Best MOSFET for output stages

[KSTR]

To late to edit above post...

.. but now I found what look to be the "official" Magnatec parts with improved oscillation stability, and better looking dataheet curves (compare the transfer characteristics vs. the ACD datasheet curves) :

Magnatec. ALFET Lateral MOSFETs

Of course it is still unclear if and how much they differ from the Class-D custom parts... to me it looks they might be the very identical parts because the ACD's have that "Alfet" tag on the Pro Audio website plus the info given on the magnatec webite.

Seems the Brits love to confuse customers (they already did with Magnatec vs Exicon brands; an application engineer from profusion.plc told me that they are absolutly identical parts to the best of his knowledge).

Klaus

[tvicol]

These are Semefab parts not Semelab.

Regards,
Tibi

[CBS240]

Originally Posted by Jon Lord

Here again my question, did anybody test these:

FQA36P15
FQA46N15 from Fairchild

These seem to be rather powerful and robust VMOS (planar stripe) type, are damn cheap, low Crss and can dissipate at least 250W.

They already worked in a prototype of mine, but I never qualified them any further (measurement and listening wise).

Hi

I use these types of fets in my amp, although not exactly these particular ones, they're all similar. As you can see they have terrible transfer characteristics compared to lateral fets but then they are really cheap. Of course all mosfets have the non-linear capacitance and variation of Vgs vs Id affecting linear operation, but with these you can see on the datasheet that Id is quite dependent on Vds. These devices are obviously designed for switching but, IMO, if you wish to realistically use these for a quality linear output stage you have to use some sort of EC. I do think that if you have no fear of the extra complexity, using a proper EC gate drive circuit is the way to go for all mosfets. I use the feedback/feedforward of HEC and, if implemented well, it works quite nicely. These fets really are very tough components even after thermal de-rating. Unfortunately they also require a bit more bias and a more robust driver circuit. Here is an example of how I used them. Obviously the IL300 is not required, I was just having fun experimenting. This first build is a success and I'm on the second build (I have 5 PCB's), but now I'm out of town on business for a couple of weeks.

[ontoaba]

Originally Posted by CBS240

Hi

I use these types of fets in my amp, although not exactly these particular ones, they're all similar. As you can see they have terrible transfer characteristics compared to lateral fets but then they are really cheap. Of course all mosfets have the non-linear capacitance and variation of Vgs vs Id affecting linear operation, but with these you can see on the datasheet that Id is quite dependent on Vds. These devices are obviously designed for switching but, IMO, if you wish to realistically use these for a quality linear output stage you have to use some sort of EC. I do think that if you have no fear of the extra complexity, using a proper EC gate drive circuit is the way to go for all mosfets. I use the feedback/feedforward of HEC and, if implemented well, it works quite nicely. These fets really are very tough components even after thermal de-rating. Unfortunately they also require a bit more bias and a more robust driver circuit. Here is an example of how I used them. Obviously the IL300 is not required, I was just having fun experimenting. This first build is a success and I'm on the second build (I have 5 PCB's), but now I'm out of town on business for a couple of weeks.

I agree. Also I simulate it like this:
QUASI is not that bad, really, (I post some proof here).
and they looks amazing, even old IRFZ44 (without N).

[CBS240]

Hi

I am always leary of simulation results used for predicting the finalized real operation for output stages because simulated devices are not subject to thermal stress and the physical changes in operation of the devices related to those thermal changes. But it will show the general operation of such devices. Here is an example of EC showing the related distortion with complementary hexfet transistor followers of relatively equal trans-conductance that I posted a couple years ago. One important thing I found with the HEC circuit is that the error amplifiers must be in physical contact with the outputs because to not have them as part of the thermal transfer fucntion creates a variable bias depending on the random temperature of the error amp components. I have since scraped the idea of a seperate Vgs multiplier. But the experiment shows the error signal required to undo most of the terrible distortion these transistors inherently create. Another point to make about this type of error correction is that not gain but rather bandwidth of the error amplifier is what's important. The higher the BW the more error can be corrected for. Because of this, I would not recommend using op-amps as the EC amps because they are generally limited in BW compared to a single transistor stage. Also a global fb loop is quite limited in this task as well because of the many stages within the loop and that the BW is purposely limited in order to maintain stability around the VAS due to voltage gain being greater than 1. Hence the issue with using hexfets as a straight up follower. Most people hate these devices for this reason and that they do not know how to stabilize them correctly but I feel those folks have not had a chance to experience them properly to see, or rather hear their true potential in overall BW and power handling ability with regard to driving reactive loads.

[Michael Rothacher]

Let me guess, this Circlotron beast sounds a little hot and edgy with lots of muddy bass ...

Actually, no. Quite the opposite, in fact with a variety of speakers.

[Samuel Jayaraj]

Originally Posted by Michael Rothacher

Actually, no. Quite the opposite, in fact with a variety of speakers.

So Michael, how do you compare the ciclotron with the F5 and other amps which have more dominant H2 than H3? Thanks.