11 April 2006

24-Volt Aikido amplifier
The 6Gm8/6N27P/ECC86 dual triode was designed to be used as an RF amplifier and as a self-oscillator mixer in a car radio, back when car radios held tubes. Unlike most triodes, it works quite well with only 12 volts on its plate. Additionally, like the 6DJ8/6922 this little triode stows a grid frame, making it doubly rare. Fortunately, it shares the same pinout as the 6DJ8/6922, so using it with the 9-pin Aikido PCB is no problem.

6GM8 Specifications
   Heater Voltage                                                6.3V
   Heater Current                                                330mA
   Maximum Plate Voltage                                  30V
   Maximum Plate Dissipation                              0.6W
   Maximum Cathode Current                             20mA
   Maximum Grid Resistor                                  1M
   Maximum Cathode-to-heater Voltage             30V
   Maximum Cathode-to-heater Resistance         20k
   Amplification Factor                                        14
   Transconductance                                           2.4mA/V
   Plate Resistance                                              5800 Ohms

At first, I planned to use two power supplies for the 6GM8-based Aikido: one for the heaters and one for the Aikido amplifier’s B+. Thus, I would need to supply 24 volts and 6.3 volts or 12.6 volts. This bothered me, as all four heaters in series would require 25.2 volts, which would also work as the B+ voltage, so why not one power supply? Surely, one power supply voltage is easier to accomplish than two. However, the 9-pin Aikido PCB only allows the four tubes’ heaters to be wired in parallel or two heaters in series, but not all heaters in series. Or does it? After staring at the heater schematic, I saw how the PCB could be used without any cutting of traces, although a few import details must be observed. Below is the heater schematic.

Note that two of the heaters are already in series, so all that remains is to place these two strings in series. Jumper J1 does just that. However, we can no longer use the two heater power supply pads, but jumper J4’s pads instead to connect the 25.2-volt power supply. VERY IMPORTANT: two of the heater shunting capacitors must be reversed from the printed orientation. Capacitors, C7 and C10, must be flipped around, positive and negative leads reversed, so that the electrolytic capacitors are not damaged by seeing a reverse voltage. Remember, only 1 volt of reversed voltage is enough to damage an electrolytic capacitor, possibly rupturing and releasing noxious chemicals (in fact, tantalum capacitors can go off like little hand grenades when their polarity is reversed).

Now that the heaters can work off the B+ voltage, the rest of the circuit can be fleshed out. First of all, I plan on using a 24-volt power supply, not a 25.2 volt one. Why? Two reasons: I own four 24-volt power supplies, two regulated linear and two switching. Second, the tubes will last slightly longer with 6 volts than with 6.3 volts across their heaters.

A 24-volt B+ voltage means that each triode will have 12 volts (minus the voltage drop across its cathode resistor) to play with. Looking at the plate curve below reveals that a cathode resistor value between 160 to 200 ohms will set an idle current of 2mA. The value I plan on using is 187 ohms (again because I have them).

Now 2mA is not a lot of current. It is about one fifth of the minimum I like to use in a line amplifier. Nevertheless, as long as short, low-capacitance interconnects are driven, the bandwidth should not suffer too greatly. (Driving capacitance at fast slew rates requires current. The faster you want to charge a capacitance, the more current you will need. A low-output impedance is nice, but without the ability to deliver current into a capacitance-laden load, it is not enough. In other words, although a line amplifier may not be voltage limit, it may current limited.)

The total gain of a 6GM8-based Aikido line amplifier will be close to +15dB, which should work handsomely in most systems. The output impedance will be less than 600 ohms. I cannot measure the distortion, as I have not assembled my unit yet. However, those who have used the 6GM8 before tell me that it a clean-sounding triode.

All that remains is to calculate the Aikido power-supply noise voltage divider resistor values. The formula is R16 = R15[(mu + 2)/(mu -2)]. Thus, with an ECC86 and resistor R15 equals 100k, R16 equals 129k.

24-Volt power supplies
After dealing with 400-volt power supplies, it is a joyful relief to work with non-lethal voltages. However, we must address a few important issues. For example, although we do need much voltage, the heaters add a heavy current burden on the power supply. The heater string requires 330mA and the four tubes require a total of 8mA, for a grand total of 338mA or (rounding up) 350mA. So, 0.35A against 24V equals 8.4W of dissipation.

A simple non-regulated power supply can be built from an 18Vac power transformer, a diode bridge, and a few capacitors. It just might sound good as well.

