Thanks for looking into this for us!@Gnobuddy - some attachments were indeed removed during the thread maintenance, so please re-post them as required.
My attachments were not too important. The one I'd like to see back is the one Merlin posted, of his experimental measurements of THD in a common cathode triode amplifier stage, with and without LED bias. But that's up to him, as it's his copyrighted work.
-Gnobuddy
You can see it here, in context too!
Designing High-Fidelity Valve Preamps - Merlin Blencowe - Google Books
There is no "knee" at any practical current.... that's an artifact of the plotting.
It is all an exponential "Knee". 
An externally hosted image should be here but it was not working when we last tested it.
Please see post #146...this exact question has been asked, and answered, several times already in this thread!
-Gnobuddy
Last edited:
I noticed across a couple HLMP6000 Red datasheets that the variation in devices at 1 mA is only 1.6v to 1.8v, with 1.7 as typical. This is quite a bit less variation that is listed at 10 mA- I referenced it earlier in this thread, something like 1.4 to 2.2v with 1.7 as typical. So it would appear these vary quite a bit less at 1 mA- each device must vary less at 1 mA than at 10mA for those to be figures. These are from two different colors of red, they don't list 1mA and 10mA on the same datasheet, but there is a graph that shows both (typical) and they are very close indeed in the .5-5mA region.
Seems to be a more predictable current range to design for than with additional current to bring the slope resistance way down.
Seems to be a more predictable current range to design for than with additional current to bring the slope resistance way down.
Some reminder:
The Red Light District - another PP EL84 amp
12 years ago, when we discussed some really serious things, without mumbo-jumbo arguments
The Red Light District - another PP EL84 amp
12 years ago, when we discussed some really serious things, without mumbo-jumbo arguments
I wonder if LED noise follows the same kind of curve as zener noise, where noise is proportional to the inverse square of current. Unfortunately when the ubiquitous Mr. Jones was measuring his beloved HLMP6000's, he was measuring dozens of them soldered together and didn't get a clear answer. If the noise does follow the same kind of curve, the 1 ma and under applications might well benefit from additional current if noise is an issue and it's more critical than distortion, say in a very input stage. The noise could be many times what it would be with a few mA. Of course you would need a very clean low voltage source to feed to the LED.
As for bypassing LEDs with capacitors for noise purposes, those would need to be some big capacitors where slope resistance might be just a few dozen ohms or less.
As for bypassing LEDs with capacitors for noise purposes, those would need to be some big capacitors where slope resistance might be just a few dozen ohms or less.
The cathode of a valve is at around 1200 Kelvin. Meantime, the LED is at room temperature, around 300 Kelvin, or one-third the absolute temperature.
Because Johnson and shot noise are proportional to the square root of absolute temperature, valves are inherently about twice as noisy as semiconductor devices like LEDs. And that's if the valve was perfect.
In reality, valves are not perfect. As I understand it, flicker noise in valves far exceeds shot noise at audio frequencies. So the valve isn't just twice as noisy as a perfect room-temperature semiconductor, it's far more noisy than that.
The end result is that the noise from the valve is almost certainly going to vastly overpower the noise from the LED, to the point where we can completely ignore the LED's noise contribution.
Designing small-signal Hi-Fi audio input stages with valves is rather like picking up ball-bearings using chopsticks; there is no logical reason to do so, only the attraction of trying to conquer a challenge.
-Gnobuddy
Possibly, but then it is more economical to bypass it with a capacitor than to burn up extra current, no?