ECN0032 (EMI and ESD protection for headset jack)

[Werner Almesberger]

Rene, we discussed using a single filter component instead of a discrete
"pi" configuration. What's your opinion ? I'm a bit concerned that we
may end up with a lot of components to place in this already crowded
area, and risk getting unexpected secondary effects.

Also, do we still need C4405 ? If I understand things correctly, it just
stands guard against general EMI badness, but since we have a real filter
now, it can only make things worse, no ? ECN0032 mentions its removal,
but the schematic still has it.

R4401 is for general DC protection ?

By the way, if we replace the pis with integrated filters, we should
have enough room to draw the various stages such that they don't
overlap, i.e., one (virtual) rectangle with the jack, one with all the
TVSs, one with all the filters, and then all the rest, including the
branch JACK_INSERT branch. That should make the whole affair look a bit
less intimidating.

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Rene Harder]

Werner Almesberger wrote:
> Rene, we discussed using a single filter component instead of a discrete
> "pi" configuration. What's your opinion ? I'm a bit concerned that we
> may end up with a lot of components to place in this already crowded
> area, and risk getting unexpected secondary effects.
>  
I totally understand your concerns those discrete pi-filters introduce a
bunch of new components into the circuit with many possible side effects.
My main idea of using discrete filters though, was to keep new
components (with different packages) as small as possible to reduce SMT
assembly cost.
Also you are more flexible with a discrete filter than with a single
component one; If we mess things up, we can just replace the caps or the
inductor and fix the problem but with a single component filter we would
need a different one and might need to change the pcb pattern as well.

However, I think we are better off using a single component filter, this
decreases complexity and risk of possible misbehavior.

> Also, do we still need C4405 ? If I understand things correctly, it just
> stands guard against general EMI badness, but since we have a real filter
> now, it can only make things worse, no ? ECN0032 mentions its removal,
> but the schematic still has it.
>  

I must have missed this one.
deleted!

> R4401 is for general DC protection ?
>  

I've no idea what the purpose of this resistor is/was. This seems to be
an ordinary pull-down resistor but for what I can't tell for sure.
Nothing makes completely sense to me.

> By the way, if we replace the pis with integrated filters, we should
> have enough room to draw the various stages such that they don't
> overlap, i.e., one (virtual) rectangle with the jack, one with all the
> TVSs, one with all the filters, and then all the rest, including the
> branch JACK_INSERT branch. That should make the whole affair look a bit
> less intimidating.
>  

Agree! Right now it looks kind of ugly, I was already thinking of moving
some sub circuits to a separate sheet.

Rene


_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Werner Almesberger]

Rene Harder wrote:
> My main idea of using discrete filters though, was to keep new
> components (with different packages) as small as possible to reduce SMT
> assembly cost.

Yes, I was hoping that we could have one "universal" filter that's
suitable for all low-frequency signals (audio, UART, etc.), given
that most of the noise lives >> 100 MHz and those low-frequency
signals are << 100 MHz.

> Also you are more flexible with a discrete filter than with a single
> component one; If we mess things up, we can just replace the caps or the
> inductor and fix the problem but with a single component filter we would
> need a different one and might need to change the pcb pattern as well.

That's true, yes. Of course, I fervently hope that we can avoid the
joy of precision-tuning those filters ;-)

> However, I think we are better off using a single component filter, this
> decreases complexity and risk of possible misbehavior.

Great !

When we discussed integrated filters, we came up with the following
components:

Murata NFL18ST (0603, "T"):
http://search.murata.co.jp/Ceramy/image/img/PDF/ENG/L0112S0119NFL18ST.pdf

Rohm MCF182CN102 (0603, "for power supply"):
http://www.rohm.com/products/databook/emi/pdf/mcf18.pdf

TDK MEM2012P101R (0805, "Pi"):
http://www.tdk.co.jp/tefe02/e9617_mem_p.pdf

Which one would be closest to the kind of "universal" filter we're
looking for ? As usual, smaller would be better :-)

> I've no idea what the purpose of this resistor is/was. This seems to be
> an ordinary pull-down resistor but for what I can't tell for sure.
> Nothing makes completely sense to me.

Are we talking about the same part ? R4401 is 220 Ohm and series, while
the pull-downs in the area (R4416, R4117) are 10 k. Looks more like a
current-limiter for problem conditions to me.

Joerg, do you remember what it does and whether we still need it ?

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Rene Harder]

Werner Almesberger wrote:

> Rene Harder wrote:
>  
>> My main idea of using discrete filters though, was to keep new
>> components (with different packages) as small as possible to reduce SMT
>> assembly cost.
>>    
>
> Yes, I was hoping that we could have one "universal" filter that's
> suitable for all low-frequency signals (audio, UART, etc.), given
> that most of the noise lives >> 100 MHz and those low-frequency
> signals are << 100 MHz.
>
>  
>> Also you are more flexible with a discrete filter than with a single
>> component one; If we mess things up, we can just replace the caps or the
>> inductor and fix the problem but with a single component filter we would
>> need a different one and might need to change the pcb pattern as well.
>>    
>
> That's true, yes. Of course, I fervently hope that we can avoid the
> joy of precision-tuning those filters ;-)
>
>  
>> However, I think we are better off using a single component filter, this
>> decreases complexity and risk of possible misbehavior.
>>    
>
> Great !
>
> When we discussed integrated filters, we came up with the following
> components:
>
> Murata NFL18ST (0603, "T"):
> http://search.murata.co.jp/Ceramy/image/img/PDF/ENG/L0112S0119NFL18ST.pdf
>
> Rohm MCF182CN102 (0603, "for power supply"):
> http://www.rohm.com/products/databook/emi/pdf/mcf18.pdf
>
> TDK MEM2012P101R (0805, "Pi"):
> http://www.tdk.co.jp/tefe02/e9617_mem_p.pdf
>
> Which one would be closest to the kind of "universal" filter we're
> looking for ? As usual, smaller would be better :-)
>
>  

