MJ Statistical Regulator PCBs

[IslandPink]

I should really ask, before too long, what this thing does - especially as it's named after me !

[jack]

Morgan Jones did a comparison of noise in various regulator designs, specificly for screens.

He ended up with a detailed analysis of noise performance in true zeners and determined that long chains of true zeners have excellent performance, orders of magnitude better than gas tubes, as the distribution spectrum of their noise (white) meant that a lot of it cancels out as the strings get longer.

He produced experimental data to reinforce the theory. See "Valve Amplifiers, 4th ed." Google will find the relevant pages for you if you don't already have the book.

Basically, an ultra-low-noise low current high voltage discrete regulator/reference.

Added bonus with these PCBs is that you can leave off the cap and zeners and just use it as a cascode current source as well...

[Nick]

meant that a lot of it cancels out

Adding together noise sources does "cancel out" as such, its can't, its uncorrelated, its just increases less than correlated sources.

So 6.3v Dc + 10mv noise (for some frequency range) times two will be 12.6v DC and 14.14mv noise, so the relative noise level will be less WRT the DC

[jack]

REMOVE!

[jack]

Indeed. Poor choice of words by me. The noise increases only as a function of sqrt (n) rather than proportional to n, i.e. with 4 times the number of zeners, you get only twice the noise.

This is important - 5V6 zeners are VERY quiet (about 300nV RMS per diode) - experimental data shows that a string of 16 x BXZ55C5V6 zeners (84V) driven at 10mA produced noise at -103dBu (5.5uV RMS or 0.065uV/V) and a string of 35 (195V) produced -99dBu (8.7uV RMS or 0.045uV/V).

Compare these noise figures with an 0A2 at -69dBu (250uV RMS or 1.67uV/V) and the best of the gas tube tested, a set of 85A2s averaging at -75dBu (138uV RMS or 1.62uV/V).

The improvement in noise performance over gas tubes is dramatic - volt-for-volt, zeners are about 36 times quieter. In addition, upping the zener current to 20 to 25mA further improves the noise figures by about another 6dBu - zener noise falls by a function of 1/root(DC current) with an asymptote at about the 25mA mark, hence the requirement to get the zener current to that figure.

[Mike H]

Daft question time...



If noise is the issue why can't the Zener chain or whichever be paralleled with a cap?

[jack]

Daft question time...



If noise is the issue why can't the Zener chain or whichever be paralleled with a cap?

It is. A 10uF, 400V polypropylene. I'm only doing this out of academic interest - it may, or indeed may not, have an audible effect on the quality of a preamp.

The excess boards from here I'll sell on eBay for a stupid amount with an accompanying load of guff about synchronising quark spin, solving Schrödinger's wave equation for quantum tunnelling and unidirectional neutrinos or some such cr*p...

You have to control the zener current too, hence the CCS.

Latest schematic, notes, and board layout. Since rev 01 I've changed the trimpot, terminal strip and FETs, plus messing with the layout to up the isolation to 400V or so. Probably going to fab this afternoon:

[Mike H]

It is. A 10uF, 400V polypropylene.

Oh yes course it is. I was distracted by the jumpers...

[IslandPink]

Oh yes course it is. I was distracted by the jumpers...

You sound like Giles Brandreth !

Thanks for the info - nice idea . Will await the results with interest .

[Mike H]

Well ta!

[simon]

The excess boards from here I'll sell on eBay for a stupid amount with an accompanying load of guff about synchronising quark spin, solving Schrödinger's wave equation for quantum tunnelling and unidirectional neutrinos or some such cr*p...

Cool! Then you can become the first member to feature in the quackery thread .

Put me down for a few boards, sure I'll find a use.

[cressy]

Put me down for a board too pleasenick, i have a phono stage to shut up

[Cressy Snr]

You can put me down for a few too Nick.

[Ray P]

Interested too Nick.

Ray

[jack]

Went to fab yesterday - ordered only 20 boards as this is a prototype.