G2 power testing and conduction angle

[Andy Le Blanc]

Never lacking in things to say about screen grids, I set up a power test
over a 8r load and took measurements under static and dynamic conditions
to explore changes in G2 voltage relative to plate.

condition 1:

static=

100.3mv plate current/bias...415v plate... screens -25 relative to plate

dynamic=

180mv plate current... 389v plate... screens -101 relative to plate

max Vout = 19.57 or 47.87w.... P=e2/r

CA= 247.7... 2x cos-1(-100.3/180)= 2xcos-1(-.5572)

condition 2

static=

100mv plate current/bias...412v plate... screens -9.1 relative to plate

dynamic=

189mv plate current... 370v plate... screens -51.2 relative to plate

max Vout = 19.3 or 46.5w.... P=e2/r

CA= 243.89... 2x cos-1(-100/189)= 2xcos-1(-.5291)

it looks like G2 voltage relative to plate also effects where in class AB the
amp rides at full power, in the equation 2xcos-1(-1) =360 or class A
I'll bet the fender circuit pushes as for to the B and max efficiency
as is possible (G2 slightly positive to plate).

[David Root]

Some old tweed Fenders do have their screens at the same or slightly higher, 1-3V, than the plates, which is to be expected when you feed the OT centertap and the screens (no screen resistors on these amps) the same voltage, and losing 1-3V in the OT windings.

What's the explanation for Condition 2 screen drop relative to plate is so much less than in Condition 1?

[Andy Le Blanc]

condition 3

static=

100.8mv plate current/bias...409v plate... screens -1.52 relative to plate

dynamic=

199mv plate current... 367v plate... screens -18 relative to plate

max Vout = 19.4 or 47.04w.... P=e2/r

CA= 240.87... 2x cos-1(-100.8/199)= 2xcos-1(-.5065)

you can begin to see the trend with the conduction angle toward "B"
just by keeping the bias condition static, the G2 relationship to the plate
is a lot more telling than first glance, increasing the difference, or making
the screens more negative than the plate would seem to push the
power side a little more towards "A"

this would seem to be what you hear when you change screen grid resistors

but there are draw backs to using series resistors, the G2 voltage in my
test was manipulated with a voltage divider placed before the regular
470r screen grid resistors.

condition 1 was 2.2k and a typical series filter cap with 220k balancing resistors.

condition 2 I paralleled the 2.2k with a 1k giving around 687r

and condition 3 the 2.2k was bypassed placing the bleeders after
a choke in a circuit that typically has none.

[David Root]

Again that makes sense, screens closer to plates under load, tends to B, screens much lower than plates under load, tends to A.

[Cliff Schecht]

Agreed this is makes sense also. As the screen drops much below the plate the tubes can operate in the more efficient pentode mode which explains the lower current draw for the same max power output. As the screens get closer to the plate they are less efficient triodes and require more current to get the same power. This in itself implies that more pentode operation is closer to class B and triode operation is closer to class A (efficiency-wise).

Before I even saw this thread, during my 6-hour road trip to Dallas today, I was thinking about Andy's past discussions about screen grid play and was thinking of building an amp where I could play around with the screen voltage easily to see its effects on sound, bias, current draw, etc.. I think since Andy has done most of his work with push pull amps I will build something that allows me to do similar studies with a class A build. In my experiments thus far I've already found that the screen grid biasing can have a very dramatic effect on the sound. This merits further investigation, once I get back to my lab in 3 months..

Thanks for more interesting observations Andy!

[Andy Le Blanc]

Assuming the input V signal for P-P or SE..

you have only a short list for design considerations...

tube type

voltage

loading

degeneration/feedback

screen grids

It's more a matter of clarifying each aspect.
each tube type is well described, plenty of reference there,
voltage is PT and PS choice, loading is OPT and speaker,
feedback can be global or a cathode resistor, nearly everything
can be measured DIY, the transformers can now be more easily custom
wound, it wasn't too long ago that you couldn't...

but I've been watching too many blather on about screen grid resistors
without knowing what they where affecting or even a basic static measurement
on the bench DIY... and it buged the heck out me...
bias is the same, when you have control the tubes don't matter

no joke... same loading and FB, same PS bias, same G2 relative to plate

the same circuit can run 6v6 to kt88 and everything in between
and you'll be hard press to tell them apart, you have to run harmonic distortion
test to keep from making qualitative statements about your favorite type

[Gaz]

Screen grids have always been the bane of my amp builder existence. Thanks for helping me make sense of the world, Andy!

One simple thing I don't get is how your 2.2k/220k voltage divider is dropping the screens 25 volts?

[Ian444]

Gaz, I often think of the screen grid voltage as an equivalent to turbo boost on a car engine. Turn it up too much, it blows up, turn it down too much, big loss in performance.

