[gta02-core] ECN0010: LDO2/LDO4 swap implemented

[Joerg Reisenweber]

[Werner Almesberger Sa  21. November 2009]:
> > *rename C3016 -> c4496 (complementary to C4406), to make mic path
> > pseudo-sym. GND of C4496 next to gnd of jk4401 
> 
> Hmm, I'm not so sure about this one. The idea is to keep the
> schematics similar to the ones of Openmoko GTA02, so that we can
> find differences more easily. Particularly the audio subsystem
> has evolved quite a bit by now, but I'd still be reluctant to
> give up on this policy.

I don't care about policy, but it showed layouters are "subconsiously" driven 
by schematics numbering and graphical layout (at least in TPE). The whole 
point is to make the mic1-path pseudo-differential and place those two caps 
close to each other, and next to JK4401
The first 2 digits seem to guide on this layout considerations


> > *suggestion: 1nF jk4401:6 to GND, to RF short HS MIC path when
> > nothing plugged in
> 
> Wouldn't MICBIAS already be sufficient to keep this path calm ?

MICBIAS is the line where RF is creeping in to WM8753 obviously. I don't want 
to use this line to scoop RF away from JK4401, I'd rather prefer it never 
enters that trace

> 
> > *r4303, c4306 buzzfix not implemented ... ?
> 
> Why fit your hens with full-body armor if you can prevent the fox from
> getting into the henhouse in the first place ? ;-) The design is still
> buzzfix-ready (see ECN0019), so if the EMI filters go the way of other
> plans of men and mice, we can implement the fix easily.

OK :-D Pleased to learn you remember my words of so long time ago.


> >   thought we go make that path symmetric like formerly done
> >   anyway, so r4303 mustn't 0R
> 
> Indeed. What happened with that ?
> 
> >   c4307 is a nonsense, even to have pads for it. Remove C4307 footprints
> 
> So you're saying RF noise on MICx per se is okay, as long as it doesn't
> creep into MICBIAS ?

C4307 is a 100uF footprint, to filter audio frequency ripple. No use whatever 
to do that at the location of C4307. We got C4301 for the purpose you 
suggest.


> >   we should replace (A)-R4305-(B) by (A)-R4305-(X), (X)-C4306_NC-(GND),
> > (X)-R4399_0R-(B)
> >   move C4303, C4304, R4305 next to wm8753, behind B3005
> 
> Ah, push the choke towards the mic and reverse the RC filter for the
> buzz fix, with the option of reversing it again. Well, let's think
> about this after considering the fate of C4307, above.

Hmm, go on. C4307 already dead ;-)


> > *make B3005 (and B4105, B3004) look like decent C-L-C filters, like e.g 
B4419
> 
> Not sure what you mean. Are you saying the symbol, the visual
> arrangement, or the circuit is flawed ?

circuit of a decent filter is C-L-C, or 
<input>-(A), (A)-C-<GND>; (A)-L-(B); (B)-C-<GND>, (B)-<output>
all those bifilar chokes are missing the Cs of one side. I sugest to make one 
of those blocks consist of one bifilar choke, plus 2 times 3 C, 3 on each end 
of choke


> 
> > *kill R3004, R3005 finally!
> 
> Okay, that will be ECN0042, "Assorted AUDIO cleanup".
> 
> > *c3017 -> 1uF ?
> 
> I have to admit that the ACOP-ACIN path is a bit of a mystery to me.
> Is there any description of what it actually does ? A search in the
> data sheet didn't help :-(

It simply creates an audio path. You could insert a DSP to the loop here, alas 
we don't have any. So it's just another joint in the audio, which needs a 
decent coupling C


> > *c3003 AND c3008??? Seems we don't need both
> >  c3002 same
> 
> Looks like the usual HF/LF pair, with the 100 nF cap right next to
> the chip and the much larger 10 uF cap at some distance.

Looks like the 10uF is double. If we really need that capacity, then make that 
ONE 22uF instead of 2 times 10uF



> >  do we need B3001??
> 
> I'm not sure. Neither side of the bead should be infested with RF noise.
> If in doubt, I'd keep it.

Ot's plain nonsense. Probably heritage from a time where that supply was from 
IO_3V3. Now that it has a dedicated LDO, we for sure don't need and even 
don't want any choke in there

 

> >  esp if we kill B3001, then c3001 AND c3005 also one of both is senseless
> 
> Agreed.

See 10 lines above. Same applies for all those double C >500nF


 
> > *suggestion: place 100pF RF suppressor C parallel to C3014, C3012 and 
C3013
> > -  we 
> >   learned RF creeps in here and causes buzz. the uF-Cs are not sufficient
> > for RF
> 
> This sounds like a good idea. Perhaps ECN0019 is a good place for
> this, even if it isn't part of the GTA02 buzz fix.
> 
> Thanks a lot !


Welcome :-)
Though almost all of that is just a second call - as you realized


/j
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
Url : http://lists.openmoko.org/pipermail/gta02-core/attachments/20091121/d7995682/attachment.pgp