However, I would prefer to be a bit more nervous. For although the Aikido boasts an excellent PSSR figure, I don’t want to tax it any more than necessary. Even if the power supply isn’t regulated, a few small tricks will deliver big sonic gains. For example, chokes perform wonders in stripping away power supply blemishes. I have found that just about any choke, of any inductance or DCR, is better than not using a choke in a power supply, even if that power supply terminates in a voltage regulator. One inductor and one extra capacitor added to the simple power supply will make a big difference in performance. In fact, a bigger difference than might be expect in normal tube gear. Why? Inductors work best working into a dead short. As the terminating resistance increases, the inductor loses effectiveness, just the opposite of a capacitor. Fortunately, for this inductor-based power supply, the heater string represents a 76-ohm resistive load, whereas the tubes alone represent a 3k load. Therefore, what started out as a liability (having to power the heater string) becomes an advantage to the inductor-filled power supply.

Because the current draw is so high, the inductor’s DCR becomes an import circuit element, as the voltage drop across the inductor will steal a much larger percentage of the available B+ voltage than it would in the normal tube line amplifier, which might only draw 20mA. In other words, a low DCR is critical. I would place a DCR limit of 10 ohms, as a 10-ohm DCR will displace 3.4V, which will demand a 22Vac transformer winding to yield 24V for the B+.

Which rectifiers should be used? I recommend ultra-fast rectifiers, such as the popular HEXFREDs or the unpopular Schottky diodes. (Developed by International Rectifier, in the 1970s, “FRED” stands for Fast Recovery Epitaxial Diode, thus the trade name “HEXFRED.” Today, manufacturers include Harris, International Rectifier, IXYS and others. Tube folk do not know about Schottky diodes because, until recently, it was not possible to buy a high-voltage Schottky diode. But these fast rectifiers also make for cleaner power supplies.) IXYS DSEI8-06A rectifiers cost less than a dollar, so no one will have to take out a loan. The worse choice, other than a WE275 rectifier (too much current draw), is the cheap and ubiquitous 1N4001. This rectifier works well in many non-audio applications, but it spurs too much switching-induced noise into an audio power supply. Still I know that many readers will opt for them, as they already own several dozen. If you insist on using them, then build the circuit below. Paradoxically, the added capacitors actually slow the diodes, while the 0.22-ohm resistors help soften the transitions between conduction and non-conduction. And the choke filters away the ripple.

Regulated 24 volts
A regulated power supply offers a low-noise, stable, and accurate B+ voltage. Furthermore, a 24-volt regulator is a breeze to design. I can imagine a few readers raising their eyebrows at the inclusion of a choke in a regulator power supply, but I stand by it. Good as the LM317 is, its PSRR figure is not infinite, and, worse, it falls off with increasing frequency, so the choke add quite a bit of improvement. The idle current through the LM317 without a load is set at 10mA by the top resistor (125 ohms) and the bottom resistor (2,260 ohms) sets the 24 volts of output. The three 1N4001 rectifiers are there to protect the regulator.

Next time
The next blog entry will cover the Aikido headphone amplifiers.

07 April 2006

PCB update
I have to admit, the demand for Aikido PCB boards overwhelmed me. Considering that at least three other fellows had been selling Aikido PCBs for many months before me, I expected the demand to be a little cooler—or, as they say in marketing, that "all the cream had been had." On the other hand, it makes perfect sense that the Aikido boards should be in big demand, as the Aikido amplifier is a great little circuit: complex enough to yield superb sonics, but not so complex that it does not get built. (Actually, having a PCB makes even complex circuits something of a breeze.) Add to this the obvious high quality of the boards (they are not the typical thin, cheesy, off-colored, naked [without solder mask], ineptly laid-out boards that give PCBs a bad name), and my not having ordered more the first time around looks even more foolish. The high demand for the three-switch stepped attenuator kit also surprised me. I guess I am not alone in hating potentiometers, as all of those PCBs sold out as well. On Friday I picked up the new batch of Aikido and attenuator PCBs (four times more boards than the first run), which is great for all those who still want boards.

Right now, I am working on getting some chassis built for the Aikido and a suitable power supply for it as well. I have a good source for front panels and knobs, but not for the simple clamshell enclosure I envision. The power supply is an interesting challenge, as it must be worthy of (and as clever and as elegant as) the Aikido amplifier.

For a lot more information on the boards, follow the link to the TCJ hardware page of the GlassWare Yahoo! store.

