Stop the sirring noise!

[EvilDragon]

Quick update:
* Sent an email to the Fairphone team
* Nikolaus has another idea that would work - using a different IC (replace the wonky chip by a different wonky chip!) which doesn't have a power converter built-in and add our own one.

The second solution would be Tantalum free and we can test it without producing full Mainboards, as such a setup is actually pretty simple.

 

[Caine]

Awesome updates Michael. Keep up the good work.

Just make sure the chip remains at least a little bit wonky

 

[Black Sliver]

* Nikolaus has another idea that would work - using a different IC (replace the wonky chip by a different wonky chip!) which doesn't have a power converter built-in and add our own one.

I would avoid drastically changing the layout/routing at this stage in a project... esp since you can't be certain that the noise will be gone in the final device (since it's probably an interference between two signals/frequencies).
Just to be sure... have you tried hot glue or epoxy (UHU plus)?

 

[Silent-Hunter]

I would also like to avoid anything like that that would possibly delay the Pyra even further.

 

[levi]

esp since you can't be certain that the noise will be gone in the final device (since it's probably an interference between two signals/frequencies).

That's an interesting idea, but I don't think it holds if Nikolaus testing in changing one cap eliminates the noise. Even if both of the sound sources are ultrasonic, I can't think of a way to combine those to create a lower harmonic, so at least one must be audible, and that must be the one that Nikolaus changed.

But I'm also doubtful of the idea of changing a chip and building your own power stage. Any power stage is going to need a fairly beefy cap I'd have thought, if the load is at all variable (or noisy, in electronic terms, but that's a difficult term to use here), and anything else that you're reimplementing from inside the chip using discrete components is going to take up more board space surely?

 

[God Ginrai]

Quick update:
* Sent an email to the Fairphone team
* Nikolaus has another idea that would work - using a different IC (replace the wonky chip by a different wonky chip!) which doesn't have a power converter built-in and add our own one.

The second solution would be Tantalum free and we can test it without producing full Mainboards, as such a setup is actually pretty simple.

If you go this route. It will need to be fully tested.

IMO, we should just go Tantalum. Aim for conflict-free Tantalum and if you can't get it, grudgingly get the conflict Tantalum and keep your eyes out for a chance to switch to conflict-free when the option becomes available.

-God Ginrai

 

[EvilDragon]

I would avoid drastically changing the layout/routing at this stage in a project... esp since you can't be certain that the noise will be gone in the final device (since it's probably an interference between two signals/frequencies).

It's no interference. It's ceramic capacitors which vibrate and that vibration moves onto the PCB which then makes the noise.
Soldering the same capacitors external makes it fully silent and it works without issues.

Also, it's no drastical change, it's just the power circuit for USB, nothing complex or spectacular.

 

[Silent-Hunter]

Won't it require redesigning the mainboard though? I'm concerned it will add more time.

 

[EvilDragon]

But I'm also doubtful of the idea of changing a chip and building your own power stage. Any power stage is going to need a fairly beefy cap I'd have thought, if the load is at all variable (or noisy, in electronic terms, but that's a difficult term to use here), and anything else that you're reimplementing from inside the chip using discrete components is going to take up more board space surely?

The problem is (Nikolaus can explain it better, I guess) that each power converter in a chip does it by chopping the input power in a certain frequency (that's also how PWMs work, and that's where the backlight flickering of some LCD come from when you lower the brightness).
The one we have sadly does it with a frequency which causes the vibration in these capacitors.

We're using a chip by TI that is a power switch with a DC/DC Power converter included.
The exact chip is available from TI as a switch-only solution. All that needs to be done is add a DC/DC power converter (there are MANY available) after that switch and that's it.

It's not a huge layout change, it's nothing that needs to be tested specifically.

