Thanks for your replies Doc & PB. First off, I want to make it clear that I'm not complaining about the L/R bias, it's entirely manageable by the moreplay balance setting. I was just curious what was causing it and had some time on my hands to dig in. I learned a bunch in the process. I hope the data I collected in the process are interesting / useful to someone.
Doc, the "20% left shift" was a perceptual shift - i.e. I perceive the center voice about 20% toward the left channel from center. On your suggestion, I'll get much more quantitative below.
Also, PB, I think I wasn't clear enough with how I was defining the 10% voltage divider bias I quoted. See the sheet I link below, "resistances" tab columns E & F for my definition. However, I also measured voltages as you suggested.
I've done a whole bunch of measurements throughout the audio signal path with my new oscilloscope. It's super interesting to see how each stage adds its own bias, some positive, some negative, and they all contribute to a total L/R bias on the speaker terminals. Here's all the data if you'd like to dig in yourself, but I'll give a summary below.
https://docs.google.com/spreadsheets/d/1FT5aIulK0keVEQmUIRlhREjrovQArGRNucm1jfJvB1g/edit?usp=sharingConclusions:
The pot adds 10% Vpp L/R bias for any setting below ~70% on. After the pot, the rest of the amp adds an additional 4-8% of L/R bias at 1000 hz and 5-10% at 15000 hz, and most of this additional L/R bias comes from the OTs. Setting the volume pot to full on provides the lowest overall L/R bias. Practically, these results are educational for me because I've never understood why I sometimes had to dial my balance pot off-center and other times I didn't. Now I understand that I should always run the sex amp at full volume to remove that source of variability and use the pre-amp for volume control.
Methods:
I used a function generator for sinusoidal test tones and dialed the input Vpp so that the output Vrms on the left speaker terminal was fixed regardless of the pot setting. I measured the Vmax, Vmin, Vpp, and Vrms on a Rigol ds1054z for all combinations of the following:
- test tones @ 1000 hz and 15000 hz
- input Vpp set to achieve L speaker output @ 1000 hz of 1.41 Vrms and 0.7 Vrms, separately. I re-adjusted the input Vpp for each pot setting. I also did one high output power test @ 3.6 Vrms
- measured 4 points in the signal path: from tube pin 10 through to the speaker outputs
- 4 pot settings: off (9 o'clock as I orient my pot), 12 o'clock, 3 o'clock, 5 o'clock, and full on.
Results:
- The Vpp L/R bias measured at pin 10 is similar to the voltage divider bias measured by pot resistance (see resistances tab): unbiased for pot at or above 5 o'clock, and +10% for 3 o'clock and below. As expected, this bias isn't frequency dependent.
- The first stage voltage amplifier adds < 1% of additional L/R bias at 1000 hz, and that's likely within measurement noise. At 15000 hz it adds 4-5% additional L/R bias, and as expected it isn't dependent on pot position
- The 2nd stage amplifier subtracts ~ -5% Vpp L/R bias @ 1000 hz and ~ -3% @ 15000 hz
- The OTs add ~ +10 to +14% Vpp L/R bias @ 1000 hz. At 15000 hz it adds between 5% and 10%. Unexpectedly, these biases do seem to be pot dependent. (larger bias for lower pot settings)
- Overall, the Vpp L/R bias at the speaker terminals are between +4% up to +18% @1000 hz, depending on the pot setting. The L/R bias is slightly higher at 15000 hz, ranging between +5% and +21%. The lowest bias is at pot set to full volume, the highest bias is when the pot is set to 3 o'clock or lower.
- The total Vpp L/R bias at the speaker terminals is roughly the same for all the output powers tested. Even up to the 1.6 output watts I tested on the "higher voltages" tab, the L/R bias is similar.
