# Quiet Diodes at Audio Frequencies

I am using diodes to bias tubes for the input stage of an audio circuit. They will hold the cathode off ground by their forward voltage (Vf). Current will probably fall between 1 and 2 mA, and the voltage will be set by the diode's Vf. I am aiming for .8V-1.6V total, and I can have diodes in series to obtain the right bias voltage (thought multiple junctions tend to raise the noise level). Peak-to-peak AC will be in the 250mV range.

I would like to know if anyone is aware of low noise, low distortion diodes that will have minimal effect on the audio signal passing through them. I've experimented with several types, Schottkeys, 4007, 914, etc. The Schottkey rectifier diodes and the 4007s are best so far. Of course, at 1mA, their rated Vf is quite different from the specs that are usually published. Only the most complete spec sheets show Vf below 10mA or so. The Cree 6010s (rectifiers) came in at .8V and the 4007s came in at .5V.

I realize that most people don't use diodes in this application, so information has been hard to come by.

imagine a triode tube, the anode is fed 155V through a 75K plate resistor, the grid is fed from the output of a microphone capsule and has 200Megs to ground. The cathode has two 4007 diodes in series to ground. Signal is taken from the anode, through a capacitor to the an output transformer. This results in 1V of bias at at the cathode, 75V on the anode, and 1.5mA of current flowing through the tube. The diodes each provide .5V of the total bias voltage and are, themselves, biased by the 1.5mA of current. The diodes never see negative voltage

• Worth re-examining your reason for wanting to avoid traditional self-biasing, with a resistor in the cathode circuit. – Brian Drummond Jan 24 '17 at 0:33
• I normally refuse to draw schematics for the people that ask a question – I feel it's the job of whoever asks to make their point clear, and that includes a schematic. I made an exception here – please check whether the schematic is what you've meant. If it is not, you must add your own – Marcus Müller Jan 25 '17 at 9:30
• @Marcus Müller: I guess the 200MOhm at input is not a resistor as component but the inherently given resistance between gate and GND. I've never seen a 200MOhm resistor. – Curd Jan 25 '17 at 9:49
• @Curd was my guess, too. I see 200 MΩ all over the place – for example, this 10cm piece of relatively dry wood here… anyway, the point is to show that the mic voltage floats comfortably around ground – Marcus Müller Jan 25 '17 at 9:51
• @Terry Setter: I still don't understand why you want to apply the bias voltage. Can you explain? – Curd Jan 25 '17 at 9:58

Diodes, pretty much by definition, are non-linear components, which means that they always distort your signal. The trick is to operate them within a range small enough to make the nonlinearity so small that you cannot measure it.

The point of why people use diodes in biasing is that if you pick a saturated operating point, the voltage across the diode will not depend much on the current passing through it. That's pretty much the opposite of what you want – you'd fix the gate at $V_{cat} + V_F$, and that's it. All signal that you apply should basically be "swallowed" by the diode.

The diode types you mention are old to positively ancient. The 400x series are rectifier diodes, ie. optimized to be nonlinear. The 914 is the predecessor of the already practically obsolete 1N4148 series – seriously, where would one even dig up such old components?

Old diode types are not preferable if you want low noise operation.

• Thanks for your input, Marcus. I was surprised to see that the rectifier diodes outperformed the signal diodes, and you are quite right, I need to try MODERN signal diodes (and maybe rectifier diodes as well). The thicker coatings of the higher current models yield higher Vf, and that has been helpful. Any recommendations as to maker/number to try? – Terry Setter Jan 23 '17 at 23:14
• as I tried to explain, it's probably a bad idea to use diodes for exactly this, since their purpose is the opposite of what you're doing, unless you're using a more advanced biasing than your question let's assume – you should probably add a schematic that explains what you're doing. If I knew what your schematic looks like, I could make more detailed recommendations. – Marcus Müller Jan 23 '17 at 23:20
• and you're mistaken: The forward voltage of a diode has nothing to do with the thickness of a coating - it's a physical function of the doting and the involved semiconductors. – Marcus Müller Jan 23 '17 at 23:33
• Could you say more about what makes "modern" diodes less noisy than "obsolete" ones like the 4148? I haven't heard of this. – Brian Drummond Jan 24 '17 at 0:35
• @BrianDrummond very sorry, I must have forgotten to answer; so: forward-biased diodes exhibit two kinds of noise; a) Johnsson-Nyquist Thermal noise, as everything else; power of that is proportional to $\Re\{Z\}$ of the impedance at working point, b) 1/f noise of carriers filling "traps", ie. contaminations, usually at the surface of the semiconductor, and then randomly relax after a time (hence the frequency dependency). Contaminiations were successfully reduced quite a bit since JEDEC standardiced 1Nxxxx. – Marcus Müller Jan 25 '17 at 9:41