TPA3255: Output filter for 100KHz bandwidth ?

Hi,

You may try with 3.3uH inductor+0.47uF cap.

BTW, slaa701A is a good reference for LC design. 

http://www.ti.com/lit/an/slaa701a/slaa701a.pdf?keyMatch=SLAA701A&tisearch=Search-EN-everything

Regards,

Hao

Hi Robert,

thanks for your remind and suggest.

customer tried 5.1uH+0.47uF, the red light of the EVM board was directly lit, and it could not operate normally.

We also tried 7.1uF+0.47uF. When the input power is small, the output characteristics are not normal, the frequency response is not consistent with the calculator,and there is attenuation at 20KHz.

In addition,test of 7.1uF 1uF the output stable,the frequency response is the same as the calculator but However, there is still attenuation at 20 kHz.

The above tests are all directly using APx525, without passing through the AUX-0025 filter.

have any idea or suggestions?

7u1H+0u47F.rarTPA3255 EVM_7u1H+1uF_192KHz.pdf

Hi Tommy,

Though a 5.1uH inductor is above the minimum, inductors always have tolerances and those particular ones likely were less than 5.1uH thus causing the immediate fault light going off. 

Can you send me pictures of the output waveform so I can see what you mean by the strange output characteristics? Also what frequency was the input signal when you saw this? What was the Vrms of the input signal? Are you using any of the AP filters? 

More information on your setup will help me better debug this. 

Thanks,

Robert Clifton

Hi Robert,

I'm Elytone electric engineer, Eric.

Actually, we have measured the inductance value by RLC meter, the inductor is 5.2uH.

We also tried 7.1uH+0.47uF and 7.1uH+1uF, we couldn't get the same frequency response curve with LC filter designer.

Below is our detail setting for EVM and AP.

EVM Setting:

Single end input

2xBTL output

4ohm load

AP Setting:

No filter, Bandwidth=90KHz(192KHz SR)

1. We've tried 7.1uH+0.47uF, the strange output is THD+N too high when the small signal input (100mVrms input@ 27.5dB gain, 4ohm load=5.5W output), but it looks normal when input=200mV(27.5dB gain, 4ohm load=22W). The situation looks like a big noise at AMP out.

At this case, the frequency response is attenuated at 20 kHz.

1. We’ve changed to 7.1uH+1uF, the THD=N looks fine. (100mVrms input@ 27.5dB gain, 4ohm load=5.5W output) But the frequency response still attenuate 1dB@20KHz.

           2. Our test result is different with the LC filter designer(30KHz no decay by calculator).

Anyway, the main problem is that we couldn’t find a set of value for LC filter to make a flat frequency response from 20Hz to 100KHz.

Thanks.

BR,

Eric

Hi Eric,

Looks like we have multiple things going on here.

Did you test the output's THD+N with the default inductors attached? If the THD+N looks better with those inductors that would indicate that the inductors that you've chosen are adding noise. Could you provide a THD+N vs Power graph? That would help me get a better idea of the curve.

It would be difficult to get a completely flat frequency response but it would be possible to get it to where it will drop less than -1 dB of gain at 100kHz. Does the system have to be completely flat or is there a certain amount of loss allowed?

Regards,

Robert Clifton

Hi Robert,

Yes, we can get the test result which similar with the datasheet after we changed to default LC.(10uH+1uF).

I have attached THD+N VS Power graph for 7.1uH+0.47uF and 7.1uH+1uF. The poor THD+N only exists on 7.1uH+0.47uF, and we tried different bandwidth AP setting, the problem only exists on the 192K bandwidth setting. I think that should not the inductor's problem. It looks like other limited factor of the LC combination when system on 192KHz bandwidth.

7.1uH+0.47uF 

7.1uH+1uF

About the frequency response, we know that is difficult to achieve flat curve to 100KHz.

Our requirement is +0/-3dB at 20~100KHz, just like the below calculator picture. But now we couldn't get the same curve when we applied the same LC values on EVM.