Small Aikido schematic page typo
Thanks to Aikido builder Sergey: he spotted a mistake in the schematic page of the Aikido PCB user’s guide. I had transposed the recommended tube types, confusing input for output tubes. I have updated the downloadable PDFs. The underlying premise is that the input tubes, V2 & V3 on the PCB, should not be as robust as the output tubes, V1 & V4. Common sense tells us that a 6H30 should not drive a 12AX7, or for octal fans, a 6BX7 should not drive a 6SL7. Here are my recommendations:

    (9-Pin outputs) V1, V4

6AQ8, 6BQ7, 6BS7, 6DJ8, 6GC7, 6GM8, 6H30, 6FQ7, 6N1P, 6N27P, 12AT7, 12AU7, 12BH7, 12DJ8, 12FQ7, 5963, 5965, ECC81, ECC82, ECC83, ECC86, ECC88

    (9-Pin inputs) V2, V3

6AQ8, 6BQ7, 6BS7, 6DJ8, 6GC7, 6GM8, 6FQ7, 6N1P, 6N27P, 12AT7, 12AV7, 12AU7, 12AX7, 12BZ7, 12DJ8, 12FQ7, 5751, 5963, 5695, 6072, ECC86, ECC88

    (Octal outputs) V1, V4

6BL7, 6BX7, 6SN7, 12SN7, 12SX7, 5692, 6080, B65, ECC32

    (Octal inputs) V2, V3

6SL7, 6SN7, 12SL7, 12SN7, 12SX7, 5691, 5692, B65, ECC32

Still, who knows? Maybe a 12BH7 driving a 12AT7 sounds great in your system. The whole point behind the Aikido PCBs is flexibility.

Major Aikido schematic page typo
I caught this next error all by myself. The Aikido amplifier uses a two-resistor voltage divider to sample a portion of the power supply noise to be injected into the bottom output tube’s grid, which causes the bottom to buck the power supply noise at the output. The ratio is equal to 1/mu + 2. Thus, the bottom resistor must equal top resistor times (mu + 2)/(mu –2). On the schematic page, the two resistors were reversed in the part value listing. In other words, in the correct version, resistor R15 equals 100k and R16 equals R15 x (mu + 2)/(mu –2). For example, with 6SN7s as output tubes, R15 = 100k, R16 = 122K, with the stock value of 121k being close enough.

To download an updated PDF of the instruction sheets for Rev 2 PCBs, click below:

     Octal Aikido PCB PDF
     Nine-Pin Aikido PCB PDF
     Three-Switch Stepped Attenuator PCB PDF

Octal vs 9-pin tubes
I have been asked which is better, octals or 9-pins. Well, my first impulse is to answer “better for what?” I prefer not to play the component partisan. Nonetheless, I can say that usually (to my ears, at least) the octals sound bigger and weightier than 9-pin tubes. And I believe that many would say that the best 9-pin tubes sound brighter or more detailed than their fatter cousins. However, there is enough variation between tube types that I am sure that you could find counter examples. So, what might be a fair shootout would be to compare an Aikido built with 6FQ7s/6CG7s vs one built with 6SN7s/12SX7s, using identical resistor and capacitor values, as the 6FQ7 was advertised as being an exact 9-pin version of the 6SN7.

Now, let me skate onto the thinner ice. I really like the sound of old octals over newer octals, say octals from the early 1940s over the late 1950s. Why? My guess is that in the old days they had to use a fatter grid wire and eventually they figured out how to use thinner wire, which allowed making smaller tubes, but to the disadvantage of the tube’s sound. Why would fat grid wire be better? Theoretically, it wouldn’t, but, practically, it might prove much less microphonic. I would like to build a test setup that amplified the internal vibrations of a tube under test, much as if it were a microphone. Then we could plainly hear the tube’s mechanical overlay.

Aikido hybrid power amplifiers
The last blog generated some good e-mail concerning building a power amplifier that uses the Aikido as its front end. (One friend asked me why I cannot just leave the Aikido as a line amplifier; why do want to play with it so? How little does he know me.) The hybrid idea has a lot of readers tickled. Which make sense, as a powerful SE hybrid amplifier can be built for much less than the cost of one good 300B output tube. Besides saving money, such an amplifier might prove to be a livable compromise. For example, if your loudspeakers just have to have at least 50 watts to make them move (ribbon and electrostatic loudspeakers for example), then no 3-watt amplifier, no matter how sweet sounding or gorgeous looking, is going to work in your system.

I have been asked which is better: MOSFET or transistor output stages. My answer is that I prefer MOSFETs, not just for their more forgiving sound, but because they are much more reliable and easier to bias correctly. Unfortunately, they are much more expensive and harder to get (well, at least the best lateral MOSFETs are).

