Posted by Wayne Parham [ 65.69.120.160 ] on March 10, 2005 at 03:42:24:

In Reply to: Re: Progress report posted by matth on March 09, 2005 at 18:17:14:

No, I still have hum. When I opened up the Stoetkit to start looking for noise, I planned to trace the source of noise with an oscilloscope, quantify it and make a plan how to reduce it. But when I found resistor R5 discolored, I got sidetracked. I found the DC levels off and eventually found capacitor C3 to be the cause. That made the voltage across R5 and C5 raise high enough that I decided to replace them even though they appeared to work. And since I must replace C3, I'm not going to replace it alone, I'm going to replace C2 as well, to retain matched sides. Just for fun, I'm swapping C1 too.

In the end, I'll replace C1, C2, C3, C4, C5, R4 and R5 on both sides to ensure each component is replaced in matched pairs. I've ordered each of these, and I expect to get them in a few days. The resistors R4 and R5 are 470 ohm, 1 watt 2% non-inductive parts made by RCA. Capacitors C4 and C5 are Solen 47uF 400v polypropylene parts. Capacitors C2 and C3 are polypropylene Kimber Kap 0.1uF 450v parts and C1 is a 0.047uF 600v polypropylene Kimber Kap.

So right now, my Stoetkit is hacked and not in its best trim. The funny thing is, it still sounds great. The damaged side has a 100uF, 100v electrolytic cap in location C5 and a 0.47uF polypropylene cap in location C3. Nothing is matched on that side, and the parts are not trimmed down to short leads. They're soldered in with full length leads, bent into position and just hanging there as a temporary fix. When I get the right parts, I'll pull the hacks off and put them back in my parts bin.

But even the noisy side still makes less hum than my Paramours, with its balance pot set for lowest noise. The Paramours have the CCS installed, which cut noise in half, but it is still prone to a bit of hum. I think I may add some additional power supply filters to the Paramours after I get done with the Stoetkit. It would be easy to get that hum down, I'll bet.

I took a moment this evening to quantify the levels of noise in three tube amps. One is the Stoetkit, one is the Paramour and another is a (really cool) amp that Frank sent made by Nelson Audio. I've been listening to it some, and it's a pretty cool amp too. More about that later.

Here is a chart of noise levels from each of the three amps:

Stoetkit           volume knob minimum (shorting input)
=====================================================================
Right side         0.12mV p-p continuous 120Hz
                   + 0.12mV   2.5 cycles 12uS damped pulse every 16.6mS
Left side          0.30mV p-p continuous 120Hz
                   + 0.25mV   2.5 cycles 12uS damped pulse every 16.6mS

Paramour           (input shorted)
=====================================================================
                   0.35mV p-p continuous 60Hz half-cycles
                   + 0.30mV 12uS damped pulse switching spike

Nelson Audio SE8   volume knob minimum (shorting input)
=====================================================================
Right side          0.20mV p-p 120Hz continuous triangle wave
Left side           0.20mV p-p 120Hz continuous triangle wave

There were a few things that stood out to me in this series of measurements.

One was how much I could raise and lower the amount of hum in the Paramour amp, and how pure the noise was 60Hz half-cycles. It has a clearly defined switching spike, much like the Stoetkit. I didn't bother to switch the time base to look at the HF spike closely, as it looked pretty much the same as the Stoetkit, which I did examine closely. The Paramour uses a dual diode rectifier, and for some reason, one cycle is filtered better. But The Paramour uses the same diodes as are used in the Stoetkit (1N4007), and that explains the similarity in the switching spike. I think I could probably reduce the noise in the Paramours pretty easily, with additional filtering.

Another thing that I noticed was the Nelson Audio amp had a very unusual triangle wave on its outputs. It was obvious that the noise had a lot of odd harmonics, and I'm not sure why. Full-wave rectifiers make a lot of harmonics, and every silicon diode is going to make switching spikes but this was something else. They make even harmonics, with higher harmonics dropping by the square of the harmonic number. So I'm not sure what formed this triangle wave shaped noise.

But my focus right now is on the Stoetkit. It's great that the quiet side has only 0.12mV noise, but the other side has double that. Maybe after the components are all matched that will come down, but I seem to recall the demo unit having slightly different hum levels between the two sides. I'm definitely looking at minutia here, because it's tiny and you can't hear the 120Hz hum unless you have horns and you stick your head in them. But I still can't help thinking if I add a 1mH LC stage in front of R1, I can reduce the hum down to an imperceptible level. And there is a threaded hole on the chassis right by R1, so It screams "coil goes here."

The 12uS switching spikes are another matter. I'm thinking that small value ceramic caps might work best at removing them. They're little 8kHz buzzing sounds, a highly damped ring at that frequency that makes 2.5 cycles. The first cycle is full, and the second is about half that, with the final half cycle being just a wiggle. I hear this tiny buzzing in a lot of amps, and it's those diodes switching open before the next one takes over.

I looked closer at the Stoetkit, and measured its power supply voltages. There is a CRC filter between the rectifier and B+ for the output section, and an additional RC filter between that and B+ for the preamp section. The preamp B+ is pretty clean DC, but the switching spike is present, about 0.7V. The output section B+ has about 2.0V p-p 120Hz ripple, and also has the 0.7V switching spike. So that's why I think a coil might reduce the 120Hz ripple, and some small value caps may reduce the switching spike. I might look at different diodes too.

That's where I'm at with the power supply noise issue. It looks to me like a pretty easy mod, but I don't think it's necessary on systems unless they have 100dB+ speakers. If that's what you're running though, I'll bet it's easy enough to add additional filtering components, maybe swap the diodes. After I get the coupling and bypass caps installed, I'll look more at adding some supply filter parts.

Replies: