Grid Current Un-Limiting?

[exclamationmark]

I've read in Merlin's book that the onset of grid current limiting is determined by the source impedance, and that by lowering it so sufficiently low values, one can actually drive the grid positive. I know this is commonly used to drive power tubes into class AB2 operation, but are there any designs that use this in the pre-amp section? Most designs (and indeed most of my own designs) use humongous grid stoppers to effectively stop blocking distortion in high gain situations (and essentially make grid current limiting rather sudden), but I'm thinking of going the other extreme and direct coupling a cathode follower to a gain stage using a MOSFET (and hope my grids don't melt...).

I'm wondering about a few things though... much grid current can actually flow during operation? I'm assuming that when the tube reaches vgk=0v the grid impedance drops to something like 1k, and significant current will have to flow to produce a voltage drop through this impedance. What do grid curves actually look like above 0v? And perhaps most importantly, given that the driving impedance is sufficiently low, will the stage ever clip on the bottom of the waveform (or top... depends how you scope it). I'm thinking the plate/grid currents would gradually approach the limit set by heater emission (with the grid voltage increasing accordingly) until it reaches the literal definition of saturation.

Just throwing some ideas out there. I find this particular area of vacuum tube operation interesting for some reason...

[R.G.]

....I'm thinking of going the other extreme and direct coupling a cathode follower to a gain stage using a MOSFET (and hope my grids don't melt...).

Been there. Search for "MOSFET Follies" and "MOSFET Heresies". I did a design for Class AB2 followers using MOSFETs instead of cathode followers. There are some people here who have built this. It's also at geofex.

You're right - if you can provide the grid drive current, you can go right up to cathode emission limits. It was appreciated in the Golden Age, but largely forgotten that the clipping on a tube grid was source impedance dependent. If it's high, the clipping can be razor-sharp. If it's low, the change in grid voltage still exerts control on electron flow, since the fraction of electrons actually captured by a grid is small, and the positive grid accelerates electrons away from the cathode.

One interesting circuit I saw had the grid of a triode tied to B+ through a very large resistor. The grid was said to be running "in clamp", perhaps meaning "clamped to Vgs=0". Don't know. But the clipping on the grid side of things like that are sharp.

It would be interesting to use some of the TO-92 500V MOSFETs available now as a follower into a triode, I think. The lower the impedance driving the grid, the softer the grid clipping as the grid goes positive will get.

[Smitty]

Where to find the current?

R.G.,

I've always wanted to try this as a mod, but I'm not sure where I'm going to find the current required in my vintage Fender-style power supply. Let's say I pull two of the power tubes in my Twin Reverb. That's only going to give me about 150 mA on a good day. Will that be enough for a couple mosfets?

Are you referring to the supertex VP2450 or VP0550 in your post?

http://www.supertex.com/pdf/datasheets/VP2450.pdf

http://www.supertex.com/pdf/datasheets/VP2450.pdf

Thanks. And thanks for everything I've learned from you in the past by lurking here. And the others, as well. You know who you are!

Smitty

[defaced]

Those are both P channel enhancement mode devices, I don't think you'll be able to get them to work without serious headache - if at all (it's too early for me to think through if it's possible to have a P device driving an N device). You'll find it much easier to work with N channel depletion mode devices like the LND150. If you're going to drive preamp tube grids positive, I don't think you'll have any issues with power requirements. If you want to drive power tube grids positive, it'll be easiest to just add a secondary small transformer. All of my questions to RG on this topic are on this board. You'll find them if you do the searches RG mentioned in his above post.

[R.G.]

I've always wanted to try this as a mod, but I'm not sure where I'm going to find the current required in my vintage Fender-style power supply. Let's say I pull two of the power tubes in my Twin Reverb. That's only going to give me about 150 mA on a good day. Will that be enough for a couple mosfets?

Plus three points to you for thinking about the power supply implications of doing things to a power stage!!

