temperature coefficient point on the Vgs/Id graph. (On your graphs it's the point where all the lines intersect). If the amplifier operates above that point it is immune from thermal runaway, without any temperature compensation circuitry. If the amplifier operates below that point, it needs temperature compensation, just like a bipolar amplifier would.
You can make just as good an amplifier with IRF530/IRF9530 or IRFP240/IRFP9240 pairs as you can with 2SK134/2SJ49.
At the same time as Hitachi released their devices, Toshiba released 2SK405/2SJ115, which were intended for audio, but they were vertical types and the circuit needed temperature compensation.
All these older devices (IRF510/520/530/9510/9520/9530 included) had smaller chips in larger packages, intended for linear operation, which gave the device low capacitances but good heatsinking.
Anything recently introduced in a TO-247 package has rather too much capacitance to make a good audio device.

 

when the Vgate threshold is reached Rds falls sharply

Consider the transistor in emitter follower mode. Its base turns on at 0.65V but is off at 0.5V. It turns on/off with a smaller voltage difference than the MOSFET. (sharper knee) I think you are worrying about nothing. (you also can make this circuit with a MOSFET)

 

Consider how a linear amplifier works built with op-amps.
A typical op-amp has a gain of >100,000.
Ideally you want an op-amp with infinite gain. You then reduce the gain and improve linearity of the circuit by applying negative feedback.

Linear amplifiers designed with MOSFET or BJT is the same thing. You want high gain and then apply negative feedback to make it linear.

 

Hi, quick question re. filtering FETs when search on, for example, rs uk. In the past I've worked on switched motor drive projects that have utilised power FETs with sharp turn on characteristics ie. when the Vgate threshold is reached Rds falls sharply. I'm now interested in audio amp outputs and have noticed one spec'ed device has much slower turn on. I'm thinking that modern devices might all have sharp transfer characteristic and this can be softened with a drain resistor. Is this right? When I search for FETs on rs uk capable of dissipating 10W or so the Vgate thresholds seem to be ~2V rather than ~0.2V for the SK134 device below. Any help in my understanding would be appreciated.

Robin

2SK134 NMOS, graual turn on:
View attachment 234090

NVD5865 NMOS (fast turn on):
View attachment 234091

To even qualify to be called a "transistor" some kind of smooth transfer function is expected at some place within the device's operation region.

Logic gates, digital logic in general, Schmitt triggers and so on are designed to have a very sharp on/off but plain old transistors are usually not, they are analog amplifiers almost by definition (I think).

 

These are great responses. Thanks a lot for taking the time to explain.

The comments on "why would too much gain be bad" makes sense in an ideal world but in my research I had stumbled across forums arguing that an amp should be vaguely linear and negative feedback should only be used to tighten it up, hence quotes such as this
"Power MOSFETs with lateral structure are mainly used in high-end audio amplifiers and high-power PA systems..."
I would have thought that overshoot in the MHz would not be an issue for any audio applications and moreover the response could be slowed with decoupling caps.

 

Let's get this straight. Decoupling caps are used for something else. They are used for reducing sharp transitions (high frequencies) in power supply rails caused by fast switching times in the signal circuitry.

For an audio amplifier, what you need is to band-limit the signal and you do this with a low pass filter. We don't call these capacitors decoupling caps.

 

Of course. The clue is in the name.

 

"Power MOSFETs with lateral structure are mainly used in high-end audio amplifiers and high-power PA systems..."

is absolutely true, because they are not used anywhere else. The converse - "all high-end audio amplifiers and high-power PA systems use lateral MOSFETs" isn't true.

an amp should be vaguely linear

The smaller input voltage you measure it over the more linear it looks. Hence with a huge amount of gain, and lots of negative feedback, smaller the voltage at the amplifier input and the more linear the output.

 

is absolutely true, because they are not used anywhere else. The converse - "all high-end audio amplifiers and high-power PA systems use lateral MOSFETs" isn't true.

Haha, good point. There's absolutely no other market for them.

So in summary there isn't a whole new world of MOSFET literature I should worry about. I can continue to build on my existing MOSFET experience

 

Another point - in most power amps, the MOSFETs are used common drain.
The voltage gain of a common drain stage is (Gm.Rload)/(Gm.Rload+1), which is just a little bit less than one.
If you change the device for one with much more Gm, the gain changes from a little bit less than one to a tiny bit less than one.

 

the speed of turn on of a mosfet is governed by the speed of its drive, this is not to be confused with the gain of the mosfet, i.e the rise in amps for a change in gate level.

 

This notion of "low feedback" audio amplifiers being higher fidelity started back in the early 70's with some papers by Matti Otala, measuring heavily lag-compensated audio amplifiers with slew rate limiting in the audio band. This gave rise to the TIM (Transient Intermodulation) distortion measurement. Unfortunately, some in the audio world misunderstood Otala, and blamed the poor TIM performance on large amounts of feedback rather than the slewing non-linearity that is the real culprit behind lousy TIM performance. In a properly designed audio amplifier without slewing limitations in the audio band, a high feedback factor improves pretty much all objective characteristics of the amplifier...

 

Hi, quick question re. filtering FETs when search on, for example, rs uk. In the past I've worked on switched motor drive projects that have utilised power FETs with sharp turn on characteristics ie. when the Vgate threshold is reached Rds falls sharply. I'm now interested in audio amp outputs and have noticed one spec'ed device has much slower turn on. I'm thinking that modern devices might all have sharp transfer characteristic and this can be softened with a drain resistor. Is this right? When I search for FETs on rs uk capable of dissipating 10W or so the Vgate thresholds seem to be ~2V rather than ~0.2V for the SK134 device below. Any help in my understanding would be appreciated.

Robin

2SK134 NMOS, graual turn on:
View attachment 234090

NVD5865 NMOS (fast turn on):
View attachment 234091

Don't worry about switching speed of FETs for a linear amplifier application. I expect you will be using feedback
in your circuit topology to control gain and bandwidth,

 

a high feedback factor improves pretty much all objective characteristics of the amplifier...

A good philosophy is to first design an amplifier to have the lowest possible distortion without feedback, and then apply enough feedback to establish the required amount of closed loop gain. The reduction in distortion at any given frequency will be reduced by the ratio of the open loop gain at that frequency to the closed loop gain (a.k.a. feedback factor).

Why do I mention frequency? Because of a little factor called “gain bandwidth product”. At some point all amplifiers will start to lose gain as the frequency increases. This is why distortion generally increases with increasing frequency in a feedback amplifier; there is simply less open loop gain available to make improvements with.