Thread: protecting output tubes- screen voltage vs screen resistors

1. protecting output tubes- screen voltage vs screen resistors

I've been experimenting with different output sections and trying to get maximum power from different tubes. It's not tough to get a lot of power- it would seem that it's tough to do it without damaging the output tubes!

Assuming that the screen voltage sets the amount of current a tube can draw at a given load and voltage I can adjust screen voltage to hit a certain power level. Clearly this varies with the tube- 6L6's don't make much power till the screen grid voltage approaches the plate voltage. I found that EL34's and KT88's can get by with much lower screen voltage and still draw lots of current. Of course they'll also light the screens up like a light bulb filament if you overdrive them when set up for maximum clean power.

How do I determine the point where I can meet my power goals but still have high enough screen resistor values to avoid destroying the screens?

jamie

2. ...there were some very "high-end" tube stereo systems that used a microprocessor with built-in A/D converter that actually "monitored" the idle current of each tube and "tweeked" their bias voltage (as necessary) in realtime.

...something similar could be done with plate current and plate voltage to calculate the instantaneous plate dissipation in realtime and then output a control signal to an adjustable high-voltage regulator "chip" that controlled the screen grid voltage to both tubes.

...not totally "simple" but obviously "do-able."

How do I determine the point where I can meet my power goals but still have high enough screen resistor values to avoid destroying the screens?

jamie
Go dual rail.

In all seriousness, if you look at the characteristic curve charts for KT88s and EL34s, you'll find one that shows screen current at different points on the load line. In every one of them you'll notice that once plate drop hits a certain point, screen current will go through the roof. It's a characteristic of the tube when it hits maximum power, not the surrounding circuitry. At some point in the plate drop, screen voltage is now above plate voltage which will cause the screens to "steal" more current. Once screen current increases to the point at which they over dissipate, you get into screen glow land.

It is this screen glow phenomenon that was one of the big driving factors that got me to start researching and prototyping dual rail designs.

4. I have the screens 200 volts below the plates- it's still not enough. Some designs simply can't make the power without a certain amount of screen voltage anyway.

I know that an amp I build isn't likely to abuse the output tubes like I can with a signal generator but I'd still like to find a little more reliable solution.

jamie

I have the screens 200 volts below the plates- it's still not enough. Some designs simply can't make the power without a certain amount of screen voltage anyway.
Correct...because the screen voltage has a direct effect on the transconductance of the tube.

I know that an amp I build isn't likely to abuse the output tubes like I can with a signal generator but I'd still like to find a little more reliable solution.

jamie
I honestly don't think you'll find one. Look on the charts in the datasheets...you'll see that every tube will start seeing a screen current increase when plate voltage drops below a certain point no matter what you set the screen voltage to.

Max power transfer occurs when the load line crosses the zero volt grid curve. Most guitar amps that operate "sliding screen" have the screen voltage and hence the curves well above the load line. As the screens draw more current, the screen resistor causes the screen voltage to drop which causes the grid curves to "slide down" the graph (hence the term "sliding screen" operation).

If the screen resistor is sized correctly, it should pull them down right to the point where the Vg1=0 line crosses the load line when the grid drive signal drives the control grid to 0 volts.

6. Shouldn't we be able to do a little simple math and calculate a resistor value then? I know it won't work for all conditions but at least for the conditions we're likely to create.

There are millions of amps out there that don't blow up- what can we do to still make big power but be just a little more certain that an amp won't blow up? Just up the screen grid resistance so that any substantially large overdrive condition causes the screen resistors to limit the current to the screens? In my experience this just makes for a change in feel as you approach an overdriven condition- sometimes desirable in a dirty little guitar amp but not so much in a bigger cleaner bass or preamp gain driven guitar amp.

Yes, I'm planning a high-gain amp build but I want it to have a worthwhile power section too!

jamie

7. The only way I can think of to get max clean power, but limit screen dissipation when overdriven, is to have a current limited supply to the screen.
In every amp I've checked in the past year (since I became aware of this issue), the screen dissipation increases to at least twice the data sheet max level when overdriven. That includes EL84, EL34, 6L6, 6V6 types, but not KTXX. With a higher than optimal load (ie 16ohm load on to 8ohm output) it goes higher still.
Is anyone aware of a simple, robust high voltage current limiting circuit available in the public domain? A pair of 6L6GC only need 20mA screen current for max clean power, but can draw well over 100mA when overdriven.
The supply would have to drop maybe 350V when limiting the screens to 20mA under heavy overdrive, assuming a B+ of 450-500V.
I've tried PTC thermistors in the screen supply, but due to the temperature variation inside an amp chassis, and the lag effect for the thermistors to cool following heating, none that I was able to use gave a satisfactory result.

