Ferrite Bead vs. LC output filter

Hey out there,

So I recently put together my first class D amplifier built around the TPA3110D2 chip from Texas Instruments. Schematic here. To my dismay, when I first fired it up, it would only work at very low volumes. When the volume to the inputs was increased to normal listening levels, the amp kept going into over-current shutdown mode. After conversing with TI engineer, it was decided that the ferrite beads that I was using for output filters were saturating at the current I was trying to drive the speakers at and causing a short through the 1000pF caps that sat between the speakers outs and ground. The solution? Remove the caps, so I did (C18,19,20 & 21 gone now), and everything seems peachy. What I'm curious about now is am I missing something by not having a proper output filter on my amp? Would it sound better with a more traditional LC output filter? Would it sound any different with no filters at all? Are these output filters only in place to mitigate EMI and the like?

I supposed my question boils down to this: is it worth my time to build a redundant amp with proper LC filters and do a side by side sound comparison?

Here's a little more info on this particular setup:

Power Supply: 14.4V NiMH battery pack
Amp: TPA3110D2 15-W Filter-Free Stereo Class-D Audio Power Amp
Speakers: 2X 4Ohm 100W rated power, 4 inch.

Thanks again!
-Chad
 
Radiated Emissions

Without those filters, you're probably radiating enough electrical noise to wipe out any AM radio reception in the vicinity. You might even be polluting the FM band with very little attenuation of the harmonics, as well as everything in between. As a neighborly gesture, I would use the recommended filters.
 

wakibaki

Banned
2008-01-08 11:51 pm
It all sounds a bit unlikely, but who am I to argue with an engineer from TI?

What seems even more unlikely is the solution. The beads are supposed to be lossy. They're supposed to absorb energy and dissipate it as heat. If they're saturating, use ones that don't, or more of the ones that do. Put the caps back on.

w
 
As unlikely as it may sound, removing the caps did seem to do the trick. It is possible that I didn't relay the solution properly. Here's what was said (the posts are here.):

"Hi Chad,
It looks like the Ferrite Beads (Q3, Q4, Q5, and Q6) maybe saturating at higher output power level. FB typically saturates more easily than inductors. In many beads, low frequency impedance falls by a factor of 10 or more at rated current. Make sure that the FB's impedance remain high enough to provide sufficient filtering at the peak current it will carry. Try remove the capcitors to ground (such as C18..etc) and see if the OC condition will stop."


This was followed by a post that was mostly me blathering on about how little I know about what I'm doing and how happy I was that this solution seemed to help. To which the response was:


"Hi Chad,
I'm glad that it worked out. The problem is caused by the ferrite beads saturating at high current level. I have attached a ppt that explains the saturation issue and FB filtering from pg 45 to pg 50. In the last few pages, there is some good info on FB - It shows how the impedance change with load current and frequency. When your FB saturates, its impedance drops and the output becomes AC short via the filtering cap. That is why the over current protection function was triggered at high output power level.
Enjoy your first Class D project!"


And that leaves us where I am now... wondering about the components I yanked off my PCB and what, if anything, I should do to replace them.


The ferrite beads I used on this amp were the same ones listed in the bill of materials for the TPA3110D2 evaluation module described by TI. All I can think is that they were intended for use with higher (than 4 ohm) impedance speakers, and that my application is "juicing them too hard."


Wakibaki, as for finding beads that won't saturate in my application, I'd love to, but I'm not really sure where to start. Except here, the d/s for the ferrite beads that I have.


When you suggest adding more of the ones I do have; how so? In parallel? Series? With the pad size on my current PCB, I could probably come up with a way to do either. I'll try anything once. And I'd even put the caps back on for good measure... to do whatever it is that they do.


Thanks again for everything. I mean it.


-CHAD
 

wakibaki

Banned
2008-01-08 11:51 pm
You would have thought TI would have specified a bead that worked.

When I said unlikely as regards the solution, I meant removing the caps as a permanent fix. As a test for the theory, fine.

The saturation is a current issue, so to reduce the current per bead, run them in parallel. I see they're SMT, you can probably stack them one on top of the other. Depending on the exact board layout, you might get 4 on, 2 side-by-side and 2 on top of those. A problem might be whether the current splits evenly between the beads, which will depend on the matching of the individual beads DCRs.

The beads are specified as having a nominal Z of 100 ohms @ 100MHz. The rated current is 8000mA. To find a bead that won't saturate, you can look for a 100 ohm, 100MHz bead with a higher rated current. This will be physically bigger in all probability, so it may not fit the existing pads.

Let us know how it works out please.

w
 

jkeny

Banned
2007-02-06 12:43 am
Dublin
It all sounds a bit unlikely, but who am I to argue with an engineer from TI?

What seems even more unlikely is the solution. The beads are supposed to be lossy. They're supposed to absorb energy and dissipate it as heat. If they're saturating, use ones that don't, or more of the ones that do. Put the caps back on.

w

You would have thought TI would have specified a bead that worked.

.........
w

I had to laugh - Wakibaki is now telling us that these engineers don't know how to specify output stages correctly. What ever happened to the bullish & adamant stance that he showed not so long ago? Wow, the road to Damascus comes to mind!

Where's your evidence for this?

I very much doubt that the engineers responsible for designing these devices are failing to terminate them correctly. They certainly know more about engineering them than the average audio buff. ...........
from here http://www.diyaudio.com/forums/digital-source/168901-rf-attenuators-jitter-reducers.html#post2221444
 
scwhiteley,

I suggest ditching the ferrite bead route in favor of a proper output filter; even if it means switching to a different IC.

I'm no expert or EE, but it seems to me that ferrite beads are only a viable solution when used with low power (a few watts) and very short speaker leads.

I've built some boards based on the TPA3106D1 and absolutely love them. I think my neighbors would call the police on me before I could turn them up loud enough to trigger any kind of fault condition.