It can be tested on a simple 2-layer PCB (same size) which you can get within a couple of days.
Testing that it's working (power conversion + limitation) is a simple test, and it doesn't affect anything else in the Pyra.
Nothing that should cause a delay (since the CPU board takes way longer for production than the mainboard and we don't need a full Pyra to test it).
[doublepost=1517676942,1517676873][/doublepost]

Won't it require redesigning the mainboard though? I'm concerned it will add more time.

No, it's a very simple change in USB-Port area.
We've got three more weeks until we need to order the mainboards so that they will arrive in time with the CPU boards (and even 4 - 5 weeks if we use express for the production).

 

[Silent-Hunter]

Well as long as you're sure! I'm just a worry wart, and I tend to get anxiety about the project losing preorders because it's taking too long.

 

[zonova]

I appreciate you working so hard to keep the project as ethical as possible ED. This is something that's so rarely seen in the electronics industry, even from small businesses. It would be fine with me personally if we had to use the Tantalum components, if it was the only way to solve the noise problem. Though, I do have a question. Would this entail replacing all the capacitors on the board with tantalum caps? Or is it just a few of them that are causing the whirring noise? If it's just a few, what is it about those few that distinguishes them from the rest that aren't creating noise?

Thanks!

 

[levi]

The problem is (Nikolaus can explain it better, I guess) that each power converter in a chip does it by chopping the input power in a certain frequency (that's also how PWMs work, and that's where the backlight flickering of some LCD come from when you lower the brightness).
The one we have sadly does it with a frequency which causes the vibration in these capacitors.

Ah, I think you're describing a switched mode type power supply. Yes, the end result is a pulsed input into a current sink (in the case of a PWMed backlight it's the backlight LEDs, in the case of a power supply it's the capacitor), but in a PWM circuit you define a period and a duty cycle and that defines the waveform, while for a switched mode power supply you switch based on the capacitor voltage. I think you could alter that frequency by putting a resistor in series with the capacitor, limiting the rate of charge it can accept, but I don't know what that would do to the voltage sense part of the switch - I've never actually built a switched mode supply yet. And that would only bring the frequency of any noise down in pitch, whereas what you really want is to be able to push it into ultrasonic frequencies.

 

[AndiTheBest]

omgomgomg! PyraPhone confirmed

 

[Djhg2000]

We're using a chip by TI that is a power switch with a DC/DC Power converter included.
The exact chip is available from TI as a switch-only solution. All that needs to be done is add a DC/DC power converter (there are MANY available) after that switch and that's it.

Have you/Nikolaus have asked TI why their IC is making a ceramic capacitor squeak then? The switching frequency of a good DC/DC should be outside audible range so I'm worried it's not operating within spec for some reason.

It would be interesting to see a scoped waveform of the voltage across said capacitor. Perhaps the component part number/datasheet as well if that's something you're comfortable with sharing at this point.

 

[EvilDragon]

Ah, I think you're describing a switched mode type power supply. Yes, the end result is a pulsed input into a current sink (in the case of a PWMed backlight it's the backlight LEDs, in the case of a power supply it's the capacitor), but in a PWM circuit you define a period and a duty cycle and that defines the waveform, while for a switched mode power supply you switch based on the capacitor voltage. I think you could alter that frequency by putting a resistor in series with the capacitor, limiting the rate of charge it can accept, but I don't know what that would do to the voltage sense part of the switch - I've never actually built a switched mode supply yet. And that would only bring the frequency of any noise down in pitch, whereas what you really want is to be able to push it into ultrasonic frequencies.

Yes, exactly, that's what I meant. Well, it's powering the USB port and the specs for those are pretty much fix, so you can't do much with the voltage :/
And yes, we don't want to LOWER the frequency

omgomgomg! PyraPhone confirmed

Only if there is enough interest for Nikolaus to do this.
It would use the same CPU board, battery and display but a different case (obviously) and mainboard.

 

[Letalis Sonus]

The switching frequency of a good DC/DC should be outside audible range so I'm worried it's not operating within spec for some reason.

I have one word for you: resonance. The driving frequency just has to be close to a clean multiple of the capacitor's resonance frequency to cause it to vibrate - with a much lower frequency.

