MOSFET Follies, Third Edition

[R.G.]

I've been doing some simulation work, and I have found that I told you guys wrong in my earlier stuff about subbing MOSFETs for triodes in tube amps.

What I was wrong about:
I said you can't use MOSFETs to replace tube gain stages because the high gate-source capacitance would eat all of the highs.

It turns out that the particular power MOSFETs I was looking at have a LOT of transconductance. Much, much more than a tube. You can trade away some of this transconductance to reduce the apparent input capacitance and still get good audio band gain.

I was wrong another way - there now exist power MOSFETs that have both high transconductance and low Ciss. There are even depletion mode MOSFETs that bias like a tube.

For grins, I simulated a Fender phase inverter but with IRF820 devices instead of 12AX7 sections. Used stock, the frequency response was absymal, even by guitar standards. It started rolling off highs at about 1.2kHz. Dull. But it had almost 60db of voltage gain.

But when I put in 1K of source resistor in each MOSFET source lead, the gain dropped to 25db and the frequency response increased out to -3db at 140kHz! Compare that to a 12AX7 doing 26db and 123kHz.

(Before I get the replies that simulators are not real world and that believing in simulators indicates naivety, I'm well aware of what simulators do and don't do well, and I'm long accustomed to sidestepping some of the worst issues.)

It looks like a MOSFET phase inverter is a practical thing to do.
1. No, I do not expect it to sound exactly like tubes do. However, if what you like to listen to is not phase inverter distortion, this circuit removes most of that. I don't contend that it's the same, just practical and maybe useful in some setups.
2. There are some low-Ciss MOSFETs that will do this better than the IRF820. Zetex makes some 450V N-channel devices with 50pF Ciss that come in a TO-92 package and will put up with the 2ma and 1/4W dissipation that a phase inverter tube section needs to do.
3. There are some low-Ciss depletion mode MOSFETs that will bias in the same manner as a tube (i.e., cathode resistor biasing) that will do the same thing; Supertex makes them - TO-92 package, 0.74W and 300V to 450V ratings.

Both of the low-Ciss families will either give you more PI gain or extend frequency response further.

Note: if you experiment with this, you must put a 100 to 1k resistor in series with the MOSFET lead directly at the MOSFET, and for MOSFETs which do not already have a built-in gate protection circuit, you must provide either a one- or two-zener gate protection to keep the gates from puncturing on transients.

The depletion mode devices are very interesting. I'll have to dig more into those.

I'll post some schematics at GEO when I get a chance.

[Don Moose]

Originally Posted by R.G.

Note: if you experiment with this, you must put a 100 to 1k resistor in series with the MOSFET lead directly at the MOSFET,

Which lead, please?

[R.G.]

Doh!!!

The GATE lead! Sorry - my mind runs ahead while my fingers slowly unload the typing buffer.

It's a gate stopper resistor to keep the MOSFET from oscillating at VHF and/or UHF where us o'scope poor people can't even see it.

[Amp Kat]

Well let me ask you this. Do you think you could make it more asymetric than a tube stage and get more even order harmonics if you did ?

[R.G.]

Originally Posted by Amp Kat

Well let me ask you this. Do you think you could make it more asymetric than a tube stage and get more even order harmonics if you did ?

It can actually be both, in a way.

First: triodes are 3/2 power devices in that the current through the plate is proportional to the 3/2 power of the voltage from grid to source. MOSFETs are square law devices; the drain current is proportional to the square of the grid/source voltage. Inherently, MOSFETs make more signal-squared and less signal-cubed distortion before you go messing with them by using linearization.

Second: MOSFETs have much higher transconductance than triodes. A typical power MOSFET has a gm of 1A/V, or 1000mS. A 12AX7 has a gm of 1200 to 1600 MICRO-mhos. The Siemen is the formal name given to the unit formerly know as the mho. So a typical power MOSFET has a bit less than a thousand times the gm of a 12AX7. So you can freely throw some of that high gain away in linearizing the MOSFET to make it much more linear than a tube would be in the same circuit. That means you can make the PI much more linear and let the distortion other parts of the amp show through, unobscured by the PI's distortion.

Then there's intentional imbalance. The stock differential PI often uses an unbalanced set of PI load resistors to deliberately get different gains from the two PI triode sections. That's because the traditional differential PI doesn't use a high enough impedance tail current source to cause equal gains on each side. The imbalanced load resistors are an attempt to rebalance things.

