Updated: 04-Sep-03
There are many philosophies, but this is the one I developed over time…

Signals and soundstage:

I believe (because I cannot proof it... and never find some one who did...) that an open pin pointed focused soundstage with lots of details and depth of image can only be achieved if the original signal is not modified in a timely matter. Amplitude or frequency response are not crucial for this. What is crucial, is the time relation of the left and right signal (we have 2 ears)   AND the time relation within the signal of the different components. This can only be achieved if the signal is treated linearly and no other signals are added, like power supply noise or feedback (!) signals. In fact, we will never be able to make sure that is no time delay at all. But, it seems to be proven, that our ears don't mind if the time delay is constant or if the frequency dependency of the time delay is CONSTANT (like in 6dB filters) and non feedback systems

With this in mind I have developed my "design philosophy" and refined it year after year

Amplifying Signals:

My whole design philosophy of amplifying signals (that is what we are talking here, so no mechanics or digital design...) is based on the substitution of amplifying elements into Current sources and passive elements in their reactive representations. Pffffffff, academically stuff, yeagh... Please bare with me for a while. I will try to make things clear by the following 3 little basic schematics:

fig 1: Basic amplifying circuit with Tube

Here you will see the basic amplifying circuit. Basically all amplifiers can be reduced to a combination of these basic circuitry's

If we now substitute all elements by their basic electrical representation, the circuit looks as follows. We assume here that Rk is much higher than the reactance of C1 and that the battery has DC resistance of ZERO

fig 2: Basic representation of the Amplifier in fig 1

It is now easy to understand how it basically works. The changing input voltage at the input (gate with a tube) is inducing a current (here anode current) which equals (S x Ug).     S is here the transconductance. Normally mentioned in mA/Volt or micromohs. Actually this current should result in a voltage across the Load Resistor. BUT, as you can see the signal is not fed straight to he load..... The signal has to pass the C1, C2, Ra in series with C3 and Rs including the Power Supply with Humm and Noise voltage. This does not look good at all you would say, unbelievable that in such a simple stage so many elements play a role. Yes they do and they can really screw up the signal.

So lets define per element what is important and what we should do to MINIMIZE the effect:

• The Ia curentsource must be as linear as possible (like old triodes.....???) Transistors for example are REAL BAD idea here
• C1 is normally the decoupling capacitor for the Cathode. 2 options here: do not use at all or use High quality. If an Elco is needed (high value...) than BYPASS with a high end capacitor like film of paper in oil or whatever. For all frequencies, we want de ZC1 to be much / much lower than the Rload (never thought of eh?). Normally this is no problem.
• C2 is straight in the signal path and should be of highest quality. Simple, no discussion
• Rload should be preferably high compared to Rp and Ra, thus minimally effecting the amplifying process
• Ra also should be preferably high compared to Zc3 and Rp, thus minimizing effects from the power supply and Plate resistance. In a MU Stage, this Ra is "bootstrapped" to very high virtual values for example; reducing its influence, so the Triode works more linear...
• C3 is also IN THE SIGNAL PATH (!!) therefore must it be large and again of high quality. Elco's parallel with film or oil capacitors can be a good choice, but these days, we see many High Quality Elkos from Black Gate for example. This is to prefer over a solution where Normal Elkos are by passed by film capacitors.  ZC3 must be as low as possible compared with Ra and Rs. Only than will the effect of the power supply be minimized. This can b also done by CASCADING multiple RsC3 sections or use high quality Chokes ... (!!)
• Us is the ripple/hummmm/noise coming from the power supply. It is probably an open door, that BEFORE the supply voltage is fed to the RsC3 secretion, the Us components should already be as low as possible. No surprise that if you use batteries you get a much cleaner sound....

So this results in the following Guidelines:

• Keep the power supply basically as regulated as possible or use PI-Stage filters of high quality
• Keep the power supply as far as possible (electrically) "away" from the Ra by applying multiple filter sections
• Use HIGH Quality Capacitors and Elkos; alternatively parallel them with Film C's if budget is limited
• Use linear amplifying elements like triodes and use them in a linear mode; so enough Ia current or stages like the MU-Stage
• Avoid as much as you can loading the voltage amplifying element (like in the MU stage .... !!)
• Ground all decoupling capacitors as close as possible to the Cathode and Rp for short signal path

For this reason, I do not care too much about the kind of power supply, solid state or with rectifying tubes. AS LONG as you keep them away!!! I need to refine this is a bit..... As you cannot always keep it really away, you do should take care in sense of good transformers, Using Tube rectifiers or at least fast diodes (please no 1N4007...) and a clean mains supply. The differences are not HUGE, but it is clearly audible....

If you do this, all the capacitors will approach ZERO Ohm and the Rs and approaches infinite, resulting in the following substitution:

Final substitution if all guidelines are followed:

As you can see, the power supply is completely gone only 3 resistors are left over (or complex resistors like complex load... but didn't we say that Rload should be as high as possible??)

Now thing become clearer and simple. Than Vout = Vin . S . (Rp // Ra // Rload). Assume Ra is high and load is high: Vout = Vin . S . Rp

Guess what?  S . Rp happens to be a tube constant called "mu" so finally we have a linear amplification called       

(Vout = mu . Vin)   .............................

And believe me, I have consequently designed this in all kind of amps (not in End Amps, to much current, I stay with Tube, C and Choke) and it works !! May be unnecessary to say that all power supplies must be separated for each stage, regardless left or right or pre or end stage.

Applying Feedback:

We can be short here, NO FEEDBACK to correct errors in the signal path. This will always lead to Intermodulation distortion and other problems. Basically it does not sound right to correct always afterwards your problems. Being pro active is much better !!! so no signal feedback please. Current feedback in the form of an un-bypassed Rk is fine. This is no problem as the feedback is applied immediately and not through a time delaying system. By doing this the Zout will increase, so make sure you do not screw up the other requirements !!

Selection of components:

This is the hard art, because normal specifications normally do not tell anything on the audio quality. But, with all of the above in mind, you can limit the group of possibilities and than ??? just take something and start listening and comparing !!!! Use different resistors, try different (combinations of) capacitors etc. I will start tweaking first with the capacitors now. Better Cap's, like Auricaps in the signalpath and Black Gate or PIO's  in the Power supply and Cathode By Pass. Important open door: do change one thing at a time only and take your time listening to it. At least over a few days!

Happy Listening and Building !!!!

Doede Douma

IMPORTANT: The information provided on this page is intended as guide for DIY activities and therefore free to copy and or publish. If any one wishes to use any of the information from my WEB site, please make sure to refer and footnote to my URL Link as source! Doede Douma