Done it, built and ear tested. The full saga
Hi folks, gave it a go. The joys of lockdown!
I thought it would be good to round off this thread. The whole 300B thing has been about 16 years in the making. Amplifiers were built as monobocs using parts I had mainly accumulated BC (before children) and a few more recent bits (1200VSiC diodes….when did they arrive? DC link PP caps, I really have been Rip van Winkling). Chassis were made from 3mm Al sheet, U channel ends and box channel sides (with the heater supplies neatly running down them). Most wiring is Ag/Cu/PTFE solid core. With the exception of the 300B filament supply filtering, all caps were polypropylene. The power supply smoothing and regulation used an etched circuit board, but this was mainly used as a convenient way to securely mount the blocky MKP capacitors. The tip of the iceberg amplifier section was point to point wired.
Power supply:
Transformers and 5H, 250mA chokes from Solartron oscilloscopes (same model, one set by ”Gresham” , the other by “Parmeko”. Both tested identically for resistances and off load voltages). The ratings are total overkill, but I’ve been hording them for several years
Rectification. Originally I was going to use GZ34, but I then had a change of mind. Several variations of design were plugged into PSU designer and came up as potentially damaging to GZ34s the others being marginal. At whatever they cost these days, I didn’t fancy shelling out for any pops.
I started with stacked 1N5408 diodes (on hand) whilst “testing” and waiting on 1200V Schottky SiC diodes (SCS205KGC) to arrive, thus safely being within the x2 voltage requirement). In the initial 1N5408 version my oscilloscope clearly showed switching spikes form the diode/transformer so the transformer secondary coils were snubbed using a [ C ||(R+C) ] combination. The values were determined by estimating the transformer leakage inductances (see Haggerman “Calculating optimum snubbers”). With the primary coils shorted the secondary was used to make a resonant circuit with a “large” film capacitor (i.e assumed to be far greater than the secondary interwinding capacitance) and performing a frequency sweep. With the snubbers (values [ 0.022uF || (120+ 0.1uF) ]) installed at the transformer outputs, switching spikes were no longer observed.
Pre-regulator filtering was a legacy of the original GZ34 plan, simply a C-L-C filter (40uF- 5H-3*40uF). Capacitors were Epcos DC link.
Regulation was directly copied from Thorsten Loesch’s original version and utilised one 6080 pass valve per amplifier.
Post regulator filtering utilised a huge 440uF 1100V Electronic Concepts polypropylene capacitor.
Any ripple in the final B+ was below the sensitivity of my ancient oscilloscope, so I couldn’t “optimise” using the Hum Buck pot. That will be looked into after “lockdown” is lifted. There is absolutely no noise detectable at the speakers (Lowther Acoustas + 16Ohm PM6a units).
Amplifier circuit
Again, directly copied from Thorsten Loesch’s original version using an El84 driver stage. B+ was adjusted to 420V.
This is where I return to the original thread….. I started out using a Silmic cap and resistor for the el84 bias. After a fortnight of listening and being stuck at home I decided to give the LED bias method a go. Using some cheap red LEDs, Everlight 2873831404, (a la Morgan Jones). The LEDS I-V characteristics were tested over 1-10mA, at 10mA the resistance gradient was 46 with a voltage drop of 1.8V. Upon installation of a string of 3 LEDs the cathode biases in the amplifiers was 5.83 and 5.90V, suggesting a current 0f 10-15mA (further testing required to narrow this down, save that for another day), but quite similar to the 12mA of the cap/resistor bias).
The output transformers are James 6123HS, set for 2.5k input and 16 output.
Listening tests
Before I go on, I should mention that I’m 54 and the ears have been subjected to quite a lot of sonic abuse over the past 40 years or so. So onto subjectivity and false memories….Valve line ups have been permutations of Mullard and Golden Dragon El84s, and Electro-Hamonix and Billington Gold (Chinese) 300Bs.
I was quite pleased with the original cap||resistor El84 bias. The sound was quite different to the ECC88 triode modded Quad IIs (see
~ QUAD II Valve Power Amplifier triode cascode phase splitter driver ~) I have been used to. They certainly seemed clearer in the upper mids, but the bass (yes I know, Acoustas….) wasn’t quite there compared to the push pull amplifiers.
Upon changing to the LED bias, there was a very definite change for the better. I’ve been listening to much the same stuff for many years and all I can say is that I can now follow individual instruments more easily and I have heard low level detail that is new to me. The bass region seems to have also improved and there seems to be more attack to bass lines and kick drums when they are there. Stereo imaging seems to be better, and some smarty pants panning effects are more evident. Some albums still sound awful.
Without measurement, whether I’ve made a harmonic generator effects box or not, I can only say that I am very pleased with the final (for a while…) sound, especially with the LED bias in place. The red light emanating from the ventilation holes is worth a smirk. Sadly the W/Kg number is vanishingly small. Ho (no) hum.