MANUFACTURERS' COMMENTS on tube amps - VAC on VTL vs RAM, As published in Stereophile, June 1991

As a manufacturer of high quality vacuum tube amplifiers, I am somewhat concerned that the recent exchange of letters between Roger Modjeski and David Manley in these pages may leave a negative impression in the minds of those considering tube electronics. I would like to offer a few comments in hopes of calming the debate, which seems to have strayed beyond engineering and sound into philosophies and personalities. There is much truth in what both have written.
The debate seems to be founded in three areas: Can one select for tube quality? Can certain tube substitutions be acceptable and rewarding? Can a single bias control for four tubes be justified?
I think that all involved with tube equipment can agree that tube sonics and reliability vary from plant to plant. Why this should be is fairly obvious. Line up a dozen 12AX7s manufactured in the U.S. during the 1960's and you will probably see at least 6 different plate shapes: some taller, some boxier, some with a pattern, etc. They also vary in ways that we can not readily see, such as coatings, chemical purity, heater construction, etc. There are actually  tremendous areas of discretion within the specifications for a particular tube type. That each different implementation measures about the same but sounds different in a particular circuit is hardly surprising. The electronic measurements we make do not always accurately reflect what we hear. This simple and annoying fact applies to vacuum tubes just as it does to cartridges, digital, amplifiers, and cables.
Quality is a vexing issue, and one which RCA, Sylvania, and others struggled with mightily in the years following World War II. Each audiophile knows what quality means to them, but how can the tube manufacturer define it? They usually work by the numbers and statistical probabilities. Is quality getting close to target transconductance and mu specifications when new? Rate of drift in use? Boundaries of drift in use? Percent lasting to 5000 hours? Microphonics or noise, and in what frequency range? To the audiophile the most important components of quality are probably sound quality (can it be measured?) and life.
Our experience indicates that sonic characteristics must be designed in, not selected for. For example, take a pair of EI (Yugoslavia) 12AX7A, one measuring correct on all parameters, and the other measuring out of spec in some ways. Also take a similar pair of Chinese 12AX7A, one in spec and one out. Compare their sound in an audio circuit. Often the good and bad Yugoslavian tubes will sound more alike than do the in spec Chinese and Yugoslavian tubes! So it would seem that a particular tube's sonic traits in a particular circuit are largely designed in, not selected for.
This does not mean that tube grading is without value. While consistency of electrical characteristics and low microphonics are functions of the design and manufacturing process, both may be selected for with a good degree of success. This may not be quite as good as a very tight design and manufacturing process, however, because the same problems that lead to variability among new tubes may also allow for greater changes within a particular tube over its lifetime. Long term reliability is harder to select for than to manufacture for.
On the issue of tube life, even supertubes, such as the RCA Special Red series, could not guarantee tube life. Some tubes will fail in the first ten hours, some after 7,000 hours, and some after 20,000 hours. (For example, see RCA Review, September 1953, pages 413-426.) A term such as "10,000 hour life" means that a certain percentage of that particular tube will reach 10,000 hours of operation without going outside of a certain window of performance. This window often includes deviation to only 85% of specified output! It also often refers to applications where the tube is not switched on and off. One of the major record companies reported back in the 1950's that 12AX7s left on continuously for two years were quieter and closer on transconductance than when new, while those that went through even one off/on cycle each day were far worse!
Now, let's change gears and look at the amplifiers that use the tubes. The contention that a power amplifier is "consummately balanced and optimized" reads well, but is not a full reflection of reality. Every choice a circuit designer makes is a judgement call, a balancing of often conflicting performance goals. Are we after minimum distortion, maximum power, good efficiency, wide bandwidth, maximum symmetry, low feedback, minimal output impedance, or any of a dozen other goals? The answer is always complex even in a no-holds-barred amplifier, and no great circuit ever completely dominates another on all parameters.
It is precisely because true optima do not exist that a surprisingly successful set of swaps can be made among types 6L6GC, 5881, 6CA7/EL34, 6550A, KT66, KT77, and KT88 if the amplifier has sufficient margins and adjustment ranges in certain key areas, the tube socket ties pins 1 and 8 together, and provided that certain voltages, currents, and other circuit values are not exceeded. Amplifier performance will undoubtedly be different, but may still be of significant sonic merit. Where possible, swaps allow an audiophile a type of choice similar to that which he or she will exercise with speakers, preamps, and cables, all of which interact with the power amplifier. It is also worth noting that swaps of tube brands, not types, can have a major impact on the sound of an amplifier. As an extreme case, we are aware of a brand of EL34 that sounds to us remarkably like a good 6L6GC in most circuits.
To demonstrate that optima do not exist, VTL would not want you to swap 6550A and KT88, since the two tubes have certain differences (see the Tube Substitution List in The VTL Book, Second Edition). Yet consider a card I recently received from VTL introducing the new "KT90" as something "better...that can simply be plugged into your amplifier" in place of either a 6550A or a KT88. Is not the notion of replacing both types with a single new tube somewhat reminiscent of Roger's contention that certain tube types may be interchanged with minimal deleterious side effects? After all, if one amplifier is consummately balanced and optimized for 6550A, and another for KT88, how can both be optimal with the KT90? Clearly, complete optimality does not exist and certain swaps are permissible in certain designs. While the additional voltage handling and plate dissipation maximums of the KT90 will do little for owners of existing amplifiers, it will undoubtedly sound different, and each interested audiophile should judge the result in their own system and with their own ears.
It should go without saying that substitutions must be done with extreme care, and strictly at one's own risk unless blessed by the manufacturer . If VTL says don't swap (except as noted above), then don't in their amps. It may well be that a certain swap is safe and sonically rewarding in an RM9, but not in a VTL. This does not mean that there is anything wrong with what either designer has done!
Some substitutions can be disastrous, including one suggested in error in The VTL Book's Tube Substitution List: 6BG6 for 6L6GC. These two types have different pin assignments, and one requires a plate cap. No one is perfect!
The debate over the RM9's single bias control is just not critical. Roger Modjeski feels that it is valuable to have approximately the same DC bias voltage on each output tube's grid. This helps ensure that grid circuit clipping will occur at the same drive amplitude on each tube, which in turn helps ensure clipping on both the positive and negative sides of the output waveform at the same power level. This requires four matched tubes, and is part of Roger's design philosophy.
And here is the ultimate point. No designer has an exclusive on the one uniquely correct way to achieve perfect sound! No amplifier will ever be perfect or immune to improvement. Roger's choice will help yield a symmetrical clipping characteristic and requires a matched quartet of tubes (but, incidently, does not require that a second set of tubes be matched with the first set, so no "octets" or "big bands" will be required). David is quite correct in noting that tubes may unmatch as they age, and this may cause the RM9 to be somewhat less symmetrical than intended, although the statistical probabilities may still favor matched quartets over unmatched tubes. Of course, if you lose one tube you do have to start over with a new foursome.
While neither our amplifiers nor David Manley's amplifiers use Roger's technique, that does not invalidate his philosophy.
We would welcome contact from either manufacturer if they feel that we have done violence to their words or have made any misstatement of fact.
There are many ways to build fine sounding amplifiers. I would like to encourage those who toil with me in this vineyard to get on with making better amplifiers and better tubes rather than clouding the field with smoke from their battles. As we collectively make better amplifiers the market for tube equipment will expand, which will in turn increase interest in the manufacture of quality tubes. In the end that is the only way that manufacturers and audiophiles both win!
.........Kevin M. Hayes
(VAC - Valve Amplification Company)