300b lovers

Hi @atmasphere ,

The bottom end is predictable. ~8K 6sn7 (260v, 9mA) internal impedance and 70-80H transformer primary inductance gives -3db around 20Hz. Actually I measured -2.37dB at 20Hz and -3.62 at 15Hz.

According to plate voltage-current characteristics, 6sn7 has very small distortion (0.5-0.7% 150 Volt peak to peak) with both 39K and 57K Ohm load. To drive 6f6 it needs 40 Volt peak to peak. I thought about trying a 27K load, but with 27K we get ~2% distortion.
But I use Linlai 6sn7 and it doesn’t meet the classical 6sn7 spec plate voltage-current characteristics. I have a bunch of classical vt231 40x productions. It will be interesting to make measurements with these tubes.

The bottom end is predictable. ~8K 6sn7 (260v, 9mA) internal impedance and 70-80H transformer primary inductance gives -3db around 20Hz. Actually I measured -2.37dB at 20Hz and -3.62 at 15Hz.

@alexberger Here's something to keep in mind. If you are not using feedback, then its likely at either end of the bandwidth of the amp that the FR will fall off on a 6dB slope. Filter theory tells us that a 6dB slope introduces phase shift to 10x or 1/10th the cutoff frequency. So a 20Hz cutoff will manifest phase shift to 200Hz since there is no feedback correction.

The ear perceives phase shift of a single frequency very poorly, but over a band of frequencies it interprets it as tonality. At high frequencies the ear also uses phase to sense the sound stage. A high frequency rolloff above the audio band can cause darkness in the presentation. At low frequencies a rolloff will be perceived as a loss of impact.

Transformers have bandwidth limits. The smaller you make the transformer, the wider its limits can be. So you have a chance using an interstage transformer to have one that has good LF response- 5Hz is a good place to be. Because you are designing an SET you're dealing with a low frequency bandwidth issue in the output already. All I'm saying here is the less phase shift you present to the output section, the less phase shift will be present in the output.

If you're planning to use a sub with this system this might not be much of a concern. But of you really want to use a sub properly, it should be cut off no higher than about 80Hz otherwise it will tend to draw attention to itself, requiring that it be in the same place as the main speakers. If a sub is used, probably best to drive it with a preamp signal rather than that of the amp due to this bandwidth issue.

@lynn_olson

Will you consider ultrapath capacitor as opposed to cathod bypass capacitor in self-bias?

It turned out the ultrapath capacitor wasn’t sonically that different than a conventional cathode bypass cap. In the context of a regulated supply with an output impedance of 3 milliohms, and equivalent to a 2000uF passive supply, the Zout of the power supply is effectively zero compared to the cathode bypass or ultrapath capacitor.

By comparison, a passive CLC supply with a Zout similar to a 100uF capacitor, a ultrapath bypass might be more appropriate, but then the exact value has to trimmed against the noise introduced by the passive CLC supply. By contrast, the active regulator has 130 dB of noise isolation, so there are no issues of noise introduced by the supply and getting into the cathode circuit.

So it all comes down to the power supply. The optimal solution for a passive CLC supply, with its distinctive noise profile, might not be optimal for an ultra quiet supply with a very low output impedance.

And then we get into the deeper waters of the sonics of the regulators themselves. Some are slow and noisy, and intermodulate with the music. Others are fast and silent. Regulators do not all sound the same, and passive CLC supplies can have a signature too, depending on the capacitors chosen. There is no one-size-fits-all solution.

I should mention the battle of active vs passive supplies has been going on for at least three decades, and is somewhat biased by prejudice against the low-quality active regulators available 30 years ago. High-voltage regulators now are far better, and far more reliable, than what we had then.

The sonics of power supplies are different for SE and PP amplifiers, so it is impossible to generalize without specifying the amplifier topology.

The modulations of supply current on the main B+ supply in a single-ended amplifier are simply the music itself, with an addition of noise from the rectifier stack. Single-ended amplifiers are entirely Class A in operation, by the way.

This is not true for a balanced or push-pull Class A amplifier. The modulation of music on the supply is reduced by 30 to 35 dB (depending on balance), and what’s there is doubled in frequency, similar to a balanced-detector in a radio. The balanced-detector artifacts are the result of symmetric nonlinearities in the balanced pair ... if they have 100% distortion, you get a balanced detector.

If the balanced-pair distortion is a small fraction of that, say, 1% or less, then you still get balanced-detector distortion but much reduced in level. If the balanced-pair are 100% distortionless, then current draw is constant, with no variation. But distortionless balanced pairs exist only in fantasy, so there is always some variation in current draw with real circuits.

In a Class AB amplifier, it is worse, with three regimes ... Class A at low levels, and clipped-off Class B at higher levels.

This has an impact on the sonics of the supply. A single-ended amp is simple ... improve the musicality of the supply, since music is directly impressed on it. Push-pull is more difficult ... the modulations on the supply are a mix of residual imbalance and balanced-detector artifacts, and significantly worse if Class AB artifacts appear in the output stage.

This is the strongest argument for stage-to-stage isolation, so distortion artifacts from a high-level stage do not modulate a lower level stage, In the Raven and Blackbird, we go the additional mile by having a shunt regulator for the input section. The shunt regulator operates by having a current draw that is the precise inverse of the audio-circuit fluctuations, so the net current draw is constant.

By contrast, in a generic Dynaco or Mullard circuit, we have several topologies, with only simple RC power-supply filtering between stages. The output stage is typically Class AB semi-pentode, the driver is Class A triode, and the input is single-ended triode. All three have different distortion signatures.