Hello all,
Having been snowed in here in Detroit, I decided to sit down and draw up a new amplifier to build. I'm dissatisfied with my EL84 amps and want a change. In any case, I recently acquired beautiful a beautiful RCA LC-1A and an RCA 515S1 in a trade. I've been meaning to put together a kick ass mono system for early LPs, 45s, and of course, 78s.
I decided I'd try to build an amplifier using what would have been near state of the art circa 1948-1950, when my speakers would have been new.
This is what I played around with:
Starting at the front end, a Van Scoyoc "cross coupled" input stage and phase inverter. There are technically superior inverters out there, but this one was period correct, pretty damn popular in the '48 to '50 time frame, offered a reasonable degree of balance at a reasonable level of distortion, and made use of one of my favorite small signal tubes: the lowly 6SL7.
This was followed by a driver stage, admittedly lifted almost component for component from a wonderful high fidelity design published by RCA in the summer of 1948. Both sections of a 6SN7 are RC coupled to the grids of the output stage delivering significant driving voltage at negligible distortion.
The output stage was again lifted almost component for component from that same design from RCA design. A single 6AS7G is used in push pull with itself. A 6AS7G contains two triodes of 13 watts plate dissipation each. The 6AS7G's only major failings are that you MUST use cathode bias with it if you expect any sort of reliability, and with a mu of only 2, the 6AS7G requires an ungodly grid drive. Other than that, the output stage is capable of incredibly good damping factor, better than a push pull pair of 2A3s even, and lower distortion than a comparable 2A3 stage. You can reasonably expect about 10 watts of real RMS power at only 2% THD into a 5000 ohm load, with the third order distortion by far dominant.
Finally, the power supply is a rather conventional type using a 5V4G. Nothing special here aside from a rather large 20 Henry choke to help quash the 120 cycle hum.
The resulting schematic is shown below:
If any of you have any potential improvements or notice any major errors on the schematic, please, let me know.
If you accept the general conclusions reached by Harry Olson at RCA, 10 watts is more than enough power, including reserve power, for the efficient loud speakers of the day. Also with the RCAs, which were revealing and low distortion drivers for their time, and are still quality full range speakers by even today's standards, it was thought by Olson, and borne out in his psycho-acoustic experiments in the late 1940s, that most people would not be able to detect ~0.5% to ~1% THD from the amplifier, as any speaker is producing much more distortion by virtue of the mechanical limitations involved.
Compared to a Williamson, which was sweeping the globe around that time, and famously offering 0.1% THD with gobs of global negative feedback, this design offers 10 watts at 2% THD, 8 watts at 1% THD, and 5 watts at 0.5%. In addition, it affords good damping factor, and something no Williamson could ever deliver, unconditional stability into near any load. Despite the inferiority on paper, I strongly suspect a real working amplifier of this type would sound pretty decent in the real world.
I promptly simulated the design in LTSpice using Ayumi's extremely accurate tube models, and it compares well with expected behaviors as found via paper, pencil, and calculator.
I then remembered that LTSpice offers the ability to feed a circuit with a WAV file, and then export the voltage signal as a WAV file from any node in the circuit. LTSpice uses the WAV file as essentially a very complicated piecewise function, for generating a voltage at a voltage source. This is conceptually no different than what any DAC does, except that LTSpice should theoretically be able to do this to a much higher degree of precision because it is just generating a table of values that represent instantaneous voltages. This is effect offered me the ability to "hear" a reasonable facsimile of the amplifier's output before even warming up the soldering iron.
I fed the simulated amplifier a short 15 second sample of Clifford Brown performing "What's New", and "recorded" the output at the 8 ohm load resistor (the "stand in" for the voice coil) as a WAV file. The simulated amplifier, under this test, clipped a bit and reached a peak THD of ~2.6% to ~2.7%
I am not claiming that simulation is an acceptable substitute for actually building an amplifier. Nor do I claim that the final built version of this amp will sound exactly like the result of the simulation. I will assuredly deviate from some of the schematic values, in particular all the coupling caps need to be tweaked as of yet. In addition, the simulation has many theoretically perfect components, perfectly matched tubes, etc. The real result will have a bit higher distortion, etc. Yet, I still find it interesting that the non-linearity of the active elements in the simulated amplifier did show a realistic level of distortion, with characteristics you would expect in such a design. This has proven yet again to me that LTSpice is an incredibly useful tool in design.
Below is a link to the two WAV files, input, and output:
https://www.dropbox.com/s/w6iq4b7ro9...Files.zip?dl=0