8. To protect output tubes, keep the amp in your living room and play quietly.

Even dual rail designs can overdissipate your screens. In a way, they can be worse, because they have smaller screen resistors, so the screen supply is stiffer and can deliver more screen-BBQing current.

One idea I once had was a light bulb in series with the screen supply. Same idea as the bulbs they use to protect tweeter horns in PA cabinets. When it kicked in, the bulb would light to inform you of a problem, and you could keep playing, it would just pump like a compressor. If things got really bad, hopefully the bulb would blow before any of the screens did.

It would work better over here, where the light bulbs are 240 volts. You now get little 240V halogen capsules that look like they'd work well. Maybe in Americaland you can use two or three 120V ones in series.

My Ninja Toaster power supply had foldback current limiting on both plates and screens, but it wasn't designed to kick in in normal operation. It originally had a screen current sensor that shut off the B+ if it thought the tubes were about to fry, but now and again it would pull the plug on you in the middle of a guitar solo! I ripped it out and threw it away years ago.

In the amps I've built so far, the screens don't obviously glow even when they're cranked into square wave clipping. They're also so loud that I rarely need to crank them that far. So I don't think it's a major issue.

9. And the new CFLs will look even cooler in that application.

10. Gah! I forgot about the lightbulb ban.

11. I just bought a box of 100w this evening... and a few CFLs too. No shortage of incandescents here.

Light bulb limiters aside, I also work on light dimmers, and my collection of porcelain bulb sockets with 100waters in them are my loads.

12. I guess this thread has gotten out of hand and drifted a bit from my original intent. I appreciate all the input.

More than anything I'm debating how to set screen resistor values. I guess there is no hard and fast rule- copying existing output sections is probably a good starting point.

jamie

13. I wonder if running a screen bypass cap would help. If I'm thinking correctly here, it would keep the grid curves from sliding down very far if at all since it would hold the screen voltage itself constant, yet the resistor would still be limiting current to the screens.

14. Originally Posted by Wilder Amplification
I wonder if running a screen bypass cap would help. If I'm thinking correctly here, it would keep the grid curves from sliding down very far if at all since it would hold the screen voltage itself constant, yet the resistor would still be limiting current to the screens.
Unfortunately no, if you did that you'd be back where you started. The screen current simply gets drawn from the bypass cap instead of through the resistor! In other words, if you're preventing the grid curves from sliding down, then you're not limiting screen current!

15. Originally Posted by Merlinb
Unfortunately no, if you did that you'd be back where you started. The screen current simply gets drawn from the bypass cap instead of through the resistor! In other words, if you're preventing the grid curves from sliding down, then you're not limiting screen current!
Yeah I see what you're saying there...didn't think about the cap becoming the new current supply.

The problem here is that under "normal" designed operation tube amps aren't supposed to clip. When you push them into clipping (which a lot of us do with guitar amps), you're exceeding the design limits.

So when designing output sections, one must decide what their main objective is..max clean power transfer or being able to overdrive it without sending the screens into meltdown. It's just like designing car engines...do you want it to be a high RPM engine while being able to live to tell about it? Or is it just going to be a "daily driver" motor where max power is not of concern? Unfortunately most guitar players all wanna be able to have their cake and eat it too, but in most cases you just simply cannot have that. Nothing's free and there's always tradeoffs.

16. How does a cap become a supply of dc screen current? It's not ac that is a problem with screen current.

17. Well, it's not a supply of DC current of course, because the average current through a capacitor is always zero.

If you add a capacitor to the screen supply, you'll have more headroom on short transients, but the screen supply will sag down and limit sustained output power.

In a few words: The output stage behaves less like a limiter, more like a compressor.

Some old Bogen PA amps had a capacitor on the screen supply just like this, and a few of them even had those cold-cathode voltage regulator tubes to stabilise the screen voltage. They would go out when you cranked the amp fully.

18. Originally Posted by hasserl
How does a cap become a supply of dc screen current? It's not ac that is a problem with screen current.
Under full drive the screen current takes the form of a square wave switching between zero and an upper limit, so although it is unidirectional and therefore technically DC, it can also be reckoned in terms of AC. In other words, the cap supplies pulses of current.

