KT66 SE amp, grid current or not?

[jhaible]

Hi,

I'm new here.

I'd like to build a SE amp with KT66 in triode mode, plate voltage around 400V, using a 5.2k primary impedance transformer.

This roughly fits a data sheet application, where a 4.5k impedance is recommended, giving 5.8W output power.

With the slight mismatch of my transformer (bought a pair of these because I wanted to use EL84 at first), I simulate approx. 3 ... 4 Watts of output power.

Now I wonder about extenting the range to allow grid current on occasional input level peaks. I don't intend to run the Tube at full power permanently (it's HiFi, not Guitar application).

(1) Is it save to do that, or will it considerably shorten the life of the KT66?

(2) I guess for this, I need a cathode follower driver stage.
I have a 100V seocondary winding on my mains transformer, which is intended for fixed bias of output tubes, and only provides 50mA (AC). Is it ok to use this to make a negative rail for the source follower driver stage, with only a few mA of idle current, i.e. not loading the negative rail too much, but providing _more_ for the case of KT66 grid current?

From theory I know that's just what a cathode follower is good at, providing a lot more current for th epositive half wave than the negative - I just wanted to check if there are any drawbacks in practice.

As I said, I'm new here. If this has been covered before, please be patient with me and just send me to the old thread.

Thanks in advance,

JH.

 

[Tubelab_com]

1) Momentary grid current on signal peaks will not adversely affect the life of most tubes since the grid current should only flow on occasional signal peaks.

2) The extra winding on your transformer is rated for 50 mA. This should be capable of supplying more than enough current for your uses. I would use SS rectification to be sure that the negative voltage is present before the positive voltage sources come up.

I have no experience with a KT66, however I use a 5K load with a 6L6GC. I run about 465 volts B+ with about 425 across the tube(cathode bias and OPT loss). I get 5 to 6 watts depending on which OPT I use (some are lossier than others). Using a 5K load will give up some power, but improve the distortion and damping factor.

I have tried using a mosfet and a negative voltage source to allow for occasional grid current. This did not offer much additional power output on a 6L6GC, but did improve the transient response. Since I was designing a simple amplifier, I chose to use cathode bias to eliminate the negative voltage requirement. None of the power transformers that I wanted to use have an extra winding or bias tap.

 

[ray_moth]

Sounds like a good idea to me. What you should achieved is some useful headroom and more graceful overload behaviour. This might make your amp seem louder than it really is.

 

[jhaible]

1) Momentary grid current on signal peaks will not adversely affect the life of most tubes since the grid current should only flow on occasional signal peaks.

Thanks for reassuring me! I think I'll really do this then.
(I would feel even saver if someone confirmed that this applies to the KT66 in particular (you wrote "most tubes"). It looks big and sturdy, but it's my first time to build something >1W in tubes, so I'm better asking one stupid question too much ...)

2) The extra winding on your transformer is rated for 50 mA. This should be capable of supplying more than enough current for your uses. I would use SS rectification to be sure that the negative voltage is present before the positive voltage sources come up.

That's my biggest concern, to blow the tube from loosing bias. I simply don't have enough HT voltage to be able to afford the almost 40V accross a cathode resistor.
In the power supply, it's SS rectification, too, for the same reason, so I'll have to take extra precautions for the negative supply to come faster.
I also read it would be important not to apply the HT before the tubes are heated up (cathode stripping?). Is this really important here? (I never thought about this in the small signal tubes circuits I'd build until now.)
In that case, my plan would be:

Both (positve) HT and negative supply will have a MOSFET source follower, with a chain of RC filters at its gate for smoothing and/or a voltage stabilisation. That way, I can control the turn-on of each supply with a rather high impedance circuit, just controlling the Gates of the MOSFETs. Different time constants in the RC filter chains should provide controlled ramp-up time, independend of the load. So there would be control of the negative voltage coming faster.

I wonder if I should use trimpots to divide-down the negative supply to set the bias then (before the cathode follower driver even - is this a good idea, or are there hidden dangers in this?), or build some servo loop. 1 Ohm cathode resistor and a diff amp + integrator servo loop looks tempting - my greatest concerns are how to build this fail-safe. The more complex the circuit, the more _unexpected_ ways it could fail ...

