Graham would be the man to ask Fritz considering he designed the Solo and is a member of this board
In the meantime to Quote Graham:
AC coupling?
"Well, this all stems from my distrust of the overuse of electronics to achieve what in my opinion is a simple task. If I were to have done this as a DC coupled design I would have had to servo the output. A servo is necessary to zero the output offset. Either that or you ac couple the bottom of the NFB potential divider with a cap. If you use a cap it's going to have to be an electrolytic because of its size, and electrolytics don't behave all that well without some DC across them. So rather than messing about with diodes to establish a DC voltage (diodes are noisy [and have a big temp co-efficient] in my opinion and should not be at such a sensitive part of an amp), just ac couple the whole thing and you've got half the DC rail across your electrolytics - problem solved! It does mean however, that you've now got an input cap and an output cap to worry about. But at least the output cap has half the supply voltage across it to make it behave as a linear device."
Q: So what about the input cap?
"Well, this is where it all gets horribly complex and deeply technical. I have a theory - in fact Analog Devices have a theory, both theories are very close, but I see it slightly different from them. It's all about low frequency stability. They see it as low frequency stability only in ac coupled amplifiers. I see it as low frequency stability in all amplifiers however they are coupled. My theory is that the power rails of an amp have a frequency above which they are as stiff as something that's stiff and below that they are as floppy as heck. In fact, this is the very theory behind power supply decoupling so I have a nerve claiming it's mine :-) "
"And this is where many designers don't put 2 and 2 together. The signal passing through goes down pretty low. If the frequency of the signal goes lower than the power rail's you're in trouble. In fact I have a safety margin of 10. The power rails must be stable (stiff) at 1/10th of the pass frequency (the -3dB point). How to arrange for this? Simple, ac couple the input calculating the cap value so the LF cut frequency is 10 times that of the power rail. The same should hold true in otherwise DC coupled circuits."
"So. I believe I've justified my reason for ac coupling at the input and the ground end of the NFB, but how can I justify ac coupling the output? After all damping factor is so important here."
Q: Damping factor? Isn't that something to do with braking the oscillations of a massive loudspeaker cone?
"My belief is that headphone diaphragms are that light they have virtually no flywheel effect and do not require damping."
Q: I've heard that capacitors degrade the signal, what's your opinion on that Graham?
"Well, I've already tackled the input and NFB capacitors and say that they, in my opinion, do more good than they could ever do bad. But, the output signal is big, on average 250mV big, and a good capacitor like a Starget should be able to do it justice. Anyway, that's what I fit, and the whole kit and kaboodle is ac coupled and I stand by my design. The input signal is handled by some simple polyester caps. Not any old cap, but made by Evox (I got the feeling that the "vox" bit gives a clue to the manufacturers intentions, whereas Wima holds no clues). I use Evox capacitors in my phono stages and both I and a thousand plus users believe they sound very good."
"The two yellow caps are the NFB ground coupling capacitors. They are electrolytics. They are low leakage electrolytics. They are as rare as RHS. They are epoxy resin coated (sealed). You can use Rubycon low leakage electrolytics here with the guarantee of a horrible sound. The ERCs are the only ones which work. Even then, if you turn the board over, you will see they're bypassed by Evox polypropylenes (1,000 V/uS types), because it made a difference. These ERCs are expensive."
Q: What made you decide to utilise the AD 826 Op Amp in your design?
"The choice of the 826 is down to speed and the fast settling time as you have already noted. I used to use these in a phono stage but found their noise too unpredictable, but the sound was really nice. Even in the Solo I get the odd one that's too noisy when I put them on the test gear (yes every Solo is measured for gain, bandwidth/frequency response, noise and distortion, not forgetting channel balance). The noisy ones are immediately squashed and thrown in the bin . I think the above may throw some light on the need for lengthy running-in. There is quite a lot of charging going on in this circuit, plus I believe the 826 needs a lot of burn-in too. After all, semiconductors do have capacitive junctions which need a similar time to stabilise."
Q: I'm sure there are other questions I should be asking but I've taken up enough of your time already Graham! Thanks for being so descriptive and informative and for giving headphone enthusiasts an insight into the inner workings of your design.
"I know I have not covered the whole circuit in the above description. I have not covered the transistor assisted power rail decoupling, the reason why just one value of tant works here; I have said nothing about the output stage, but suffice it to say it is class B biased to what people like to call class A at low headphone volumes. The transistors are small, but they're fast which is important to keep up with the speed of the voltage amp. Oh! And the PSU1 uses those nice 105 degree Panasonic FC caps, 4 x 1,000uF. It also uses a frame transformer. Yes, I should be using a torroid, or should I? Torroids don't hum! Well, try telling that to the torroid in my power amp. No Mike, everything is chosen on it's merit to provide the best possible sound quality."
Changes to the SOLO
ELN / Class A output stage: "The Emitter Lift Network serves a bidirectional benefit. The output stage emitters assist the op-amp during large output swings and the op-amp assists the output stage during small signals. Although Current Dumping is explained like this, the Solo is not a current dumper. Its output stage is always biased-on and its output will remain within the class A envelope even at elevated listening levels. 100,000uF simulated voltage stage power supply smoothing completely eliminates power supply feedback."
CDO: "Current Drive Output allows the Solo to drive headphones up to 1000 Ohms at realistic volumes. No matter what the headphone impedance is, the Solo output stage produces the same high output which is converted to a current and automatically adjusts to give the same power into any headphone impedance. You can swap between 32 Ohm Grados and 300 Ohm Sennheisers without having to alter the volume control."
A few porn snaps (my modded Solo 2007) but should give you the idea Fritz:
Graham is still burning the R&D candle at both ends and is currently experimenting further "I've taken out the NFB grounding capacitors completely by altering the bias to make the NFB direct coupled to ground. The offsets within the 823 make the thing work harder and the removal of the cap and its bypass for a piece of wire remove a low frequency pole."