MV, Powerscaling in 50W, 100W Amps

[Steve Conner]

You're not supposed to have HV on a pot. For one thing, most of them have a maximum voltage rating below what the B+ on a tube amp can dish out. But in practice, it makes the circuit design so much easier! For my last amp build I gave in and did it. I used a huge old 1M pot from the junk box, hoping that the large size implied a higher voltage rating.

I published this circuit years ago and it was good enough to get Kevin O'Connor to sign up to MEF and comment on it. (He does have a valid point: the circuit needs a capacitor to filter out wiper noise from the pot.)
Open Source Power Tweeking - AMPAGE Archive

[frus]

What are you talking about Tom? I don't own an amp that has B+ on a user interface. That's one of the first bad design features that I eliminated from my amps.

What about the B+ fuse holder?

[defaced]

Relevant discussions regarding pots and safety here: http://music-electronics-forum.com/t24885/
Some other B+ regulating circuit schemes were offered as well.

[bob p]

What about the B+ fuse holder?

0. I don't build my amps that way. I think that sometimes it's better to think outside of the box, rather than to just blindly copy what everyone else has already done.

1. The fuse holder comment sounds like an end-user comment rather than a designer's comment. Designers know that a properly chosen fuse holder is designed for that application and is rated to handle the applied voltage. Steve already pointed out that a pot is not rated for the voltage or the application, though some people will misuse a pot and take the risk anyway. As others have already noted, B+ on a user control interface would not meet any safety standard. It's just not a good idea. My preference is to not blindly copy bad ideas from the 1950s if there's another way to do things that makes more sense.

2. A fuse holder is not part of the user control interface. It's part of the maintenance interface. Good design separates things that are supposed to be touched while playing the amp from things that aren't supposed to be touched while playing the amp. This brings up an important point. Leo Fender made some lousy design choices in the Tweed era (like putting the fuse holder on the user panel). I'm not a fan of the twist-off fuse holders that Fender used, and I don't think they have any need to be on the front panel user interface. I think Leo put them there because it was a cheap and expedient way to build an amp, with every hole drilled in a straight line, not because it was the safest way to do things. Electrical safety was in the dark ages in the early-mid 20th century. We had knob and tube wiring, 2-wire non-grounded appliances, polarity reversal plugs, death caps, etc. At least things got better in the BF era when Fender started putting fuse holders in a better location -- out of the way on the back of the amp.

3. I don't feel the need to have a fuse holder on the front panel of an amp. It just clutters up the faceplate with something that the user isn't ever going to want to use, so why put it there? The back panel would be a better location. Better yet, anything that handles HV should be totally isolated from the user, on the inside of the amp where a drunk or stoned musician can't easily make contact with it. Think about it -- when a fuse fails, what are you supposed to do? Stick in another one and see if it fails too? Or put the amp on a bench and determine why the fuse blew? To me it doesn't even make sense that there is an external fuse holder. When a fuse blows, I'm going to pull the chassis to see what's wrong. In that circumstance, and internally mounted snap-in fuse holder works fine, and it has the added advantage of keeping B+ inside of the amp where it belongs, and away from the user where it doesn't belong.

There are lots of different ways to do things. It's not absolutely necessary to build things exactly the same way that things were done in the 1950s unless you're trying to create a reproduction. After 50-60 years there might be some new ways to do things that are worth considering. Just some things to think about if you're interested in user safety.

[Wombaticus]

I implemented a Merlin-style VVR circuit in a 100W amp, which you can check out the schematic for here. The VVR portion is the bit lying between D26 and D25 in the schematic. Note that D27 and D29 were not in Merlin's book, but were necessary to ensure reliability in the case that the R14/R29 divider isn't set just so. I may have ended up tweaking that divider's values subsequent to this schematic to adjust the voltage range a bit. Note that this also used his bias tracker scheme, which is immediately under the LTP in the schematic. Note that it was necessary to add a 220pF cap between the collector and base of the MPSA42 pass transistor in the bias tracker to ensure HF stability, but after I did this it worked great.

