Meanwhile I scanned my archives and found my attempts of full audio range PFFB around TPA3118 in 2.0 stereo BTL configuration. This has been a trial and error procedure while playing with LTSpice and doing real world measurements. I did not find a solution working without a snubber - which was the no-go for me, so finally I dropped that idea. Anyway there certainly exists room for improvement so I upload some simulations and the power amp circuit to play with.
Last edited:
Actually I am prototyping and so stumbled over another issue:
Turn-On Pop Noise with TPA3118
Powering up the TPA3118 powermodule produced some small audible pop noise. Not really disturbing but enough to look for a possibilty to improve it. At this time the inputs had been connected to an AFE-PCB containing a symmetrical input interface. So I disconnected the AFE and that noise disappeared completely. Not sure whether the AFE was the cause or the TPA itself I replaced it by a dummy connector that simply shorted the TPA-inputs to GND. The plop was back again! So let us focus on the TPA3118.
It should be mentioned that I use SDZ-input to power-up and the mute pin is permanently enabled. Not the nicest method as this activates the analog inputs and from that moment on input bias voltage rises from zero to about 3 Volts. During that phase coupling caps must be charged from zero to 3V as well, and that is the problem here. Without any additional delayed muting charging results in an input voltage transient that will be amplified. After all this must be a tiny spike in reality so there is no point in trying to measure something with an oscilloscope.
Keep in mind that both inputs of the TPA are driven symmetrically - but the rising input voltage during power up is a asymmetric disturbance that should be cancelled out by nature - in a perfect world. To make a long story short I removed both input caps and measured 1084nF and 1057nF respectively. Probing my stash of 1uF-caps I selected 1036nF and 1037nF for best symmetry obtainable. After the replacement - the popping on power up was definitely gone!
This outcome is interesting as it shows that the level of popping noise is directly related to input caps mismatch - giving a wide range of noise level depending on individual samples distribution. Furthermore it reveals a very high grade of symmetry inside the TPA-3118. Otherwise cancellation would require some individual caps mismatch. Or an extended power-up sequence using the mute input.
The input capacitance mismatch may be a widespread cause for power-up popping noise in many other class-d-designs. It may be a good idea to select perfect input capacitance matching for your DIY projects. And a bad idea to use electrolytic caps for this purpose with their big capacitance tolerances. Needless to say this is a DIY approach and useless for commercial applications as no factory will select caps that tight for little money.
This is my very first contribution to support caps rolling!!!
Turn-On Pop Noise with TPA3118
Powering up the TPA3118 powermodule produced some small audible pop noise. Not really disturbing but enough to look for a possibilty to improve it. At this time the inputs had been connected to an AFE-PCB containing a symmetrical input interface. So I disconnected the AFE and that noise disappeared completely. Not sure whether the AFE was the cause or the TPA itself I replaced it by a dummy connector that simply shorted the TPA-inputs to GND. The plop was back again! So let us focus on the TPA3118.
It should be mentioned that I use SDZ-input to power-up and the mute pin is permanently enabled. Not the nicest method as this activates the analog inputs and from that moment on input bias voltage rises from zero to about 3 Volts. During that phase coupling caps must be charged from zero to 3V as well, and that is the problem here. Without any additional delayed muting charging results in an input voltage transient that will be amplified. After all this must be a tiny spike in reality so there is no point in trying to measure something with an oscilloscope.
Keep in mind that both inputs of the TPA are driven symmetrically - but the rising input voltage during power up is a asymmetric disturbance that should be cancelled out by nature - in a perfect world. To make a long story short I removed both input caps and measured 1084nF and 1057nF respectively. Probing my stash of 1uF-caps I selected 1036nF and 1037nF for best symmetry obtainable. After the replacement - the popping on power up was definitely gone!
This outcome is interesting as it shows that the level of popping noise is directly related to input caps mismatch - giving a wide range of noise level depending on individual samples distribution. Furthermore it reveals a very high grade of symmetry inside the TPA-3118. Otherwise cancellation would require some individual caps mismatch. Or an extended power-up sequence using the mute input.
The input capacitance mismatch may be a widespread cause for power-up popping noise in many other class-d-designs. It may be a good idea to select perfect input capacitance matching for your DIY projects. And a bad idea to use electrolytic caps for this purpose with their big capacitance tolerances. Needless to say this is a DIY approach and useless for commercial applications as no factory will select caps that tight for little money.
This is my very first contribution to support caps rolling!!!
Last edited:
from TPA3118D2 datasheet, "Differential Inputs"
"The impedance seen at the inputs should be limited to an RC time constant of 1 ms or less if possible. This is to allow the input dc blocking capacitors to become completely charged during the 10 ms power-up time. If the input capacitors are not allowed to completely charge, there will be some additional sensitivity to component matching which can result in pop if the input components are not well matched."
With Rseries=4.4kOhms and Cseries=1uF time constant becomes 4.4ms. This explains the need for input caps matching as well.
All in all it is not the TPA3118 that is notorious for pop-up noise - but its poor implementations.
"The impedance seen at the inputs should be limited to an RC time constant of 1 ms or less if possible. This is to allow the input dc blocking capacitors to become completely charged during the 10 ms power-up time. If the input capacitors are not allowed to completely charge, there will be some additional sensitivity to component matching which can result in pop if the input components are not well matched."
With Rseries=4.4kOhms and Cseries=1uF time constant becomes 4.4ms. This explains the need for input caps matching as well.
All in all it is not the TPA3118 that is notorious for pop-up noise - but its poor implementations.
Last edited:
If you are in for simplicity, you can do this. Sadly THD degrades with asymmetric drive.Regardless of the technical aspects, it would be very nice if the inputs could just be asymmetrical. Would save 2 caps in each amplifier too
If I understand your question the right way - yes, it will oscillate above some critical point like any amp will do.
To answer this question I gave it a try today. Added 2 inner layers with KiCAD and - from now on I will no longer stick to 2-layer exclusively
As I do not have a clue how the internal auto-bias works I doubt it can be done that simply. Did you ever try this in reality?
Low gain is not so much of an issue, although having a least a gain of 2x (6dB), would be nice because otherwise you're asking a lot from the previous buffer stage.
You would like to keep using something like ±12-15V for the power supply.
(or like 24V split or something)
I have not tested an auto-bias on this amp specifically, but on similar situations.
I think it's worth a shot. I have to also dig into my archive for that.
edit: I also don't see a reason why the input stage of a TPA311x would be so fancy.
So probably it's just stuck at analog VCC / 2
edit 2: If I remember correctly, the input bias is a set voltage of 3Vdc
You would like to keep using something like ±12-15V for the power supply.
(or like 24V split or something)
I have not tested an auto-bias on this amp specifically, but on similar situations.
I think it's worth a shot. I have to also dig into my archive for that.
edit: I also don't see a reason why the input stage of a TPA311x would be so fancy.
So probably it's just stuck at analog VCC / 2
edit 2: If I remember correctly, the input bias is a set voltage of 3Vdc
Last edited:
I forgot but isn't it always just set to 3V, no matter what the power supply voltage is?
In the end the TPA311x have an internal voltage regulator for the analog side, if I am not mistaken. (can be seen in the block diagram)
Nice idea, but how is the stability?
I can imagine a bit tricky with so much feedback.