Discovery #3 consists of what I ended up with as my output power supply for
my 6336A/B based OTL amp in M-60 chassis. My design objectives were: high
current capability, low noise, good speed, efficient, and without being
regulated. The only form of regulating that I think would create any gains
would be a shunt regulator. But it being less than 50% efficient makes it
prohibitive for high power supplies like this. The job of the P.S. is to
take 117 Volts AC (which swings about 330 Volts from peak to peak) and turn
it into about 140 Volts DC without any AC ripple riding on the DC. As
described in my previous article the stock M-60 transformer could not
maintain proper voltage levels for the 6336A/B tubes and their current
demands, so a 470VA output supply transformer is being used.
Since I needed four caps to get the value needed (about 10,000uf) for the
6336A/B modification, I decided to use the multiple capacitors to my
advantage. I don't like the idea of the transformer and rectifiers hitting
10,000uf. It seems like it would feel like hitting a brick wall. It felt
better if the transformer and rectifier would feed only one cap and that
cap would feed the next cap etc. etc.
I am using a triple pi filter that originally used chokes (L) and caps (C) like this: CLCLCLC. This will reduce noise at a rate of 42 dB/oct being a 7th order filter. But the 250uH @ 2.5A chokes added a slight metallic flavor to the music. Now my triple pi filter uses resistors in place of the chokes like this: CRCRCRC. This will reduce noise at a rate of 24 dB / oct now being a 4th order filter. Still not bad being that the stock supply is this: C. Which is a first order filter @ 6 dB / oct. The beauty of the Atma-Sphere amps is that they are so well balanced that they can reject a large amount of P.S. noise. So they sound great even with a simple supply. Below is a simplified schematic showing the basic idea.
With the amp biased up at 600mA (for comparison to a stock M-60), I measured the total ripple voltage across each cap. This shows that even a fraction of an ohm makes for a good filter with large capacitors. The impedance of a 2200uF cap is 0.6 ohms @ 120hz, 0.3 ohms @ 240hz, etc. The output supply ripple dropped a substantial 17.3dB from the first cap to the last cap. Keep in mind that this is mostly made up of 120Hz and the spectrum analyzer showed the harmonic components dropping at a much higher rate due to the 4th order nature of the filter.
I found a simple way to make sure that the Hexfred devices would survive the power up current surge. The Hexfreds are known to blowup from excessive current, It is quite possible that the 25 amp parts will do just fine on a stock M-60, but why risk it. With larger than stock transformers and three times the stock capacitance to charge up, I did not want to take any chances.
The inrush current limiter (CL-30) is a device with a negative temperature coefficient, so as it heats up, its resistance drops. The higher "cold" resistance limits the current to the transformers when the switch is flipped on. Then the CL-30 quickly heats up and drops resistance.
This is all fine and dandy but the M-60 kit instructions had me install the CL-30 current limiter between the AC plug and all three fuses for all three transformers. This means that the current limiter will do its job when the main switch (amber lamp) is first turned on. To me, this also means that when the output supply switch (red lamp) is turned on, the current limiter is already hot and is not able to limit the current to the output transformer. Ouch.
To test this (before I installed the Hexfreds) I placed a 0.1 ohm resistor in series with the AC side of one of the bridge rectifiers. Watching the voltage across the resistor with a scope I was able to see the peak current through the bridge rectifier. With the CL-30 in its stock location I was getting 40-45 amps on the first few peaks of the 60hz cycle. Keep in mind that I was not measuring average current during the first moments of power up, but rather the instantaneous current showing as a peak of current every 8.33 mS (120 times per second) as the voltage hit the caps. There is dead time in between the peaks as the AC cycle alternates, so the average will be much less than the peaks.
Next I connected the fuses of both the filament and driver transformers directly to the AC, so that the CL-30 current limiter is connected only to the fuse for the output supply transformer. Now I was only seeing 20-25 amps on the first few peaks. Not seeing the need to limit the current to either the filament or driver transformers, I left the CL-30 connected to only the output supply transformer fuse and installed the 25 amp Hexfreds into the output supply circuit with confidence. I do plan on installing a second CL-30 an each amp to limit the inrush current for the main power switch (filament and driver transformers). I understand that this is how the MA-1 is wired. Additional CL-30 Inrush Current Limiters are available from Mouser Electronics for $2.17 each. Order Mouser #527-CL30 at 800-346-6873.
I did not consider the stock bias/balance procedures convenient enough to be acceptable for me so I built up my amps with a totally different set up. The bias current meter resides in the middle of the power supply filter so that the 3P3T switch that I used is well away from the signal path. With the switch in the up (normal) position, the meter is out of the circuit. With the switch in the middle position the meter is connected across the negative sides of both power supplies to show balance. With the switch in the down position the meter is connected in series with one of the power supplies to show bias current. With this arrangement bias current readings are not affected by load, the meter will show accurate bias current from the speaker being disconnected to having the speaker output shorted. The bias meter can now also show the extra current demands to the power supply while music is being played with the switch in the down position.
As described in my previous article, I placed the fast blow fuses (required for 6336 type tubes) between the positive output of each output supply and the plates of each tubes phase. A single fuse in series with the speaker output will not protect excessive current flowing in the outside circle of the circlotron circuit.
It was not intentional, but there is nothing left that was original to the stock M-60 kit except the standby power switch in this tweaked out power supply. I enjoy the ease and speed of setting bias and balance. I will usually keep the amps biased down to 400 mA for background & ambiance listening and for my wife and son's use; then fully bias them up for myself when the house is empty, dark and quiet.