Overall, the circuit (which here scales only the PI and power tubes) worked just as advertised. As the amp was shipped overseas, I didn't take any chances, and I used one of those humongous (and very expensive) ISOTOP package MOSFETs, attached directly to a generous heatsink, the hole for which was cut into the aluminum chassis with the fins protruding above.

Other reader's comments about the suitability of a 100W amp for VVR should be taken seriously. I found that it was more-or-less tonally transparent in the B+ range of about 140V-430V. However it was still quite surprisingly loud with only a 140V B+. The VVR is NOT a direct alternative to a MV! However, with the right combination of MV setting and VVR, it's possible to get the amp sounding quite nice at reasonable volumes. But a 100W amp will never be a bedroom amp, no matter what.

Note that Merlin's design can still result in a couple hundred volts on the "watts" pot, but at such a high impedance that it's hard to imagine that it poses any substantive risk..

[bob p]

Relevant discussions regarding pots and safety here: http://music-electronics-forum.com/t24885/
Some other B+ regulating circuit schemes were offered as well.

Well, it looks like this ground has been covered already. Thank you for the link. I particularly liked your post which summed things up:

It's not about data sheets, it's about best practices. As I'm sure you know, best practice does not abide by data sheets or any published literature, it's about objectively looking at the design and thinking "does this make sense". I'm a welding engineer, and pretty much everyday the text book gets left in the office while I play "best practices" making test welds.

Or to put it simply, my amp (or really test rig in this case), my rules. I've felt 500v, I don't ever want to again. And if that takes a more complicated solution that one already out in the wild, then I'll develop it with the help of people who know what they're doing.

thanks again.

[Gregg]

I did some PS experiments with my power amps and I couldn't hear much difference between when plate and screen voltage are decreasing simultaneously and when only screen voltage is varied. In both cases the measured power level corresponding to the PS pot position is very close if not the same.
I decided to build an auto bias unit based on the AFD100 architecture and drew an experimental PCB but now it comes time for the firmware and some questions popped up.
AFD100 has a bias pot at the back panel which sets the bias current for all tubes. When implementing the "easy" PS method using a dual ganged pot the plate/screen voltage is decreasing simultaneously with the bias voltage which results in lower plate current at lower power settings. I was wondering if I should keep the current set by the bias pot at the same level for lower voltages as well or after the initial bias at full power/voltage to "instruct" the unit to behave like the usual linear dual ganged pot? In other words how do you calculate AB class of operation at 200V or at 100V if at all?
I would appreciate to hear your comments and suggestions. Also if you have in mind any other features for such unit I'll try to implement them as well.

[Gregg]

On second thought since the plates will have the full voltage when lowering the screen voltage the negative bias voltage should be made more positive keeping the same current (plate dissipation) set by the bias pot.

[Steve Conner]

I got good results just scaling the screen voltage on a cathode-biased amp. As the screen voltage is turned down, the cathode bias voltage automatically tracks it. In the single ended amp I used to test it, the plate current decreases with decreasing screen voltage, but the power stage clipping seems to stay more or less symmetrical, and I think that is the important thing: that the tube stays biased in the correct place relative to cutoff and saturation, whatever those might be at the current screen voltage.

I guess if you did the math, you would find that the cathode voltage varied as the 2/3 power of the screen voltage or whatever. I think linear tracking would be plenty good enough, but if you're using a uC, you could implement a table of whatever power law you wanted.

Keeping the same plate current is wrong, because the saturation current will decrease as the screen voltage is turned down, so the bias point will get nearer to saturation.

[Gregg]

These are some quick measurements showing the correlation between Plate and Screen voltage and the resulting current. Initial plate voltage was 450V, bias -44V in a 2x6L6GC amp. First 3 columns showing what happens when only the screen voltage is decreasing. 4th column when both plate and screen voltages are lowered. Last column approx. knob position. It's a "regular" PS unit - a dual ganged 1M linear pot going to a 100k resistor to ground.


