I've been mulling over a few performance enhancements for my KT120 Mk. 2 amp. The major addition I wanted to add was an Enhanced Fixed Bias (EFB) output stage regulator. While I was in there, might as well throw in a few power supply upgrades I've been contemplating also. The basic design of all the iterations of my "KT120 amp" is: Two channels on one chassis with one power supply for both L and R channels. Mullard-like topology: triode voltage amp, 6SN7 cathode coupled inverter, UL connected output stage. KT120 power tubes. Triode Electronics clones of the Dynaco A431 output transformers (these have 33% UL taps) UL operation. B+ voltage of 475V supplied to center taps of output transformers. Fixed bias operation. The Mk. 3 is identical to the Mk. 2, except for the following enhancements: The Mk. 3 utilizes an Enhanced Fixed Bias (or EFB) output stage regulator whereas the Mk. 2 does not. The Mk. 3 utilizes a snubber on the high voltage secondary winding for some power supply noise suppression whereas the Mk. 2 does not. The Mk. 3 swaps the positions of first and second banks of capacitors in the high voltage supply before and after the smoothing choke to reduce transient ringing within the L/C interaction of these components. In the Mk. 2, the first filter cap was an equivalent 275 uF and the second filter cap (after the choke) was an equivalent 110 uF. In the Mk. 3, these capacitor banks are reversed. The biasing approach of my original KT120 amp, which has carried forward through all iterations, is a fixed bias output stage circuit with individual bias adjust pots for each output tube. In these individual bias adjust circuits, there is no DC "balance adjust" pot per se. Rather, each pot acts as both the mechanism to adjust bias and to achieve DC balance. It has worked well if you know the bias target you want--just set all four tubes to the same bias target and all four tubes are both DC biased and simultaneously DC balanced. Simple--until you want to change bias without upsetting balance, which is exactly what I want to do with the EFB regulator so that I can find the best bias point that minimizes distortion. Sliding bias up or down without changing DC balance is doable with individual bias adjust pots, but each change in bias requires resetting DC balance--a tedious and cumbersome adjustment. So with the addition of EFB, I decided to add a bias adjust pot on the EFB board itself, where one single pot is used to set bias for all four output tubes. Then, each individual adjust pot that previously was used for both DC bias and balance is now only needed for DC balance. This is working well in that I can set DC balance once via individual adjustment of each of the four pots, and then slide bias up or down by adjusting the bias pot on the EFB regulator board. The test procedure then is basically the following: Set bias to 60 mA (that's what I used on the Mk. 2), set power output to 50 watts both channels driven, then measure distortion. Find the lowest distortion bias setting at 50 watts output, both channels driven. For giggles, substitute in new production Tung Sol 6550 tubes (because I have a quad laying around), and find its lowest distortion bias setting at 50 watts output, both channels driven. Here are the results: As stated in the Mk 2 thread, the Mk. 2 amp (with no EFB), measured with the following distortion numbers at 50 watts output, both channels driven, and at 60 mA quiescent current per tube: 20 Hz: 1.35% 1 KHz: 0.42% 20 KHz: 2.0% The KT120 Mk. 3 measures with the following distortion numbers at 50 watts output, both channels driven and at 60 mA quiescent current per tube: 20 Hz: 0.95% 1 KHz: 0.25% 20 KHz: 1.25% Comparing these two sets of distortion measurements clearly show that EFB is doing its handy work. Percent THD improvement between Mk. 2 and Mk. 3 under identical bias and output conditions is the following: 20 Hz: 30% decrease 1 KHz: 40% decrease 20 KHz: 37% decrease But how much lower THD can the Mk. 3 produce if adjusted optimally? Lowest distortion readings were obtained at 100 mA quiescent current (per tube), at 50 watts output, both channels driven. Under these bias conditions, quiescent plate voltage clocked in at 455V, and this represents idling each KT120 tube at approximately 76% of its max plate dissipation. Here are the THD readings: 20 Hz: 0.8% 1 KHz: 0.15% 20 KHz: 0.71% This represents really great THD performance. In fact, several years ago before I sold my Conrad Johnson Premier 11 amp, I measured its distortion at max power output. It was roughly similar to the Mk. 3 at 1 KHz, but my CJ amp could only deliver less than 1% THD up to 15 KHz. But, I don't like idling the KT120's at 100 mA each. That's a little too hot for my tastes. So I think a reasonable compromise here is to find the idle current that will deliver < 1% distortion at full power output from 20 Hz to 20 KHz. That idle current was determined to be 85 mA per tube. Plate voltage was measured at 465V, which idles each tube at 66% of max plate dissipation. Distortion at this bias setting is: 20 Hz: 0.82% 1 KHz: 0.17% 20 KHz: 0.95% Which is still very respectable for a Push Pull amp. Now, since I have a quad of new production Tung Sol 6550's available, I thought I would measure distortion with them in the amp. Minimum distortion reading was found to be at 77 mA and 475V plate, at 50 watts output, both channels driven. Quiescent dissipation is 87% of max. Distortion readings are: 20 Hz: 0.97% 1 KHz: 0.165% 20 KHz: 1.15% Still really quite excellent performance but again, this is a little too hot for me so I'd probably back that off to 60 - 65 mA at the expense of a little higher distortion. My gut says that if I were to convert the output stage to pentode mode (instead of UL) I would be able to achieve the same distortion numbers at lower static dissipation levels. But there is a tradeoff...usually more feedback is required with pentode output stages to obtain the same damping factor as with less feedback but in UL mode. And more feedback represents its own set of issues that need to be dealt with. So overall, the numerous iterations of this amp I've built over the last 6 years have lead me to what I believe is about the best this topology and configuration has to offer. For sure this is the best overall performing amp I am personally able to create. I keep saying that as I build each amp, as I learn a little something new with each build. But I think also because of the off-the-shelf parts used, this near world class performance within the stereo chassis configuration is within reach of the normal DIYer. My thanks again to Dave Gillespie for inventing the EFB regulator and making it public! I will leave you with the Mk. 3 schematic as well a few pics of the amp and the EFB board. This is a really pretty amp with a polished aluminum chassis that is clear anodized and with wood accents. The unique thing about this chassis is the transition from front face plate to top plate is a clean 90 degree angle made from one solid piece of aluminum. It makes for a very clean look. Amp: EFB and negative bias board: Schematic attached as PDF document.