The TPA32XX seems like a good upgrade from the TPA31XX series with lower noise/THD and higher power. In addition, the power rails are now separated between left and right channel, potentially enable higher quality design.
One catch is the chip requires 30V supply to achieve the best results. The goal of this project is to design a power supply unit around the TPA3244 such that the amplifier always gets a clean 30V power in regardless the main power supply is used (input voltage should be >=14V and <=24V).
The main idea is to use a boost converter to step up the input voltage to ~32V then use an LDO to clean it out and generate the 30V supply. The parts I chose is following:
Boost converter: TI LM3488, at 1MHz switching frequency with a 10uH inductor and ~1mF output cap the ripple is about 70mV @4A output.
GVDD_X: This is the supply for the gate driver of the amplifier, 12V is recommended. 2x ADP7142 is used to generate the 12V directly from the 13-24V input. This LDO can achieve high PSRR all the way to 1MHz with low noise
PVDD_X: This is the supply for the output stage of the amplifier. 2x LT3086 are used to generate 30.4V from the 33V output of the boost converter.
The standoff holes are placed such that it can be stacked on top of an RPi. 21mm heatsink space is also left for the boost converter's diode and NFET and all the LDOs.
The schematic, layout is attached. The 3D figure can be downloaded from the following link:TPA3244_3D.pdf - Google Drive
The BoM can be viewed from TPA3244 BoM - Google Sheets
PCB top:
PCB bottom:
So, what do you think? Any suggestions are welcome.
One catch is the chip requires 30V supply to achieve the best results. The goal of this project is to design a power supply unit around the TPA3244 such that the amplifier always gets a clean 30V power in regardless the main power supply is used (input voltage should be >=14V and <=24V).
The main idea is to use a boost converter to step up the input voltage to ~32V then use an LDO to clean it out and generate the 30V supply. The parts I chose is following:
Boost converter: TI LM3488, at 1MHz switching frequency with a 10uH inductor and ~1mF output cap the ripple is about 70mV @4A output.
GVDD_X: This is the supply for the gate driver of the amplifier, 12V is recommended. 2x ADP7142 is used to generate the 12V directly from the 13-24V input. This LDO can achieve high PSRR all the way to 1MHz with low noise
PVDD_X: This is the supply for the output stage of the amplifier. 2x LT3086 are used to generate 30.4V from the 33V output of the boost converter.
The standoff holes are placed such that it can be stacked on top of an RPi. 21mm heatsink space is also left for the boost converter's diode and NFET and all the LDOs.
The schematic, layout is attached. The 3D figure can be downloaded from the following link:TPA3244_3D.pdf - Google Drive
The BoM can be viewed from TPA3244 BoM - Google Sheets
PCB top:
PCB bottom:
So, what do you think? Any suggestions are welcome.
Last edited:
Not the case when you need to catch fleas.
That is, the purity of food certainly matters. But reducing the length of the conductors to the inductors will be more important, because of the decrease in the effect of parasitic pickups from the output to the input.
Sorry for my bad english)))
That is, the purity of food certainly matters. But reducing the length of the conductors to the inductors will be more important, because of the decrease in the effect of parasitic pickups from the output to the input.
Sorry for my bad english)))
The wire length is minimum. The output inductors sit right next to the output of the amplifier. The track length is about 7mm.
The output filter cap is film capacitor. The EMI filter uses C0G type ceramic, should be even better than most of the film cap.
If TPA3244 is internally pretty much the same as the TPA3255 then based on my experience with the latter I'll suggest the following.
Don't bother too much with the 12V rail in terms of regulation, that's because internally there's another regulator taking this 12V for its input and putting out 7.8V for the analog signal processing circuitry. Its this AVDD rail (brought out for decoupling at pin9) which is the crucial one. I found a 1000uF/16V Nichicon HZ to make a considerable improvement to clarity when hung on the equivalent rail of the TPA3255.
Don't bother too much with the 12V rail in terms of regulation, that's because internally there's another regulator taking this 12V for its input and putting out 7.8V for the analog signal processing circuitry. Its this AVDD rail (brought out for decoupling at pin9) which is the crucial one. I found a 1000uF/16V Nichicon HZ to make a considerable improvement to clarity when hung on the equivalent rail of the TPA3255.
Be sure your twin output inductors are NON coupled, special class-d-output types.
Until now I could not find your pcb layout
Are you using kicad?
I am designing a TPA3255mono plate amp. The actual batch of pcbs will arrive within a few days.
Yes, this inductor is not coupled. They are specifically designed for TPA3244 by Coilcraft. I am using Altium designer, I am not sure how to upload the layout since it depends on the footprint library. The PDF I have attached shows the layout but it won't help too much.
PCB layout is crucial, so it would be beneficial if you post some screenshots here.
looking at your 3D image...
I strongly recommend coupling the terminal blocks in a row just to strengthen against mechanical torque when fixing the screws.
I strongly recommend coupling the terminal blocks in a row just to strengthen against mechanical torque when fixing the screws.
Attachments
thanks. I recommend studying the pcb-layout guidelines found in TI-appnotes.
You do not need more than two bulk low ESR elelectrolytics in the range of 470~2200uF.
Separate the power-side (pins23~44) strictly from the input side (pins1~22)
Locate the bulk electrolytics on the power side.
Minimize the both current loops VPP-pin-elcap-GND return pin.
Pushing the amp you will have ground plane currents that induce non neglible voltage differences. This can easyly spoil the excellent THD performance that these chips can achieve with a good pcb layout.
So my suggestions are
-stop the bottom power ground plane in the region of pins 1~23.
-Continue with a separate AGND plane exclusively used by the analog path.
-connect this AGND-plane with a single zero-Ohm resistor to the power GND-plane
-connect all TPA-GND-pins with top GND-plane, whereever possible
-Do some via stitching to connect top with bottom power GND-plane
btw - did you reserve space for the heatsink?
Last edited:
The L3 inductor seems to belong to a DC-DC converter? I am not sure what is the power level but the output trace looks thin to me.
Another thing I have noticed is that the input signals seem to be routed all the way around the board and pass the output stage. The output stage of class-D is very noisy, very high voltage overshoot before inductor because of the switching. Normally the inductor will block these noise from reaching the output but it can be coupled into the input trace and corrupt the (much weaker) input signals.
If you have to route the input signals, maybe put them on the bottom side and put a lot of vias along them to create a GND shield?
Thanks a lot for the suggestions about the ground plane. For the heat, TPA3244 is not as powerful as TPA3255 (only 1/5th power) so it is not a big problem. The chip is supposed to use PCB as the heatsink. There is enough space to put a small 10mm by 10mm heatsink on the top of the chip, it should stay cool with a quiet 12mm fan.
One question regarding separating the ground plane, all the gnd pins are internally connected together in the chip, so if the power gnd level has been changed, the internal analog gnd plane will be changed accordingly. In this case, is it still useful to separate the AGND and PGND?