In the process of experimenting with resistorless-capacitorless amplifiers, I found that vacuum tube diodes could be used to provide effective biasing of amplifiers. The lower dynamic resistance (lower dynamic resistance than a resistor used to obtain the equivalent bias) provided effective biasing without the need for a bypass capacitor. This eliminated a potentially sonic troublespot in the electrolytic capacitor needed to bypass the cathode biasing resistor, and still did not substantially harm the gain nor the output resistance of the stage.
Since the amplifiers built using this biasing technique sounded good, I decided to do a check to see if there was some measureable difference in the distortion produced.
The test circuit
This is, as you might recognize, the no R, no C 845 amplifier already described. For the test, the 845 was removed, and the output from pin 4 of the Lundahl 1660 was connected to the distortion measuring equipment (total load is about 150k ohms).
There are 4 different conditions I tested (B+ was about 124 volts for the tests, and about 22 mA current):
For the columns below "a" is 50VRMS output level, "b" is 70VRMS output level. Input frequency=1004Hz.
Harmonics below -110dB not reported. Note that -60dB is equivalent to 0.1% distortion, -80dB is equivalent to 0.01% distortion etc. Numbers are dB. below fundamental.
|Harmonic||Ckt 1 -a-||Ckt1 -b-||Ckt 2 -a-||Ckt2 -b-||Ckt3 -a-||Ckt 3 -b-||Ckt 4 -a-||Ckt 4 -b-|
There are several interesting things to note in this table. First, any of the topologies listed produce pretty good results, maximum distortion is still less than 0.1% at 200 volts peak to peak with no feedback! The choke in series with the B+ feed reduces distortion slightly without any apparent side effects. It also reduces higher order components moreso than lower order components. (Compare "Ckt1 and Ckt2 results).
The 6AL5 diode bias arrangement actually performs better than a resistor bias network, even in this differential stage. (Compare "Ckt2 and Ckt3" results.). Again, the higher order components are reduced more, particularly at lower levels. This ought to sound better, and in listening tests, indeed it seems to sound better. There is another subtle effect of using the 6AL5 as bias "resistor". As the "mains" line voltage varies, the heater of the amplifier and the "bias" part are similarly varied, somewhat compensating for line variation.
By comparing "Ckt3 and Ckt4" you can note a slight degradation of performance, due to the effect of the electrolytic capacitor.
It should be noted that the gain of the 4 variations listed above is approximately the same. This is modtly due to the differential amplifier topology chosen, rather than any "magic" from the biasing networks.
Notice that the odd order distortion is greater than the eve order distortion products. This is typical of a differential amplifier. Raising the current (by using a higher supply voltage) lowers the odd order components substantially. For instance, the 3rd order distortion product at 70VRMS output with 150V supply is about -70dB (about 0.03%), and higher order products are virtually non existent.