Using diode for biasing

[dosilegecko]

What are the advantages of using, say, a red LED for cathode biasing for a 12AX7.

What properties would one associate with using diodes instead of the typical resistors?

[mach1]

A 12AX7 normally biased will not pass enough current to run a normal LED in its linear region. You will have to supply extra current (approx 6 - 8 mA) from a stable source.

Advantages: you lose a cap and a resistor, and some say it sounds better.
Disadvantages: you require an additional source of stable current, and some say it sounds worse.

[dsavitsk]

Originally Posted by dosilegecko

What are the advantages of using, say, a red LED for cathode biasing for a 12AX7.

Linearity, lower cost.

Originally Posted by dosilegecko

What properties would one associate with using diodes instead of the typical resistors?

Good sound

Originally Posted by mach1

A 12AX7 normally biased will not pass enough current to run a normal LED in its linear region. You will have to supply extra current (approx 6 - 8 mA) from a stable source.

Yes. A CCS should work very well here.

[TheGimp]

There are some High Efficency LEDs that will work at much lower current levels than previously specified.

For instance the LGT67K ("Hyper Bright Low Current LED", Yellow-green, Osram) is specified for operation at 2mA and data is available down to 1mA.

The LGQ971 is specified at 20ma, but if you read the fine print that is pulsed for 25ms. Typical drive is 2mA (or less) continuous. These LEDs begin emitting light as low as 50uA in the linear regeon of the curve starting as low as 1.1V forward bias.

Similarly the LST670 (Super-Red, also Osram) is specified at 10mA, Luminousity data data is provided down to 1mA, and If/VF data down to 100uA.

The 20mA spec for the LGT67K puts them in the knee of the curve above the linear regeon.

I stick with operation in the linear portion of the curve at much lower current levels.

I'm using the LGT67K LED in my "FireFly" SET (Single Ended Tetrode amplifier, Schematic soon to be posted) I'm designing for my wife to use as a replacement for her present self powered speakers on her computer.

Unfortunatly many new inovative components are available in surface Mount only, and rely on heat spreaders on the PCB to handle thermal dissipation. The LGQ971 only dissipates 4.3mW at 2.9mA (6NP1 bias for the input stage) so it is not an issue with this device.

[bocka]

Originally Posted by mach1

A 12AX7 normally biased will not pass enough current to run a normal LED in its linear region.

AFAIK a diode (whatever it is) is always a non-linear element. Current through the LED is

i = is * [exp (U / Ut) - 1], where Ut = kT / q

It should not make any difference what kind of LED (low-current or standard) you are using, assuming all other parameters are equal.

[Arnulf]

I believe they meant "constant" (as in: constant Vf) region, which appears as an almost straight vertical line at V = Vf. Granted, it's hardly linear, and besides, what use would it have to replace the most linear element out there (the resistor) with another ?

Back on topic: I use the cheapest LEDs I can find in local electronics stores, I specifically ask for "cheapest, low luminosity" ones, red for 1.75-1.8V and green for 2.0-.21V. These are usually the dull-glowing type with diffused bubble and work well from few mA upwards.

High brightness types usually have much higher Vf and require more current to reach the region where Vf is relatively constant.

[bocka]

Originally Posted by Arnulf

I believe they meant "constant" (as in: constant Vf) region, which appears as an almost straight vertical line at V = Vf.

Yes, I know what you mean. Unfortunately there is no "linear" region in the diode characteristic. Because of the linear scale of the i-v curve it looks only like linear. When you are using a logarithmic scale there is no bend regardless of the current or voltage.

When biasing a LED higher than 5 to 10ma the internal resistance has to be taken into account. In this region the LED becomes more and more linear as the internal resistance linearizes the current dependant conductance of the LED. Maybe this is the intention for the additional CCS.

Originally Posted by Arnulf

Granted, it's hardly linear, and besides, what use would it have to replace the most linear element out there (the resistor) with another ?

I cannot see the reason, too.

[kevinkr]

Originally Posted by bocka

Yes, I know what you mean. Unfortunately there is no "linear" region in the diode characteristic. Because of the linear scale of the i-v curve it looks only like linear. When you are using a logarithmic scale there is no bend regardless of the current or voltage.

When biasing a LED higher than 5 to 10ma the internal resistance has to be taken into account. In this region the LED becomes more and more linear as the internal resistance linearizes the current dependant conductance of the LED. Maybe this is the intention for the additional CCS.


I cannot see the reason, too.

The LED actually replaces a cathode bias resistor and its bypass capacitor. Obviously there is no -3dB corner in the low end response due to RC rolloff. Overall the LED generally behaves more linearly than most electrolytic caps in this application which is why the use of LEDS for this application has become quite widespread.

I was sceptical, but have used LEDS in applications where very large cathode bypass capacitors (1000uF) would be required at very low polarizing voltages (say 1.5 - 2.5V) - something that I have found not to work well. (Think of 7788, D3A, 5842, C3G running at high currents resulting in very small cathode bias resistors and very large electrolytic caps)

[SY]

Originally Posted by mach1

A 12AX7 normally biased will not pass enough current to run a normal LED in its linear region. You will have to supply extra current (approx 6 - 8 mA) from a stable source.

If you use a CCS plate load (an excellent idea for a 12AX7), the LED will work fine as is. See my recent "Guess the Tube" post.

[mach1]

Sy,

very interesting that your usual 'resistor trick' made no difference at that low a current. I don't discount your result given that it was based on careful measurement, but rather counter intuitive based on log-lin I-V curves (linearity definitely appears better and dynamic impedance lower at higher currents for the ones I can remember). I wonder if the result may have been slightly different for a lower resistive plate load?

Were you using HLMP6000 ?