Help identifying an issue and a fix

[boodle]

Hi all,
I have an amp build that started as Sluckey's DualLite, but with 5879 and ECL82 tubes and only the Voxy channel.  It is a success.  (Schematic attached in second post due to file # restrictions.)

When the build started, I was going to use a quad of ECL84's, but it was suggest that I consider using a pair of ECL82's instead.  Sounded like good logic, so that is what happened.  I had already put in sockets for the quad, so just left two unwired.

Played a Christmas gig with it last week in a big room with lots of hard surfaces and the mic'd amp could barely be turned up past 10 o'clock on the volume dial.  This is just below where the amp really comes into its own, so it worked for the gig, but left me wanting in the tone department.  So my 10W amp was too much for the room!  This got me thinking...what if I wired the 2 unused sockets for the ECL84's and had the option of running with either tube.  So, 5 watts or 10 watts.

Besides power output of the pentode sections, the other significant difference between the tubes is the ECL84 triode has a transconductance of 4mA/V and the ECL82 has 2.2mA/V.  mu's are almost identical at 65 and 70, respectively.

The only thing I changed about the circuit was to have the grid stops on the ECL84 be 47k  vs. 8k2 for the ECL82's.  I did this for no other reason than that seemed a similar value to other amp designs and it was easy to implement.  Perhaps a poor choice?

The issue that I am having with the ECL84's installed is VERY low headroom and some very displeasing overdrive at VERY low volume levels.  As can be seen by the oscilloscope pics, the distortion is very asymmetric.  The scope measurement was taken at the speaker jack with a 10 ohm 100W resistor attached to the jack.  The input signal was 1k2Hz at the instrument input.  ECL84's are NOS RCA's.

Pic 1 is max "clean" ~ 9 on o'clock on volume
Pic 2 is onset of distortion ~ 10 o'clock on volume
Pic 3 is at 1 o'clock on volume
Pic 4 is just shy of dimed at 3-ish o'clock on the volume.  5 o'clock is dimed and when dimed, the positive side of the wave flattens out.

Is the asymmetry due to very mismatched triodes in the PI?  The asymmetry seems way too large to be due to the intentionally different PI plate resistors of 82k vs 100k.

Will crack open the amp to get further scope pics as necessary.

Thoughts?

[boodle]

Amp schematic.  Note that the voltages shown are with the ECL82's installed.

[Lectroid]

Did you consider wiring the 5879 with a pentode/triode switch to lower the overall output?

[boodle]

Lectroid,
I love that idea!  I will test that.  It should give me another tone relatively simply.

Still wondering what is causing such significant asymmetry (and thus terrible tone) in the overload of the PA.

[boodle]

Looking back at my voltage measurements, with the screen about 100V lower than the plate at idle in pentode mode, I wonder if putting the tube in triode mode with the plate, g2, and g3 all at ~166V will lower the output? 

Hard for me to discern from the datasheet. 

The datasheet seems to say that the current will increase significantly in triode mode, however, plate dissipation increases by only ~20%, so that must mean voltage gain decreases into the plate load when in triode mode?  VA doesn't change by much, but I goes up significantly, thus, V out is lower?

But, there are many things for me to still learn about tube curves.

[Lectroid]

I likely know less than you so I’ll let wiser heads answer your questions. One thing, I wonder if your screen voltage is high enough on the ECL82?  Is it really 100V lower than the plate?

Most schematics I’ve seen show pentode screen voltages just a few volts below plate voltage.  Your own schematic shows 243V and 233V for plate and cathode respectively.  Did you mean to write "10V lower than the plate"?   If it's only 10V lower, looks right in the ballpark to me.

Either way, that's all I can suggest.  I’m interested to hear some of the smarter brains weigh in on this.

[sluckey]

I likely know less than you so I’ll let wiser heads answer your questions. One thing, I wonder if your screen voltage is high enough on the ECL82?  Is it really 100V lower than the plate?

Most schematics I’ve seen show pentode screen voltages just a few volts below plate voltage.  Your own schematic shows 243V and 233V for plate and cathode respectively.  Did you mean to write "10V lower than the plate"?   If it's only 10V lower, looks right in the ballpark to me.

He's not talking about the power tubes. He's talking about the preamp tube. It's quite common to see the screen much lower than the plate in that 5879 circuit.

