You can see from the equations in the previous post that there must be an optimum: equivalent at the input, the noise voltage goes down with increasing collector current, but the noise current goes up with increasing base current. As increasing or decreasing the collector current also increases or decreases the base current, there must be some value where the total noise is minimal.

Noise optimization of moving-magnet amplifiers is too complicated a subject to cover in a diyAudio post, but you can read my opinion about it here: "Noise and moving-magnet cartridges",

*Electronics World* October 2003, pages 38...43,

https://worldradiohistory.com/UK/Wireless-World/00s/Electronics-World-2003-10-S-OCR.pdf Mind you,

*Electronics World* drew one of the sections of the gain switch in the wrong state in figure 5 and I mixed up the terms spectral density and power spectral density. I later extended the calculations to other noise weightings. The 3852 Hz becomes 5179 Hz when you prefer ITU-R 468 weighting, see

*Linear Audio* volume 8, September 2014,

https://linearaudio.nl/grammophone-preamplifier-noise-calculations-3852-hz-rule-revisited (behind a paywall).

The 51.56 μA I calculated for a bipolar transistor with

*h*_{FE} = 600 and a 500 mH cartridge is for a single transistor, not for a differential pair. In a differential pair (a normal one, not one that is as asymmetrical as the one I used), both transistors contribute to the noise voltage, which therefore gets √2 times as large. You have to increase the current per transistor by a factor of √2 to correct for this, so the optimal collector current becomes about 72.92 μA per side, total tail current 146.1 μA. That's also the optimum sum of the emitter currents through all the transistors when you use multiple differential pairs in parallel, or a differential pair made with multiple transistors in parallel, provided

*h*_{FE} stays 600 and the assumed cartridge impedance is correct.

That's the optimum, another question is what happens when you are not precisely in the optimum. Noise increases, but quite slowly.

Suppose that, for simplicity, you want to realize the 47 kΩ input resistance by simply connecting a 47 kΩ resistor across the input. That resistor will then contribute about 0.5869 pA/√Hz of noise current when its temperature is 20 °C. If base shot noise is allowed to contribute as much as the resistor, that requirement is met as long as the base current of the input transistor (or of all the input transistors together, if you use several of them) stays below 1.075 μA. With

*h*_{FE} = 600, that corresponds to 645 μA per side, tail current 1.292 mA.

Record surface noise is not included yet. If you only care about the noise with a record playing, and only look at the integrated noise and don't care about changes in the sound of the noise, then you can go much further from the optimum.