12AU7/12BH7 Headphone Amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Jump to Latest
sorenj07

sorenj07

Member
Joined 2006
Hokay. Major update. After a boatload of hours inhaling sweet sweet solder/lead fumes, I came up with something. Not too unexpectedly, it makes almost ZILCH sound, in either headphone or line out. But, it looks kind of cool while doing it :)

I'll post full schematics as soon as I write them out nicely. In the meantime, it's a White CF per channel driven by a single triode grounded-cathode. All tubes are 7AU7's. Why 7AU7's? they're cheap. Why two filament transformers? I had them lying around when I thought I needed them for monoblocks. The big cap in the middle is also used, if you recall my first somewhat awkward 6SN7 preamp, that was one of the three. Also, please don't laugh TOO hard at my very "special" looking homemade cap clamps for the output couplers...

The two adjustable feedback pots are allegedly 50K linear but I never managed to measure anything higher than 25K from them.

In short - what in God's name do I do now to get this thing working? I'm thinking of subbing in some 6N23P's (they sim better in TubeCAD.) I'll also mess with the cathode resistor and bypass caps, because TubeCAD told me to use a value of less than 32uF...? This thing will be a bit of work especially since it's the highest-density soldering I've ever done.

Some pics:

Outside
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.

Inside
An externally hosted image should be here but it was not working when we last tested it.

Schematic
An externally hosted image should be here but it was not working when we last tested it.


(and yes, that is Deep Dish's classic "Penetrate Deeper" album. Cheesy deep house at it's best :D

If anyone wants detail pictures, let me know..
 
EC8010

EC8010

Ex-Moderator
Joined 2003
That circuit's not going to work as drawn. Start by removing the 0.47uF coupling capacitor between the input stage and the White cathode follower. Also remove the 1M grid leak. That should bias the output stage so that it can almost work. Then change that 14k (weird value) in the anode of the White cathode follower to 1k and you might even get sound.
 
sorenj07

sorenj07

Member
Joined 2006
Wow. I guess White CF's aren't quite as simple as TubeCAD makes them out to be :D The thing is, if I snip the .47uF coupling capacitor, how do I get signal to the output stage? Should I DC couple it? I'll also cut the 13K resistor I had in series with 1K for the 14K value.. I think I'd read somewhere that you want the plate resistor to be the inverse of the gm of the tube or something but I didn't really internalise it :(

Here's the article. I can't follow the guy's math though. From what I could find, the transconductance of a 6DJ8 is 12,500 micromhos or just "12.5" In what possible way could one invert that figure and end up with a plate resistor of 100? 1/.0125 gets me 80.

edit: should I try cutting the NFB connection? why does it give me DC offset?
 
Brian Beck

Brian Beck

Member
Joined 2004
About the DC offset: I don't know exactly where you'll have the first stage biased, but there could be 4 or 5 volts across the 910 ohm cathode resistor. The feedback resistor connects this voltage to the output. Without headphones plugged in, the 47K ohm output resistor will make a divider with the feedback resistor and you'd see about half of the cathode voltage at the output, maybe a couple of volts. When you plug in phones, the current will shunt through the voice coils dropping the voltage. The current is limited by the 50K feedback to maybe 0.1mA, so it's unlikely that the offset would damage the phones, but it would be a steady DC current flowing through the voice coil, pushing the diaphragm off its center point a bit.

Yes, I would try it without the NFB loop.
 
sorenj07

sorenj07

Member
Joined 2006
I changed my WCF's cathode resistor to 280 ohms to suck a bit more current. Plate resistor is now 1K 1W. I also cut the 1M between the bottom grid and ground, and both NFB connections. Unfortunately I'm still getting the same as before - really loud hum, and only the tiniest faintest bit of music when plugged into my EL34 Williamson with gain all the way up, from either output. Nothing through 32 and 63 ohm headphones on the headphone out.


STUPIDITY EDIT

I wired the input to the potentiometer wrong. It don't get much stupider than that. As soon as I fix, I anticipate at least SOME music.
 
sorenj07

sorenj07

Member
Joined 2006
I get heavy distortion at mid-to-high volumes on low-impedance headphones, but the line output works pretty well. There's still a bit of hum but it seems to be at acceptable levels. I had reconnected feedback when fixing the line output problem. Is there something else I could try to reduce this horrible distortion?

edit: changing the headphone out feedback pots from 25K to 1K just seems to decrease volume - no effect on distortion is detectable.

of course, when I pick up my oscope probes, I can have a better look.. 20Hz, 2KHz, 20KHz sinewave input should help demonstrate clipping, right?
 
Brian Beck

Brian Beck

Member
Joined 2004
OK, first of all leave the feedback disconnected until we can get this playing, and even after that probably. I hope you left the 1M grid "leak" resistor on the lower tube of the WCF. I think that EC8010 might have misspoken when he suggested removing the 1M grid leak along with the 0.47 uF cap. I think he meant for you to remove the 100K grid leak on the upper WCF tube along with the 0.47uF cap. You did then connect the upper tube's grid to the first stage's plate, right? Now, after making sure all this correct, can you post a new schematic with revised values and with measured DC voltages at the plates and cathodes, as well as measured values of B+1 and B+2? I don't think we're ready for music or even square waves just yet.
 
sorenj07

sorenj07

Member
Joined 2006
Brian Beck said:
OK, first of all leave the feedback disconnected until we can get this playing, and even after that probably. I hope you left the 1M grid "leak" resistor on the lower tube of the WCF. I think that EC8010 might have misspoken when he suggested removing the 1M grid leak along with the 0.47 uF cap. I think he meant for you to remove the 100K grid leak on the upper WCF tube along with the 0.47uF cap. You did then connect the upper tube's grid to the first stage's plate, right? Now, after making sure all this correct, can you post a new schematic with revised values and with measured DC voltages at the plates and cathodes, as well as measured values of B+1 and B+2? I don't think we're ready for music or even square waves just yet.
Click to expand...

Thanks for all the help! We did indeed cross-post.

I actually DID disconnect the 1M grid leak, and left the .47uF/100K RC. However I just clipped the 1M at one point and bent the lead away so I can use a spot of solder to reconnect. I'll DC couple the gainstage to the top triode in the WCF. My MM actually gives me the low-battery sign now and then, but it seems to be working other than that.
 
sorenj07

sorenj07

Member
Joined 2006
Unfortunately, these beefier output valves (I have a quad of 6CG7's in my EL34 Williamson, for example) will probably draw more than .3A of heater current. As the PSU schematic I posted earlier shows, it's a contest between heaters and B+ headroom. I'll pop those 6CG7's in but hopefully B+ won't sag too much..

In order to provide some extra voltage swing to feed some NFB, might I try dropping a 6DJ8/6N23P into the first stage? I'd do a quick bias recalculation, but the extra gain might help, right?

also, what value resistor might I try for NFB? I took out the pots already because I might want less NFB than 25K would afford..
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Top Bottom