Hi Guys. I'm hoping someone can help me to make sense with a problem I am having. I'm looking into simulating moving coil head amp models and they are all showing very high noise levels. That wouldn't be odd but the level is always the same for different schematics and also for different values for components in each sim. The level is always just below 30nV/√Hz.

I have attached an .asc file for the known Huygens schematic which shows this behaviour for me here. AI (sic) tells me that this is a common problem which can often be spotted by the noise figure being flat across the frequency range. Mine is flat above around 200Hz, however I can't make head nor tail of what it tells me is the fix! It talks about LTspice defaulting to an internal opamp mode but there are no opamps in this circuit.

With that last point in mind, is it possible that this is simply because the .MODEL entered as the statement for the JFE150 contains no noise data? There is a parameter KF and AF which I think may be noise related. I should say I'm an experienced LTspice user but by no means an expert on the whats and hows behind the scene.

Thanks for any advice you can offer.

 

I've done a bit more work on this and I think I may have an answer.

I found that it definitely isn't anything in the circuit setup. I took another simpler known head amp circuit, (the Gamow design from the Hifisonix MC Head Amp collection .pdf), and simulated that exactly as they show in the document. The document says noise should be 844/866 pV/rt(Hz). With exactly the same schematic and .noise analysis definition that is listed in the document I see 10nV dropping to 4nV at 100kHz. This does not change if I declare every resistor noiseless, (there are only basically five). If I change the original BF862 JFETs for types like the J201, my JFE150, and LSK389C I do see the curve change shape but it remains very high for all those types.

However, looking at .noise in the LTspice Wiki I found a reference to V(onoise) and V(inoise). V(inoise) is the "input-referred noise voltage density". If I plot V(inoise) rather than my specified output node the noise plot drops down to a sensible level, it still doesn't seem to match the stated value but, as it is trying to put a value on a curve, that is flexible anyway.

Maybe that is something like an answer to my problem.

 

The voltage noise densities in the hifisonix paper are referred to the input and entail the noise generated by the source resistance Rs and that of the amplifier. I tried the Gamow design with 3 parallel 2SK3557-6 at a total drain current of ~20mA. 2SK2394-6 are also very low noise and a good alternative. Flicker noise (AF, KF parameter) is not modelled but the white noise of ~715pV/sqrt(Hz) (@ Rs = 10 Ohm) at and above 1kHz looks plausible.