About RMS
Audio Design
Kits & Parts
Contact & Links


Measurement equipment

The measurements were done with the following equipment:

  • Standard PC
  • Motu Microbook professional USB sound card. Own distortion <0.003% and -130dB noise floor (24bit, 96kHz sampling)
  • ARTA software.
  • Custom made attenuator probe for high power output signals.

The measurements were done at one terminal of the balanced output reference to ground. This means that the mentioned voltages should be multiplied by 2. For instance 10V AC means 20VAC RMS over the total load which equals 50 W at 8 Ohms. This level is just below the clipping level. For comparison also a measurement is done at -20 dB relative to clipping. This level corresponds to 0.5W and is representative for average listening on approximately 80 dB sound level with regular loudspeakers.

I have to apologise for the difference in scales and colours, due to getting my first experience with ARTA. I first measured the global feedback version and after changing the hardware I found out how to get rid of the dark background and get a B/W version. Also the noise floor is different because of averaging in the case of the local feedback version.This has no effect on the harmonic distortion figures but makes the harmonics more easily perceivable.

Only the measured data of one channel is shown (Left channel). The differences were not significant enough to show them here. The only exception was the harmonic distortion with 1kHz near clipping as mentioned below.

Harmonic distortion

First the distortion with 1kHz 1VAC output is shown with at the left side the global feedback version and at the righ side the local feedback version:

Harmonic 1kHz 1V AC globalHarmonic 1 kHz 1VAC local

The distortion figures are not very much different and very acceptable (~0.1%)

At higher levels, 1kHz 10V AC the difference becomes more clear:

Harmonic 1kHz 10V AC globalHarmonic 1kHz 10V AC local

With 2% compared with almost 0.2% there is little doubt about the benefits of feedback for measuring results. Possibly the level of the first measurement was just a little bit more away from the clipping level as the right channel showed 0.64% harmonic distortion in the global feedback situation. Next time I will do the measurements 10% below clipping. It can at least be concluded that the typical tube non-linearity near clipping (soft-clipping) is clear from these measurement data.

At 10 kHz almost the same effect is observed:

Harmonic 10 kHz 1VAC globalHarmonic 10kHz 1VAC local

Here the harmonic distortion with local feedback is almost double, 0.24% instead of 0.11%. Not a big deal!

At 10V both distortions are high:

Harmonic 10 kHz 10VAC globalharmonic 10kHz 10VAC local

Also these values are a factor two different, 1,5% and 3% which underlines the conclusion that the measurement at 1 kHz was not correct for the global feedback version.

Intermodulation distortion

Two measurements were done with two-sines. One at 0.25 & 8 kHz, magnitude ratio 4:1 and one at 13 & 14 kHz, magnitude ratio 1:1, both at 1V AC and just before clipping (≈8 V AC) (remember in reality the voltages over the load are a factor two larger).

First the two-sine 0.25 & 8 kHz @ 1VAC:

Twosine 0.25-8kHz 1VAC globaltwosine 0.25&8kHz 1V AC local

Like the harmonic distortion, the IM distortion is also a factor two larger: 0.36% vs 0.18%

At ≈8 VAC the same is observed although dramatic numbers appear as can also be directly seen with the high amount of hair in the graph:

twosine 0.25&8kHz 8.7V AC globaltwosine 0.25&8kHz 8VAC local

1.5% IM distortion with the global feedback and more than 6% IM distortion with the local feedback. This is really tube amplifier behaviour!

Now finally the IM distortion at 13 & 14 kHz. First again at 1VAC:

twosine 13&14kHz 1VAC globaltwosine 13&14kHz 1VAC local

0.14% vs 0.29% is not too bad.

At 7VAC we get:

twosine 23&14kHz 7VAC globaltwosine 13&14kHz 7VAC local

1% vs almost 3% which is less dramatic. Also with the latter measurements the effect of distance to the clipping level must be taken into account.


Like mentioned in the introduction, local feedback does not work beneficial for the distortion of the power stage. This power part has to purely rely on the kathode feedback and the "ultralinear" configuration. The distortion figures are much worse than with almost any transistor amplifier which automatically means that the addition of harmonics adds "sound" to the music. With some music this sound is perceived as very pleasant, with other music this is less the case. Especially with multiple recorders like the "Amsterdam Loekie Stardust Quartet" the sound might become a bit harsh (Dutch: knoerpen!) but some will say it are my ears that fool me!