You are reading the older HTML site
Positive Feedback ISSUE
Kismet 3: the Last Generation
In the more than five years that have passed since the publication of my article on the Kismet amplifier (Glass Audio 6/96, "Kismet, a simple 2A3 SE amplifier") I have never stopped developing the design: I tried to improve it, exploiting the experience gained from my top designs. The Kismet is now in its third generation, probably the last and ultimate. But never say never....
I repeat the maxim published with the original article:
"Everything beautiful has a limit. Everything that goes beyond that limit is
bound to be ruined."
Before describing the Kismet 3, I would like to explain my audio philosophy. First Id like to remind you that the purpose of audio is MUSIC. If audio is merely an end in itself, I see no interest in it. If you dont like music, if you dont like live concerts, if you are interested in technology only, please dont read this article.
An audio system is the way to listen to music at home, but too many people listen to their systems and not to music. So, first of all, go to as many concerts as possible. Then buy LPs and CDs! This is not so obvious. Many people ("audiophiles"), even with expensive systems, have only a few test discs. They play a disc and listen to the system, trying to hear defects, then spend time searching for improvements. This is very common behavior, and not only among rich audiophiles.
There is a very important difference between music at concerts and music at home. The latter is a reproduction, while the former is a production. From this point of view, any live music event is unreproducible, unique, enclosed. On the contrary, at home, you can repeat any LP or CD all the times you wish. A concert is imperfect because it is different every time you repeat it, while at home any repetition is perfect because it is always the same. In a live concert there is perfection in a imperfect event!
Reproduction is totally different from production. Its impossible to reproduce live music at home through an audio system, so we must look for aspects common to production and reproduction. In other words, we must choose what we want from a reproduction. We cant talk about reproducing reality, but we can talk about "believableness" and how to reach it. I think that the only common aspect between the two kinds of music is emotion. It is, in my opinion, the only thing that really matters. Emotion may be totally different from person to person. Agreeing with my Audio Master, Mr. Kiyoaki Imai of Audio Tekne, Inc. of Japan, I think that there are only two results: RIGHT sound or WRONG sound. When I listen to music from an audio system, I can hear within a few seconds if the sound is right or wrong. I can hear it because I go to concerts and because the right sound is instinctual. Women can easily distinguish whether the sound is right or wrong because they are not conditioned by audio and by audiophile standards. Women cant explain why the sound is right or no rightt, but they hear the difference.
Why do so many audiophiles have more than one system, or "need" a different system? And why do many audiophiles stop the music and turn off the system after a short time? Because the sound is not right. When we go to concerts and like the music, we never become tired of listening! Please visit my web site (www.audionautes.com) for further info.
I developed the second-generation Kismet in 1977. The mods were two: the working point of the 2A3 and the biasing. These were very important mods; in fact, they changed the result of a few orders! First, I have to explain why I modified the working point of the 2A3. The original working point was 275V and 60mA, with a 3Kohm load. Its a very good point, with low distortion and sufficient power, and I used it for a long time without any problem. Using Chinese triodes after more than 5000 hours, distortion and power remained the same. So, why did I make the change? I had to test a pair of NOS RCA 2A3s, and could test them only in my Kismet amp, having no curve tracer at the time. I put them into the sockets and turned the amp on, and after making a few measurements (the main purpose of my test was to verify the matching), I listened to them. I was very surprisedwith a warmup of only a few minutes, the NOS RCA 2A3s were better than the Chinese ones. You may think that its obvious, RCAs are better than Chinese triodes, of course, but there was something strange. After this test, I installed a pair of brand new Chinese 2A3s and listened to them, and they, too, were better than the "old" pair! No differences in distortion and power, square response, etc., and I found the same results using another Kismet amp. This was the starting point, and the second-generation Kismet was born.
What happened? I spent time thinking about it, and talked to Kiyoaki Imai. He teaches: Do not exploit the device at its maximum. If you use a tube at its maximum, it will be under stress, and you can hear the difference, even if you cant measure it in terms of distortion or output power. If you use use devices far from the limits, not only will they survive longer, but their performance will be the same for their entire lives. Starting from this, I researched a new working point, and after a lot of hours spent on the curves of the 2A3 triode, found a very good one. The output power is lower, but the distortion is much lower, while the anode power is only about 12.5W! The only "limit" is the high load, 5Kohm. To tell the truth, I think this is a strong point, and not a limitation. 5Kohm is a top-quality load for the 2A3 triode, but its difficult to design a very good 5Kohm transformer. Once this problem is solved, you will hear music! With this working point, you will listen to more music, and for a longer time.
The other mod of the second-generation Kismet was the biasing. I chose fixed bias for this version. I think that this typology has a more correct sound, and gives more balanced results with this design. Its a little more risky, because if the connection is wrong (positive biasing) the power tubes are gone. More attention is needed.