19. Hi Guys

"Is anyone aware of a simple, robust high voltage current limiting circuit available in the public domain?"

The simplest protection is to use a minimum screen-stop value of 1k for each tube. Make this a 5W to assure it is flame-proof. This is amply demonstrated in the first TUT volume.

Screen voltage has the effect of increasing the current capability of the tube, as others describe above. Limiting plate current by controlling screen voltage alone requires large voltage reductions to effect proper protection. What you really should concern yourself with is the screen circuit resistance. This reflects through the tube controlling plate current much better. The drop across a 1k resistor in a 4-500V environment is negligible, but the protection to the tube is immense.

You definitely can have low screen voltages and high power output with standard tubes. Where the published plate curve for Vg2=250V may now apply to your amp, what you don't see is that driving the grid positive gets you the same plate current as if Vg2 were much higher. To drive the grid this way requires low-impedance drive, either a cathode follower or a transformer. RDH and RCA both point this out in various articles and spec sheets for tubes.

The use of non-existent or low-value screen-stops is an "ideal". They are needed in real circuits to maintain stability with variations of the tubes themselves and with layout. Increasing the values above what is needed for stability affords great protection to the tubes. In an upside-down-tube amp and/or combo amps and/or any amp you want to clip frequently and/or any amp you want to not eat tubes, the 1k standard should be adhered to.

For EL-84s, 1k5 to 2k2 should be used instead.

Have fun
Kevin O'Connor

Stabilised power supply

Maybe Merlin's new book on power supplies has something in there too.

21. I recieved Merlin's newest book for Christmas. There are a quite a few circuits depending on your end goal. Most are somewhat similar to your design- a big power fet or transistor for the heavy lifting with a small bipolar current limiter.

Kevin, thanks for the reply. I have your books on my list for my birthday- hopefully my wife will get them for me! You verified what I thought from destructive testing of a few tubes- a large screen resistor is largely transparent until screen current rises and things start to get crazy.

jamie

I've been experimenting with different output sections and trying to get maximum power from different tubes. ...
How do I determine the point where I can meet my power goals but still have high enough screen resistor values to avoid destroying the screens?
A lot depends on what your goals are.

It may not be POSSIBLE to get X watts from Y tubes. Or it may be possible only with perfect conditions, which don't exist for long in the real world. For a given set of power tubes, there is a maximum power output you can get without killing them. As you approach that point by twiddling this and diddling that, you make the whole setup more fragile - any variation kills the tubes, and more mercilessly as you approach the "perfect" point. It gets less forgiving.

The comments you've received are worthwhile, but I suggest that you reexamine your goals. Getting 10% more power out of a pair of X tubes is a neat trick, but it is that - a trick. It's like being able to balance a basketball on your index finger. It's a feat of skill and derring-do, but not very practical beyond being able to do it, and perhaps win bets and beers on the result. In the real world, you have to do something like 10X the output power to double the loudness, and you're very, very unlikely to be able to do that if you're near the heat-death edge of the tubes' endurance, and even if you do, how will it sound? Trashing the quality of the sound and going class C will produce more power, but it will sound terrible.

By the way, as one poster mentioned, it is possible to pull tubes back from the brink with a protection circuit. I once did a circuit that looked at peak tube currents and shut both (or all four) tubes down if any one exceeded a preset threshold. It's easy to extend that to cutting volume if the tubes are too hot (with a temp sensor) or sensing screen current and cutting back screen voltage, and/or volume, or all or a mixture of the above.

But reliability in the real world is worth more than dancing ever nearer the edge of the cliff.

23. Hi Guys

"Reliability" in a tube power amp does not require computers to monitor the output stage, or temperature sensors of the modern kind, or any great expense. These modern techniques are interesting unto themselves and are sometimes skillfully employed, but high power audio amps have been around since there have been tubes.

Keeping the discussion within normal operating classes, there is nothing obvious to most observers to suggest that one push-pull amp is more reliable than another, other than the actual value of the screen resistors. The "protective" value is essentially independent of the amount of power one wishes to derive from the tube set. Looking at manufactured circuits does not always help.

For example, the SVT and V-9 used 22R in parallel with a diode for each 6550 screen. Vg2 was set to about 350V with around 700V on the plate. Despite having the screen voltage 100V below the "rated" value by the tube manufacturer, SVTs and V-9s are notorious for "eating" tubes. The diode protects the circuit board from being burned when the resistor rating is exceeded. When the tube pulls a lot of screen current, the 1/2W resistor burns up but then the diode can conduct well over an ampere of screen current. This is no protection at all.

