GA-20 Voltages Dropping Too Much

[AHeck]

Good Afternoon,
I just powered up a new build and I am wondering if anyone can advise regarding the second dropping resistor on B+.  I have not found any but one one schematic for the original GA-20 with the octal front end and had a hard time reading it.  I was hoping that I'm mistaken and the 47K resistor between B+1 and B+2 is really supposed to be a 4K7.  Any help would be appreciated.

[Dave]

Never built one and never played through one, but 47k sounds ridiculously high. 4K7 would be a normal enough sounding resistance for a dropping resistor. 47k would be more than I've ever heard of.


Dave

[Dave]

Had a look at the schematic... certainly looks like 47k. Also I remembered that I have, in fact, seen a dropping resistor that high before. Blonde Bassman's have a 56k.


Dave

[Dave]

The dropping resistor values are not set in stone. If you don't like the voltages, change them until they look right or make you happy.


Dave

[sluckey]

The original schematic shows 47K.

[AHeck]

If you don't like the voltages, change them until they look right or make you happy.


Dave

Thanks for the quick reply, and of course you're right.  It would help me considerably if I had an idea where to start. I'm not familiar with the way that the 6SJ7s are biased and I'm not sure what plate current I'm aiming for.  The datasheet puts the normal operating conditions at 2.9mA for a plate voltage of 100V.  Is that dependent upon bias technique?

[vintasonic]

Here's a fairly clean schematic if it helps

[PRR]

RDH4 recommends 10%-20% of the plate resistor for economical decoupling and filtering.

Taking 470K||1Meg||470K||1Meg= 160k suggests 16k-32k.

As a first-dart you expect plate to be about halfway up the supply voltage so it can swing both ways. So for 192V you expect 80V to 100V.

> 2.9mA for a plate voltage of 100V.  Is that dependent upon bias technique?

Of course; and it may be a "show-off" condition. There is NO WAY there is 2.9mA here: 200V and 500k is AT-MOST 0.4mA, and likely closer to 0.2mA when plate sits halfway.

7V is just wrong. That tube is bottomed. And it isn't one bad tube because both show the same?? Is it really wired that way? If grid 1 were grounded it might sit low, but the plan shows 10Meg and a capacitor.

That schematic is not authoritative. It probably was not drawn in the lifetime of the designer. Question everything!  EDIT- ok, the hand-drawn one just posted is contemporary, and does match the machine-draw job pretty near....

Use the RC tables painstakingly compiled by some junior engineer to encourage tube sales. Reject the grid-leak bias because that went out when hot pickups came to market. I've hi-lighted a condition of similar supply and plate loading. Add Rk, change Rg2, change Rg, short input cap.

[bmccowan]

I'm pretty certain that the 47K is correct. Maybe there are exceptions, but I think all octal 2x6V6 GA-20, GA-25, GA-30 have a 47K in that position. I have a 20 and a 25 - both sound good with the stock value.

[HotBluePlates]

... 47k sounds ridiculously high. 4K7 would be a normal enough sounding resistance for a dropping resistor. ...

... If you don't like the voltages, change them until they look right or make you happy.

... I'm not familiar with the way that the 6SJ7s are biased and I'm not sure what plate current I'm aiming for.  The datasheet puts the normal operating conditions at 2.9mA for a plate voltage of 100V.  ...

High plate current makes other data sheet numbers "look better."  But preamp pentodes in guitar amps were typically run at low current, which some information suggests is a higher-gain implementation.  That is, high-gain in this case happens at low screen voltage, which also leads to lower plate current.

If the tubes operate at lower plate current (and smaller current-variations) then the plate load resistors need to be made larger to achieve the same voltage-output-swing.  That's an application of Ohm's Law: Same Voltage / Less Current = More Resistance. 

So the 6SJ7s have 470kΩ plate load resistors rather than a 12AX7s 100kΩ.  As PRR points out, tube plate current is very low compared to the 1mA typical value of a 12AX7.  So Ohm's Law makes an appearance again in the power supply dropping resistors; to get same-voltage-drop, the resistance again needs to increase:  Same Voltage / Less Current = More Resistance.

Of course, PRR pointed out there is a rule of thumb about making the decoupling/dropping resistor a percentage of the plate resistor values present.  That is no different than what I described above, but is thinking about the problem from a different direction, and applying a time-saving "rule" rather than making a longer series of calculations to arrive at the same answer.



Which means what?  That 47kΩ is perfectly reasonable in a power supply when the currents of individual tube stages are low.

[AHeck]

Use the RC tables painstakingly compiled by some junior engineer to encourage tube sales. Reject the grid-leak bias because that went out when hot pickups came to market. I've hi-lighted a condition of similar supply and plate loading. Add Rk, change Rg2, change Rg, short input cap.

