Wednesday, June 13, 2012
Ju-Jutsu: Ultimate Monoblock (Pt 3) First Stage
What is going on here?
First of all, I chose balanced stages all the way through the amp.
We have to get to balanced anyway, for a Push-Pull at the output stage,
so at some point I'd need an interstage transformer or a splitter, with all those problems:
added stages, another layer of distortion etc.
Well I learned a lesson working with professional studio gear, which is balanced all the way from the microphone.
It works best, and gives the highest noise immunity and lowest distortion.
All the advantages of the Push-Pull (harmonic cancellation, Powersupply noise cancellation)
are now transferred to the whole amp.
If the Mu-Follower wasn't clean enough, its now bullet-proof.
What you are looking at is two Mu-followers, one for each side of a balanced input.
One is drawn upside down for symmetry and beauty.
The input transformer is a Jensen.
Simply put they make the best input transformers you can buy.
We use them in everything in a studio environment, for isolation, balance, and frequency-response, noise immunity, and low distortion.
They are a cost-effective solution and replace an interstage transformer, which,
to get the same specs would cost a fortune, and they don't make them like this anyway.
Jensen can make near-perfect transformers because of the signal amplitude and size.
Nuff said. No one in their right mind would prefer an interstage to an input transformer.
That leaves the resistor values.
The input stage here could be any tube, but lets pick a 12AX7A
just to keep it easy to get good quality cheaply available parts.
Well, we'd like about 250 volts across each tube.
We'd like to bias them according to my little method.
But people have designed plenty of 12AX7 Mu-Followers,
and you can use tested circuits that are known to deliver high fidelity.
Some things never change.
The input grid resistor for each triode should be 500k to 1Meg R.
This provides an appropriate input impedance for the transformer,
and a path to ground to (self-)bias the grid.
The cathode resistor should be about 1k R to give a volt or two of self-bias.
It's positive voltage at the top (due to the idle current) makes the grid relatively negative.
Input tubes should pretty much always be self-biased unless you have a good reason not to.
If you are using a 6922 or some other more sensitive tube, don't forget a grid-stopper (shown).
Otherwise you risk hidden parasitic oscillation and poor performance.
You'll notice I have a three-resistor network to ground for the two bottom tubes (one upside-down).
That's Hafler's minimum IM distortion trick, discussed in the literature.
You can just as easily use a couple of LED diodes to ground on the bottom tubes (NOT the top ones!!!),
as some people are doing these days (they weren't available in Hafler's day).
The top tubes are set up similarly, but the extra resistor on the cathode is another parasitics-stopper (horizontal, optional).
Again a 500K resistor (or 1 Meg R) is a grid-leak or input resistor (and a horizontal grid stopper if needed: try 1k).
The top resistor is not really a load resistor, its again a 'parasitics stopper' resistor. (500 R would do, or even skip it).
The top tube is operating as a cathode follower.
The input cap for the top tubes are .22 uF, but you could use a .1 uF, its not critical.
What's left? The input is grounded to the chassis- via a star-ground at the bottom of the input tubes.
This is for shielding, but it can be skipped, and you could just use an XLR input for a balanced line input.
You could even eliminate the input transformer entirely if you have balanced lines.
Any questions? No, then we can move on to the next (last) stage.