I'd firstly like to thank Allen Wright for his appreciation and his criticism on my (p42 design. I like this kind of constructive debate and, therefore, I'll express some thoughts about the points Allen made, one by one.
2) Two points: Number one is that high transconductance tubes like the 417A are very attractive at first sight, but they pay a big price for their apparent ability to “resolve charge levels on the order of an electron"[?!]. The first one is that their linearity is rather poor. Also, they can only stand very limited plate voltages on the order of 100V. The only point about high current is that you must use a very low value/high dissipation plate load resistor. That makes for distortion in both the resistor and the tube — low level operation is no excuse!
The second one is that, in order to obtain 25 mA/V, these tubes have their electrodes extremely close together and their frame grids are wound with micron-thick wire. The problem with this is that, despite their very accurate and sophisticated construction, due to inevitable mechanical tolerances, the sample-to-sample consistency of their parameters (which is critically dependent on the inter-electrode distance) is very poor and varies widely with temperature, age, and vibration — a complicated way of saying that they're also very prone to microphonics. Not to mention mediocre cathode-to-heater insulation resistance and relatively short life. To me, the highest transconductance tube that is really useful for audio signal is the 6118/6922, but due to insufficient amplification factor and microphony, it is at its best as a power amp input stage, and I discarded it for phono input.
Conversely, the I 2AX7 and — even better — the 65L7 and the 12AY7/6072 may only have 1/20th the Gm of a 417A, but they are usually loaded with a twenty to forty times higher value load resistor, dissipating fractions of a watt with less thermal noise and no temperature coefficient problems.
Most importantly, these are audio-specific tube designs and they have better heater/cathode insulation, “infinite” life ~systematically exceeds 50,000 hours and easily reaches 100,000) with satisfactory consistency and way better linearity.
Identifying high transconductance with high resolution tout court is nothing but a commonplace in my opinion, and substantially wrong if you look at loading requirements. There are other factors that matter more.
I regret that I took it for panted in the p42 article, because the high transconductance guys seem to have missed it completely. While it is true that the input impedance of the EQ network should not overload its driver, it must load it! Otherwise, the current that the tube modulates will do nothing but heat the plate resistor, no matter the Gm. Nobody seems to care about this. In the (p42 (SP#14:p.30), you will notice that R4 is only about 1.6 times R2, while in Allen's design (SP#15:p.39), this ratio is close to 10, and in the Loesch (SP#3:p.26), it is close to 20 in the 75 network and around 6 in the 3180/318 network. This way, the advantages of a high Gm stage are largely thwarted.
3) I do get RIAA paper caps matched to my specs, of course, and I can supply matched sets as well. Having said this, after quite a few years of screwing around with RIAA stages, chasing perfect accuracy like an idiot, I have come to the conclusion that +/- 0.5 dB from 30 Hz to 15 kHz and +/-1 dB above and below is all that's needed. Better accuracy than this, all other circuit parameters being the same and channel tracking being accurate, is of no interesting consequence sonically.
4) Load a high mu triode first stage to the criteria explained in point 2 above and you'll get ‘air and naturalness” with no need to resort to this trick.
5) I have seen too many star-wired amps turn into FM transmitters. Star-wiring is maybe OK on PCBs, but it is a big pain in the butt when it comes to hard-wired circuits of substantial size like the (p42. The physical size of the components forces one to make ground returns of some length, introducing a number of parasitic inductances you can't prevent from coupling together in a way that is virtually impossible to control, no matter how you mess with the layout. Even when the thing doesn't oscillate at 140 MHz, as recently happened to a friend of mine, the risk of “cat piss sound” is right around the corner. If you want to build my circuit, stick to buss grounding please.
6) It seems nobody has ever had any luck with cathode followers except myself and Kondo-San. When he came to Italy last year to visit Audio Note ltalia, I asked him why. His very wise answer was that there's nothing wrong conceptually with cathode followers; it's the application that is tricky. I entirely agree.
The cathode follower has a low output impedance but it is not more load tolerant than common cathodes. So what? The whole point is using it properly and, if it doesn't work, this is the fault of the designer, not the circuit, period. You roust look at the circuit as a whole, not as a pile of Lego bricks. In this specific case, I purposely optimized it to work in a certain way and for a very specific purpose.
Allen's suggestion to remove it is extremely awkward, takes for granted that an inter- connect is a ‘high-quality” capacitance (which is miles away from being true com- pared with a good cap of the same value, especially regarding mechanical factors and series inductance), messes up the loading on the second stage (which is guaranteed to mess up the sound), and is entirely point- less overall.
The biggest threat to all these theories that the cathode follower itself is supposed to sound evil — a point, by the way, which I have never found to be supported by any positive argument but the usual “It's a 100% feedback circuit and sounds bad” bullshit — is the sound of Audio Note Japan gear. Funnily enough, despite this, even the Audio Note UK people love to spit on the CF...
Please disregard the gossiping and leave the CF where it is... or make your own design.
7) You must not use the components to “find the sonic flavor you like.” This is equivalent to destroying all the work you've done to try and get the circuit OK. It's already bad enough to set up the circuit “conceptually” in order to have the least sonic flavor of its own; components should be chosen to add as little as possible. If you want to intentionally color the sound, it's cheaper to bugger around with circuit parameters than components.
8) Removing electrolytic capacitors is not a cure-all, and pretty often does more harm than good. Cathode caps in particular are, in my opinion, often not replaceable. Unbypassed cathode resistors effectively rob transconductance, and all alternatives that have been proposed are guaranteed to drop you from the frying pan right into the fire. Diodes and LEDs put in lots of distortion resulting in a horrible gritty sound, batteries have inconsistent performance and cancel out lots of low-level information, and film caps of the required values are often worse than electrolytics. Much the same holds true for decoupling caps. They act as “local batteries” and an electrolytic does it better than an actual battery. Use Black Gates or Cerafines if you want the best, but look at the design as a whole and leave blind integralism to the Middle East terrorists.
