D1080 MkII 08 actives - power supply mod details
Posted 17th December 2010 at 04:34 AM by abraxalito
Updated 5th March 2012 at 03:54 AM by abraxalito
Updated 5th March 2012 at 03:54 AM by abraxalito
When considering how to hot rod a particular piece of electronics, my first attention always goes to power supplies and layout rather than the somewhat more popular approach of swapping out components for boutique variants. This is because so far I'm not sure that I notice the differences between audiophile grade caps (for example) and the bog standard ones. But I am sure of the differences brought about by improved grounding - to my ears these aren't subtle changes at all. So if I turn out to be dissatisfied with the sound of my layout mods, then I'll turn to tweaks on individual parts.
On examining the layouts of the two PCBs (XO layout already shown in a prior post) it turns out the amp PCB is the one with the relevant power supply components. In more detail, it has two independent supplies coming from separate windings on a standard EI-core transformer. The higher current one is unipolar, unregulated and feeding only the TDA8947s, the small-signal supply is bipolar, regulated and goes only to the XO board via a 3-core cable.
So first up - improving the regulated supply. It uses LM78/79L types which have poor HF rejection, so I normally install an RC filter on the input side to deal with this. Here I used 22R (actually two 47R paralleled) and 47uF as I had them to hand. Already there are 0.1uF on the input side, so these go in parallel with the added electrolytics. Snubbers are included across the diodes, which here are all discrete. Values I used are 470R and 470pF in series. These serve to soak up the RF emissions on diode turn-off.
Finally I hack the ground traces to provide clean feeds from a star ground to get ground references to be as quiet as possible. Once a decoupler is connected to a ground, I consider that a dirty ground, even if the power supply is regulated upstream. The regulators here have fairly high output impedance at the lowish output currents needed for a couple of quad opamps, so there's plenty of potential for cross-coupling via supplies.
The results of all the snubbers, filters and star grounding can be seen in the photo of the underside of the board.
The unregulated supply is reserved solely for the chipamps, it turns out that the negative side of this is connected to 0V of the bipolar supply. The chip amps have separate power supply grounds for the inputs and outputs, but here they've been joined. So its necessary to separate them which requires a bit of hacking with the craft knife and strapping with the ubiquitous yellow wires. Input grounds are lead off to the star point with their own dedicated connection. An integrated bridge rectifier is used, each internal diode has its own snubber.
The interconnection between the two supplies has been re-routed too, to minimise common impedances. The PCB shows the designer had some inkling of star earthing, but he didn't quite take it far enough for my liking...
On examining the layouts of the two PCBs (XO layout already shown in a prior post) it turns out the amp PCB is the one with the relevant power supply components. In more detail, it has two independent supplies coming from separate windings on a standard EI-core transformer. The higher current one is unipolar, unregulated and feeding only the TDA8947s, the small-signal supply is bipolar, regulated and goes only to the XO board via a 3-core cable.
So first up - improving the regulated supply. It uses LM78/79L types which have poor HF rejection, so I normally install an RC filter on the input side to deal with this. Here I used 22R (actually two 47R paralleled) and 47uF as I had them to hand. Already there are 0.1uF on the input side, so these go in parallel with the added electrolytics. Snubbers are included across the diodes, which here are all discrete. Values I used are 470R and 470pF in series. These serve to soak up the RF emissions on diode turn-off.
Finally I hack the ground traces to provide clean feeds from a star ground to get ground references to be as quiet as possible. Once a decoupler is connected to a ground, I consider that a dirty ground, even if the power supply is regulated upstream. The regulators here have fairly high output impedance at the lowish output currents needed for a couple of quad opamps, so there's plenty of potential for cross-coupling via supplies.
The results of all the snubbers, filters and star grounding can be seen in the photo of the underside of the board.
The unregulated supply is reserved solely for the chipamps, it turns out that the negative side of this is connected to 0V of the bipolar supply. The chip amps have separate power supply grounds for the inputs and outputs, but here they've been joined. So its necessary to separate them which requires a bit of hacking with the craft knife and strapping with the ubiquitous yellow wires. Input grounds are lead off to the star point with their own dedicated connection. An integrated bridge rectifier is used, each internal diode has its own snubber.
The interconnection between the two supplies has been re-routed too, to minimise common impedances. The PCB shows the designer had some inkling of star earthing, but he didn't quite take it far enough for my liking...
Total Comments 0
