The sound of VMOSFET

Status
Not open for further replies.
Disabled Account
Joined 2003
here is Soundstage's review of Ayre V5x, with Charles' commentary. Sounds like a pretty interesting amp. I also read somewhere else that the V5x uses current mirror and T-driver stage.

enjoy.

http://www.soundstage.com/revequip/ayre_v5x.htm

Charles, can you give some background on your comments about the VMOSFET? It is the first time I have heard of them.

=============from the Soundstage interview===========
But you're using MOSFETs, right?

"Nope."

But everybody agrees that MOSFETs are more like tubes and simply sound better. Don't they?

"You've got to be careful about making generalizations about MOSFETs because there are vertical MOSFETs and lateral MOSFETs and they don't behave exactly the same. Almost no one uses the Hitachi lateral MOSFETs, though -- and while the vertical ones are very linear in the lower regions, they have a wonky, non-linear input impedance that results in some really nasty high-frequency distortion characteristics. Well, nasty-sounding to me, at least."

============the end============================

I am curious if anyone else has the same observation.
 
All power MOSFETs have a capacitive input impedance.

What is important however, is that this input impedance is comprised of Cgs which behaves pretty much as a simple capacitor (at least to a first order approximation), and Cgd which is the problem.

If you have a common-source stage the Miller feedback from drain to source via this capacitance causes the input impedance to dynamically change. This is especially noticable in SMPSUs where Vgs (when the gate is driven from a non-zero source impedance) shows a distinct 'plateau' due to the NFB from drain to gate. This reduces the switching speed of the FET quite noticeably.

Lateral power FETs show the problem to a lesser extent because Cds tends to be a lot lower than for vertical types.
 
Disabled Account
Joined 2003
I guess my puzzle is to reconcile between the excellent reviews those Pass Labs amps got and Charles' take on the mosfets. Since Pass uses exclusively vmosfets, and the reviews for them seem to emphasise how good they sound.

I am at a loss as to how the two are miles apart.
 
Ouroboros said:


What is important however, is that this input impedance is comprised of Cgs which behaves pretty much as a simple capacitor (at least to a first order approximation), and Cgd which is the problem.

If you have a common-source stage the Miller feedback from drain to source via this capacitance causes the input impedance to dynamically change. This is especially noticable in SMPSUs where Vgs (when the gate is driven from a non-zero source impedance) shows a distinct 'plateau' due to the NFB from drain to gate. This reduces the switching speed of the FET quite noticeably.
But in the case of a output stage the mosfets are usually used as source folowers...so this parasitics capacity Cgd is not amplified by Miller effect as the stage doesn't have voltage gain..actually a litle loss..
The Cgs (Gate source) also become diminished by the bootstrap action of the folower...

In a nutshell ...in a mosfet folower the parasitics capacitys are less restrictive than in a commun source gain stage...:)
 
In MOSFET's the reverse capacitance is really a problem, it looks small in datasheet but it's magnified when the Vds changes and is very noticeable even in a source follower, the Cgd impact is far greater because with an amplifier with for instance a rail voltage of +- 50 V the Cgd is working against a voltage change of up to 100 V swing related to the Gate comparing to the Cgs which is working against a voltage swing of just some volt.

From my experinces working with SMPS's is that the reverse capacitance effect can be seen very clearly although many SMPS IC's have a peak current drive capability around +- 1 A like UC3845 and even more and still the voltage plateau can be seen on the gate.

I tried to compare 2SK135 (Old classic Hitachi FET)with IRF630 (Obs, not 630N version, just 630 for clarity!) that seem to be quite close in performance, the datasheets are not really comparable because they use diffrent parameters when stating for instance the Crss but the 2SK135 have much smaller capacitance.

In the datasheet for IRF630 we can se that Qgs is 7 nC but for Qgd is clearly greater with 23 nC for a certain condition, so the conclusion is that the Cgd (or Miller capacitance) has a clear impact.

Edit: Vertical FET's have a distictive behaviour around the Vgs threshold and probably why Charlse Hansen don't like them, they are just hard to drive "linearly" and it's not so much depending on the topology, eg. CD or CS.