9) I never spent much time experimenting with voltage regulators, mainly because I never came across a regulated preamp that I liked. I have tried Counterpoints, Conrad-Johnsons, Audio Researches, a Lazarus, a Convergent, maybe others I don't remember, and I wouldn't live with any of them.
Maybe this has to do with the fact that all fancy voltage regulators are very high feed- back amplifiers and often rely upon cheap IC OpAmps running open-loop or nearly so, a situation in which their distortion is NOT low, despite the good specs on the datasheets. High-gain error-amplifiered preamps I have heard all had a sort of mechanical/artificial, very “audiophiliac” presentation which is exactly what I am trying to avoid, and that made me skeptical about them.
However, I would be curious to give Allen's Super Reg a try and I have the diagram and setup instructions, but no time to draw up a PCB for it. Hey, Allen, couldn't you just send me a photocopy of the PCB layout so I can etch myself a board? My address is in SP#1 4.
A few more general notes: Cascode-configured input stages do offer some advantages — essentially very high and easily adjustable gain — but their gain and distortion characteristics are more load dependent, which makes them less suited than high-mu triodes for the loading strategy that I like best. I happened to use a cascoded 6922 as the input stage for a small power amp (Costruire Hi-Fi #299 and in this application I did get good results. Besides, the very first phono preamp I ever designed used a 6DJ8 hybrid-cascoded with a transistor, but by now cascode is no longer my choice for a phono input. More gain than you can get from a p42 is of no useful interest anyway. Direct amplification of signal from MC cartridges in order to save the expense of a good step-up trans- former is a loser strategy to me, as I have already argued. The only worthwhile MC cartridges are very expensive and so deserve a transformer. Budget MCs have too many flaws and they're battered by similarly priced fixed coils very easily, so why look for an MC direct preamp?
Direct drive??? Yes — and direct sound. No more audiophile bullshit, i.e. vibrating motors, irritating start-up times, slipping belts, wow~ rumble, awkward suspensions, motor and platter tweaks. Just music, positive, strong and pure. Enough said. The arm does require a bit of tweaking to work properly, however.
The Monophono offers a choice of six treble cuts (flat, 20, 50, 64, 75, and 80 microseconds), six crossover frequencies (318, 400, 450, 530, 630, and 880 microseconds), and two bass cuts (3180 and 1590 microseconds), for a total of 72 combinations. Although the 1590 bass cut is rarely used (only some Columbia LPs require it), the other time constants allow you to compensate for virtually any recording characteristic ever used. You get the RIAA/New Orthophonic (318+75), the CCIR(450+50), the AES (400+64), the London 78 (530±75), the Decca FFRR 78 (880+20), the EMI 78 (630+flat), and so on.
Don't worry about records whose record- ing characteristic you're not sure about: try switching the bass lift first, then the treble cut, and you'll easily find a setting at which the sonic balance is obviously most correct. It's easier than you might think. You'll find that the majority of LPs are to be played with either the RIAA or the AES, and most pre-1970s 45s sound best on AES too. 1950s 78s are generally played back with CCIR, while the situation is more varied with older ones. Most often, the optimum setting will have to be determined experimentally.
The circuit has two parallel inputs and two parallel outputs in order to be plug compatible with a stereo system. You don't need to bugger around with cartridge connections since the generators will be paralleled at the preamp input. This will cancel out the vertical modulation output, with the side advantage that all of the components of surface noise, impulse noise, and, most importantly, mistracking distortion engraved on the grooves by previous poor tracking that are modulated vertically will be rendered inaudible also. Many old records that sound badly battered if played in stereo become much more listenable if played back in mono.
The use of a stereo cartridge for 78s also offers the possibility of playing back ancient “hill and dale” vertically modulated records. To do this, just reverse the connections of one channel at the cartridge pins— choose the channel with no ground tab, usually the right channel. In this way, the two coils are connected out of phase and, therefore, when paralleled, they output the vertical modulation. Note however that these records can have grooves of strange shape and may be difficult to track.
Another advantage of using a stereo cartridge to play mono records is that it'll track ‘em much better.
The circuit uses only two tubes, one 6SL7 and one 6J5. If you can't find a 6J5, you may use one section of a 6SN7 instead: it makes no difference apart from heater consumption.
In order not to add an extra tube, I gave up to SRPP in the second stage, but I biased the two 6SL7 halves in such a way that input acceptance is still ample. Please don't cultivate the illusion that common cathode sounds any better than SRPP. This is not true and it is claimed by those who make SRPPs with two sections of the same tube and/or a common heater supply. This is what really degrades the sound of SRPP. The circuit itself is excellent if properly biased. Please note that I don't agree with those who “load it down” to decrease static measured distortion. This robs dynamics.
This circuit is designed to run from the same power supply as the $2. I suggest that it is a very convenient idea to construct a flexible phono preamp with both the $2 and MONOphono boards sharing the same chassis and power supply. I found that this solution does not in the least degrade the sound of either, provided that signal switching is avoided. Route the stereo and mono outputs via individual interconnects to two inputs of your line preamp or integrated amp, and connect the turntable leads to whichever circuit you intend to use.
Have fun! -- Diego
Get the entire collection of Sound Practices now!
Enjoy the Music.com highly encourages our readers to buy the Sound Practices CD filled with all 16 issues in high quality format from publisher/editor Joe Roberts' eBay store by clicking here.