TPA3110D2, strange problem when power re-cycle

Hi, Erik
This is a strange behavior. Are this phenomenon can be repeated in other machine? And both two TPA3110D2 on your board has the same behavior?

Hi Peter


Thanx for prompt feed-back...and now...a long story...

Yes we have this problem on all 10 boards/units made so far. We have in the past, made 25 boards without the LC filter which we have introduced on this version to cut away the 345KHz switch noise ( can be found in another post made some moths ago...you helped us on that one too)

So it seems that the problem is a outcome of the LC filter which of course load our power rail more than before. In our next rev. (04) we will change the coils to 22uH to reduce the load ( current drawn)

1.st we added 2 x 470uF , one on each TPA3110 from L47 to ground. We found some ripple on the 24V and C186 became too hot. After 2 x 470uF was introduced...C186 (low ESR type) back to a OK level on temperature. The 470uF are mounted on L47 (PVCCL side) and L11(PVCCL side) and to GNDA

This was made on all 10 units.

This did however (so we learned) not help on all units, unfortunately, our time schedule did not allow us to be fully focused on in depth testing all 10 units.

We hooked up 4 units ..and they all had the problem with the FAULT/SD problem. It can be noticed as a "click" sound going on. All voltages are up and running, so the "click" noise must be from a coil..we have not been able to identify it..
(click sound is due to TPA's going in to FAULT and cleared by SD..and over again))

With a speaker attached, this "click" noise id also easily heard in the speaker..kind of noise "click" with same duration as the "click" from the on-board power rail.

The "click" duration can be anything from 10-25 times or continue..unless the reset of the ALC622 will "clear" it. So if no boot of the platform..then in some cases the "click" will continue forever.

On these 4 units plus the 4 others we made a extra modification, we removed L28 and L26, and replaced by 0 ohm resistors, this seemed to do the trick..no click sound.
So we rushed them through with a short test to verify that all audio was still operational (boot win7, and playback music on the 4ohm speakers attached during testing)

Our customer still had 2 units (from our 1.st initial modification) only modified with 2 x470uF. Plan was to modify L28 and L26 on these 2 as well.

So we started out with one (just to be sure that what we modified yesterday in a rush still was a "final" solution to the problem)

BUT..sadly 1.st one modified today does still click..mostly 5-25 times and then it stops, so it stops before OS via ACL662 will "reset" the "click" problem. On a normal booted system our current is apprx 690mA (including N2600 Qseven module) When the "click" is going on the current flow jumps up and down from 550mA and close to 620mA (read out made on the Lab PSU powering the platform during test) So it never reaches steady 690mA.

OK..what if we lower the input voltage 18V??...well it turns out..no matter what we do..we cannot provoke the "click" sound (due to FAULT /SD so kick in...) on the same unit which clearly has problems on 24V DC.

On the last out the 2 platforms from today..with only 2 x470uF modification....we cannot make it go "click"
On this one we found a earlier modification we have tried out earlier. The 5VA was modified a bit to avoid N3 getting to hot.
So the drop resistor R19 was changed to 2 x 270 ohm 1206's i parallel. This could also maybe explain why this one is OK.. we wonder if the overall problem we see, is a "current flow limit" which sometime makes the TPA 3110 entering FAULT.

So the problem with the FAULT/SD is not 100% identical on all 10 units, some er worse than others..one seems to be worse than the rest.

The one that seemed the worst..this one got the R19 modification as well...and now it has so far been impossible to get this into the "clcik" problem again (TPA's going FAULT /SD) The one working OK on 18V DC in.

it's a long explanation...we admit we have not been fully focused in our trouble-shooting process...anyway I hope that you based on your experience could share would you believe could be the root cause to our problem. We seem to have solved it..but not quite understanding why. Next week we will " roll-back" on some more units and do a more focused analysis , trying to figure out what is going on.

best reg
Erik

Hi, Erik
Please provide the latest TPA3110D2 schematic to me, Thanks!

Hi Peter

Sorry, attached please find the updated TPA3110 schematic.

Please notice the control on FAULT using V78 and V88 are not used at the moment, R97 is removed until we are able to control it from our GPIO  (sw control not made yet)

Best reg Erik

TPA3110 latest.pdf

Hi, Erik
Have you already confirmed the fault is still exist after replace C174/C175/CL41/L35 with 0ohm resistor?

Hi Peter

No, these are still mounted. We have tried to shorten C174 and C175, no effect on the problem. We will try to remove all of them and come back to you. As informed in previous post..we have changed some components (mostly removing some coils..and this seems to have helped..but without really understanding why. As informed..we will roll-back and start over again to find out what has the most effect on solving the FAULT /SD issue.
best rg
Erik

Hi Peter

I think we have found the problem and that we also understand it.

1) Problem is often seen on a system that has reached operating temperature.

2) Problem is only seen using 24V DC supply, never on 18V

3) Problem has been introduced with the new LC filter.

4) Problem is a "shut-down" problem (FAULT) which is most often a temp issue or a output stage overload 

5) Problem "is cleared" when OS kicks in and reset the ACL622 codec.

What we have learned: ( I hope it makes sense)

During powerup on a system that has reached working temp, the inductance is lower  and the thereby the load on output stage has increased due to the higher current. (current is also higher on DC24V compared to DC18V..which then again explains it is a "load" problem.

During powerup the ACL622 is outputting some "noise" (we have not been able to fully identify it) but this "noise" going to the TPA3110's L INP and R INP and the load on the output stage makes TPA3110 entering FAULT. The TPA3110 tries to playback the noise into a too low load.

It stays in this state until the ACL622 receives the reset which lowers the signal level on L INP /R INP, then SD can clear FAULT and WIN OS can playback normally.

We have fixed the problem on the ACL622 line out noise by adding 330ohm on C165 and C163

In our next revision of the PCB we will change to 22uH, as suggested from your application note, and thereby lower the load etc.

From our reference design also using ALC622 we will investigate in the "noise" ..we also have high click sound during power up we would like to get rid of.

Best reg

Erik