diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Tubes / Valves (http://www.diyaudio.com/forums/tubes-valves/)
-   -   Hafler/Keroes 6L6 UL (http://www.diyaudio.com/forums/tubes-valves/84475-hafler-keroes-6l6-ul.html)

sorenj07 9th August 2006 08:55 PM

Hafler/Keroes 6L6 UL
 
I'd picked a schematic to build, and lo and behold it's one of the earliest (if not the original) incarnations of Ultra-Linear topology, as invented by Hafler and Keroes of later Dynaco and Acrosound fame.

Here's a nice copy of the schematic:
http://www.dc-daylight.ltd.uk/Valve-...L-H&K-Fig3.pdf

I've got a matched quad of Sovs (the WXT+ kind), and matched pairs of Sov 6SN7GT's and 6SL7GT's. My power supply is a well-regulated (~10mV) CLC, where the C's are two 200uF sections of a big 500V LCR can cap, and the L is a Hammond 193K for 2.6H and 21 ohms. I also stuck a 470uF 400V Rubycon cap parellel to the first C. I get about 364V 1/2A out of this (I have a beefy AnTek 300VA toroid at 275-0-275). Chassis is a Hammond aluminum 16x8x3.

Anyway, the output transformer that should arrive sometime next week is from Edcor, 6600 primary impedance and a nice healthy 60W power tolerance, only has an 8-ohm tap, besides ground. Seeing as the NFB on this amp is connected to the 16-ohm tap, what should I do? What effect would connecting the NFB to the 8-ohm tap do? I'll post pics of my layout as all the stuff arrives.

EC8010 9th August 2006 09:07 PM

Hafler and Keroes published their "An Ultra-Linear Amplifier" paper in Audio Engineering November 1951, but the topology had already been patented by Alan Blumlein in BP 496,883 (1937).

sorenj07 9th August 2006 09:49 PM

well yes, but blumlein's patent was close to running out by the time our friends hafler and keroes got to work, and besides, he doesn't seem to have built any amplifiers that used it. i'll keep y'all posted on how things work out when i start building it.

SY 9th August 2006 09:51 PM

The NFB resistor value will set the gain. Most "classic" amps have gain that is much too high, so you might think about reducing that resistor's value. Now, you'll probably have to adjust the compensation cap value to make up for this, but with a new output transformer, you'd have needed to do that anyway.

Double-check that this schematic is accurate. The coupling RC time constants don't look right- they should be staggered at least 5:1, if not 10:1, for stability.

sorenj07 10th August 2006 01:06 AM

What I'd heard about the NFB was that with an 8-ohm tap, I should use about 8.5K instead of 12, and raise the capacitance to about 170pF. I have no clue what you mean by the coupling contants, though :D.

SY 10th August 2006 02:28 AM

Quote:

Originally posted by sorenj07
What I'd heard about the NFB was that with an 8-ohm tap, I should use about 8.5K instead of 12, and raise the capacitance to about 170pF. I have no clue what you mean by the coupling contants, though :D.
That will get you in the ballpark- in reality, the exact gain is not that important. That's why you have a volume control. The capacitor value will not scale- since you're using a different output transformer, you'll have to adjust that value on the bench.

Without getting into the Mysterious Bode and his Imaginary Poles, stability with feedback depends very much on how the open loop gain rolls off on each end, most especially the slope.

Now C2 and R6 (also C3 and R7) form a high pass filter (i.e., roll off low frequencies). The time constant is C2*R6, in this case about 0.1 s. Likewise, C4 and R13 form a time constant, C4*R13, about 0.067 s. That's a 1.33 ratio. If there are no other time constants around, that would be stable. A third rolloff nearby would cause things to get rather tipsy. But unfortunately there are a bunch of other LF rolloffs: the output transformer, the bypass cap on the input stage cathode, the power supply decoupling... you really want the RC time constants to be further apart.

The best way, IMO, to do this is to reduce C4 (and C5), to perhaps 0u047 or 0u033. This will restrict the bottom end, but the feedback will partially make up for this. And you've got no business using a 25 watt amp to drive a subwoofer anyway.;)

A perfectly acceptable alternative would be to bring C2(3) up by a factor of 2 or 3, then reducing C4(5) by a factor of 2 or 3. That might also require stiffening the supply a bit, but you ought to have a good solid supply regardless.

sorenj07 10th August 2006 03:44 AM

Hmm. First off, if I were to stick a volume control in there, what rating should it be? I have a smallish 6SN7-based SRPP linestage that I could use to adjust volume (it has 4 input selection too) but I wonder if it'd be too much gain with what I assume to be an integrated amp. I don't suppose I could could simplify things by cutting out one of the gain/phase split stages? Anyway probably the easiest solution for me would be to slap another .22uF cap alongside c2 and c3 as I have a couple left over. It _is_ all right that I'm using .22 and not .2, right? I couldn't really find any of those.

Thanks for the info on the filtering, however, as it's provided me with some sweet insight into how schematics work. They used to look like random lines and symbols a few months ago, and already I'm beginning to make sense of them :).

sorenj07 10th August 2006 04:42 AM

updated schematic:

http://img81.imageshack.us/img81/848...copyoj7.th.jpg

SY 10th August 2006 12:15 PM

A few things.

First, a general note- you might want to use one of Hafler's later designs as your starting point.

Specifics: the LF rolloffs from the RC coupling are still only spaced by a factor of 3. You really want them much further apart. Five times should be a minimum, ten times is best. For my first shot, I'd set the second rolloff at 20-30 Hz, the first one at 2-3 Hz. It's better to make the second rolloff the dominant one because of the way this phase splitter operates.

Also, you've got almost 700u as the input to the pi filter in the power supply. That will cause unneccessarily large ripple currents, and very likely cause the 4007s to pop. Something like 60u would be more appropriate.

You'll likely have to revisit the decoupling. The time constants are too short for comfort. Bigger decoupling caps would help; regulation would be even better.

Check out the later Hafler article for his adaption of the Williamson. It's a much better starting point.

BTW, my volume control comment was directed at your preamp.

corbato 10th August 2006 12:46 PM

Quote:

Originally posted by sorenj07
updated schematic:
Ouch...I would move that stand-by switch before the PSU caps.


All times are GMT. The time now is 07:51 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2