The MOSFETs themselves and the current into the grids of power tubes will not necessarily pull a lot of current, if you do the design correctly. Remember that they're only driving the output tube grids, not any real load current themselves. That means they only supply whatever grid current flows, and that's a lot lower than the resulting change in output tube plate current. But the power tubes definitely will draw more current as you can drive them to bigger currents. That is going to have an effect on how much current comes out of the power supply, power supply ripple, and all the rest of what power supplies do, including get hot. I hadn't thought about it before, but in grid conduction territory, tubes do definitely have a "current gain" like bipolars - change in current out per change in grid current in. Hmmh.

The MOSFETs themselves don't necessarily use a lot of current. They are a Class A (i.e. always conducting, never off) stage, and only need to be biased with enough current so they are still conducting when they are at maximum voltage to the grid they're driving, including the current into the grid and any current in their source-pins loads. This is dependent on how much grid current the driven tube(s) need to be driven into positive grid conduction. Some experimentation is needed here; I don't have a good store of how big the grid currents really are.

This can be minimized by going to a source load that's an active device, not a passive resistor. That could be a P-channel device set up as a complementary source follower, or another N-channel set up as a constant current source load. The N-channel is probably easier and more predictable. I don't think I've drawn that up on the schemos at Geofex, but I can if it's needed.

The small add-on transformer to make, say, +/-75 to +/-100V for the MOSFET drivers is a good idea on several fronts. It eliminates the issues of what happens in the existing power supplies for driving the MOSFETs, and especially keeps you from having the idea to use the bias supply for a negative source voltage supply. Stay out of that bias supply unless you have good protection circuits on the output tubes!
It makes any changes more readily reversible and manageable from a conceptual standpoint. It just leaves you with the concerns of what happens now that your amp is pulling more power out of the power supply. That could be addressed directly with an "overtemp" sensor and indicator/alarm/whatever on the power transformer and perhaps power tubes. The excess power only happens on peaks that drive the amp to drink... er, into grid conduction so it may not be a 100% duty cycle increased power, depending on the playing style and power level demanded.

There are better MOSFETs these days than when I did the first stuff on this. The IRF800 series I mentioned first were dramatically more than needed to do the actual job. They were picked because they withstood the voltage.

[Smitty]

Source Follower to replace Cathode Follower

R.G.

Thanks for your wisdom and guidance on this subject. I've been having some real fun tinkering with direct coupled cathode follwers in guitar preamplification stages and have wanted to try the "Source Follower to replace Cathode Follower" recipe from your site.

You mention above that there are better mosfets for the job, can you point me in the right direction? Do you have a recommendation for a specific direct coupled MOSFET source follower for a 12ax7 that will help soften the transition into grid current limiting on the next stage grid and not draw too much additional current from the power supply? In my amp the gate of the MOSFET will likely see between 450 and 500 volts (unloaded B+ no tube sockets populated condition) when your circuit is energized and with no tube loading down the common-cathode-tube-stage plate resistor and about 200 volts with a tube drawing current.

In my all tube circuit, I'm using a 47K resistor and a 1n4007 connected to the cathode resistor of the follower to protect the grid of the follower. In my circuit it never sees north of 200 volts. Should I keep that? Modify it for use with a MOSFET?

You're kind to offer to design and post a recipe for a source follwer grid driver using "...a source load that's an active device, not a passive resistor. ...or another N-channel set up as a constant current source load. The N-channel is probably easier and more predictable. I don't think I've drawn that up on the schemos at Geofex, but I can if it's needed."

I'm not sure I need it but I sure WANT IT!!! I'll be happy to post results to the community with some snaps of the o-scope trace showing the grid north of zero volts.

Smitty

[exclamationmark]

I would think that the grid would have the potential to draw a lot of a current, given that the cathode emission is something like 10 times the usual operating current. Exactly what happens when the grid voltage is driven hugely positive hurts my brain though... Furthermore... What happens if the grid is somehow driven more positive than the anode?

Anyway...RF tubes that used to be run in A2 operation 100% of the time had specialized grids (presumably to stop them from melting), and I know that it is possible to damage grids this way when too much current is drawn (although hard to do). There are a lot of A2/AB2 audio amps out there, especially in triode strapped mode. Easy to find info if you look around. IIRC I was reading one thread and the mosfet supplied around 50mA peak at 60-70 volts on the grid of a (I think?) a 4-65A RF transmitting tube.

