While we are on the subject of my Dynahi, I'll talk a little about the wiring and grounding scheme I used. Since the PSU is in a separate chassis than the amp, it offers some advantages and also a slight complication.The advantages are:
1. Increased distance between the PSU power transformer and the amp circuitry means that there is minimum of hum induction.
2. Separate cases provide an opportnuity to optmize the chassis grounding.
The first advantage is obvious and I have been advocating this arrangement for some time now, so there is no need to elaborate.
The second one is interesting from a performance and safety point of view. In a single-chassis scheme, one must choose whether to connect the chassis to either AC earth ground, or to signal ground, or to both. Since AC mains voltage is present in the chassis, for safety reasons it is a good idea to tie the chassis to AC earth just in case a live wire becomes loose and touches the chassis, which could in turn lead to a shock hazard. It would also be a good idea to tie the chassis to the signal ground for RFI shielding.
But doing both can lead to a ground loop (if another component in the audio chain also has the AC earth connected to signal ground via the case, then there are now multiple paths for the ground), which then causes hum and noise.
The usual solution is the use of a "ground breaker" circuit, often with a resistor/capacitor parallel combination, or sometimes with back-to-back diodes. The chassis is tied directly to AC earth ground but signal ground is connected to the chassis through the ground breaker. This solution works but is not always optimal for the best noise performance, because AC earth ground is itself noisy, and such noise can leak into signal ground through the ground breaker, polluting the audio signal reference. I've observed this phenonmenon while testing and experimenting with some amps.
The two chassis scheme offers the best of both worlds. In the PSU case where live AC mains voltage exists, the AC earth ground is tied to the chassis for safety. Since the PSU case has no audio circuitry, there is no need to tie signal ground (or power supply ground, which in the dynahi is electrically the same node) to the chassis. In the amp case, there is no live AC mains, so its chassis need not be tied to AC earth, and can be tied to signal ground for shielding. No ground breaker is required anywhere, and there is no opportunity for AC earth noise to leak into signal ground. Presto! The best of both worlds.
The only thing to be mindful of, is not to allow the two chassis' metal parts to touch each other (not that there is any danger in that for this amp, but the noise benefits of isolating the two grounds is lost when the two chassis touch). In my build, the aluminum bottom plates and some screw heads are the only places where the chassis metal is electrically accessible from the outside (unless the heatsink fins' anodizing is scratched). The front panels are oversized and made of plastic, and rubber feet on the bottom of the chassis prevent the two cases from electrically contacting each other under any normal placement circumstances.
Now, the complication and solution.
As many of you are aware, I have been an advocate of a three-channel active ground topology for headphone amps (e.g., M³, PPA, Pimeta). In that topology, the ground channel amplifier sources or sinks the return current from the transducers, which would otherwise have been dumped into signal ground or power supply ground. This shifts responsibility for the high current reactive load of the headphones from signal ground to the tightly regulated power supply rails, thus removing the primary source of signal ground contamination. This is not just speculation; the benefits are real and measurable (especially with low-impedance loads where the output current is higher). It is most readily observed in improved stereo crosstalk with static test signals. With dynamic signals and reactive loads where the output voltage and currents are not always in phase, the benefits extend beyond just improved stereo crosstalk.
The Dynahi is not a three-channel design and Kevin Gilmore is apparently not a fan of that topology. Nevertheless, the importance of minimizing signal ground contamination is there. On the Dynahi amp PCB, the ground plane is the signal ground. It would be bad to connect the headphone jack ground return to this point because that would cause the return currents to modulate the signal ground (because the ground plane, terminal blocks and wiring all contribute to impedance, and when there is a current flowing through non-zero impedance, voltage is developed across that impedance).
Indeed, notice that Kevin's amp PCB layout provides a pad for the output, but no pad for the ground return. It is intended that the headphone jack ground be wired directly to the PSU PCB (which has extra ground terminal block connections for this purpose). This doesn't completely eliminate ground contamination (because, after all, the two ground are still tied together at the PSU), but short of making a three-channel Dynahi this is the best one could do.
Side note -- Dynalo builders: the same care should be exercised with respect to the output ground. The Dynalo rev. C PCB actually has pads for the output ground return but it simply connects to the ground plane (signal ground). I suggest that the headphone jack ground return not to be connected here, but wired to the PSU ground instead. This is easy to do with something like the dynahi PSU or two TREADs, but hard to do when an Elpac wallwart is used.
Back to my Dynahi. In a two-chassis configuration, this means that the headphone jack ground return should actually be carried over the umbilical cable into the PSU case and connected to the PSU board. The connectors and cable I used between the two cases is a heavy-duty locking Amphenol connector with four circuits, which are used for V+, V-, PSU-G and OG respectively. The PSU-G and OG are both connected to the same ground at the PSU board, but PSU-G carries the low-current reference for the amplifier boards while OG is for the headphone jacks' ground return.
This is an example of a level of attention-to-detail that one must consider when the aim is to squeeze every bit of performance possible out of an existing design. Hope this long-winded discussion would make you think and help you get the best out of your build too.
posted 02-20-2006 04:06 AM CST (US) 
