Ping salas re: LED baising

[rdf]

Hey salas, I got to thinking about alternatives to the LED biasing discussed in the 6DJ8 thread. It might have been the last ETF that held a shoot-out between various methods but the fave from memory was a simple unbypassed resistor driven with extra current. For example, my headphone amp's first tube is biased ~2 VDC on the grid with around 1.5 ma plate current. This would normally require a ~1300 ohm cathode resistor and cap bypass. The second stage however is a cathode follower with CCS tail running about 40 ma. By returning that CCS through the first stage cathode resistor the required resistance drops to about 50 ohms for the same 2 VDC grid voltage. A resistor that small provides almost no degeneration in this application so no bypass is required. To my complete surprise it sounds terrific and for you provides the option of a feedback return path.
I haven't done a spectral analysis yet but if it's typical of what I've seen in the past from very low level feedback most of the impact will be at the highest harmonic orders with 2nd and 3rd unaffected. Worth a look.

[Salas]

Thanks rdf. Problem is, that I use just 6mA of a 12AT7 Shmidt drive stage tail CCS and this gets me down to a no good 750R for 4.5V needed if returned from there. But this is a very good hint you gave for developing other stuff in the future.

[SY]

It might have been the last ETF that held a shoot-out between various methods but the fave from memory was a simple unbypassed resistor driven with extra current.

That was probably three years ago. I wasn't at that one, but I've been to other ETF shootouts and, to put it charitably, I would not weight the results very heavily. One demo a couple years ago involved different wires; the crowd favorite (by voting) was a strand of solder. I was sitting next to John Broskie for that "test," and he observed throughout both the wire tests and the preamp shootouts that the voting seemed oddly sociological (in a better group test, voting is silent, not "show of hands").

[rdf]

Thanks SY. The concept was more interesting than the listening test. I usually consider the latter, unless performed by a Nortel, are almost universally bunkum.

It's been my consistent finding that tiny amounts of local degeneration can have a 10-15dB impact on high harmonic components and yet under 1dB on the 2nd or 3rd. I have no explanation for it but the 'biased resistor' technique promised a way to get the grid voltage required and have small local degenration with a more linear cathode load than possible with LEDs. Since the following stage of this circuit was already burning up ~40ma in a CCS CF tail it was a natural combo.

And it sounds great. Measurements may tell a different story, maybe tonight. One minor downside, avoiding warm resisitors means high wattage ratings or many in parallel. Hey salas, don't forget the source of electrons isn't important. I had them handy in a CCS tail but a CCS or resistor from B+ is just as valid if you don't mind the extra dissipation, or from a separate LV supply if you do.

[SY]

That's an interesting observation since Baxandall predicted that higher order distortion would go UP with small amounts of degeneration. Clearly, there's a gap between models and reality.

[rdf]

A parallel pair of green LEDs gives within 10% the same Vgk with a different residual cathode load. A capacitor bypass for 50 ohms is another test. I'll try to give it a look tonite.

[Salas]

Originally posted by rdf
Hey salas, don't forget the source of electrons isn't important. I had them handy in a CCS tail but a CCS or resistor from B+ is just as valid if you don't mind the extra dissipation, or from a separate LV supply if you do.

I would need to bleed 96mA trough a 47R cathode res for my 4.5V bias. My output stage consumes 220mA stereo. So asking my PSU for about 200mA stereo extra bleeding will bring it to its knees! Plus the resistors will consume circa 40W each, and must be metal cast and finned, and put externally on chassis. Brother of Zen stuff. Anti eco. Even G.B. cares for the polar bear these days.

[Sheldon]

If the extra current doesn't come from a signal modulated source, then isn't it just a form of fixed bias?

Sheldon

[rdf]

Essentially, but one without capacitors in the direct signal path and without the potential (if trivial) non-linearity of a LED. In this particular application where the following load is a CF grid the dynamic plate current is so miniscule the corresponding cathode voltage swing will be almost nonexistent.