Yet another Volume controlers and source selections

[danzup]

Originally Posted by Vikt0r

In the "off" state the PGA is set to mute and source selector relays are unpowered, only the power rails stay connected on. The output remains connected to the poweramp's input, which is also off.
As I mentioned before, I had a similar behaviour once - that time the sound went suddenly cranked (with a huge amount of "digitally sounding" distortion) and the PGA started heating up. Only powering the whole device off and on helped. The chip naturally "hung", and what I noticed - immediate and powerful overheating. So, my "speculation" comes out of a logical analysis, not just outta pointing my finger into the sealing...and the "actual cause of overheating" seats inside the chip's behaviour. If you have ever read the datasheet - it's mentioned inside, that the chip is basically a resistor ladder switched by digital switching network. It's more or less obvious for me, that once the data in chip's register for any reason get's wrong, there is a POSSIBLE STATE, where switches get open the way they are shorting power rails to the ground. This should not happen during the normal operation, of course, but the PROBLEM is that there is no protection within the chip against a mentioned uncontrolled condition.
Or have you got any better idea, not just speculation?

Ok : this is something that must be investigate very carefully !
Maybe you discovered a bug or a flaw in TI design !!!
Or maybe not .....
Time and investigation will tell us .

[Vikt0r]

Originally Posted by danzup

But if you forgot to connect digital gnd to the analog gnd then the PGA start burning !

I think you have a problems with GND traces .

Dan, your assumptions are wrong. The AGND and DGND polygons are connected beneath the PCB in a single point through a ferrite bead inductor. This is hardwired, no chance they aren't connected.
Moreover, the device was working quite fine for more than a month, delivering me a big amount of pleasure, until yesterday

[jan.didden]

Originally Posted by Vikt0r

In the "off" state the PGA is set to mute and source selector relays are unpowered, only the power rails stay connected on. The output remains connected to the poweramp's input, which is also off.
[snip]It's more or less obvious for me, that once the data in chip's register for any reason get's wrong, there is a POSSIBLE STATE, where switches get open the way they are shorting power rails to the ground.[snip]

Is it possible that the input is left unterminated and that the chip oscilates because of that?

Anyway, the 'obvious' thing is not obvious to me as the switches are NOT connected to the power lines - so how can they short the power lines??

Originally Posted by Vikt0r

[snip]Or have you got any better idea, not just speculation?

No I'm still thinking loud, but I don't think you should start to point to TI engineers before you really know what's going on. Just to be fair.
The mentioned ground issue is also a possibility imho.

jd

[danzup]

Originally Posted by Vikt0r

Dan, your assumptions are wrong. The AGND and DGND polygons are connected beneath the PCB in a single point through a ferrite bead inductor. This is hardwired, no chance they aren't connected.
Moreover, the device was working quite fine for more than a month, delivering me a big amount of pleasure, until yesterday

Ok , here one thing to do : use a relay that is connected in parallel with the power relay to short the input of PGA to ground when in stand-by .
Possible oscillation as suggested by @janneman can destroy the PGA .
Or you can attached a 100k resistor on input of PGA to the ground for left and right input .
And tell us if there are more problems .

[Vikt0r]

Originally Posted by janneman

Is it possible that the input is left unterminated and that the chip oscilates because of that?

Okay, then mute doesn't mean anything? Shouldn't it be CONSIDERED BY DESIGNERS, that for a VOLUME CONTROLLER IC such a state can, and will most probably happen? WHY this did not happen before? Also, your assumption simply ignores my previous experience I just described...

Originally Posted by janneman

Anyway, the 'obvious' thing is not obvious to me as the switches are NOT connected to the power lines - so how can they short the power lines??

Any switch is just a MOSFET which has to be fed by power, isn't it? What happens if they get ALL suddenly open? Normally they shouldn't...

Originally Posted by janneman

No I'm still thinking loud, but I don't think you should start to point to TI engineers before you really know what's going on. Just to be fair.

Okay, just to be fair - your'e right. But also, just to remain fair enough I'll go back to study the datasheet more precisely. Worth to add, I'm NOT POINTING TI's engineers (really I don't understand why you take this in such an offensive-resisting manner... I am also an engineer, since about 15 years...) - but I am raising a definitely existing problem, which could be caused by POTENTIAL DESIGN ISSUE, this may at least be taken into attention by TI. As I mentioned before, the PGA is a volume controller IC - and such "unforeseen" cases should be foreseen by designers, imho.

Originally Posted by janneman

The mentioned ground issue is also a possibility imho.

jd

Indeed, this is just your imho, because it's not the case, sorry.

[danzup]

Viktor : also I will do some extensive test about what you suggest ,
to see what is happening with this chip .
I will do some testing with the oscilloscope when the chip is in mute status to see if oscillation occur if the input are not connected .
Keep the finger cross !

[Vikt0r]

Originally Posted by danzup

Ok , here one thing to do : use a relay that is connected in parallel with the power relay to short the input of PGA to ground when in stand-by .

This can be done easily, as the input selector relays are DPDT. So, if I connect the unused position (of any selector relay) to the ground - it will short the input of PGA to the ground, when the device is off.
But again, out of MY EXPERIENCE, this happened already with input's ON - see my previous posts, so I still don't think oscillation is the case.

[Vikt0r]

Originally Posted by danzup

Viktor : also I will do some extensive test about what you suggest ,
to see what is happening with this chip .
I will do some testing with the oscilloscope when the chip is in mute status to see if oscillation occur if the input are not connected .
Keep the finger cross !

Keeping

[jan.didden]

Originally Posted by Vikt0r

Okay, then mute doesn't mean anything? Shouldn't it be CONSIDERED BY DESIGNERS, that for a VOLUME CONTROLLER IC such a state can, and will most probably happen? WHY this did not happen before? Also, your assumption simply ignores my previous experience I just described...

Any switch is just a MOSFET which has to be fed by power, isn't it? What happens if they get ALL suddenly open? Normally they shouldn't...


Okay, just to be fair - your'e right. But also, just to remain fair enough I'll go back to study the datasheet more precisely. Worth to add, I'm NOT POINTING TI's engineers (really I don't understand why you take this in such an offensive-resisting manner... I am also an engineer, since about 15 years...) - but I am raising a definitely existing problem, which could be caused by POTENTIAL DESIGN ISSUE, this may at least be taken into attention by TI. As I mentioned before, the PGA is a volume controller IC - and such "unforeseen" cases should be foreseen by designers, imho.


Indeed, this is just your imho, because it's not the case, sorry.

I said it's a possibility, I didn't say it is the case. And a possibility it is.
The point I'm trying to make is that nobody has an idea what is going on. I only wanted to add some possibilities. Doesn't anybody measure and test anymore these days??


In your earlier post, without any information on what caused the problem, you immediately pointed to TI engineers. If we all are engineers, we should do what engineers do: measure, test, collect information, THEN conclude what the problem is and solve it.

jd

[Vikt0r]

Originally Posted by janneman

In your earlier post, without any information on what caused the problem, you immediately pointed to TI engineers. If we all are engineers, we should do what engineers do: measure, test, collect information, THEN conclude what the problem is and solve it.
jd

Your'e right, I wasn't precise enough. Just to add - I didn't have an intention to blame anyone, especially not TI's engineers - rather I wanted to highlight a possible problem. Honestly, I didn't think the subject will cause such amount of feedback, therefore I didn't struggle too much to share all the prerequisites of my conclusions.
Again, I would be happy if I'm wrong and something went wrong due to my specific mistake, but, based on "test, collect&conclude" I had to conclude, that everything's fine with my design, the reason is somewhere behind.