Eliminating Cathode Bypass Caps In SET Amplifier

[regal]

Its the funniest thing...

Everyone talks about caps as if they are the culprits in "the sound" (problem). After much esquisitely precise testing ... I have found that the "cathode cap" is not in itself the "problem", but rather tightly controlled fixed bias (which is what a resitor + capacitor makes). In essence, the resistor and capacitor form a tiny, compact power supply in series with the cathode, raising its voltage relative to the grounded grid. To all the well-informed here, this is nothing at all new, or even "newly revealed". It just is.

For instance ... so long as one's using a separate winding of the transformer for filament power, it is pretty straight forward to make a low current negative power supply ... a few rectifiers, capacitors, a very low current choke, and so on ... With nothing more sophisticated than a quality pot, one can get 0 to -15 volts of bias for the grid. Grids just aren't connected by a high-value resistor to round, but rather, to the wiper of the pot sitting across ground and the C- supply.

The effect of this kind of supply is exactly the same as using a cathode-resistor-and-capacitor (of sufficiently high quality capacitor and low-value low-impedance double-bypass caps, of course). Using a DPDT switch and careful calibration of the ampere flow through the cathode ... one can do A/B comparison trivially. Amplification remains a constant. Whatever "problem" the fixed bias has given the system, remains in both cases.

Not so... when one uses the bias-in-cathode resistor WITHOUT the capacitor. Now suddenly the tube will behave much less like a Gm (transconductance) amplifier, and more like a cathode follower (with multiplicative gain on the anode load). Almost all nonlinearities that the transconductance-as-a-function-of-current curve produces with fixed bias are removed. Not all, must most. In simple circuits without fancy constant-current or cascode loads, the net effect is reduced gain and increased linearity. It is hard to do the "a/b" switch ... requires a 3PDT switch to do it right and an OUTPUT pot to reduce gain from the cap-bypassed higher-gain configuration. Doable, but not easy. The sonic difference though is quite remarkable. "Floating bias" (or loosely, cathode-follower) response is more linear - but the effect seems to also be a slight muffling of the sound. Overall - when the whole system gain is pepped up just a bit to accomodate that ... the sound becomes lovely and more rich due to increased symmetric linearity.

So... there you are. Stuff wrought from hundreds of hours of fiddling.

GoatGuy

I think this is why led's and filament bias sound good, because there is still some cathode denegeration due to the impedance left.

A good trick is the partially bypassed cathode, use two resistors instead of one with a 1:10 ratio, bypass the bottom bigger resistor, best of both worlds.

 

[ErikdeBest]

Not so... when one uses the bias-in-cathode resistor WITHOUT the capacitor. Now suddenly the tube will behave much less like a Gm (transconductance) amplifier, and more like a cathode follower (with multiplicative gain on the anode load). Almost all nonlinearities that the transconductance-as-a-function-of-current curve produces with fixed bias are removed. Not all, must most. In simple circuits without fancy constant-current or cascode loads, the net effect is reduced gain and increased linearity. It is hard to do the "a/b" switch ... requires a 3PDT switch to do it right and an OUTPUT pot to reduce gain from the cap-bypassed higher-gain configuration. Doable, but not easy. The sonic difference though is quite remarkable. "Floating bias" (or loosely, cathode-follower) response is more linear - but the effect seems to also be a slight muffling of the sound. Overall - when the whole system gain is pepped up just a bit to accomodate that ... the sound becomes lovely and more rich due to increased symmetric linearity.

In my understanding removing the bypass capacitor actually makes the output tube work more like a transconductance amplifier as the total plate resistance increases with an unbypassed cathode resistor. An unbypassed resistor causes negative feedback, which explains the lower gain and maybe the increased linearity? Anyway, the output impedance of the amplifier will increase (due to higher plate resistance) which may work quite well, but not really with complex crossovers designed with assumption of a 'zero' impedance output amplifier. MJ arpeggio loudspeaker (article at diyaudio) and NP's article on 'transconductance amplifiers' describe this much better!

 

[FRUGIVO]

Fixed Bias is by far the best
test and hear
that simple

 

[andyjevans]

Fixed Bias is by far the best
test and hear
that simple

No it isn't and it's not simple. In all my tests filament bias sounds the best, but this depends on the quality of the filament bias supply, and it can only be used with a small range of DHTs. After that may come fixed bias, but again that depends on the quality of the bias supply and how it's used, for instance with or without a capacitor preceding it. And a LED supply in the cathode or a some similar combination of devices giving a bias voltage can be very good again depending on the implementation and the tube.

 

[FRUGIVO]

this is my empirical opinion
removing the cathode capacitor
always has to be the goal
fixed bias with possibility of adjustment is the way
is proven that
release the signal path passive components reduces signal compression (NFB) and
dynamic range is gained and hearing confirms this

 

[juanitox]

no fixed bias with negative voltage on grid is not the best you will add a coupling Cap working at low voltage where it works not as it"s best.. filament bias as andy do works fine or battery cathode bias works fine too with small signal tube .. my amp has a 47000uf source resistor bypass and i prefer it to fixed bias .. no rules

 

[porkchop61]

I've built Mikael Abdellah's SE KT-88 amplifier which is very similar to George's design you are looking at. I too was originally using a cap bypassed resistor for bias for my driver tube (6N1P in this amp), and didn't really care for the sound. I switched to using two red LED's in series and never went back, I'd definitely give it a try. I also run EL-34's in this amp mostly for Rock music. I also have mine set-up for UL, and Pentode modes via a switch.

Glenn

 

[daqvin_carter]

You can always go half fixed and half cathode bias. Use standard fixed bias supply and smaller cathode resistor, or use one or more 9 volt bias batteries and some cathode bias.

 

[Hearinspace]

Yes, easiest thing to do - unsolder one connection to the bypass cap. Put in a temporary switch if you like to do A/B.
More advanced option, for fun (ala Wavebourn), is to convert the ccs to a gyrator (fixed anode voltage, but high impedance at audio frequencies). This allows the current to adjust as the tube ages, just as it would with cathode bias.

Sheldon

+1

http://www.diyaudio.com/forums/headphone-systems/144360-tube-headphone-amp-designs.html#post1840492