 

[pyrat]

Thanks for the updates. I haven't fully understood how is there place for a DC/DC power converter where there is no place for diferent ceramic capacitors, but it must be so. I also don't understand how one can test this without a full mainboard, but I don't know electronics. I don't know what I'm saying, here, but can one add some inductance between the capacitors to break the resonance ? or would that hinder EMC ?

I think I wouldn't mind a little audible noise, although the effects on the device life it might have worry me a little.

I appreciate efforts to avoid conflict tantalum. Let's hope you find a solution.

If no other option remains than tantalum, I would still try to use KEMET unless some other option is known more ethical.
I can't tell whether any guarantees are reliable enough, but if nothing else can be achieved at least buy from suppliers
that take the trouble to advertise ethics (even if it was a lie it might help show market interest in ethics, counterargument:
encouraging all suppliers to lie).

Since I can't help in useful ways, I made some searches and I'm dumping here what I found, I don't think it's terribly useful.

European legislation against import of conflict minerals (effective 2021?)
Regulation (EU) 2017/821 (2017-05-17)

Press release (2016) by NGOs about the european legislation. Includes
contact info for those NGOs. Summary: weak but better than
nothing. Requires European mineral (gold, tin, tantalum and tungsten)
or ore importers above certain thresholds to comply with OCDE
regulations or join industry associations to be created to address due
diligence. I tried to contact one of those ONGs in September to ask
about KEMET but I got no answer.

Governmental cooperation (leaded by NL) to reduce conflict minerals

iTSCi joint industry traceability and due diligence programme
auditing and certification

iTSCi member list with the status they consider them to have.
I've only found one capacitor provider, KEMET, which was
already mentioned in the forum by Alec and Black Silver.
I'm not at all sure how reliable this is. It could be
good or just greenwashing.

The link about KEMET Alec sent

 

[Djhg2000]

I have one word for you: resonance. The driving frequency just has to be close to a clean multiple of the capacitor's resonance frequency to cause it to vibrate - with a much lower frequency.

That's definitely true. Getting enough amplitude for the resonance to be audible suggests it's being driven quite hard though.

According to a very experienced electronics engineer I know, TI provides simulation tools for almost all of their DC/DCs where you can test your design to make sure you're in spec.

He also said when ceramic capacitors makes audible noise it's usually because you're asking too much of them in terms of power, you need to calculate the load and check the spec of the cap as well.

I have no reason to doubt his word, because he's usually the guy who gets called in when nobody else can't figure out why a board won't behave like it's supposed to.

 

[Letalis Sonus]

According to a very experienced electronics engineer I know, TI provides simulation tools for almost all of their DC/DCs where you can test your design to make sure you're in spec.

Unfortunately this is one of the things that are not entirely being covered by specs, though. Capacitors have a fairly large tolerance on their specs and things like resonance frequency are being influenced by the stuff those components are being glued to as well. The circuity still works as intended, the noise is just a side effect.

He also said when ceramic capacitors makes audible noise it's usually because you're asking too much of them in terms of power, you need to calculate the load and check the spec of the cap as well.

A DC/DC converter surely is a usage scenario that produces quite some spikes, but this isn't a problem that only happens under load. Resonance tends to amplify the resonating frequency's amplitude by quite a bit.

 

[Djhg2000]

Unfortunately this is one of the things that are not entirely being covered by specs, though. Capacitors have a fairly large tolerance on their specs and things like resonance frequency are being influenced by the stuff those components are being glued to as well. The circuity still works as intended, the noise is just a side effect.


A DC/DC converter surely is a usage scenario that produces quite some spikes, but this isn't a problem that only happens under load. Resonance tends to amplify the resonating frequency's amplitude by quite a bit.

Those are all good points, but the core question still remains; what are we actually asking the capacitor to do?

Running the simulator might not give us a definitive answer but at least it will hopefully tell us we're within spec of the IC. Worst case we're back to finding a different IC or tantalum caps, right?