Ping Sy: LED for Cathode Bias

[SY]

Let's look at the numbers. 70H plate choke? At 10 Hz, the inductive reactance is 4k4. For the 20mA you chose as plate current, plate resistance is about 2k for a grounded cathode. Mu is about 75. Gain = (4.4/6.4)*75 = 52. With a 5R resistance in the cathode circuit (typical for surplus LED, higher than HLMP), the plate resistance is 2k + (76*5) = 2380. So gain is then (4400/6780)*75 = 48.6. 20 log (48.6/52) = -0.6dB difference at 10 Hz between an LED and the cathode tied directly to ground. If you're seeing more difference than that, you've got something else going on besides the LED.

Let's look at this a different way: your original circuit had a 470u bypass cap. At 10 Hz, the cap has a reactance of 33 ohms. That's in parallel with a 100 ohm resistor and the looking-in impedance at the cathode (4k4/75 = 59R) for a total cathode impedance of 17.5 ohms, a higher number than the LEDs. Hmmmm.

[audiodesign]

Here the measurements in the same conditions.

[SY]

Like I said, it might be worthwhile to figure out why you see this; it may be enlightening (bad pun, sorry). You've got something odd going on and getting to the bottom of it will likely improve the design and give all of us a better predictive tool. Many of us have used LEDs for cathode biasing with great success and no frequency response or distortion anomalies.

It's also possible that you picked some really odd and poor LEDs- a measurement of AC impedance is easy (I give some details in the Red Light District article on my website) to confirm or refute this possibility. One fun thing to try is to put a series resistance in the cathode to simulate the higher impedance of the RC bias and see what happens to the frequency response. The 1dB or so midband gain difference is suggestive. Could it be that you used LEDs with a built-in series resistor?

[audiodesign]

I have tested many new led coming from different stock, the green led is better than red and blue in my case.

Probably the led ac impedance is too high when used with high gain vaccuum tubes like the D3a, 6C45 and 5842.

[SY]

Clearly that's not true given the relative impedance of the RC circuit and the basic circuit analysis.

[Gary P]

In looking at the measured frequency responses it looks like you are getting peaking from the 470uf cathode cap.

The smooth LF roll off response with the LED shows that you don't have a resonance between the power supply final cap and the interstage transformer. When you swap from the LED to the RC cathode circuit the frequency response actually rises a bit before rolling off, and the roll off does not look like a simple single pole response. To verify the theory that the RC cathode circuit is causing a LF resonance you can move the resonance around by changing the value of the cathode capacitor.

[SY]

That makes perfect sense. Thanks, Gary.

[dave slagle]

The sharp corner of the roloff and 12dB slope sure does look like resonance as Gary suggested. My problem is that assuming a 2500 ohm Rp and 70 hy's the rolloff corner with the LED is still off by an octave. This suggests either a 5K Rp or only 35hy's of inductance (or possibly a combination of higher than expected Rp and lower than expected inductance.)

what was the signal level of the tests?

dave