You can do the same thing in search of imbalance. This will drive one output tube with a bigger signal than the other. In an AB amp, this will give you obvious asymmetry.

Which brings us to control. Engineers like controls, even if they hard wire them in position. The use of MOSFETs in a PI gives you several more controls to shape the way an amp sounds, whether you use them or not. There is a crowd of musical techies who wrap themselves in the flag of "nothing but tubes, ever!" who will be dismissive, even offended at the thought of a bit of silicon in a tube amp. In doing that, they cut themselves off from the new range of adjustments and controls that can give an amp new sounds without losing the tube character.

There is not going to be one sound of a MOSFET used in an amp. There will be good ones and bad ones, just like there are good tube amps and bad tube amps. There are more controls to be adjusted, more sounds to be made.

[bob p]

Originally Posted by R.G.

Second: MOSFETs have much higher transconductance than triodes. A typical power MOSFET has a gm of 1A/V, or 1000mS. A 12AX7 has a gm of 1200 to 1600 MICRO-mhos. The Siemen is the formal name given to the unit formerly know as the mho. So a typical power MOSFET has a bit less than a thousand times the gm of a 12AX7. So you can freely throw some of that high gain away in linearizing the MOSFET to make it much more linear than a tube would be in the same circuit. That means you can make the PI much more linear and let the distortion other parts of the amp show through, unobscured by the PI's distortion.

Needless to say, my interest is piqued, especially following the discussions we've had in that other thread about using SS devices in support of vacuum tubes. I'm thinking that I'm probably going to take advantage of the idea of using the depletion mode MOSFET in support of the differential amp PI so that I can minimize the adverse "qualities" that a tube-based PI circuit offers.

Even though this is a case where the MOSFET surely helps a tube circuit to perform better, I would be remiss if I didn't look into this more closely, to see what other types of advantages can be gained. R.G., you've got my attention.

Based on the modeling that you've done so far, can you offer any opinions about how a MOSFET-only PI might compare to the MOSFET-enhanced tube-based diffamp we've been discussing in the other thread?

[Amp Kat]

Originally Posted by R.G.

It can actually be both, in a way.

First: triodes are 3/2 power devices in that the current through the plate is proportional to the 3/2 power of the voltage from grid to source. MOSFETs are square law devices; the drain current is proportional to the square of the grid/source voltage. Inherently, MOSFETs make more signal-squared and less signal-cubed distortion before you go messing with them by using linearization.

When Marshall first came out with a MOSFET amp I bought it and I hated that dang amp. It was very noisy and maybe the reason but your winning us over in the preamp section.

Originally Posted by R.G.

Second: MOSFETs have much higher transconductance than triodes. A typical power MOSFET has a gm of 1A/V, or 1000mS. A 12AX7 has a gm of 1200 to 1600 MICRO-mhos. The Siemen is the formal name given to the unit formerly know as the mho..

Right and I knew of the change but always wondered about the conversion unit from milli to micro as my tube data file is in Siemens and the 12AX7 as 1.6Si. I've also seen gfs for transconductance as gm on MOSFET's

Originally Posted by R.G.

So a typical power MOSFET has a bit less than a thousand times the gm of a 12AX7. So you can freely throw some of that high gain away in linearizing the MOSFET to make it much more linear than a tube would be in the same circuit. That means you can make the PI much more linear and let the distortion other parts of the amp show through, unobscured by the PI's distortion. Then there's intentional imbalance. The stock differential PI often uses an unbalanced set of PI load resistors to deliberately get different gains from the two PI triode sections. That's because the traditional differential PI doesn't use a high enough impedance tail current source to cause equal gains on each side. The imbalanced load resistors are an attempt to rebalance things.

You can do the same thing in search of imbalance. This will drive one output tube with a bigger signal than the other. In an AB amp, this will give you obvious asymmetry. .

Very awesome Rg will you be able to throw some values at us for this arrangement to try it out ?

Originally Posted by R.G.

There is a crowd of musical techies who wrap themselves in the flag of "nothing but tubes, ever!" who will be dismissive, even offended at the thought of a bit of silicon in a tube amp. In doing that, they cut themselves off from the new range of adjustments and controls that can give an amp new sounds without losing the tube character.