[shooter]

Is the asymmetry due to very mismatched triodes in the PI?

Pic 2 is onset of distortion ~ 10 o'clock on volume          Pic 3 is at 1 o'clock on volume

the asymmetry is there before the PI, my guess the PI adds some since it's getting beat pretty hard, then the 84's in Class A, really don't like being overdriven.
my "rule of thumb" to limit OD at the PA, set the max drive signal (Vac) to the PA grids ~~~~== to the vdc at the PA cathodes.

[Lectroid]

He's not talking about the power tubes. He's talking about the preamp tube. It's quite common to see the screen much lower than the plate in that 5879 circuit.

My mistake. Thanks for the correction.

[boodle]

Thanks, Lectroid, Sluckey, and Shooter for your input.  I appreciate the help.

Had some time today to do more experiments.
1. tried other ECL84's
2. tried the original ECL82's

Both cases still had that very strange asymmetry, so I decided to try pulling the 86k and 0.005 plate resistor bypass cap on the + side of the PI and replace them with 100k to match the - side plate resistor.

Magic!  This amp may now be the best sounding amp that I have owned/made.  Having the symmetrical PI also cleared up some other issues with the amp: Brilliance switch under performing my expectation and making the Cut control much more functional across the entire turn. 

The Brilliance switch is essentially a switchable hi-pass that had original knees of 33 Hz, 76 Hz, and 118 Hz.
Worked OK, but didn't cut enough of the low end to easily fit in a dense mix.  Correcting the PI allowed me to hear its full behavior.  While I was in there, I replace the fixed 0.0027 cap with 0.0016 (2 - 0.0033 in series) to reset the knees to 38, 103, and 199 Hz.  Much better range.

Here is the scope image for 2 ECL84's with the volume close to dimed.  Fairly symmetrical top and bottom with expected cross-over distortion.

Thanks again!  Updated schematic added.  The amp is called Thorsdatter in honor of my daughter (who is a beautiful strong young woman!)

[kagliostro]

Did you consider wiring the 5879 with a pentode/triode switch to lower the overall output?

if you decide to experiment I'll prefer Merlin's Morph instead a switch



Our friend Tubeswell (and others) tried it and were satisfied about
(remember to use a Grid Leak Resistor)



Franco

[boodle]

Kagliostro,
Thanks for the tip!  It looks very interesting.  I look over the forum for tone descriptions.

[boodle]

Here is something that I am puzzled about...from my original post:

"The only thing I changed about the circuit was to have the grid stops on the ECL84 be 47k  vs. 8k2 for the ECL82's.  I did this for no other reason than that seemed a similar value to other amp designs and it was easy to implement. "

For grins, I tried the ECL84's in the sockets with the 8k2 grid stops and the amp was SILENT, zero sound from the speaker.  I am assuming that the pentodes were in cut-off with those grid stops?

I put the same tubes in the sockets with the 47k grid stops and they worked as expected (wonderfully!)

In both cases, the grid leak resistors were 470k to ground.  Seems like the grid leaks should have been able to bias the tube for operation in both sets of sockets. 

The ECL82's work fine in the sockets designed for them (the 8k2 grid stop sockets).

Could there be something else going on?

BTW, my underlying intention with that experiment was to see if I have all 4 sockets be identically wired for the pentodes, but have only 2 wired for the triodes to handle the PI job.  This would give me 3 different power tube arrangements:  2XECL82, 2XECL84, and 4XECL84.  The OT can handle impedance matching pretty well in all 3 cases.
I don't think I am losing anything major by not having the 4xECL84 option.  It is mostly a sonic curiosity.  Not sure I want to undo the 2 triode circuits to satisfy it or not.

[boodle]

OK, I'm a bonehead.  I can't use the ECL82 sockets for ECL84s or vice versa BECAUSE THEY HAVE DIFFERENT PINOUTS. 

Wow.  Just...wow.

[PRR]

> DIFFERENT PINOUTS

I believe you; just had to see it. Several pins do the same, but that's not a simple changeover.

I'd kinda assumed you would double-up on sockets.

[Wonder2121]

Same power section Wem Westminster MK9... https://www.vintagehofner.co.uk/britamps/watkins/schematics/westminmk9.html
he used a conjunctive filter over OT....

I could never get good bass sound from el82s..