Still, even the best MOSFETs, in the absence of a global feedback loop, are not as clean sounding as the average power tube in a feedback-free amplifier. As soon as I typed the previous sentence, I began to argue against it in my mind. What is missing is the qualifier “typically,” as a single-ended MOSFET amplifier can sound just magical. Unfortunately, it is so easy to become power drunk and design for the maximum voltage output, not the maximum refinement in output. I know, as I have fallen into that trap many times: “…but 30 watts have to sound better than 15 watts.” After all, isn't “more is better” the American national mantra; why else would we have invented the 64-oz cup of soda. In other words, just because it is so easy to get over 200 watts from solid-state devices, be happy if you can make a great-sounding 50-watt hybrid amplifier.

Therefore, my advice is to design relatively small; to do it elegantly but sneakily; to experiment wildly; and, most importantly, to begin cheaply. Why cheaply? Even if you have cash to burn, there is something aesthetically repulsive about buying a $2,000 power cord or $1,000 solid-silver binding posts or $700 coupling capacitors for an amplifier that sits on a Bud box (especially for an amplifier that all too often is—in terms of intellectual competency—downright mediocre, if not worse). Like the ridiculously jacked-up pickups that short men like to drive, a smoldering hot credit card betrays a towering inferiority complex.

Besides, I have seen too many DIYers paint themselves into a corner by being forced to use some disgustingly expensive parts. For example, a reader once wrote to tell me that his preamplifier sounded terrible and could I help. When I saw his schematic, I knew instantly why the sound was off: instead of 1k cathode resistors, he had used 20k. I asked how he had come up with such odd resistor values and his answer was that he had many tantalum resistors, but that was the lowest value he owned, so he used them. Now, I bet of a third of you are thinking,“What a moron,” while another third are thinking,“What’s wrong with that? That’s what I would do.” And the last third are thinking, “Tantalum resistors? Surely, he must mean tantalum capacitors.”

Once your project is built and tested, then discern its single greatest failing.  Take that information and modify the operating points, then the topology, if need be; finally, upgrade the offending parts as needed. Imagine how full of pride you would be if your $200 line amplifier, with nickel-plated brass RCA jacks, $1 coupling capacitors, and Radio Shack carbon-film resistors beat the pants off your friend’s $2,000 commercially-made line amplifier. On ther other hand, if your line amplifier cost $3,000 to make, then you would have to back off the smug attitude quite a bit.

Bias schematic typo
Reader Per from Denmark, points out that the biasing arrangement shown below is intrinsically dangerous, as adjusting the potentiometer too far will blow the output stage, as too great a bias voltage will be presented to the output devices.

Well, maybe or maybe not, as the voltage division might not exceed the safe limits of the MOSFETs’ operation. With a transistor output stage, however, I agree wholeheartedly. Below is a safer version.

Good design

Aikido hybrid SE transformer-coupled power amplifier
Let's stretch a bit. The idea here is to use high-voltage MOSFETs as output devices, directly coupled to the Aikido gain and buffer stage, and transformer-coupled at the loudspeaker connection. With such an arrangement, both tubes and MOSFETs share the same power supply, which makes its construction much easier (and allows the use of tube rectifiers). In addition, the output transformer will limit the harshness at clipping, as the transformer limits the high-frequency bandwidth, where much of the clipped signal’s energy resides.

Of course, we have not altogether eliminated the dreaded coupling capacitor, as the output transformer must be capacitor coupled to the MOSFET output. Moreover, the output transformer adds its own limitations into the mix. Still, this circuit might prove to be a great little amplifier. I would start cheap, by using a large-valued Solen or Dayton polypropylene coupling capacitor and a cheap 70V, 15W, audio power distribution transformer. Such an output transformer would reflect about 600 ohms back to the MOSFET output device (this value is a guestimate on my part, based on mathematically working backwards from specs, as Edcor does not specify the transformer’s winding ratio). Alternatively, 600 ohm to 8 ohm output transformers do exist.

Assuming that all the idle current can be delivered to the transformer’s primary, the power output would equal 12 watts. Not bad compared to $3,000, 2.5 watts, 2A3-based single-ended amplifiers, but comparatively grim considering that the output stage must dissipate 60 watts at idle. Note that 60 over 12 equals 20% efficiency, not the potential 25%. In order to achieve 15 watts of output, the primary impedance would have to equal 750 ohms and the tubes and MOSFETs would have to be supernaturally perfect.

Next time
I wanted to post much, much more today--alas it wasn't meant to be. The next blog entry will cover the 6GM8-populated Aikido line amplifier; I promise.

//JRB