I have no experience with a KT66, however I use a 5K load with a 6L6GC. I run about 465 volts B+ with about 425 across the tube(cathode bias and OPT loss). I get 5 to 6 watts depending on which OPT I use (some are lossier than others). Using a 5K load will give up some power, but improve the distortion and damping factor.

Are you running the 6L6 as triode or as pentode?

I have tried using a mosfet and a negative voltage source to allow for occasional grid current.

A MOSFET for the drive stage (source follower), or as a current source for a tube driver / cathode follower?

JH.

 

[jhaible]

Sounds like a good idea to me. What you should achieved is some useful headroom and more graceful overload behaviour. This might make your amp seem louder than it really is.

Yes, that was the idea. Graceful overload is what I really want, when using tubes.

So I was kind of shocked how the waveform clipped (in simulation) with a high impedance driver. Ok, maybe the tube model isn't good enough to really examine the effects of grid current.

Having seen this, now I'm puzzled as to how so many tube amps do gracefully overload without a low impedance driver. Like, all the guitar amps?

JH.

 

[Tubes4e4]

Hi,

Having seen this, now I'm puzzled as to how so many tube amps do gracefully overload without a low impedance driver. Like, all the guitar amps?
JH.

They just don´t overload "gracefully". Except you consider THD% numbers very deep into the two digits range still to be "gracefully"

Tom

 

[Tubelab_com]

They just don´t overload "gracefully".

Many guitar amps are terrible in the overload recovery depertment, and they often see overload. There is a phenomenon known as blocking distortion (guitar players call it farting out) where the grid goes positive enough to allow the charge to be drained off of the coupling cap, causing bias upset. In severe cases the output tube can be cutoff as soon as the overload is removed. The charge must be restored through the grid leak resistor which is large. This can cause audible distortion for a second or longer.

I have no KT66's to test, but I have tortured several (similar) 6L6GC tubes. I doubt that you could pull enough current through the grid of a 6L6 to cause any damage by using a cathode follower.

Cathode stripping is a real possibility. The severity of this effect has been debated on this forum before, without a clear cut answer. I have adopted this strategy. I ignore it on any amplifier built using cheap tubes (under $10USD each), and use delay circuitry or a standby switch where pricier tubes are used.

I would use all of the available voltage for the negative supply. Your supply must be negative enough to keep current flowing through the cathode resistor on the most negative signal excursion (signal and bias voltage).

A MOSFET for the drive stage (source follower),

The circuit that I use is here:

http://www.tubelab.com/powerdrive.htm

This circuit was designed for an 845 output tube which needs nearly -300 volts to achieve cutoff with an 1100 volt supply.

I currently have the 6L6 tube set up with a switch so I can operate in triode or ultralinear mode. I tried pentode mode, but I didn't like the sound. I am operating the tube near maximum ratings, which you may not want to do if you use good tubes. I got some Chinese 6L6GC's off of Ebay for $3 USD each, so I don't mind running them at their limits.

 

[ray_moth]

They just don´t overload "gracefully".

Compared with SS amps, they do. Ask any guitar player who's compared the two. When an SS amp overloads, it clips vey sharply. This is largely due to the heavy negative feedback used. The attendant distortion is very harsh and unmusical. That's why some people say tube watts sound louder than SS watts, because if overdriven an SS amp sounds dreadful while a tube amp behaves better and sounds a lot better - more "graceful", if you like. Tubes clip too, of course, but with a rather more rounded waveform. There is not the same harshness and so the effect of transients causing momentary overload is less unpleasant.

What can make a tube amp sound nasty is if overloading causes grid current in the OP stage but it is capacitively coupled to the driver. When grid current flows, the coupling cap is discharged and then takes some time to charge up again when the transient is past. This means there is a finite recovery time, when the coupling cannot do its proper job and during that time there is severe distortion, which is exacerbated if there is significant negative feedback. Having a direct-coupled driver eliminates this problem because there is no coupling cap to become discharged and then recharged. Overdrive conditions are handled by relatively gentle clipping and recovery is fast.

