Hallo Tony,
thanks for your comments on the design.
I have some questions though (cronological order):
Other comment regarding R101 and I am not quite sure of the final arrangement of course.
Perhaps an extra pad just below the lower pad. This would allow us to cut the return track and use the extra pad for a different implementation on the return for Vref.
I got to thinking the Vref return would be modulated by the Sen output current, so a dedicated Vref return may be in order.
I am not quite sure what you mean here, although extra pads can always be included.
First of I am not an expert as you know and we have not had a ply with the circuit yet.
I think the Vref needs to be very stable.
I can see were you bring in the positive end of Vref, then trough the two dividing resistors and back to the analogue ground terminal of the Sabre output. This ground line potential will be modulated 16mA DAC output signal and hence the Vref will be modulated. If modulated enough then its not going to be good.
To this end the whole Vref circuit supply may be better if fed directly from the +tve & -tve terminals of the AVCC regulator, not sharing the ground return the the Sabre current loop. So the +tve feed is fine, but the negative return from the lower dividing resistor could be better going back on a separate trace and wire to the AVCC regulator.
Even better may be taking the Vref -tve back to the Placid -tve terminal.
This is were some trials will be useful.
One dirty way to achieve this is to put an extra pad in.
I was looking at top left R101, lower end of this resistor.
If there was an extra pad at this point we would have the option of wiring the extra pad back to Vref -tve and cutting the vertical part of trace from the existing lower pad.
I may need to do a sketch.
I also have a little worry about the physical position of the Vref capacitor, maybe a more direct and ordered approach would be useful.
For example AVCC +tve & -tve wired straight to the dividing resistors, the capacitor right on the resistor at this point.
And then the traces route Vref +tve to the gate pins.
If its tricky we may be better to have a separate board for Vref, or just a dedicated area on the main board.
The Vref return current from the dividing resistors is nearly 2mA and should be exactly constant at 1/2AVCC.
The Vref return current from the gates is practically zero.
The Sabre return current could be 16mA heavily modulated by the audio signal.
You picked up on the clue of maybe just using one quad per channel?
I don´t know. Did I?
I remember you writing something about it. Just for single-ended output, if I am right. Can´t it be just used like this, with only half of the parts populated?
Yes I think we are talking of the same and only half the board populated.
I may well be real dirty with my evaluation board and just start of with one board, two quads, two battery packs and use it as a base point.
The quad bottom left matchs the the quad heat sink and the mirror for the other side of bottom right.
Both the quads at the top are reversed upper pair to lower pair.
I realize this, is this too confusing? I guess the gate-lage of the FET is always in the middle, so can be bent either way.
No its fine you came up with a nice layout.
But if you glue up your fets and fit to the evaluation board would you then be able to pull the quads from the evaluation board and fit them to the pinky board. The fets are glued to the heat sink and partly bug wired to allow removal.
I am not quite clear in my own mind
Some were along the way we may / I may have to accept new fets and sinks will be needed when moving from evaluation board to pinky board.
Its a matter of trading costs against practicality of prating around with a layout forever and a day.
Only having to account for five legs may help your layout perhaps.
I will have a look into this. If I omit 7 of the 12 holes, maybe additional instructions are needed to bend the legs under the heatsinks correctly, to "aim" at the right spot with the connecting legs. But otherwise: more elegant, no doubt.
Yes I am about to solder up my bug legs, will post a picture when done.
Trying to aim for a leg in each corner and lots of space between but practically it need to be a compromise.
It is difficult to visualise, a case of needing the quad in your hand.
Probably not worth omitting pads unless you really need to pull a trick with your layout. Right now its looking doable.
I am just a little bit concerned about activating the charge cycles. Usually, I tend to forget that my hifi system is on. My hackernap, for example switches on through signal sensing, and switches off again after a few minutes silence.
I have to figure out a way to adopt this to the battery-charging circuit...
I am sure we will sort this out.
One way is easy if you lose lock when you have finished playing, just hook the Buffalo lock Led into the charging relays coil circuit.
As I use Arduino and mute the volume when finished to turn off the display back light I will also drive the charging relays.
Hope I am clear enough, I may need to do some sketches.
Tony
