Single Ended Parallel Lead - review my planned overhaul

[Bergheim]

The 5k screen node resistor will add quite a bit of overdrive compression, and reduce power output somewhat. I believe the 5k resistor is chosen to keep ripple as low as possible, or to keep screen current to a minimum. With only 10uF of filter capacitance, and the need for low ripple on the screen(s) of a single ended amp, the resistance must be high. The -3dB frequency of a 5k/10uF is 3.2Hz, not far from the -3dB frequency of a more conventional 1k/47uF filter which is 3.8Hz. When capacitance is decreased, resistance must be increased to keep the same cut-off frequency.

In the schematic, the 5k/10uF is connected to B+2, and if I understood you correctly, you have changed the 1k feeding the B+2 for a 4.7k? If so, all screen current must pass through almost 10k worth of resistance. I'd try changing the screen node to 1k/47uF, and connecting that to B+1 instead of B+2. Might get you a couple of watts more, but the audible loudness difference between 15 and 20 watts is very subtle. IMO, the reasonable parameters here is tone and character, which surely will change when changing the screen node setup.

[GlideOn]

The 5k screen node resistor will add quite a bit of overdrive compression, and reduce power output somewhat. I believe the 5k resistor is chosen to keep ripple as low as possible, or to keep screen current to a minimum. With only 10uF of filter capacitance, and the need for low ripple on the screen(s) of a single ended amp, the resistance must be high. The -3dB frequency of a 5k/10uF is 3.2Hz, not far from the -3dB frequency of a more conventional 1k/47uF filter which is 3.8Hz. When capacitance is decreased, resistance must be increased to keep the same cut-off frequency.

In the schematic, the 5k/10uF is connected to B+2, and if I understood you correctly, you have changed the 1k feeding the B+2 for a 4.7k? If so, all screen current must pass through almost 10k worth of resistance. I'd try changing the screen node to 1k/47uF, and connecting that to B+1 instead of B+2. Might get you a couple of watts more, but the audible loudness difference between 15 and 20 watts is very subtle. IMO, the reasonable parameters here is tone and character, which surely will change when changing the screen node setup.

100% agreed.

The preamp doesn't pass through the 5K/5w though. That's purely for the power section. The B+ line does a very funky bend upwards on the board to allow a starting point for the 5k/5w connection, then goes back under board diagonally towards the preamp tubes. I can't forsee a need for it doing this when I remove the 5k/5w. May just send it purely horizontal as I change it like other amps I have.

I figured choosing 4.7k dropper was close to the figure the preamp was seeing before:

300/*/300 x 1.4 = 415 raw B+

415 - 5v drop for 5H/150ma Choke = 410v

Using voltage drop calculator, to get low 300s for the V3 PI node, around low 20ma-ish, ballpark of 4.7k/3w was chosen.

Stock 10k/2w for the V2 Cathode Follower and V1 Gain Stages is perfectly fine, get me likely high 170v-180v ish. Kinda brown but I like a slightly brown preamp. It certainly sounds like it is.

As-is currently, the tone is nearly friggn' perfect. I do run out of steam when the Master Volume (10% audio taper at that) gets around noon, then instead of it getting louder it just compresses, get slightly mushy so bass gets rolled back, sounds great once again.

If I can eek just a but more headroom so that power amp clipping is more around 1 o'clock, I'll be happy.

Yes, by removing the 5k/5w the filter node I'll need to beef up the capacitance to achieve near or better the same filtering node here. More than happy to swap out the dinky 10uf/500v for a 47uf/500v which is larger, better filtering (better tone arguably) but now with the 5k/5w removed and B+ like going straight across - plenty of room and nicely configured the power section for tone and filtering.

One last thing is like to do: get more bias than 42ma per EL34.

I have 390r/100uf Cathodes now, per tube.

I'm wondering I could get at least 95-100% bias out of them being Cathode based.

That says somewhere in the range of 330r-360r I could use.

Other lingering thought is I'm wondering if I should beef up the power amp Screen Grid Resistors hanging off the pins 4 and 6. With the 5k/5w gone, that means the brunt of the B+ is now coming at them. I do see 1k/5w in 99% of EL34 power amps regardless of wattage, they all can't be wrong. Nothing wrong with the sound of them either IMHO.

However 1k is a significant voltage drop. Higher R means higher P.

Would 10w units the same value be better served here?


In an effort to be a good student, I have shown my math work below:

(New calculations updated 2.0)

[Bergheim]

What is the DC voltage drop across the 5k screen node resistor during full overdrive compared to quiescent/no signal?

[GlideOn]

What is the DC voltage drop across the 5k screen node resistor during full overdrive compared to quiescent/no signal?