[R.G.]

You mention above that there are better mosfets for the job, can you point me in the right direction? Do you have a recommendation for a specific direct coupled MOSFET source follower for a 12ax7 that will help soften the transition into grid current limiting on the next stage grid and not draw too much additional current from the power supply? In my amp the gate of the MOSFET will likely see between 450 and 500 volts (unloaded B+ no tube sockets populated condition) when your circuit is energized and with no tube loading down the common-cathode-tube-stage plate resistor and about 200 volts with a tube drawing current.

In my all tube circuit, I'm using a 47K resistor and a 1n4007 connected to the cathode resistor of the follower to protect the grid of the follower. In my circuit it never sees north of 200 volts. Should I keep that? Modify it for use with a MOSFET?

You're kind to offer to design and post a recipe for a source follwer grid driver using "...a source load that's an active device, not a passive resistor. ...or another N-channel set up as a constant current source load. The N-channel is probably easier and more predictable. I don't think I've drawn that up on the schemos at Geofex, but I can if it's needed."

I'm not sure I need it but I sure WANT IT!!! I'll be happy to post results to the community with some snaps of the o-scope trace showing the grid north of zero volts.

That's a very nice offer. It may take me a little while to draw it up. All of my experimenting has been done with followers driving the power tubes where bias can be adjusted. I'll have to introspect a bit on the smaller but more fiddly job of driving a 12AX7 preamp tube. I would be very interested in how this works out. The trick is to figure out how to wrap the self-biasing into it.

What's different about modern MOSFETs is that their gates are not only completely insulated but also that the conduction mode is not just depletion mode, like tubes. They can be enhancement, depletion, or both. It's not a big deal if you can afford to hand-tweak each bias point, but that's too big a PITA to do.

That's what I get for opening my big mouth. Now I have to think. That is both exhausting and painful.

I would think that the grid would have the potential to draw a lot of a current, given that the cathode emission is something like 10 times the usual operating current. Exactly what happens when the grid voltage is driven hugely positive hurts my brain though... Furthermore... What happens if the grid is somehow driven more positive than the anode?

It comes down to electron ballistics. The electrons are boiled off the cathode into a cloud that hangs around the cathode, the cloud repelling some of them back into the cathode until equibrium is achieved between boil-off and repelled-in. The (relatively) remote distance anode attracts some of them with its positive charge, just as a positive screen grid does in a tetrode and pentode.

In a triode/tetrode/pentode, the signal grid is closer to the cathode than the screen grid, so its voltage exerts a bigger effect on electron movement than the distant plate. To keep electrons from traveling to the plate, you have to reduce the plate's voltage by the full B+ if you do it from the plate. If you leave positive voltage on the plate, the much-closer grid may only have to be a couple of volts negative to convince electrons to not go to the plate in spite of the big attraction of positive B+ voltage.

But if you change the grid voltage to exactly the cathode voltage, electrons are no longer repelled from the grid, and head for the plate, accelerating as they go. Most of them literally miss the grid, flying through the spaces between wires. A few hit the grid, so grid current starts. As you make the grid more positive, the grid wires attract more electrons; still the vast majority miss the grid wires, like a marble thrown at a wire fence.

It's this "miss the wires" thing that limits plate current. If an electron gets a little past the grid, there is that big, fat, hugely positive plate to go to, and most of them do. The positive voltage on the grid can only warp the path of the cathode electrons so much. Most of them miss.

Anyway...RF tubes that used to be run in A2 operation 100% of the time had specialized grids (presumably to stop them from melting), and I know that it is possible to damage grids this way when too much current is drawn (although hard to do). There are a lot of A2/AB2 audio amps out there, especially in triode strapped mode. Easy to find info if you look around. IIRC I was reading one thread and the mosfet supplied around 50mA peak at 60-70 volts on the grid of a (I think?) a 4-65A RF transmitting tube.

Yep. The grid wires are very thin, and only have so much emitting surface to radiate away the heat generate by capturing electrons. Melting could be a problem in severe circumstances.

However: the tube data books do tell you an approximate value for the grid resistance when it's driven positive. For the 12AX7, I've seen values around 5K. It's nonlinear, of course.