But its not that easy, its the power (voltage x current) that kills screen grids, so you can't just say "x" amount of volts is too much. The screen grid resistors and seperate screen supply (if it exists) are the major factors. The grid resistors limit the current by dropping the voltage during peak demands of power when the tube is in cut-off and the plate voltage is minimum and most of the electrons are heading for the screen instead of the plate, since the screen is now more positive than the plate. These days its standard to use 1K screen grid resistors on EL34 tubes, to help them live longer, but 40 years ago the standard was 220 ohms or none. Something has changed, tube quality I guess?

In the conditions 1, 2, and 3 above, same power output, swap more current for less change in voltage, or vice versa. In some datasheets you get a graph for 2nd and 3rd harmonic distortion vs load impedance. Usually with heavier loads (lower impedance) there is higher current drawn through the tubes, more 2nd harmonic distortion, less 3rd harmonic, and a bit less power, and more chance of exceeding tube power dissipation limits.

[guitarmike2107]

Andy, Did you stick a big ass filter capacitor after the voltage divider? if not changing the series resistance on the screen grid would account for the varying compression you are seeing on the screen grid voltage.

[Andy Le Blanc]

Apologies... here's a little clarity

Amps cool and the load is static, It all looks familiar and simple...

your utilizing a dropping string, letting the bleeder work for you where you
need the regulation the most, and able to manipulate the G2 voltage

The 2.2k can be replaced with a diode, the bleeders are low wattage
I'd still keep some value of screen grid resistor, its the voltage difference
that's important, and you can bypass, or not, to taste.

your decoupling the G2 from the power rail as well, giving independent
control, to a degree, on G2 voltage from the rest of the pre voltages

simply clipping in to tube socket to measure the difference means you
can set it up statically, and even make the control variable like a bias

[Andy Le Blanc]

I posted a simple mod using a 100r/220k divider a couple days back
might be what Gaz is thinking of..

getting a real control over screen voltages is a eye opener, no assumptions

just two resistors to back an amp away, slightly, from the "B" end of operation

a real measurement, the difference at the tube socket between plate and G2
at set up that can be repeated, with accuracy is fantastic.

Like a favored bias setting for the tube type and load of a specific amp

[Gaz]

Andy, in your drawing, you're simply just adding a series resistance with the choke. I don't really see how there's really a "voltage divider."1 The 2.2k/440k divider (formed by the series bleeders) is only dropping a little over 2 volts itself. That drawing is unlike the one you posted in the other thread http://ampgarage.com/forum/download/file.php?id=16989, where it looks like you have a separate divider/cap coming off of the existing screen node. Let me know if I'm even further confused here, Andy. Sometimes your posts resemble modern poetry. LOL.

[Andy Le Blanc]

their the same... the earlier is simplified

the 2.2k in the last, is the G2 set, I was farting around on the bench
with common elements to see how well it would function.

I started at the power rail in the last with around 431V, and after the 2.2k
it was around 408V, and the screen grid resistors dropped 2v

the 2.2k and bleeders form the divider, it looks like a common power rail
because its the same components your used to seeing in a fender schematic
but here they are used differently

[Gaz]

Andy, I just tried setting this up again with a 80W 6L6 amp. Here's the schematic so there's no confusion. I tried to do it just like you did in the picture. [img1131]http://i55.tinypic.com/28icnjl.png[/img]

I was able to get that 18 volt drop with the 6L6s, but output power was reduced to 20vac, or 50 watts RMS. I even tried bypassing the 1Ks, but that did not increase power whatsoever.

I still don't see how this setup is doing any regulation with the bleeder at such a high value like that. To me it's just another PS node, not really a voltage divider. I also don't understand how your power output increased, or how you were even able to get near 50W output with a 415vdc B+ to begin with, that sounds totally impossible to me. The most I can get with that B+ and two power tubes is about 35-40W RMS, and that's with the screens a few volts under plate voltage.

This just isn't adding up...

[Andy Le Blanc]

Thats ok... the resistors do form a divider. You can set it any way you want.
the voltage drop is whats important.

Test criteria I think is what the difference we are seeing....

P = e2/r... yes? volts squared divided by the load resistance?

I ramped up a 1k signal until I observed a peak volts reading at output
on my meter, then did the math.

the PT I used is a #022798 200ma... I maintained the bias at 100ma
through out all the test conditions

The tube set in my tester however where kt88

I should be able to repeat the conditions with 6l6, 6550, el84, etc

your schematic is nearly identical to my tester.

I more than appreciate you replicating a test and refuting my test numbers
That is truly what a real discussion is about, this is not a clone

your not wrong as seeing a voltage divider as just another node..

I guess the most Important question is what test criteria can we agree upon?

how about going backwards a bit and running a simple divider, un-bypassed
keep the 2.2k, run a range of bleeder values, and maintain the bias at
a constant and known current?

what tube types do you have on hand? and what value of screen grid resistor?