Modding the Melos SHA-1 Headphone Amplifier

 It was Corey Greenberg’s Stereophile review of the Melos and hearing an all Melos system at my local dealer that first piqued my interest. Indeed, a desire in the early 90’s to own one of these is part of what got me into headphones in the first place. At the time, unable to come up with the cash, I bought some Grado SR-80’s and a Headroom Little instead. I finally bought my Melos in 2003. But, it has been sitting unused in a closet for a number of years. 



The Melos has some good things about it. But, both in terms of parts quality and in terms of design, it has some decided issues. Quite frankly, it is a bit of a wonder that these things work at all, and as they age, they do not necessarily do so gracefully. So, it seemed like doing a few mods was a good idea.

I started by doing the standard Melos mods -- replacing all direct signal path caps with Sonicaps.

Additionally, the final HT power supply cap was replaced with a Blackgate. The low voltage power supply caps were also replaced -- pre-regulator caps are Nichicon Muse bypassed by Wima films and the post regulator caps are Nichicon Muse Fine Gold bypassed by Sonicaps. I also rerouted the input signal wires to not run directly beneath the circuit board.


After doing this, while the amp did indeed sound better, it was still noisy. This turns out to be from two main sources. The first is that the power supply in the high voltage section is terrible, and the second is that it wasn’t working right, anyway. But, first the terribleness …

The Melos high voltage power supply consists of some solid state diodes and a p filter formed by two capacitors and a resistor. Now, I have no qualm with solid state diodes (I often use tube rectifiers, but this is primarily for the slow startup characteristics, not because I think they sound better – they don’t). However, there are diodes, and then there are diodes. The Melos uses, so far as I can tell, some off the shelf cheapies that can withstand the voltage. The problem with this is that bad diodes impart switching noise into the power supply that does not sound good.

Second, the pi filter consists of a 470u cap, a 150R resistor, and a 100u cap. As can be seen, this arrangement results in both high ripple (over 9mV), as well as ripple containing all sorts of high frequency harmonics.

The usual solution to this sort of thing is to add a choke. However, while adding a small choke reduces the ripple to a reasonable level, due to the capacitor arrangement, adding a choke here does not really do much to solve the sharp peaks. 

Moreover, also due in large part to the backwards arrangement of the caps (typically you want the caps to increase in a high voltage supply), the addition of the choke causes ringing.  

After futzing around with the original board, the solution here was to just tear the whole mess out and redo it. To this end, I first used super low switching noise avalanche diodes. These have about 25db less switching noise than standard rectifiers due to what are essentially built-in snubber caps. Additionally, I reduced the cap size (I find large caps in tube power supplies tend to make things muddy) and added a choke. I also added an extra filtering stage to make up for the smaller caps. While this is still not perfect, it is way better – the ripple is smooth and small – more than an order of magnitude smaller than it was before. 

… and it doesn’t ring. 

The reason this is all so important is that the regulator in the Melos is just awful. It is a TI TE783 which is a three pin regulator similar to the LM317 except for higher voltages. While the datasheet suggests it rejects ~60db if ripple, this turns out to be pretty optimistic.

Looking more closely, one can see that at the voltages where the Melos runs (~106V) the ripple rejection is more like 50db. Moreover, even this spec is misleading as it is only at low frequency. As frequency increases, it drops precipitously. And it rings quite a bit.

In mine, this combined with the ringing choke from before the rebuilt pre-regulator filter created an oscillator.

However, without the choke, all that nasty HF garbage and switching noise was going right past the regulator.

At any rate, this is a part from the 1980’s, and even simple new three leg regs seem to work better. Additionally, it only takes a few parts to make a better supply. For this, I built a little regulator board using a Supratex LR8 regulator, a pass transistor, and importantly a cap from adjust to ground that the TE783 does not allow, and that the Melos board had no spot for. The Supertex part, as can be seen, does perform a bit better, and the additional parts help even more.  

  
So, here is what this all looked like. 

Soundwise, bass is better, more punchy, more there, and the amp is way quieter with all the remaining noise being tube noise. Most interestingly, it is a lot less sensitive to tube rolling. My best guess here is that the Melos required super tubes in order to cut through all the noise and cruft. Now, even the cheapies work well. Somewhat unfortunately, the Melos does use 6922’s which are notoriously noisy, so the best bet now is tracking down quiet tubes, not necessarily fancy ones.

Not one to leave well enough alone, I thought I'd tinker a little more. The obvious next thing to do was to swap out the last PS cap. I had used a Blackgate Std, but these aren't my favorite caps, and I figured a film cap would be better. To that end, I removed the Blackgate and added a big Motor Run film cap. These are just big polypropylene caps dipped in oil of some sort.

This improved clarity a little, but unfortunately it also lost some bass. The usual solution to not enough bass is bigger caps. So, I got a bigger cap.