Mysterious interaction between EF tube placement and channel balance

[ccmccull]

Hi folks,

I have a fun conundrum for you. First off, I have the amp running beautifully so there's nothing to solve exactly, but I did notice this weird behavior and wanted to see if anyone could explain it.

All resistance and voltage checks were well within spec. I can provide them if they're useful. And photos too if you want.

When I first hooked up the eros, I had a distinct feeling that the phantom center channel was pushed ~30% toward the left speaker. So I started taking measurements as suggested in https://forum.bottlehead.com/index.php?topic=13072.0.

I recorded them in the following google sheet: https://docs.google.com/spreadsheets/d/1WUQ_Z-eWfubSRfNxxPuCLwYB4QuHu0oBYpv9H1EhPws/edit#gid=0

Indeed, Left / Right was consistently ~1.25 , between 20 to 10,000 hz. So it didn't seem like an eq issue. Next I swapped the EF86's. The L/R reduced to ~1.05! I assume that's within tolerance, and it audibly places the phantom center right in the middle of the soundstage.

Then I cleaned the EF86 pins and re-tested, and swapped them back and forth for more measurements. Same results!

So I have a "good" pairing EF86s to the two sockets, and a "bad" pairing. What could cause this?

 

[Paul Joppa]

The two triodes in the 6922 have slightly different gains, and so do the two EF86 pentodes.

 

[ccmccull]

Thanks Paul.

This comment helped me realize I can use the data to estimate the gains in the 6922 and the EF86 separately. Or at least I can estimate their individual relative gains (Left / Right), since I don't have an accurate input voltage measurement (it's lower than the minimum 10mV reading on my DMM). And the EF86 swapping allows me to generate two replicate estimates at each frequency.

I added my math and estimates to the sheet if anyone's interested.

Summary:

Tube      Left Gain / Right Gain
6922              ~1.15
EF86              ~1.09

So by swapping the EF86 tubes I can almost cancel out these two gain imbalances. (1.15 / 1.09 = 1.06)

For my education, is this amount of variability in gain typical for EF86's and the two sides of the same 6922 tube?

 

[2wo]

In my experience, 6922's  tend to be reasonably matched shouldn't be hard to find. The EF86 is a pentode, considereabley more sample to sample variations are normal...John