> a 5534 op amp preamp noise is minimal when a 150:5K transformer is used.
The optimum audio-band noise impedance for 5534 is I think 6K, and broad enough so that 2K or 15K isn't much different. 5K is a nice standard number.
> With tubes a 30 to 50K secondary is typical.
In the audio band (RF is different), tube current noise is nearly zero, so optimum noise impedance is nearly infinity.(*) However a near-infinity impedance transformer winding will have near-zero treble response. Over 100K is done in some PA amps, 20K to 50K is more reasonable for hi-fi response.
(*) It isn't always necessary to aim for the "optimum" noise impedance. When current noise is small, we merely need to step-up the source votlage noise to be "much greater than" the amplifier voltage noise. We can just make them equal and get a 3dB noise figure which is plenty good enough for rock-n-roll, better than most studios' background acoustic noise. If amplifier noise is half of source noise, the noise figure is 1dB, and any further "improvement" is inaudible.
The noise of a simplified tube is like the total cathode resistance, including 1/transconductance (1/Gm). For a typical preamp tube the Gm is about 1,000 microMhos, so 1/Gm is about 1,000 ohms. The random thermal noise of a 1,000 ohm resistor across the audio band at room temperature is about 0.5 microvolts. But this "resistor" lives at the cathode surface, where it is MUCH hotter than room temperature. Noise voltage will tend to be 3 times higher, or 1.5 microvolts. The random thermal noise of a 200 ohm microphone is about 0.2 microvolts. We want a step-up to make this about twice the tube's noise, or about 3 microvolts at the tube. 0.2/3= 1:15 step-up ratio. But this implies (200ohms)*(15^2)= 45,000 ohm secondary. If we just assume 100pFd capacitance (winding and grid), then we will be down 3dB at 35KHz, -1dB at 17KHz, a marginal performance. WHen we actually wind it, the leakage inductance will probably ring at the top of the audio band. Tricks to reduce leakage inductance also increase capacitance. 45K windings are tough audio problems. We can get better response specs with maybe 1:10 turns ratio. Noise figure is 2dB, not too shabby. Secondary impedance is 20K, which is a little easier than 50K. And the lower number of turns allows use of fatter less-breakable wire, which can be a major cost issue. Don't say "money is no object": it always is a factor. The bucks saved on a less expensive winding can be put into a fatter power supply, allowing use of a fatter tube at higher current, more Gm, lower noise voltage.
Oh: and if you use condenser mikes with built-in amplifiers: none of the above noise-calcs should matter. While a dynamic mike's noise may be 0.2 microvolt, many condenser mikes spit out 1 microvolt of noise. They can get away with this because their output level is 5 times higher than a dynamic. But that means the "mike preamp" in your board or rack is NOT the input of the system, is NOT the noise-critical point; the amp inside the mike is the critical point. Your board/rack is now a "line receiver". Working at lower levels than a standard line level, but far above the levels of a dynamic mike. A 10dB noise figure in the board/rack "mike preamp" will not raise the system noise.
> It's really hard to tell the difference with out actually setting them up on a piano or a really good singer and A/B them side by side.
Theory is not enough. There are many subtle poorly understood details in transformer behavior. The only way you can judge the relative merits of two trannies that don't suck is to try them on a variety of real sounds.