What is wrong with TPA3255?

[chermann]

😀

 

[29285]

Would it be possible to do the same test with a board where you have replaced the power line decoupling capacitors with better ones? That would show how important the quality of the power line decoupling capacitors is.

What exactly do you call power line decoupling caps?
The measured audio band ripple on the supply rails is mainly determined by the decoupling electrolytics on board, their capacitance and ESR. MLCCs with several uF with their higher impedance at audio frequencies do have little effect here.

 

[chermann]

He is writting about the decoupling caps near the chip.
normally its recommended low esr and about 2220µ/63V is enough.

the YJ use elna or sanwah 4700µ/63V (100) labeld. look at the pic(original board...first switch on)

as i wrote i measured after weeks of operation no esr and the sound is better with new caps 😉...maybe because of the noisy regulator (LM2575)

 

[Kay Pirinha]

Got it. A really big heatsink.

...or an actively convected CPU cooler...
Best regards!

 

[chermann]

...or an actively convected CPU cooler...
Best regards!

generally CPU cooler are "loud"

 

[Kay Pirinha]

Yes. It depends on the device's purpose. A PA power amplifier surely will benefit from an active cooler, an amplifier for your home environment most probably not. OTOH, usually it won't be driven that hard here.
Best regards!

 

[chermann]

Yes. It depends on the device's purpose. A PA power amplifier surely will benefit from an active cooler, an amplifier for your home environment most probably not. OTOH, usually it won't be driven that hard here.
Best regards!

yes yes you are right... i forgot that sweetperfume want to push 400W
so the colling fans are not the problem...😉

 

[doctormord]

yes yes you are right... i forgot that sweetperfume want to push 400W
so the colling fans are not the problem...😉

The EVMs heatsink is available predrilled at digikey for example. Otherwise you may look at Fischer Elektronik:


Standard extruded heatsinks, Heatsinks f.cool, Fischer Elektronik


Feed the search-filter with "<=2" K/W to get some ideas.


I.e.:


SK 453, Standard extruded heatsinks, Heatsinks f.cool, Fischer Elektronik

 

[chermann]

As requested by FF at post #636😉😀


LRS 150-24....28,7V set


i try a loop test with a burst of 40Hz, periode is 200ms, 1 Cycle. i try 800mVrms in or 1300mVrms into the amps input. at 4R load its more visible. first test did with 8 R.


original amp and caps.


pic 1 8R 1300mVrms in
pic 2 8R same but single burst view


pic 3 4R 800mVrms in
pic 4 4R same but single burst view


pic 5 4R 1300mvrms


pic 6 4R with 2 cycles 800mvrms...omg
pic 7 4r with 2 cycles 1300mVrms..ooooomg




...continued..

 

[chermann]

other board (amp6) with GPD caps...quicker but with less juice

just 4R output


pic 1 800mVrms in
pic 2 same but single burst view
pic 3 1300mvrms
pic 4 4R with 2 cycles 800mvrms...omg
pic 5 4r with 2 cycles 1300mVrms..ooooomg



the smps is directly connecte no additionaly caps are installed! (for the post before too)

 

[chermann]

playing around with the original board/caps i found something interesting:


i set the periode smaller= 40hz burst with 600mVrms but the periode of this burst is shorter. 100ms... so i can see the "recovery time" for the smps is not really existing 😀
+ there is an overshoot ! this is not happend with the test setup before

 

[chermann]

if something is wrong tested or some other ideas...please let me know!


thanks
chris

 

[chermann]

does high frequency burst the power supply? yes....


4R load burst is 5 cycles, periode is 20msec, 800mVrms in the original amp/caps



pic 1 6khz
pic 2 same but single view


pic 3 12khz
pic 4 same but single view


pic 5 16khz
pic 6 same but single view


pic 7 +8 is the 1 cycle burst




..continued...

 

[chermann]

4R load burst is 5 cycles, periode is 20msec, 800mVrms in the amp6 changed decoupling caps = GPD

pic 1 6khz
pic 2 same but single view


pic 3 12khz
pic 4 same but single view


pic 5 16khz
pic 6 same but single view


pic 7 +8 is the 1 cycle burst




good night

 

[doctormord]

As you can see, it is not much important if you have 1000uF or 10000uF. 😀

The overshoot is totally normal and will rise with more capacitance

You may add 500000uF, that should help at low frequencies but will increase overshoot if long bursts are run. (If the supply will start)

 

[chermann]

As you can see, it is not much important if you have 1000uF or 10000uF. 😀

The overshoot is totally normal and will rise with more capacitance

You may add 500000uF, that should help at low frequencies but will increase overshoot if long bursts are run. (If the supply will start)

hi doc

thank you for your feedback. i will try some extra caps.
as i can see on the pics there is no marcant event happend that can be markeble for SQ between the original and the new caps. in real life it is.....any hint to see that?

 

[chermann]

generally


is the response time of the LRS-150-24 (28,7V) ok???? too slow...normal...What can better supply do in respect of response time?

or should i try with less voltage set...etc.

chris

 

[chermann]

....doctormord;5636771]The EVMs heatsink is available predrilled at digikey for example. .........]

I go into the BOM of TPA3255 and found the heatsink of advanced Technologie...its labeled for TAS...but is it the correct one????????...i guess yes

ATS-TI1OP-519-C1-R3 Advanced Thermal Solutions Inc. | Lufter, Warmemanagement | DigiKey

 

[doctormord]

Correct.

 

[29285]

4R load burst is 5 cycles, periode is 20msec, 800mVrms in the amp6 changed decoupling caps = GPD

pic 1 6khz
pic 2 same but single view


pic 3 12khz
pic 4 same but single view


pic 5 16khz
pic 6 same but single view


pic 7 +8 is the 1 cycle burst




good night

looks sane to me