The fix for the above amp is to remove the 22Rs and diodes and install a 1k-5W for each tube. The amp still produces its full power but never again eats a tube.

Fender made a similar decision in the PS-400. They used higher R values for each screen with a similar voltage ratio for Vs and Va. They left out the diodes and let the resistor act more like a fuse. Over-all, the PS-400 was much more reliable than the SVT despite producing 435W vs. 330W from the same tube set. However, in the PS-400s I worked on, the 1k-5W solution was installed and the player was happy with the result and reliability.

A detailed output stage design process is presented in "The 400" chapter of TUT6. Maximum power available from the common tube types is explored, along with the implications to the power supply design.

You may wish to get "more" power or "maximum" power from a given tube set for practical purposes or merely for an experiment. As long as heat is managed - a fan is added - then maximum power is available reliably. You have to derate things if the tubes are to be upside-down or if the amp is to be a combo. The latter pushes reliability WAY down.

Have fun
Kevin O'Connor

24. How do you calculate screen dissipation under signal (say at maximum output before clipping)? I'm thinking you would measure the DC voltage drop across the screen resistor and divide by the resistance to get average current, then multiply that by the screen-to-cathode DC voltage to get dissipation. Is this correct?

25. At high signal levels there will be a large ac voltage on the screen, as well as the dc voltage. Plus the ac waveform won't be any kind of sine wave. So a true rms meter that will measure the combined Vac + Vdc is needed, to obtain the true effective voltage across the resistor and between screen and cathode.
And as this is a guitar amp, it must cope with heavy overdrive (when screen current increases significantly), so drive it to get the max square wave output.
Expect results well in excess of the data sheet absolute max. Pete.

26. Correct, just multiplying averages won't work, you need to do some RMS trickery.

I've never bothered trying to calculate it, I just look through the holes in the plate and go by the colour of the screen wires under maximum overdrive. In pentodes, a little red is OK but orange is worrying. If they light up bright yellow, that is bad.

In beam tetrodes, the screen wires probably shouldn't glow at all. I've never seen them glow except when abusing 6V6s well beyond their ratings, and reducing the plate load impedance fixed it.

27. Has anyone worked with the arrangement shown on the early Marshall schematics
http://www.drtube.com/schematics/marshall/jtm45tr.gif
in which the screens are fed by a shared 1k which then splits into individual 470R (no smoothing cap at that junction / node)?
The supply to it is well smoothed by a CLC arrangement (though there's a schematic error reported, as all examples see have the OT CT fed from the reservoir cap).n
To my thinking, the shared 1k may cancel out the local ac negative feedback.
So get the benefits of screen supply current limiting / voltage reduction during heavy overdrive, but without the loss of gm that equivilantly large individual screen resistors (eg ~2k5) would cause.
But there must be a reason that this arrangement was subsequently dropped; even the JTM45 / Bluesbreaker RI just had individual 470R screen resistors.
http://www.webphix.com/schematic%20h...__45w_1962.pdf
Pete.

28. Thanks, guys.

29. Hi Guys

The screen dissipation calculation above is essentially correct.

The tubes used most commonly in guitar amps have very similar plate to screen current ratios, so you can make an approximation of screen power assuming Ig2 is about Ia/10.

For example, a typical 50W amp operating at 400V and 4kaa, say we idle the tube at 30mA. Ia=30mA, so Pa=400Vx30mA=12W. Similarly Ig2=3mA soPg2 will be 3mAx400V=1W2. Peak signal current is about 300mA and peak screen current is then 30mA. For this discussion, we don't actually care what Pa is over-all, so won't go into that. What is important is peak-Ig2=30mA occurs with the same 400V on g2, so peak screen power is 12W. Average screen power is half of this, 6W, for continuous full output.

The shared Rg2 method was used in hi-fi amps to steady the voltage on the screens of push-pull output stages. High-R values were never used. Initially such an arrangement was just the shared R, then the fan-out of small Rs for each tube were added to assist service and improve reliability. Now five resistors had been added where most people wanted none - or had historically tried to get away with none. So, economics and easier wiring lead to independent screen resistors in amp designs where anyone cared about reliability.

An Rg2 of 2k5 is not needed except for EL-84 or similar.

Looking in the tubes to assess the screen R value correctness is not recommended. It won't work with opaque or coloured glass.

Have fun
Kevin O'Connor

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•