And then they put a stomp box right in front or a noise gate and effects in the loop that are full of nothing but SS devices. I was tube anal for awhile until I realized that all the great recordings and even experimental tones were a combination of the two not a whole lot of one of the other.
What kind of supply voltage are we talking on the drain and is this basically a diff driver or complimentary ?

[bob p]

That means you can make the PI much more linear and let the distortion other parts of the amp show through, unobscured by the PI's distortion.
...
Which brings us to control. Engineers like controls, even if they hard wire them in position. The use of MOSFETs in a PI gives you several more controls to shape the way an amp sounds, whether you use them or not. There is a crowd of musical techies who wrap themselves in the flag of "nothing but tubes, ever!" who will be dismissive, even offended at the thought of a bit of silicon in a tube amp. In doing that, they cut themselves off from the new range of adjustments and controls that can give an amp new sounds without losing the tube character.

There is not going to be one sound of a MOSFET used in an amp. There will be good ones and bad ones, just like there are good tube amps and bad tube amps. There are more controls to be adjusted, more sounds to be made.

As it turns out, I have a little listening experience as far as MOSFET and tube based HiFi gear goes, and my experience with MOSFET-based HiFi gear has been disappointing. As it turns out, I've just convinced a friend to replace a set of boutique MOSFET-based Magnum components in his den (his secondary listening system) with an Audio Research 6550-based Class AB amp. Although the Magnum amplifier sounded quite good on its own, the tube-based AR system has a dramatically different tone, and the Magnum paled in an in-home comparison.

That's not to say that the MOSFET based amp sounded at all bad -- It was powerful, it was musical, and it was accurate, though I personally had the opinion that the overall character of the amp was somewhat cold, analytical and sterile in comparison to a top quality tube amp. To me, that represents a significant hurdle for MOSFET based amps.

My experience is that MOSFET amps can sound very good, and HiFi tube amps can sound very good, but I think that there is some very fertile ground in the area of the hybrid approach. The AR amps do use MOSFETs to support vacuum tubes in performing difficult functions, and the amps do sound good. The natural question that follows then, is just how and where MOSFETS can be used to really improve the sound of a tube amp.

Of course, we're not the first people to think about this -- there are HiFi people who are already making hybrid amps, and even hybrid CD players. Tube based line-stages are becoming popular as well. R.G., right now I'm thinking that the 100W Class A1 tube amp project is definitely going to use the MOSFET-enchanced diffamp that we've been talking about. Since I'm not really an anti-SS zealot, maybe it would be worth looking further into the concept of the MOSFET based PI. It would certainly be interesting to compare the MOSFET-assisted tube diffamp type PI to a wholly MOSFET based PI.

[Gregg]

Correct me if I'm wrong but Marshall already used current source in a PI circuit in their 9xxx series of rack amps:

http://www.schematicheaven.com/marsh...9100_2x50w.pdf

The FB circuit is also interesting.

And then they put a stomp box right in front or a noise gate and effects in the loop that are full of nothing but SS devices. I was tube anal for awhile until I realized that all the great recordings and even experimental tones were a combination of the two not a whole lot of one of the other.

+1. What's the point of a tube based FX loop if 95% /or more/ of all effects are SS based. One or two more transistors wouldn't change anything drastically especially if they are so linear and neutral in sound as claimed.

[Steve Conner]

Great stuff RG! I never thought of that, but I guess it stands to reason that the input capacitance of a MOSFET will be mostly Miller capacitance, so degenerating the source will make the input capacitance smaller.

Having read this, I think I may well replace the 12AT7 reverb driver in my new amp design with an IRF820, as I'm pretty short of room in there. And then to avoid microphonics, I could replace the EF86 front end with a JFET and MPSA42 cascode, where will it all end!

For hi-fi applications, I still swear by my current feedback MOSFET stereo amp. I managed to prove to my satisfaction that it doesn't create any audible distortion products provided that it's not driven into clipping. Then it has the most vile sounding clipping behaviour, and a bright yellow warning light to discourage you from doing that...

[R.G.]

Originally Posted by Steve Conner

Having read this, I think I may well replace the 12AT7 reverb driver in my new amp design with an IRF820, as I'm pretty short of room in there.

