This is the definitive phono stage:
And here’s the line stage, aka “The Plate Follower”:
And this is the completed high voltage supply. You can see I’ve implemented a hybrid regulator consisting of a LM317 (3 pin voltage reg) and a common power mosfet, the IRF840. This last one needs an heatsink, or you could also use a metal chassis (insulated!). Instead the LM317 doesn’t even need one. Now I’ve got clean 220V: but the voltage could be tweaked by acting on R7.
Here it is… sorry for the bad photos 😉
Ehm… the volume knob isn’t definitive, real one is still in the building process…
Some details of the internal construction: I’ve used both point to point and PCBs.
– line gain: 11 (or 21dB)
– phono gain: 63 (or 36dB)
All the resistor that don’t have a power rating are intended to be 1/2W. Note that 1/4W also will work, as I tend to be very conservative with power ratings. But in general 1/4W resistors aren’t rated to those voltages. Use only high quality 1% metal film resistors (or ceramic power resistors where needed).
About capacitors: for the electrolytics, voltage ratings are shown in the schematics, use the best caps you can afford (strongly suggested 105° switching applications ones, low ESR). For other caps use polypropilene / mylar or whatever flavour you like, but please use HIGH voltage ratings, at least 400-600V. Higher voltage film caps tend to sound better due to better construction.
Use star ground system. It really helps in a sensitive low-level design like this. You can read about star ground over the internet, just use Google.
The gain of the preamplifier is about 57dB, of which 36dB in the phono section and about 21dB in the line. The line section does measure flat (±0.1dB) from 20Hz to 30kHz, and the phono section follows the updated IEC RIAA curve within ±0.1dB (actually less, but there is measurement error to remember) from 50Hz to 20kHz, and ±0.5dB from 10Hz to 30kHz. This, without special components or any selection (neither resistors or capacitors) is a sign of excellent performance. The ±0.5dB differences in the extremes of the audio band are justified because:
- in the high end, some roll off is better, since it will avoid noise from the LP surface to enter, and better rejection of EMI-RFI interference. Actually, most LP doesn’t have much output over 15-18kHz. Also consider that a typical MM cartridge has a resonance in the upper audio band, determined by its inductance and the capacitance of the load (both the cables, 100-200pF, depending on the type, and the input capacitance of the first amplification stage, about 100pF) , and some compensation for this resonance isn’t bad.
- in the low end, since I don’t have an over-the-top turntable, I preferred to loose RIAA accuracy in favour to better rejection of infrasonic noise. In fact there isn’t a roll-off in the low end, as you might think: the 30-50Hz zone is a bit accentuated, by about 0.5dB, but the gain is abruptly cut under 10Hz. This is a “mod” from the original design (I updated the schematics in this page, now they reflect exactly the circuit I’m using), since before I was following closely the RIAA curve, and got some infrasonic rumble and a poorly defined bass. With this additional shaping, I got much less noise and a cleaner, punchier bass. Of course this may be tuned to suit your speakers or your ears, you can play with C1 or C4, the latter is easier to deal with.
THD of the line section cannot be measured with my instruments, we’re under -80dB up to 5-10Vrms output. The phono stage has a SNR ratio of about 70dB, this could get much better by smarter mechanical construction of the preamplifier, and maybe by shielding the tubes.Of course let me know if you mod the circuit, and what do you think about, I will add mods and suggestions here.
Update 15/02/2007: hit “refresh” in your browser when you’re looking at the phono section’s circuit. I’ve updated it, notice C8. Use it (a 400V silver mica part) to implement the fourth time constant in the RIAA curve. That’s not generally needed with high quality vinyls, but some less-than-optimal ones need it. I think it makes an useful upgrade.
You could make it switchable, also you can use different values (not exceeding 200pF, or the like) to tune the upper end of the spectrum. Note that 82pF is the best simulated value to obtain a flat response.