Maybe some input from EVA who sholud know this issue too. :cool:

Cheers! ;)
 
Mosfet= Hexfet Mosfet= Vmosfet

Millwood,

The IR MOSFET or HEXFET is has a hexagonal geometry so there are a bunch of the hexagonal cells in the devices in the MOSFET. The more cell the lower the on resistance and the higher the capacitance.

The V channel MOSFET was a Siliconix deal, where it is similar to a planner device with a V grove structure. Therefore, the VMOSFET is like a Big FET with a V channel. So you can't get a V grove MOSFET device with the kind of low on resistance like the IR HEXFET.
 
Re: Mosfet= Hexfet Mosfet= Vmosfet

jewilson said:
Millwood,

The IR MOSFET or HEXFET is has a hexagonal geometry so there are a bunch of the hexagonal cells in the devices in the MOSFET. The more cell the lower the on resistance and the higher the capacitance.

The V channel MOSFET was a Siliconix deal, where it is similar to a planner device with a V grove structure. Therefore, the VMOSFET is like a Big FET with a V channel. So you can't get a V grove MOSFET device with the kind of low on resistance like the IR HEXFET.



But the old Siliconix V and U grooves, the IR 'hexfet', the On-semi square cell, the Philips 'trench' structure, etc all have essentially vertical current flow between source and drain. The Hitachi (and Magnatec) audio FETs have horizontal current flow. The lateral geometry is what keeps the capacitance down.
 
Ouroboros,

That all true, still there are major differences in the geometry of these devices. I don't pretent to know which one sounds the best, however the IR parts lead in sales in the major markets.

One thing I do like about the L MOSFET is the biasing.
 
The one and only
Joined 2001
Paid Member
The sound of any device is very dependent on all the other
details of the circuit.

Responding to an earlier comment, the gain device doesn't
know what mode (Common Drain, Source, or Gate) it is being
used in. It only sees the varying voltages and currents, and
these can easily be identical from the device's point of view.
 
Re: Commentable Thoughts

amp_man_1 said:
We only use Vertical IRF Mosfets because their performance is True to Life Professional Standards.

Hello -

That's exactly the opposite experience to what I've had. Our first power amp (the Ayre V-3) was 100% vertical MOSFETs, including the discrete zero feedback power supply regulators. When we were selecting the devices to use in that design we found a funny thing with all of the International Rectifier P-channel devices we tested.

If you put them on a curve tracer, the forward trace and the return trace did not overlay at all. After running several experiments we found that their transconductance would shift in the first few milliseconds after Vgs was changed. This shift was not small, but on the order of 10% to 20%. You will note that the time constant of this effect put the misbehavior right in the middle of the audio band.

This was behavior was found on every single IR P-channel device we tested, using at least 5 samples each of 5 different part numbers. No other manufacturer's devices exhibited this behavior, including Siliconix, Motorola (now On Semi), Harris (later Infineon and now Fairchild), and at least one other I can't recall now.

The original tests were performed around 10 years ago, in 1992 and 1993. I revisited this issue about 5 years later thinking that IR might have corrected their parts, but there was no difference.

I spoke with the application engineers at IR about this and they refused to help, saying that the parts were designed for switching and not linear applications. However, an apps engineer from a competing manufacturer knew exactly what I was talking about and even knew the mechanism that caused the problem in the IR parts.

As long as we were using vertical MOSFETs, we refused to use the P-channel parts from IR. (Their N-channel parts seemed to be fine, although we only used these in the case of an emergency since they had been so unhelpful on this issue.)

Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
 
Disabled Account
Joined 2003
I think the jury is still out as to how badly vmosfet sounds. Charles is in the camp of "they suck big time", and others seem to disagree, and by surveying the pass forum, there are quite a few of them. not to mention the commercial success of pass products and the fact that so far we have not been able to come up with a mechanism that would suggest poor sounding vmosfet.

so for now, I will continue to use my IRF devices until more experts come out and present a more comprehensive view on this.
 
Charles,

interesting note about IR P-ch. FET's, is it possible to point out that effect in a datasheet??

I picked out the following transistors below as an example, I couldn't find any clear sign that there should be such a thing and so even when compared N & P-ch in between.
Just the common delay thing for Fet's but they are quite the same for both.

IRF540N

IRF9540N


Thanks
 
Status
Not open for further replies.