I have no idea, it would be tricky to measure during full overdrive but highly likely lower than I'd care to have as far as headroom goes. I'm not as confident in my high voltage multimeter reading and I'm working up towards that.

5k is kind of a stupidly high value for screen. If I had to guess as to why it was chosen, the stock Hammond transformers were comically undersized and needed to keep the current/voltage demand in check. The 6550 it was meant to drive was criminally underpowered as a result, likely also why a comically high 1M Grid Leak ala Master Volume 1M pot was used to get saturated sound without stressing the transformers or tube.

It was originally an AX84 SEL build kit from Doberman amps that came with parts to meet the price point of $300 back in 2010. The chassis and board are arguably the best things about it, even still. I was cheap then, I think there was an option for $500 which included an upgraded Heyboer 600v PT and 25w OT. Took me 14 years, but in a way I finally upgraded to it, lol

[Bergheim]

Just hook up across the resistor with crocodile clamps, hooks or something of the sorts. Or get someone else to strum the guitar while you take the measurements. The readings don't need to be super accurate, +/- 5 volts is close enough, but without measurements you're really fumbling in the dark.

[GlideOn]

Friend, the resistor is going to be removed one way or the other so the voltage across it is irrelevant. More power is needed.

My primary goal at hand is safety. I need to know if anything needs beefed up as a consequence of higher B+ going to the screens. I was thinking 20w 1k Screen Grids for starters. Also, my B+ wire from the Mains Filter is Gavitt 22awg rated for 300v insulation. Current is in milliamps, yes but I could go 600v there for safety.

The tone happening in the power amp is secondary concern to output at this time, plus any modern EL34 by nature is gonna saturate and compress at some point whether sooner or later. It's what they do.

[Bergheim]

Yep, I know it is gonna be changed or removed, but if the amp is still functionable to the point of just turning it on, it would be nice to know straight away where the screens are at, dissipation wise. Then it will also be easier to estimate what resistance to change to afterwards, to extract as much output power as possible while also keeping the screens safe. The output impedance you're running is fairly high when looking at the load line, so the screens will try to draw considerable current when overdriven. Therefore I'd be catious to lower it too much without some beforehand measurements.

[GlideOn]

Impedance isn't really something I can change of course.

I selected 5k on the graph as that's the closest selectable estimate for a single tube. We are running two of them in parallel thru a 2.5k OT now.



EL34 nominal impedance is 3200-3400ohms.

*Maybe* slightly higher for EL34-B variants (like the Shuguang EL34-Bs I am using)

Parallel load, which halves the impedance.

Approx 1600-1800 ohms nominal.

That's why a 2.5k OT was chosen.



Yes, current draw may be a bit higher than calculated due to this slight impedance mismatch. It's still a reactive load after all. But this is a very common phenomenon, no?

Moreso to the point that the Screen Grids need all the protection they can get as more current = more heat. In fact the present 5K/5w in place may be already be stressed under increased current draw from parallel power tubes, so even more reason to be rid of it and not mess around with it or measure under heavy load!

Hence my thoughts to increase Screen Resistors to at least 20w of handling. Any larger they would need to be mounted to the chassis! This seems like being plenty cautious to me!

[Bergheim]

Friend, the resistor is going to be removed one way or the other so the voltage across it is irrelevant. More power is needed.

This is my point. How do you estimate or calculate the value of the new resistor if you don't know how much current is passing through the existing one?

My primary goal at hand is safety. I need to know if anything needs beefed up as a consequence of higher B+ going to the screens. I was thinking 20w 1k Screen Grids for starters.

No offense, but it seems this assumption is based on completely wild guessing.
The EL34 screen grids can dissipate max 8 watts, and with 415V (let's ingore the voltage drop across the PT secondary winding, the choke and the cathode bias voltage itself for the time being) it means current must be 19.3mA. Two tubes = 38.6mA. Voltage drop across the 4.7k B+2 resistor will be 4.7*38.6=181 volts.
Ditching the 5k resistor, we have 415-181=234 volts at the screen grid stoppers. 19.3mA through 1k ohms gives 19.3 volts drop. So 234-19.3=~215 volts directly on the screen grid pins.
Now you'll have to do this math and loadline work over again, since the screen voltage is obviously not 415V anymore, and therefore the current needed to dissipate 8W has increased.
You'll find that you can do this over and over again, and get new results each time, everytime getting a little closer to the point of equilibrium. But still not a definitive answer to how the amp will perform in the real world.
As I think I've mentioned before, calculations and loadline work is nice for ballparking the needed specs for transformers and such, but once the amp is up and running, the way forward is through measurements and testing.

Impedance isn't really something I can change of course.

