TPA311x Output Filter

[mlee]

Hi All,

Not sure where the best place to post this is so hope this is the correct spot. Just wanted to share a small side project, amplifier built around the TPA311x.

After reading TI application note SLOA119A and testing out a number of different LC components with load (both resistor and driver), I found different LC components for 4 and 8 ohm speaker loads. I'm sure most of you knew that already but the I wanted to share my results and hope they agree with the consensus. The goal was relatively low peaking at 20k and reasonable suppression of the carrier signal.

8 Ohm = 33uH and 680nF. Used bourns 2200HT-330-V-RC and Wima MK series.
4 ohm = 10uH and 3.3uF. Used sumida CDEP147NP-100MC-125 and Wima MK series.

They ICs sound pretty good Using them in an active setup with 4 channels per box. Phone camera fuzzy pictures.

Cheers

 

[DUG]

There are a couple of threads running for the TPA311x...pix of class D is one.

I like the box.

I would not recommend the bourns 2200HT series for class D filtering...just not enough data on the cores for me to put them on the list.

Whose boards?

 

[mlee]

Hi DUG,

Thanks for the tips and comments. Maybe I could move this post to an appropriate thread? Do you have any recommendations for a 33uH Inductor? I Could drop it in and see if there is much difference. I rolled my own boards for this one as I was able to sneak a 2 layer onto on another panel order. Would like to do some distortion testing on different filters but time and access to reasonable test equipment is not one my side. Any suggestions are welcome. I did some thread reading but not too many 33uH recommendations.

 

[irribeo]

both your filters have lowish Fo, 27khz, so yes they will attenuate 400khz or higher more than 50khz Fo, but when load is higher the peaking is more likely to be within audioband too.

3.3uF/10uH already is peaking with 2 ohm load, more standard for 4 ohm would be 1uF/10uH
the 33uH/680nF is like 10 ohm version of 33uH-1uF, so that will dampen a little more

 

[DUG]

Here are some 10uH types I have looked at:

COILCRAFT RFS1412 10uH 6.2A(10%L drop) RFS1412-103ME
WURTH WE-PD SMD 10uH 10.5A(10%L drop) 7447709100
SUMIDA CDEP 10uH 11.5A(10%L drop) CDEP147NP-100MC-125
COILCRAFT SER2900 10uH 13A(10%L drop) SER2915L-103KL
COILCRAFT SER2900 10uH 18A(10%L drop) SER2915H-103KL
WURTH WE-HCF SMD 10uH 22A(10%L drop) 7443631000
COILCRAFT SER2900 10uH 28A(10%L drop) SER2918H-103KL
WURTH WE-HCF SMD 10uH 34A(10%L drop) 7443641000

There are others.

 

[mlee]

Hi All,

Thanks for the information and confirmation. Sounds like the numbers are in the ball park at least.

irribeo, you are right about the heavier loads. I'm ensuring a strict "flat" impedance over frequency for both the 4 and 8 ohms filters. Seems like a widely varying impedance while running low filter fc would effect the audio band too much.

 

[irribeo]

I checked with filtersimulator, 33uH/680nF with 8 ohm load has oscillation frequency of 16khz and is a little overdamped for 8 ohm
10uH-3.3uF is ideal for 2.5ohm load, so doesn't peak with 2 ohms like I thought It does however oscillate at 19Khz and with 4 ohm load at 25Khz, not sure what that really means for ampbehavior

 

[mlee]

Hi irribeo,

Thanks for checking. I do see some of that ringing on the bench but when connected with the driver, the effects are a bit less pronounced, bit still there. Further optimization can definately be done. Seeing the Q effects on the oscilloscope are a bit scary, as the higher frequency peaking looks bad. A better 33uH inductor is in order as the response is not ideal. I found the higher I pushed the fc, the higher carrier oscillation amplitude was. This only makes sense as it is a LP filter. I also found the driver load vs the resistive load makes a difference in the peaking so I picked my drivers to have low impedance rise. I didn't realize the output LC of class D was that critical. Definitely not a connect and forget kind of thing.

 

[gmarsh]

If the goal is sound quality over efficiency, I think your best bet is to move up the corner frequency of the filter, and up the switching frequency of the amp. This way any peaking/droop caused by load mismatch will be pushed further out of the audio band...

 

[mlee]

gmarsh,
That is a good idea. Will try that next.
Thanks!!

 

[irribeo]

TI sloa119a looks at 250khz attenuation because their amps at that time switched at that frequency, higher Fo even with 400khz setting is possible with same attenuation for switching frequency. For 8 ohm that is 22uH/680nF instead of 33uH/1uF, 41khz vs 27khz, but with higher freq even higher Fo will result in same attenuation and distortion gets down a little with higher switchfreq or so says TI, but sloa119 also says distortion gets down with higher inductance, but that might be related to quality inductors???

Hypex claims 600-700khz is ideal range for audioamp at the moment, don't know why or didn't understand explanation, result is forgetting why

btw: your Bourns do have good color core I think, some other in big thread seem to have cores made for microwave frequencies I think??

 

[mlee]

Hypex claims 600-700khz is ideal range for audioamp at the moment

Great! Will try an increase in fs and another round of filters. Will report back.
Thanks!

 

[mlee]

Hi All,

Finally had a chance to get this on the bench. As suggested buy others, pushed the corner frequency up around 50khz and increased switching to around 1Mhz (did not worry about EMI yet). Resulted in a decrease in carrier amplitude and audio band looked better (less peaky in audio band). Tested with 10uH and 1uF filter for both 8 and 4 ohm loads. Might be some distortion around 9kHz ish. Not sure if thats the signal generator or the TPA or output filter.

Be sure to bring out the AM modulation lines on the TPA311x. Not sure what effect this has on EMI generation but signal looks better overall. Thanks for the tips, this makes it easier to setup!