The grid current in a 12AX7 won't be huge, but it may cause problems. As I noted, I have not played with signal tubes doing this. But I think the grid current is going to be a couple of orders of magnitude smaller than the plate current, as befits the relative physical areas facing the cathode. I'd be amazed if the grid of a 12AX7 pulled over 1ma for any practical positive grid drive. I think you'd go into cathode emission limit first. But I don't really know.

I'll think about it.

[R.G.]

FWIW, I just looked at Mouser's supply of MOSFETs. There are now some 600V/180ma/1W devices in TO-92 packages. This is mildly amazing!

I'm thinking about the offset+bias thing. I believe that the simplest thing to do may be to use an opamp servo to force the MOSFET to have 0.00V DC on its source. The opamp would be a DC-only offset. I'll have to look at the gate capacitance and what that does in a source follower mode, and what servo offsets do until the MOSFET gate gets "warmed up" to the right operating current. Maybe nothing, if the servo start at most-negative at power on.

Interesting question.

[diagrammatiks]

ahhh mosfets .

definitely an area I wanna look more into myself.

That being said...there's ways of doing this stuff with tubes as well as long as you willing to do some weird and alternative topographies.

[exclamationmark]

I've actually gotten a 12AU7 to draw about 1mA of grid current in a dc coupled cathode follower arrangement in a low-ish voltage circuit at quiescent conditions (I intentionally chose circuit values to maximize the compression effect). The cathode follower actually began to draw more grid current than the gain stage ahead of it! It sound pretty horrible though... Given that the B+ was only 48 volts, I'd imagine the grid current that it would of drawn at something like 300 volts would be nothing to scoff at.

[jmaf]

I've actually gotten a 12AU7 to draw about 1mA of grid current in a dc coupled cathode follower arrangement in a low-ish voltage circuit at quiescent conditions (I intentionally chose circuit values to maximize the compression effect). The cathode follower actually began to draw more grid current than the gain stage ahead of it! It sound pretty horrible though... Given that the B+ was only 48 volts, I'd imagine the grid current that it would of drawn at something like 300 volts would be nothing to scoff at.

The grid isn't made to withstand current and 1mA is already a lot of current for a 12AU7 grid, never seen anything like it. If you had a linear relationship there and at 300V Vp it attempted to go approximately 6x as high, you'd probably destroy the grid instantly.

[back-from-thefuture]

Hello,

I came to this site to post some questions about the mosfet follies page. I'm new and not so knowledgable... so excuse me for jumping into the conversation.

i planned on building a 100watt EL34 based guitar amp (I have built a couple in the past). Originally I was going to use the idea of having cathode followers after the PI for each pair of output valves. These would be AC coupled before the usuall 22nf's feeding the grid stops. However, I'm using an existing chassis and don't have the space to add an extra preamp tube.

Orginally I planned on just doing a straight swap for 2 mosfets after reading the follies, but might be better overall to use the source follower driver for the output tubes.

would this require 4 caps from the PI into 4 individual mosfet (one on each EL34 grid) or can each pair of tube be fed from 1 mosfet?
The other thing is getting the negative voltage supply. I've been trying to find an adequately spec'd transformer that will fit in my chassis to provide the -V. the HT on my PT is 360-0-50-360@400mA. the 50v tap is the existing bias tap.

Is there a working schematic it would be possible to take a look at? sorry if that is cheaky.

[defaced]

This works, though my approach is a little more complicated than what you're describing, but can be easily adapted to what you're trying to do. http://music-electronics-forum.com/t16854/#post148551 The MOSFETs I discuss in the last posts were indeed dead and the circuit worked after installing new parts.

[back-from-thefuture]

Thank you. That was amazingly helpful. It's really helpful to see your design and read the conversation. It goes a long way to helping me figure out exactly how its working and to understand the design. A lot of FET based ciruits I've seen have 12V zeners added for safety. Does that not really apply in this situation, does the IRF820 have the zener built into it?
Just out of interest how did it sound in the end? An improvement over the original design of your amp? I don't have any spare parts to prototype it yet so I'll pick up some IRF820's and other bits soon and try it out.