I selected 5k on the graph as that's the closest selectable estimate for a single tube. We are running two of them in parallel thru a 2.5k OT now.



EL34 nominal impedance is 3200-3400ohms.

*Maybe* slightly higher for EL34-B variants (like the Shuguang EL34-Bs I am using)

Parallel load, which halves the impedance.

Approx 1600-1800 ohms nominal.

That's why a 2.5k OT was chosen.

Perfectly fine, all I meant was that with a loadline passing that far below the knee means that the screens are going to need a fair bit of series resistance to keep screen current in check.

Yes, current draw may be a bit higher than calculated due to this slight impedance mismatch. It's still a reactive load after all. But this is a very common phenomenon, no?

Moreso to the point that the Screen Grids need all the protection they can get as more current = more heat. In fact the present 5K/5w in place may be already be stressed under increased current draw from parallel power tubes, so even more reason to be rid of it and not mess around with it or measure under heavy load!

Hence my thoughts to increase Screen Resistors to at least 20w of handling. Any larger they would need to be mounted to the chassis! This seems like being plenty cautious to me!

These statements kinda confirmes why you really need to do the measurements of the real world (amp), your assumptions are way, way off. And again, no offense, but if you really insist on NOT taking the measurements, we're both wasting our time here.

[GlideOn]

This is my point. How do you estimate or calculate the value of the new resistor if you don't know how much current is passing through the existing one?

I'm stumped. What new resistor are you referring to?? My Preamp dropper? I've already explained 2 posts ago how I came up with that.

I had planned on leaving the 1ks at the Screen pins. At least the value. Same value as 99.99% of EL34s out there. I think higher value the tone suffers greatly.

The wattage is indeed suspect and as you say, the impedance will be on the high side, so more current draw and even more reason for high wattage resistors.

You haven't yet given feedback on the power handling/wattage of these components so I presume my suggestion is correct...

No offense, but it seems this assumption is based on completely wild guessing.
The EL34 screen grids can dissipate max 8 watts, and with 415V (let's ingore the voltage drop across the PT secondary winding, the choke and the cathode bias voltage itself for the time being) it means current must be 19.3mA. Two tubes = 38.6mA. Voltage drop across the 4.7k B+2 resistor will be 4.7*38.6=181 volts.

Maybe a layout refresher would help us here:

AX84 SEL V2.4.png

As we see, the 4.7k dropper is downstream of the 5K/5w Shared Screen. Is it relevant here?

The Shared Screen is technically not in series with the B+. Or at least it isn't physically in series. Is it electronically? I don't quite know. That's part of why this thing puzzles me as it is unique to the original design and I cannot find a similar instance in other designs.

I admit that my G1 dissipation @ 15w is high. We can play around this I'm sure. Maybe some tubes can handle this at momentary loads. Maybe a lower value B+1 dropper would be better?

Ditching the 5k resistor, we have 415-181=234 volts at the screen grid stoppers. 19.3mA through 1k ohms gives 19.3 volts drop. So 234-19.3=~215 volts directly on the screen grid pins.
Now you'll have to do this math and loadline work over again, since the screen voltage is obviously not 415V anymore, and therefore the current needed to dissipate 8W has increased.
You'll find that you can do this over and over again, and get new results each time, everytime getting a little closer to the point of equilibrium. But still not a definitive answer to how the amp will perform in the real world.
As I think I've mentioned before, calculations and loadline work is nice for ballparking the needed specs for transformers and such, but once the amp is up and running, the way forward is through measurements and testing.

Yes, it does indeed change. The Screen voltage has to be absurdly low to get the Screen dissipation below the 8w rating. Something in the range of 120-130 volts.

The cathode resistance goes way down, bias goes up and Screen values go high, something in the order of 2.2k @ 34w. Yuck!



As far as ditching the 5k resistor...

More than a handful of Single Ended schematics show a very small value dropper before the Screens/B+2 and B+1.

Or they do not have them at all! They instead opt for a 22uf, 33uf, 47uf Screen Cap Filter just as I suspected I could do too.

5k OT for one, 2.5k OT for the another.

So this is hardly fumbling around in the dark, as you say. Others have done it in the real world.

Here two examples:

SEPlex_v2.png

Plexi+6v6Se.png

These statements kinda confirmes why you really need to do the measurements of the real world (amp), your assumptions are way, way off. And again, no offense, but if you really insist on NOT taking the measurements, we're both wasting our time here.

Communication is hard. Especially with engineers, ha!

We can't read each other minds and we get easily frustrated when something doesn't translate.

I don't know what you know and you don't know what I know. Let's try to be patient with one another.

I am learning so much and I do appreciate your points thus far. Keep it classy.




So to simplify this, the solution is either one of two things:

1) Change Cathode Bias resistance.