Thank you for your help.  I did change one of the inputs, adding a 1K8 at Rk, changing the 1M to 2M2 at Rg2, but keeping the .05uF to ground, added a 270K from input to ground and shorted the .05uF input cap.  I'm now getting 40V at the plate, 10V at the screen and .5V at the cathode.  I was actually getting less gain than the other channel, from 3.1V p-p from .1 V input to 2.7V self biased. I went ahead and bypassed Rk with 10uF and it bumped up to 7.1V. But I also went from approximately 8W of clean output to 6W.  I dropped the bias resistor from 250R to 200R and I'm still only getting about 37mA each out of the 6V6s.  I'm still working on it though.  Again, thanks. 

[AHeck]

I made some changes as per advice given, which I was very glad to have been given.  I've attached an updated schematic and D/C voltage chart. Now I have more questions and observations.
 
First A Question: Why might the screen voltage on the newly self biased 6SJ7 be so low? I went from a 2M2 to a 1M and only gained a few volts.

Next An Observation: The other input, the original one, sounds heads and shoulders better, which I found strange.

Again An Observation: I have very little clean headroom as the signal is clipping very early in the PI

And Finally, Question: Even though the power tubes are properly biased at 100% dissapation, I'm only getting ~5W clean output (4.5VRMS^2/4).  Any obvious reason or would an A/C voltage chart be needed to get to this?

I want to thank those who have showed out to help so far. I was at this all day today and I'm a bit frustrated. 

[PRR]

> Why might the screen voltage on the newly self biased 6SJ7 be so low? I went from a 2M2 to a 1M and only gained a few volts.

You don't need high screen voltage with a huge 470k plate resistor.

You don't want high screen voltage if you want high gain.

Twenty or twentyfive volts is ample.

[AHeck]

> Why might the screen voltage on the newly self biased 6SJ7 be so low? I went from a 2M2 to a 1M and only gained a few volts.

You don't need high screen voltage with a huge 470k plate resistor.

You don't want high screen voltage if you want high gain.

Twenty or twentyfive volts is ample.

So, with the intention of getting half of the supply voltage at the plate, I dropped Ra to 100K, a 470K Rg2 now has me at ~156V supply, 80V at the plate, and 46V at the screen.  Biased -1.7V.  10uF bypass cap andI'm getting a gain of about 55, which is still overdriving the PI very quickly.  It sounds a whole lot better though.

[HotBluePlates]

... still overdriving the PI very quickly. ...

How do you know you're "overdriving the PI"?

Without analyzing the circuit, I would expect the 6V6s to distort (by having a grid-signal input larger than bias voltage) before the phase inverter tubes actually distort.

Watching this on a scope by probing PI plate outputs, when the 6V6s receive that too-big input signal their grids will clamp the output of the PI plates.  That makes the PI plate-output look like it has flat-topped (or clipped), but it's not the phase inverter that is distorting but the 6V6 grid that are clamping/distorting the signal.

[dwinstonwood]

First, this is an EF86 Pentode, not a 6SJ7.

But, it's an example of how low a pentode's screen voltage can be, and still have all the gain I need.

[Why the drastic 5.6M screen resistor? I tried various other resistor values, and I could only get the Tremolo to work by using the original AC4 values. Maybe Vox kept increasing it until they got it to work; we'll never know.]

[AHeck]

... still overdriving the PI very quickly. ...

How do you know you're "overdriving the PI"?

Without analyzing the circuit, I would expect the 6V6s to distort (by having a grid-signal input larger than bias voltage) before the phase inverter tubes actually distort.

Watching this on a scope by probing PI plate outputs, when the 6V6s receive that too-big input signal their grids will clamp the output of the PI plates.  That makes the PI plate-output look like it has flat-topped (or clipped), but it's not the phase inverter that is distorting but the 6V6 grid that are clamping/distorting the signal.

Okay, I wasn't aware this was a thing.  Thanks for the information.  I was going by probing the output of the 6SL7 outputs like you said.  I have about 10% of volume potential before I see the PI outputs clip and thought that it was a bias issue there.  The cathodes of the 6SL7s are only sitting at 1.2V.  I'm going to have to learn about the grids clamping the output of the driver stage.

[bmccowan]

Not to short stop this exploration, which is quite interesting, but if you have not, you may want to look at some other Gibson circuits of this time period. The BR-6, GA-25, and GA-30 all use the 6SJ7 in the V1 position, but use different PIs. You will see different plate load resistors: 100K, 220K in addition to the 470K, and of course changes to the screen and cathode resistors to match. Just food for thought. Those schematics are here on Doug's pages.
BTW, do not expect to get a lot of clean power from a Gibson circuit. That's not what they do best.