• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Matisse Line Amp

Refering to the attached schematic, I am wondering if there's a fraud with this circuit.

The value of R14 is 47K in the above schematic. I just feel that it is rather weired having such low impedance at the 12AT7 grid. It reflects the same impedance to 12AX7's plate load. I am wondering if it was a mistake in the design or not. The plate load resistance of 12AX7 is about 70K. The plate load resistor should be 3 times of the plate resistance which is approx 210K. If we load it with 47K, surely it would have higher distortion. May be that's what the designer wants. Please correct me if I am wrong.

I have changed the R14 to 500K and C51 to 0.1uF. It improves the quality and dynamic of the sound dramatically.

I wish I could have the Tube CAD software to check this out.


  • matisee line stage.jpg
    matisee line stage.jpg
    35.5 KB · Views: 935
Overloading of the first stage 12AX7 would cause severe distortion. However, the heavy feedback could reduce the amount of distortion.

May be the designer want to make it this way to create some magic. I still think this is not logical at all. I remember my friend said the guitar amps were designed with lots of distortion. Possibly, the designer of Matissee intended to apply the same magic to this line amp.


Paid Member
2003-06-12 7:04 pm
Maine USA
> I am wondering if there's a fraud with this circuit.

Fraud is a very strong accusation.

> I am wondering if it was a mistake in the design or not.

47K is not the value commonly used, but there are no "right answers" and this is sure not wrong.

> surely it would have higher distortion.

What distortion? We are asking for 1V-2V signal from 400V power supply. Distortion is going to be very low even if we "screw-up" the "right answer for lowest distortion".

Also the 12AT7 has significant gain, which means the 12AX7 works at much lower signal level (about 50mV output) than the 12AT7 (1V output), and the 12AT7's distortion will dominate.

Also the two distortions are in opposite directions, and can cancel. I don't believe it gives large cancellation, but it might.

> I wish I could have the Tube CAD software to check this out.

There was a time when we did not have TubeCAD, SPICE, computers, or even LED calculators. Somehow we managed, scribbling on napkins.

A resistance loaded triode will approach 5%THD when peak signal is 20% of supply voltage. So the 12AT7 will run 5%THD at about 80V peak, 50V RMS. THD does not go down exactly with signal voltage, but that's an easy approximation. At 1V RMS output, the 12AT7's THD will be around ~0.1%.

With hi-Z loading, the 12AX7 working at 0.05V signal level will be down around ~~0.005%. With 47K loading, signal current is high, and Gm/Ik nonlinearity will raise THD around 10 times, to say ~~0.05%. Indeed it may semi-cancel the 12AT7 distortion.

Working hi-Z and bypassed, a 12AX7-12AT7 pair can have forward gain of around 2,000. We often want closed-loop gain around 10, and that gives a LOT of negative feedback. No matter how you feel about high-NFB audio, 200:1 feedback can be hard to stabilize.

Even with the heavy 47K load, and unbypassed cathodes, this pair can have forward gain around 200-400. Fed-back to closed loop gain of 10, this is 20:1-40:1 of feedback. If we assume that distortion does not cancel between stages, and is around 0.1% open-loop, it will be below 0.01% closed-loop.

> my friend said the guitar amps were designed with lots of distortion.

The classic (not fuzz-box) amps were originally designed without regard for distortion. They use hi-Z loading for high gain, and tend to run very high signal levels, without any feedback except a trice around the speaker-jack for slight damping.
PRR said:
Unretouched this-moment snapshot of what is in front of me:

Nice set of persuaders (vice grips) there too... i've a couple slide rules kicking around too... haven't used one much since i got the HP25C in 1976... and now it is either the RPN calc on my 'puter or Trapeze for the more complicated tasks. Would be nice if TubeCAD got ported to OS X, i'd buy it in a flash.



Paid Member
2003-06-12 7:04 pm
Maine USA
> that's a much nicer slide rule than mine.

It isn't a nice sliderule. It was about the cheapest pear-wood (not bamboo, and certainly not polystyrene) ever made. It says Charvoz-Roos and it once looked about like the 105 pictured here. But my aluminum end-braces are buckled and straightened. Being a little tight, and wood, and damp climate, there is a LOT of pencil lead rubbed into the tracks, and even so I don't use it on rainy days.

There is a 1970s plastic Sterling in the desk but I have not touched it in years. My dad had a fine bamboo log-log-decitrig and also one of the very first practical aluminum rules ever made(*) and after those a plastic rule just feels cheap. (* Conventional wisdom was that paint would not stick to Al well enough for a practicing engineer. One guy found a way to make it stick, and was going desk-to-desk in engineering offices trying to sell them. Throwing it around, stirring coffee, and getting precise answers. He had the trick: my father's rule still looked good a decade later, and by ~1960 the Big Guys were producing Al rules.)

> it's ages since I've used mine

As you see: I used it more recently than the vice-grips but less recently than that stupid memo on top of the sliderule. One way it still beats a digi-calc: you have a ratio and you want to find standard-value resistors to suit.

> I once had a book of 7 figure log tables for accurate calculations...

If you run short of log-books, haunt the older used book stores. Some sellers are still trying to get money for these. I've used log-tables, but I admit the last time I did it was to get a semi-random yet easily looked-up series of numbers for setting 3/4-digit combination locks. And of course, there is nothing in audio that justifies even 2-place logs. As noted in post #3, I do most of my figgering on my thumbs.