2) Change Screen Resistance. Before or at.

Or a bit of both.


I know we're already in the ballpark of this. Let's bring it home.

[Bergheim]

Sorry, I was referring to the AX84 SEL schematic regarding the 5k/10uF node coming off of the 4.7k B+2 node resistor.
With only a 5H choke, and that high of a output impedance, you'll probably want more than 1k in series with the grid. Removing the 5k/10uF altogether might be risky.
It doesn't matter very much to the screens if the resistors comes before or after the filter cap. I've found that higher resistance in the individual screen grid stoppers gives a more defined, modern character to the overdrive, whereas placing that same resistance (of course halved if running two tubes and so on) before the filter cap, gives the overdrive more of a typical Fenderish overdrive with more attack. Probably due to the filter cap holding the voltage a bit longer before dropping due to increased screen current draw.
For example, having 500R before the filter cap is basically the same as 1k individual resistors on the two tubes, albeit two different overdrive styles.

Regarding the topology of the 5k/10uF branch, it's nothing special or problematic. I've done this myself in one of my amps that has a B+1 voltage of only 260V and a large screen node resistor. This keeps the preamp voltage from following the drop of the varying screen filter node voltage during overdrive.
In a normal topology, your 4.7k preamp dropper would be connected to the downstream side of the 5k/uF, and the preamp B+ voltage would follow the voltage of the screen node. In your amp, the designer possibly concluded that preamp voltage would drop too low because of the 5k, therefore connecting the preamp upstream of it instead. There's nothing wrong with the design itself.

You haven't yet given feedback on the power handling/wattage of these components so I presume my suggestion is correct...

There are lots of ohms law calculators out there, if you aren't familiar with the formulas already. You're unsure if 5w is enough on the 1k screen grid stopper.
To dissipate 5w over 1k, current has to be 70mA, which means 70V drop across it. Which in turn means screen voltage will be 70V lower "after" the drop. Which probably is gonna reduce screen current quite a bit, so voltage across it will be lower. So you'd have to do the maths over again to figure if the value is sufficient to achieve your design targets. New results -> new loadline -> new calculations -> new results..

You're showing schematics of two different amps using different tubes and probably different OT impedances. What they don't show is voltages, which would have shown how the amp will operate.

You're right about communication via text, misunderstandings come easy. And we don't know exactly the skill level of the other person.

What we do know is what we don't know:
- We don't know the B+1 voltage, idle and overdriven
- We don't know the screen node voltage, idle and overdriven
- We don't know the cathode voltage, idle and overdriven

I just now saw your comment about not being confident with the voltmeter, but you have to understand and accept that we're really not going anywhere without these measurements.

[GlideOn]

Thank you for stating that and reinforcing the points that are true thus far.

We don't know the dynamics under load and as you say, the multimeter voltage reading is key to that. I'll try to work up towards that. I do think I need better than a $20 Amazon thing to do that reliably, clip to where I put it instead of sit loosely like it does. I've never really needed more than that until this build.

I think the filtering is a moot point. 10uf to 47uf will do the same thing tighter or looser.

The key thing you did being up is the Screen dissipation.

I completely missed that in my load line, being in excess of 15w! I really don't know if an EL34B can tolerate that for extended period with some special circumstances, but I do have the mindset to error on the side of conservative for sake of reliability and long-term usability.

I suppose more or less voltage drop at the screens is more a matter of taste so as long as the current and voltage doesn't exceed the 8w rate we see on virtually every spec sheet. Not just EL34s mind you, but KT77, KT88, 6L6. They all have delicate 8w Screens or below, so we have to deal with this fact.

The problem as you also stated is impedance.

Voltage goes up, impedance goes up. I do have a high voltage for Single Ended Class A too.

I chose a 2.5k OT anticipating a lower Z due to paralleling the output section.

Perhaps this was a mistake?

Would it fare better with a 5k?

Otherwise it'd have to be some stupidly low cathode R or high Screen R to keep the dissipation under 8w.

And a combination of a stupidly high wattage component. At least that's what my erroneous calculators are telling me this far.

What to do now...

[GlideOn]

I think I may have a simple solution to try:


My OT is 2.5k

8 ohm and 4ohm outs.

300/0/300 PT




Time for some Merlin math:


300 * 1.4 = 420v

-5v approx for Choke

= 415v B+

415 ^ 2 / 25 = 6889

6889 / 2 (parallel halves load)

3445 total load.


945 ohms over of OT primary.


However, if I plug into the 4 ohm...

1723 ohms

775 ohms under the OT primary, no?

That give me a bit more breathing room to run my bias higher, get more power and stay under the 8w screen max.

[Bergheim]

Yes, halving the load will be easier on the screens.