 

[cerrem]

One thing to keep in mind about going into A2.....
You need to change the plate load to do this....
If you look at your 5K plate load right now and plot it at your B+, you will see where it Y-intersects the current axis, which is is not much above the 0-bias curve.... if you want to reach up into the positive curves, you will need to make the load line smaller....such as 3K ...ect...depending on where you want to reach to... Then re-bias to be well centered in the full AC swing....
A follower is the ticket for your situation as a driver.... A 6V6 makes a great follower for pushing into A2 in your application...
Keep 6V6 in tetrode mode for follower...


Chris

 

[vax9000]

How is a 6V6 follower compared with tubelab's power drive? I guess power drive is much better. The only draw back is the big heat sink.

 

[Tubelab_com]

The big heat sink is required on the 845SE amp because I need to swing about 300 volts P-P at the grid of the 845. This requires at least 400 volts across the fet (headroom) and the quiescent current is about 10 mA. The 845 is notoriously hard to drive.

The requirements for a 6L6 type tube are much smaller, and the grid current capabilities of the 6L6 are lower. In this case I would even try a cathode follower if I was against using silicon. Choose a tube with a decent peak current capabilities. If using a mosfet look for low gate capacitance and low "reverse transfer capacitance".

 

[richwalters]

Many guitar amps are terrible in the overload recovery depertment, and they often see overload. There is a phenomenon known as blocking distortion (guitar players call it farting out) where the grid goes positive enough to allow the charge to be drained off of the coupling cap, causing bias upset. In severe cases the output tube can be cutoff as soon as the overload is removed. The charge must be restored through the grid leak resistor which is large. This can cause audible distortion for a second or longer.tubes are used.

I came across this problem..... an emergency stage swap from a duff EL34 to 6550...........and had to turn level down as overhang (I call it lock-up) was so bad. Changing the output stage fixed bias/grid resistor from 100K to 47K did the trick........the thd went up from extra stage loading but who cares for MI !

richj

 

[jhaible]

Wow, what a wealth of information here!

(for some reason, my email notification didn't work; what a pleasant surprise to see all these replies now that I came back!)

JH.

 

[jhaible]

The big heat sink is required on the 845SE amp because I need to swing about 300 volts P-P at the grid of the 845. This requires at least 400 volts across the fet (headroom) and the quiescent current is about 10 mA. The 845 is notoriously hard to drive.

The requirements for a 6L6 type tube are much smaller, and the grid current capabilities of the 6L6 are lower. In this case I would even try a cathode follower if I was against using silicon. Choose a tube with a decent peak current capabilities. If using a mosfet look for low gate capacitance and low "reverse transfer capacitance".

I read your pages about the powerdrive and the amp for which you built it. Very interesting stuff.

I frankly admit that I don't really have a clue about where it makes sense to use semiconductors and where I should better stick to tubes. Coming from silicon and just starting to explore tubes, that's also something I want to learn.

My first intention was to build an all-tube amp. (But I'm already convinced that the 1st stage for my 1.5mV output phono cartridge demands a silicon input stage - but that's a different topic.)

I wanted an all-tube amp, intill I noticed there will be this nasty clipping from grid current in the SE amp I had planned. (I haven't heard anything anything yet - it's all just simulation and maths by now.)

Now the next step was considering a cathode follower (that's why I started this thread). But I've read a lot of negative stuff about cathode followers, most of it rather vague. And from reading your story of the powerdrive, you're not too keen about a cathode follower either. I accept this as common knowledge (about everyone seems to agree on it), and yet it's hard to understand _why_. Given that it doesn't ocillate, has a fairly constant cathode current: shouldn't the distortion be increadibly low, just from the 100% local feedback?

Speaking of fairly constant cathode current: It's clear that the transition from A1 to A2 will change the cathode current significantly, unless I over-design that stage immensely. (BTW: how much, approximately, of grid current to I have to expect for a KT66, or 6L6 ?) But _that_ change is not what worries me, as _everything_ will be better than the clipping. (Nevertheless, out of curiosity: Morgan Jones recommends running a tube in a mode where it doesn't get _out_ of grid current, if you choose to have frid current at all. How is that supposed to work ??)