Screen only Pl+Scr o'clock
Scr1 P1 I1 I2 knob
400 450 43 41 2
350 455 41 38 1
300 456 38 33 12-1
250 460 33 26 12
200 465 27 19 11-12
150 472 20 12 11
100 478 13 8 10
45 485 7 2 8

[Gregg]

Hi,

I've got some more data about power scaling the screens only.
A 2xEL34 power amp was used for the test. The sound path was 100mV/400Hz sine wave - audio pot - triode section - PI.
I replicated the AFD/YJM100 power scaling (EPA) unit (analog section) but since my digital part is not ready yet I used a 4 gang 50k log pot to control the unit. 3 of the wipers went to opamps inputs - 2 of them to the 18k resistors at the bias circuit, one to the screen voltage controlling opamp and the last one was used for the feedback. You can see the results from the table below. The last column indicates that pot's sections don't track each other very well towards the end. Instead of STW9NK90Z I used a STW12NK90Z. The transistor wasn't placed on a heatsink but stayed cool at all times and I was able to hold it without any problem.
testing was done the following way:

1/ Adjust screen voltage
2/ Measure both bias voltages and currents
3/ Adjust amp's input signal for clean sine output near onset of clipping.
4/ Measure clean power output.


AFD100EPA1.xls

[soundguruman]

Hi,

I've got some more data about power scaling the screens only.
A 2xEL34 power amp was used for the test. The sound path was 100mV/400Hz sine wave - audio pot - triode section - PI.
I replicated the AFD/YJM100 power scaling (EPA) unit (analog section) but since my digital part is not ready yet I used a 4 gang 50k log pot to control the unit. 3 of the wipers went to opamps inputs - 2 of them to the 18k resistors at the bias circuit, one to the screen voltage controlling opamp and the last one was used for the feedback. You can see the results from the table below. The last column indicates that pot's sections don't track each other very well towards the end. Instead of STW9NK90Z I used a STW12NK90Z. The transistor wasn't placed on a heatsink but stayed cool at all times and I was able to hold it without any problem.
testing was done the following way:

1/ Adjust screen voltage
2/ Measure both bias voltages and currents
3/ Adjust amp's input signal for clean sine output near onset of clipping.
4/ Measure clean power output.


[ATTACH]26741[/ATTACH]

That's just what I got. The tracking is not accounted for.
the bias is always off at some point in the scaling...
There needs to be a "track" and a "scale" adjustment.

[Gregg]

Fortunately my pot is a two axis one and I can adjust the bias independently from the screens voltage for the sake of this experiment.
Better tracking audio pots like the blue ALPS for example are available but they are expensive and I don't think it's so critical.
If you want to go the easy way you can just use the screen control section and use only pots for the bias which solves the "pots at B+" problem.
In the case you have 4 tubes a 6 gang pot will be necessary unless of course you use one bias node for all tubes.

[soundguruman]

Fortunately my pot is a two axis one and I can adjust the bias independently from the screens voltage for the sake of this experiment.
Better tracking audio pots like the blue ALPS for example are available but they are expensive and I don't think it's so critical.
If you want to go the easy way you can just use the screen control section and use only pots for the bias which solves the "pots at B+" problem.
In the case you have 4 tubes a 6 gang pot will be necessary unless of course you use one bias node for all tubes.

Since I have tried this myself...
The tracking of bias VS high voltage is critical.
It makes a huge difference.
What is happening is bias is correct at higher power, then bias is too cold at lower power. This causes crossover distortion at lower power levels.
This is where the design of power scaling is goofing up. The tracking design is too primitive.

There is no pot, alps or otherwise, which corrects this problem.

[Gregg]

I agree with you. It just doesn't seem right to have these running at 1 or 2mA. And yes they go into crossover distortion very easily.