I wouldn't use the IRF820 there. Even with degeneration, the Ciss is too high. You could drop in a Zetex ZVN0545A which is a 450V/90ma device with max Ciss of 55pF in a TO-92 package; or a Supertex DN3545 depletion mode device; or a Supertex TN2540 enhancement mode. These all have low input capacitances, small leaded package, and enough current and power rating to run a reverb. Watch the drain voltage running a transformer though, and clamp it below the breakdown of the device.

Originally Posted by Steve Conner

And then to avoid microphonics, I could replace the EF86 front end with a JFET and MPSA42 cascode, where will it all end!

Indeed! We have a saying here that I'll update for the UK from American usage. "This axe used to belong to King George (Geo. Washington here). It's had three new handles and two new heads."

[Steve Conner]

Hi R.G.,

I think considering how hot the 12AT7 runs as a reverb driver, I'd like a device with a heatsink tab. I also would have trouble getting the parts you recommended over here. I looked for the ZVN0545 the first time you suggested it, but couldn't find them.

I guess if the worst comes to the worst I can cascode a TO-220 MOSFET with something else, or drive the gate off a high-gain BJT. I've also seen a MJE340 used to drive a table radio OT in a hybrid ham radio receiver, because the designer ran out of current capacity on the low-voltage rails.

In my circuit, the reverb driver grid is driven off the wiper of a 1M pot, which itself is driven from a source impedance of about 50k.

About the drain voltage thing, I heard modern MOSFETs are avalanche rated. I wonder if old table radio OTs are, though

[Gregg]

Farnell UK stocks ZVNs:

http://uk.farnell.com/jsp/Semiconduc...sp?sku=9525424

[rocket]

Originally Posted by R.G.

The Siemen is the formal name given to the unit formerly know as the mho.

Working for the Siemens company I must insist that it is Siemens even in singular. Ok - might also be a typo. ;-)

[Amp Kat]

If you don't mind me asking what do you do for them Rocket and how and why did all this come about or why the change from mho's to siemens ?

[Steve Conner]

I wouldn't be surprised if the European Union passed a law requiring every physical unit to be named after a dead scientist.

[R.G.]

Originally Posted by rocket

Working for the Siemens company I must insist that it is Siemens even in singular. Ok - might also be a typo. ;-)

"Typo" was kind of you.

Being a product of the USA public school system, I know all the letters - it's just arranging them in the right order and rrright nummber that is a problem.

[rocket]

Originally Posted by Amp Kat

If you don't mind me asking what do you do for them Rocket and how and why did all this come about or why the change from mho's to siemens ?

I am doing SW development for mobile communications.

Unfortunatly quality audio stuff for studios is no longer produced by siemens, anyway I started here only little more than 1 year ago.

Apparently the SI unit "siemens" was introduced in 1971.

http://en.wikipedia.org/wiki/Siemens_%28unit%29

[bob p]

Originally Posted by Steve Conner

I wouldn't be surprised if the European Union passed a law requiring every physical unit to be named after a dead scientist.

Steve, you should do standup comedy.

I think that its wasted effort to re-name a unit of measure that is already universally accepted (Mho), and to replace it with an eponym that honors a famous industrialist (Siemens). Typically, eponyms are reserved to honor people who made a signifcant scientific discovery, or discovered a principle to which their name becomes attached. I don't think its the case that Ernst Werner von Siemens discovered the Mho or was the first to describe it. It would seem that getting rid of the Mho and replacing it with Siemens is just the case of honoring one of the world's weatlhiest industrialists just because his company is one of the big players in world manufacturing and has enough clout to attach their corporate name to a unit of measure. Of course its easy to understand why Siemens would want to do it -- That's more prestigious than having a single letter ticker symbol on the NYSE.

The abbreviation for the unit, "S", does lead to some ambiguation when it comes to performing calculations. I'm not the first person who's mistaken the letter "S" as the number 5 when performing a calculation. But them I also find it ambiguous to have three homonyms that all sound the same: seamen, semen, and Siemens.

[rocket]

Originally Posted by Steve Conner

I wouldn't be surprised if the European Union passed a law requiring every physical unit to be named after a dead scientist.

There is no need for a law: All SI units regarding electromagnetism are already named after dead scientist: VOLT, AMPERE, OHM, MHO, (which is OHM backwards, and I can easily understand that this awkward construction has been replaced), TESLA, GAUSS, FARAD, HENRY

PS: Will you ever forgive Edward Heath for getting into to EU?