But back to what worries me: If I use a direct coupled cathode follower to drive my KT66, I must go _down_ to negative voltages quite a lot. Rough calculation: Biasing the KT66 grid at -38V is recommended for normal (A1) operation. There, the positve input voltage swing is limited to 0V by grid current, which means a negative swing of -76V. I presume with A2, I'm supposed to keep the bias point at -38V, but then swing up to something like +10V, and down to -86V ? Or up to +20V / down to -96V ? (is that right?)
Then there is not much room with a negative rail of ca. -100V ...

That's why I was thinking cathode follower + silicon constant current source instead of a cathode resistor. but even then (and I still don't know if that silicon part is "good" here or not!), won't this mean that sooner or later I just hit the negative rail, and now get nasty clipping on negative peaks, if on a (slightly!) higher level than th epositive clipping that occured before ...

Questions, questions ....

JH.

 

[jhaible]

One thing to keep in mind about going into A2.....
You need to change the plate load to do this....
If you look at your 5K plate load right now and plot it at your B+, you will see where it Y-intersects the current axis, which is is not much above the 0-bias curve.... if you want to reach up into the positive curves, you will need to make the load line smaller....such as 3K ...ect...depending on where you want to reach to... Then re-bias to be well centered in the full AC swing....
A follower is the ticket for your situation as a driver.... A 6V6 makes a great follower for pushing into A2 in your application...
Keep 6V6 in tetrode mode for follower...


Chris

I see ...

Too bad that I already have bought these output transformers, which have 5.2k primary.
They were originally intended for a EL84, but then I noticed their impedance and DC current specs (70mA) would fit for something better, when I read the KT66 datasheet.

Why use a tetrode for the driver?

JH.

 

[cerrem]

The reason you want to put the 6V6 in tetrode mode when using it as a follower is for the lower output impedance...
1/gm is roughly what you get "looking in" the cathode terminal ..
If you wire it as a triode then you loose a little drive..
Either eay it would make a good driver for your application....
With the plate load you have...your not really going to make it into Class A2...SO whay not make a nice Class A1 amp and forget the driver thing...

Chris

 

[Tubelab_com]

But I've read a lot of negative stuff about cathode followers, most of it rather vague. And from reading your story of the powerdrive, you're not too keen about a cathode follower either. I accept this as common knowledge (about everyone seems to agree on it), and yet it's hard to understand _why_.

I am not totally against cathode followers. In that application (driving an 845) I couldn't get one to deliver the sound dynamics that I could get from a mosfet. There is a lot of negative stuff printed about cathode followers. Some of it is based on real experience, and some is pure conjecture. There is even more negative stuff printed about puting any silicon in the signal path. Again, this is based on the fact that a transistor is inherently non linear when used as a voltage amplifier. It can be very linear as a follower. I have been told of possible effects caused by the capacitance change as a function of applied voltage. How significant these effects are, I don't know yet.

As mentioned, the transition from no grid current to grid current is abrupt. The grid impedance goes from megohms to a few killohms with a grid voltage change of a few volts. This change will cause a nasty distortion if the source impedance of the driver is not very low. A mosfet follower can source a lot more current than the average cathode follower, and offer an output impedance of a few ohms.

It has been suggested to operate the output tube such that grid current always flows. This is only practical with a few tubes. The 811A is one of them. It was designed to operate with grid current and can be quite linear in the region of positive grid bias. Many popular output tubes were never designed to operate in the region of positive grid bias. They are not always linear in this region. I looked at the Tung Sol 6L6GC curves and there ARE triode mode curves extending up to +10 volts on the grid. The curves also look reasonably linear in this region. Grid current is listed at 10 mA for +5 Vg and 30 mA for +10 Vg.

