A Direct-Input Triode EL34 Amplifier
Richard's updated amp; that power transformer butch enough for you?
|
The schematic shown is for the current version, for my friend Richard Tadaki. My original prototype used a 12AX7 mu-follower as the input amp rather than the 6SL7 SRPP, and 12AU7's in the driver diff-amp rather than 6SN7's; so there will have been a few different circuit values, too. Mine also had 5K Scott outputs, while these are the very nice 6K Audio Note TRANS-185; and the coupling caps in mine were SCR film-foils, rather than the Jensens. Either setup works fine, this one shown sounds better; remarkable, actually, for a dirty ol' push-pull EL34 amp. I just had this amp on the bench to fix the power supply, and did some work in the audio circuit as well. The power supply got a custom transformer from Jack Elliano at Electra-Print Audio, which is of very high quality, completely silent, has a bias tap on the B+ so I don't have to have a bias transformer, and is impregnated with very yummy smelling varnish. This amp was built around 1994, and I've learned a few things since then. The B+ supply lost about half of its capacitance and gained a small (same as a Dyna ST70) choke. Each supply circuit starts with a small-value film-foil capacitor to shunt diode noise. The small-signal tubes now have AC (no supply) heater sources, both heater supplies shunted at the transformer lead connection with a 1µF RadioShack 'chicklet'. The addition of this cap drops heater-coupled hum from about 1mV to .5mV, and the character of the leakage signal changes from a 'buzz' to a very soft hum; almost inaudible with the ear next to the speaker.
I made several small changes in the audio circuit. All the grid-stop resistors were removed, and the global negative feedback reduced from ~10dB to 8dB (feedback resistor value increased from 16K to 22K). This was just a lucky first cut: my original plan was to remove all grid stoppers, and then put the circuit on the scope and adjust the feedback to give less than -1dB at 20KHz and no resonant behaviour (no wiggly tops on square waves). With the first-try 22K Riken, -1dB is at 19KHz with an 8
Other changes to the audio circuit were aimed at reducing the higher impedences. Richard likes GE fatboy 6CA7's in this amp (so do I; they can sound chunky and opaque in pentode, but in this kind of pseudoTriode they're very natural), and they can draw remarkable amounts of grid current, so the bias-connect resistors (which actually define the Z of the bias-grid circuit of the output tubes, being the highest resistances in the neighborhood) were reduced in value from the standard 270K The result in this case is very gratifying. The reduction in hum+noise in the circuit, global NFB, and the number of elements in the signal path, results in a substantially blacker background, stronger dynamics, and a noticeably higher and more natural color of timbre. Back wall, breath in the horn, rosin on the bow, et audiophile cetera. I have had guys ready to buy this amp, and was developmentally reticent; now, I'm ready to go. Is this the ultimate WilliamsClone? How would an auld hippie in Maui know? Maybe, though; it sounds really, really transparent. And, it gives ~16-18 watts out; enough to drive lots of real-world speakers. There is a small developmental problem here. This amp was designed when I was using about 12dB of negative feedback, and reducing the NFB from 10dB to 8dB has resulted in a circuit with a lot of gain. With a typical line-level source, the volume control is at 9 o'clock or below, and that's too low. Noise and any heater hum from the first stage start to build up, and the signal is passing through a lot of the pot resistive element, so the sonic signature of the pot is increased. Since the first stage is a SRPP rather than a Mu-Follower, it's not perfectly balanced, and heater hum from the two devices is not completely nulled out. The hum+noise starts to climb a bit. As a quick try to address this problem, I simply put a spare 6SN7 in the first position, which drops first stage gain about 60% (to about 40% of the 6SL7). Okay, the hum+noise+ gain problem is addressed. The gain is reasonable, and the amp is spooky quiet; hum+noise at a measured 0.5mV. Now, there's another small developmental problem; overuse of the same device. There's a good piece of conventional wisdom that says, 'Don't use the same device or component, etc. more than once, because the sonic signature of the etc. will not add, but multiply.' No less a light than J.C. Morrison said, 'The 6SN7 is the laboratory standard, what goes in is what comes out; but use it twice in a row, it will rip your ears off.' Well, it doesn't even rip my ears off, but it does start to sound just slightly dry, referenced to the 'lacy, frilly' (J.C. again) sound of the 6SL7, which as an auld hippie card carrying, I prefer.
The Solution? I don't know, yet. The amp is now back in the owner's rig, and he's posting me every few hours with reports of stuff he never heard before in his reference recordings. It sounded just slightly dry to me, but who am I to argue with my favorite client? Besides; when I heard it, it had about ten hours on it after a layoff of over a year, so it was pretty cold. The problem that I hadn't noticed before with using the very nice 6SN7 and 6SL7 is that there are no other devices in the series to try out. If it were 12A*7 tubes, there are a dozen in various combinations of characteristics. Even the 6DJ8 / 6922 / 6N1P has half a dozen pin-compatible flavors. Making octal / 9-pin socket adapters is a real chore, but I may have to do it. I'm trading amps with him this weekend; he wants to hear my 6V6 amp at leisure, so I may sneak a couple more experiments under the hood while he isn't looking. One thing that's happening here, with a 6SN7 plugged right into a 6SL circuit, is that the 6SN is operating pretty lean, about 3mA; which can give a reticent sound. If I can find the right resistors, I'll reduce the cathode resistance of each device from 1K Last installment for awhile; I've been listening to this amp all weekend, and taking gain and bandwidth tests which are reflected in some slight changes in the spec list above. Hearing it with a little time on it, I think I'll just leave it alone. I'm not so sure that anything I could do to it would be an improvement. I'm much tempted to build myself one (my original got cannibalized), and if I do I'll use the current side-by-side layout and put a 9-pin socket in the first position so I can do some rolling. Meanwhile, I'm going to put this one back in service, so all these guys can see how smart I am. Like to build one? Like to buy one? Got questions? Drop me a line at poinxie@yahoo.com. Poinz |
resistor; 22KHz with my speakers as the load. I never put the circuit on the scope; with either load, the output drops from the steady state on down, displaying no peaks up to 200KHz, where I stopped testing. No wiggly square waves or transformer resonances here. Too bad I didn't get to drag out my scope; I just got a new 1x : 10x probe from PartsExpress I'm hot to try out.