After more than two years with the second-generation Kismet, a problem had to be solved. The 5Kohm output transformer of the Kismet 2 was a custom design of very good quality, it bevame unavailable. Since I had to find a new device, I thought it was a good time to make a few improvements to the design. As I mentioned at the beginning, the best way to reach music is through fully transformer-coupled amplification, from source to speakers. When a limited budget wont let me use transformers, I use the SRPP circuit, so I changed the first stage of the Kismet to an SRPP circuit. It requires one more double triode, but is worth the effort by far.
The double triode was chosen after the experience gained from my top-of-the-line, fully transformer-coupled preamps. I chose the 5687. Its not a very linear tube, so I do not think it is good for common cathode circuits (RC or transformer-coupled), but its very good in SRPP, since SRPP is a very linear circuit of its own. Of course, its low impedance is very important. The gain is really low, so be careful and use high gain preamplifiers with high voltage output swing.
Taking advantage of the opportunity, I also designed a new power supply. Its not really new, as I changed the values of the capacitors, nothing more, but it mean everything! Have you ever tried to use small capacitors? The smaller they are, the better, but using this approach with a power supply may cause several problems. First is noise and poor energy, but with the new working point of the output tube, the current is lower and a smaller capacitor is required. Starting from this, I tried to lower the values of all anode power supply capacitors. With small values, the bass is faster and has more harmonicsin short, there is more music. There is also less distortion, of the kind you cant measure but hear as unnatural, and a sound you can call RIGHT.
Schematic and suggestions
Only a few comments. First of all, dont waste your money using very expensive passive components. I made lots of tests, discovering that almost all very expensive parts are not balanced. They may have a few very good characteristics, but not in every aspect of audio reproduction. This means they are not good, in my opinion, and do not have the right sound. They do not have music. They are good for business only.
I use carbon resistorsstandard types 1 and 2W are good enough.
I use good quality, general-purpose, industrial electrolytic capacitors for the power supply and cathode bypass. In inexpensive amps I dont bypass electrolytic capacitors with polypropylene/polyester caps of small value. Its a critical technique with standard amps (such as the Kismet), and its not easy to use if one wants balanced results. For the capacitor between the driver and the output stage, use ERO MKT1813. For the right sound, dont use silver, but pure copper cables only, litz if possible.
As I wrote in my previous article on the Kismet amp, the RCA 2A3 is the best 2A3 I know. Its price is too high for the original purpose of the Kismet, but its up to you. A matched pair of Golden Dragon 2A3s is my suggestion. I prefer to ground the centre of the secondary winding supplying the heaters because its a cheap solution, and the 2A3 usually is not noisy. (This is the same solution I used in the Kismet 1.) If you have noise problems, use a pot.
Regarding the double triodes, the Sovtek 6922 is the choice for good gain (even if still a bit low), but NOS 5687s made in the U.S.A. (Sylvania and Philips ECG are rather common, and cheap) are a very good choice. If your preamp has high output voltage capability, please use 5687s. If you want to use 6922s, use 680 to 820ohm cathode resistors.
You probably know that SRPP may be critical about the filaments power supply. If no precaution is taken, the tube might be damaged after a few hours. Thats why I did not referred the filaments to ground, but to A. The Russian 5U4G is a must. And the output transformer? I solved my problem using a standard Tamura trans, model F-475; dont worry about the maximum DC current of the primary winding (you can set it for 45mA if you prefer). I love some Tamura transformers, and think they are simply perfect for the right sound. No further comment is needed.
Then a question was born in my mind: What preamp is now perfect for this special power amp? This is not only from the output voltage capability point of view, but above all for the right sound. The perfect solution was to design an integrated amplifier, and the Kismet 3i was born. I changed the double triode, and after a few tests decided to use the 12AV7 (or 5965). It is perfect for this purpose, in particular in SRPP. The total gain is now enough for an integrated amplifier. Of course, you can also use it for the Kismet 3 power amp version, and I suggest this solution, even if the gain will be high. I dont like high gain circuits, because they have problems, such as noise. I like "right gain" circuits, with the gain just high enough for the purpose and not higher, but in this case, the 12AV7 (5965) can be used anyway. The ALPS Black Beauty is a wonderful pot and it is perfect for this application. I made no more mods.
The Kismet 3 (and 3i) is the third and possibly the last generation of this amplifier. Everything about this amp is designed for music, and not for impressing audiophiles. I hope you enjoy it.
Dont waste your money on very expensive parts. Buy LPs and CDs! And go to many concerts. Remember that you must listen to music, and not to your system. Good listening.
Id like to thank Mr Kiyoaki Imai, my audio Master (www.audiotekne.com).
Fabio Camorani, 34, is a regular contributor to the Italian publications Costruire Hi-Fi and Fedeltà del Suono. He received a degree in electronic engineering at the University of Bologna with a thesis on OTL-OCL tube amplifiers. He works as Production Director for Air Machine fitness device, but has his own audio business, AudioNautes (www.audionautes.com). He lives in Forlì, a small town on the east coast of Italy.
Kismet 3 parts list
N.B. The schematic shows only the left channel, but the power supply is common. The components of the right channel are indicated with the sign (like R1, R2 etc.).