Hi Robert,

Sorry that I attached the wrong picture for 7.1uH+0.47uF testing, please refer it as below.

7.1uH+0.47uF

Hi Eric,

I was a little confused on what the issue was! This is very strange indeed. It's almost as if the signal is getting buried by noise but then abruptly returns to what we would expect to see. I would start by double checking the signal chain to make sure that everything is clean. I noticed that the low frequencies had a rise in THD+N. Was this observed with the 7.1uH + 0.47uF filter?

Are you generating the input signal with the AP? If not what are you using? If you are using the AP to generate those signals would you be able to try testing the input by switching to a differential input signal to see if this removes what you are seeing when the signal is small? 

Regards,

Robert Clifton

Hi Robert,

Yes, we used AP to generate the input signal, I will try the differential input then back to you.

There is another urgent question need to be check. Can we get a set of LC values to achieve 20~100KHz +0/-3dB frequency response?

Although we solved the THD+N issue for 7.1uH+0.47uF, it still couldn't meet our FR requirement.

If you have the TPA3255 EVM, could you please test and provide a set of LC values for 20~100KHz +0/-3dB frequency response first?

Thank you.

BR,

Eric

Hi Robert,

I did some testing about 7.1uH+0.47uF differential input, please refer below pictures.

Base on EVM board default setting, the gain of the differential input is only half of the single-ended input. The test result is still very similar with the single-ended.

Hi Eric,

This is strange. I think we will need to do some testing on our side to determine the root cause. Can you tell me the EVM configuration? I want to know all of the jumper settings as well as the part numbers for the inductor you are using and the capacitors. 

Thanks,

Robert Clifton

Hi Robert,

We confirmed that the setting is the same with EVM BTL(2.0), and we adjusted J4 and J19 for single-edned input and differential input.

Hi Eric,

Can you also provide the LC part numbers you used? 

Thanks,

Robert Clifton

Hi Robert,

Inductor: Coilcraft MA5172-AE 10uH (We reduced the wire round then measured the inductance through Chroma LCR meter for current testing.)

Capacitor: EPCOS B32652A3474K000(0.47uF)  and Kemet PHE426HB7100JR06 (1uF)

We found the 4ohm dummy load which with around 18uH inductance, that is why we couldn't get the same curve with the LC calculator table.

We have changed another 4ohm dummy load(inductance =0.3uH) and re-test again, now we can get the correct curve. But the THD+N issue is still exist on 7.1uH+0.47uF.

We also tried to change back to 5uH+0.47uF. It can be powered on normally, but the fault LED always lighted when I did THDvsPower testing.  

Hi Eric,

Well it looks like at the lower signal levels, the THD+N vs Power curve looks normal than jumping up high. This was all as a result of changing the load to be less inductive. 

Are the frequency responses of both filter configurations within the range that you are wanting for your system? 

Regards,

Robert Clifton

Hi Robert,

The frequency response of 7.1uH+1uF looks fine for us, and this setting without THD+N jump to high issue, thanks.

Now we have another problem, I couldn't reproduce the THD+N test result as datasheet even though we use the original filter(10uH+1uF).

Could you help to check anything we may missing? Our requirement is under 0.01% for 0~100W.

TPA3255 Datasheet Page 10

EVM ( BTL2.0, 36Vin, Differential input, L=10uH, C=1uF, AES17+AUX-0025)

Hi Robert,

We have two TPA3255 EVMs, and we got almost the same THD+N curve as last post.(75W couldn't under 0.01%)

We would like to check that even if we use the original inductor of EVM, it still might not to achieve under THD+N<0.01% curve?

Thanks.

Hi Eric,

Just noticed even on datasheet curve, THD+N could go above 0.01% at 25W and 75W. It may be very hard to achieve THD+N<0.01% in 0~100W...

Regards,

Hao

Hi Hao,

Got it, we will discuss with our customer for the THD+N specification base on what we can achieve now.

Thanks for your reply.

BR,

Eric