I'm still in favor of the pi-filter but let me do another search. Maybe
I'll find a smaller one with similar characteristics  ;-)

>> I've no idea what the purpose of this resistor is/was. This seems to be
>> an ordinary pull-down resistor but for what I can't tell for sure.
>> Nothing makes completely sense to me.
>>    
>
> Are we talking about the same part ? R4401 is 220 Ohm and series, while
> the pull-downs in the area (R4416, R4117) are 10 k. Looks more like a
> current-limiter for problem conditions to me.
>
>  

Ups my bad. Unfortunately, I mixed up R4401 with R4404!
Yes this one looks pretty much like a current limiter.


_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Werner Almesberger]

Rene Harder wrote:
> Yes [ R4401 ] looks pretty much like a current limiter.

Joerg and I had a quick discussion on IRC:

- he agrees that R4401 is probably a current limiter to protect the
  codec's MIC1 input. Can't hurt to have it.

- he also agrees that C4405 should go. Also, R4404 can safely meet
  the same fate.

- he suggests to switch to a symmetrical configuration for the
  microphone, like in earlier GTA02 variants or in GTA01.

  The change to single-ended in GTA02v6 was an unsuccessful attempt
  to solve the buzz problem. A symmetrical design would have the
  advantage of better common-mode rejection.

  In the symmetrical design, R4301 and R4302 would be useless, so
  it's not any more complicated than the asymmetrical design. Plus,
  it would actually be a bit clearer what things do.

Regarding component choices, I looked for filter arrays which would
save even more space, e.g., Murata's NFA series. Alas, they all
seem to be difficult to source.

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Rene Harder]

Werner Almesberger wrote:

> Rene Harder wrote:
>  
>> Yes [ R4401 ] looks pretty much like a current limiter.
>>    
>
> Joerg and I had a quick discussion on IRC:
>
> - he agrees that R4401 is probably a current limiter to protect the
>   codec's MIC1 input. Can't hurt to have it.
>
> - he also agrees that C4405 should go. Also, R4404 can safely meet
>   the same fate.
>  

C4405 is already gone! I'll add the removal of R4404 to ECN0010.

> - he suggests to switch to a symmetrical configuration for the
>   microphone, like in earlier GTA02 variants or in GTA01.
>  

Okay, as discussed on IRC we'll change R4305 and R4303 to 1 kOhm. (ECN0022?)

>   The change to single-ended in GTA02v6 was an unsuccessful attempt
>   to solve the buzz problem. A symmetrical design would have the
>   advantage of better common-mode rejection.
>
>   In the symmetrical design, R4301 and R4302 would be useless, so
>   it's not any more complicated than the asymmetrical design. Plus,
>   it would actually be a bit clearer what things do.
>
> Regarding component choices, I looked for filter arrays which would
> save even more space, e.g., Murata's NFA series. Alas, they all
> seem to be difficult to source.
>
>  

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Wolfgang Spraul-3]

Werner,

> Regarding component choices, I looked for filter arrays which would
> save even more space, e.g., Murata's NFA series. Alas, they all
> seem to be difficult to source.

How do you determine that?

Wolfgang

On Thu, Oct 08, 2009 at 11:09:32AM -0300, Werner Almesberger wrote:

> Rene Harder wrote:
> > Yes [ R4401 ] looks pretty much like a current limiter.
>
> Joerg and I had a quick discussion on IRC:
>
> - he agrees that R4401 is probably a current limiter to protect the
>   codec's MIC1 input. Can't hurt to have it.
>
> - he also agrees that C4405 should go. Also, R4404 can safely meet
>   the same fate.
>
> - he suggests to switch to a symmetrical configuration for the
>   microphone, like in earlier GTA02 variants or in GTA01.
>
>   The change to single-ended in GTA02v6 was an unsuccessful attempt
>   to solve the buzz problem. A symmetrical design would have the
>   advantage of better common-mode rejection.
>
>   In the symmetrical design, R4301 and R4302 would be useless, so
>   it's not any more complicated than the asymmetrical design. Plus,
>   it would actually be a bit clearer what things do.
>
> Regarding component choices, I looked for filter arrays which would
> save even more space, e.g., Murata's NFA series. Alas, they all
> seem to be difficult to source.
>
> - Werner
>
> _______________________________________________
> gta02-core mailing list
> [hidden email]
> https://lists.openmoko.org/mailman/listinfo/gta02-core

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Werner Almesberger]

Rene Harder wrote:
> Okay, as discussed on IRC we'll change R4305 and R4303 to 1 kOhm. (ECN0022?)

Maybe a new ECN ? It's the conversion of the MIC from asymmetrical
to symmetrical. At least conceptually, an important change.

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core

[Werner Almesberger]

Wolfgang Spraul wrote:
> How do you determine that?

I search at the main western distributors whether they list them,
whether they have stock, and - in the absence of stock - their lead
time.

My usual list is: Digi-Key, Mouser, Arrow, Farnell.

Note that, for gta02-core, we're not very flexible with sourcing,
because each distributor we buy from costs something like USD 100+
for the transaction, plus additional handling overhead on our side.

So it's better to have a very short list of suppliers, and
alternative parts even at the same supplier.

- Werner

_______________________________________________
gta02-core mailing list
[hidden email]
https://lists.openmoko.org/mailman/listinfo/gta02-core