I played around with my simulator (SE amp cad from TubeCad) using a KT-66 tube and a 5.2K transformer (assumed 300 ohms DCR) and 400 volts B+, I can get 3.5 watts at 3.8% THD (3.7 % 2nd harmonic)without grid current (-31.5 volts bias, 60 mA plate current). If I crank up the drive (until the 3rd harmonic begins to rise) with all other conditions constant, I can get 6.6 watts at 5.2% THD ( 5.1% 2nd). The grid voltage is now swinging from +11.5 to -74.5 volts. These simulations show that a significant power gains are possible without a serious distortion penalty if a suitable driver is used. I have had good results with this low cost simulator. See if you can match these numbers with your simulator.

I will be away from my computer until Monday, so I probably won't respond until then.

 

[jhaible]

With the plate load you have...your not really going to make it into Class A2...SO whay not make a nice Class A1 amp and forget the driver thing...
Chris

I start to think I better should ...

Grid current is listed at 10 mA for +5 Vg and 30 mA for +10 Vg.

30mA ! I don't think my PSU will provide these extra 60mA for two channels. except I put some extra storage capacitor in there, at a lower voltage, to act as a reservoir for the driver, so my PSU will only see the average current, not the 60mA peaks.
Would be fun to experiment with. But I'll probably build the whole thing in A1 first, and see if power is enough or not.

I played around with my simulator (SE amp cad from TubeCad) using a KT-66 tube and a 5.2K transformer (assumed 300 ohms DCR) and 400 volts B+, I can get 3.5 watts at 3.8% THD (3.7 % 2nd harmonic)without grid current (-31.5 volts bias, 60 mA plate current). If I crank up the drive (until the 3rd harmonic begins to rise) with all other conditions constant, I can get 6.6 watts at 5.2% THD ( 5.1% 2nd). The grid voltage is now swinging from +11.5 to -74.5 volts. These simulations show that a significant power gains are possible without a serious distortion penalty if a suitable driver is used. I have had good results with this low cost simulator. See if you can match these numbers with your simulator.

My simulator is PSpice, and the Koren tube models. No KT66 model, unfortunately. So I'd made roung approximation using the 6L6 model. Apart from a slight discrepancy in Vgk, I get the bias points of the following data sheet application pretty well:

http://frank.pocnet.net/sheets/086/k/KT66.pdf

referring to Page 2, Triode Connection Class A, single Valve, Cathode Bias. Of the 440V, roughly 40V appear at the cathode resistor.
My transformer has a DC resistance of 600 Ohm, so I'll have roughly 40V across the primary winding. Now, when I use constant bias instead of the proposed cathode bias, I should have the same 400V across the tube as in the data sheet (from 440V HT).
Now the data sheet gives 5.8Watt, presumably _without_ going into class A2. (Load is 4500 Ohm, compared to my 5200 Ohm, though. So I expect slightly less power, not not much less. (?) )

JH.

 

[jhaible]

I've started building!

It goes slow - I'm putting a lot of effort into safety, and into doing the wiring right. (It's not my first tube project - but the first with voltages where 450V electrolytic caps aren't good enough, the first to use AC heating, and the first with huge tubes that stick out of the enclosure ...

I had the tubes glow last sunday, and found a place & position for the output transformers to have the lowest hum. 1mV(eff) on the primary from mains transformer coupling (with only the tube heaters connected). That's -97dB, refered to 1Watt output power.
I hope this is enough, with some reserve, as I expect another couple of mV from the PSU after the regulator. And I'm curious how much (or how little) the heaters will contribute.

Anyway, back to the original topic & question:
I'll take this slow and build it step by step now.
First thing will be cathode bias (sacrificing power, but giving more safety not to fry the tubes).
If this works fine, I'll try grid bias.
And then, I might experiment with a cathide follower / grid current driver. If not for much increasing power, then for giving better rounding on overload.

Speaking of safety: I was looking for wires with good insulation for the high voltages. Then I was surprised to see that all the flexible wire I was using for low voltage projects all along, is specified for 1000V AC. I'm still sceptical, with these wires having an outer diameter of just 1mm ...
But I've double checked, the data sheet says 1000V AC operating voltage! (not test voltage, which would imply a much smaller operating voltage.) Is this possible?? It's not even Teflon ...

JH.