I have read through most of this forum, and I have found some issues which come close to explaining what I am seeing. As there was no real resolution with those posts, I am creating a new one here :)
I am attempting to use the TPA3111D1 amp in BTL configuration to drive an 8Ohm. PVCC is 12V. My audio input is referenced to a virtual ground at 4V. I have placed large 10uF X7R caps on both input pins, with INN connected through its cap to my virtual ground to take advantage of the chips CMRR. I am using an LC output filter, with component values chosen for an 8Ohm load. I also have the option to tap the output before the reconstruction filter to compare functionality with and without the output filter. SD pin is tied to FAULT. My input capacitors are both between 3-8mm from the pins (but not equal distance!). Also, I have 2 identical boards that are showing the same issue, i.e. I don't think it is simply a faulty IC
The problem I am seeing is the following:
Not always, but increasingly more frequent (recently every time), the chip will fault during power on. Along with the IC shutting down, there is a large current spike and a loud pop-POP (quiet-loud) from my speaker. This fault occurs with and without a load attached to the outputs, also with and without an output reconstruction filter. I suspect it is a DC detect fault, as it requires a hard reset of PVCC to resolve. Comparing input pin voltages with the working EVM board, there are clear discrepancies found which I would like some guidance with.
My start up procedure for measuring the input pin voltages is always the same:
1. Hold down SD switch on the amp
2. Turn on main power
3. Wait some time (1-10s) for all voltages to stabilise.
4. Release SD switch and measure DC pin voltages before the fault occurs (fault occurs in expected time of ~500ms for a DC detect fault)
I see the following discrepancies between the functional EVM board and the problematic custom board. Note the EVM and my custom board are not identical
Property EVM Custom
Cin 1uF 10uF
INN ref voltage GND 4VDC
I see the following results comparing the two
Functional EVM Board:
SD pin held low, PVCC applied: INN = 5.18V, INP = 1.28V
SD pin forced to 1, PVCC applied: INN = 6.92, INP = 2.93V
Chip powers-up without fault
Problematic Custom Board:
SD pin held low, PVCC applied: INN = 0.63V, INP = 0.00V
SD forced to 1, PVCC applied: INN = ~1.4V (before shutting down), INP=0.00V (no change)
pop-POP and chip faults
This INP result is concerning. it should not be held at 0.0V. I have replaced the caps on the input but no improvement is found.
What about the different trace lengths from the INN/INP pins to their respective caps?
Is 10uF too large? Can there be problems at the inputs due to discharging caps?
Should I design a buffer stage before the class-d input to take care of impedance mismatching? The INP signal is currently taken from the output of an active filter stage, INN is simply tied to Virtual GND. A buffer would give a similar low impedance output closer matching the virtual ground impedance
So there are my issues, I will attempt to sort them out on my own, yet if someone has seen this sort of behaviour and has figured out a solution, I would be most grateful for some guidance.
Tim van Boxtel
Some further information . . .
The INP pin was clearly shorted to GND. I was worried that the thermal paste that I chose to connect the PowerPAD to the heat plate was not optimal, so I went and got the TI recommended Arctic Silver Ceramique 2 compound and replaced the amp chip with the new thermal paste. This should be ok, no? I intend to use a solder paste stencil and reflow techniques to connect the PowerPAD in production runs, but I want an easier method for prototyping.
The new amp still faults on start-up, the double pop are strangely more spread out than before, perhaps due to the chip input not being shorted to GND, so the DC detect trips slower? The voltage on both input pins is now around 1.2V when the SD pin is held low. When pulling the SD pin high, both pins go up to around 3.5V before faulting out.
When I disconnect the input caps from the rest of the circuit, the amp turns on without fault. This is clear evidence of a problem on the input. Any ideas?
It does sound like the DC detect function of the amp is protecting the speakers due to the large DC offset at the inputs caused by the impedance difference between the INN and INP during startup. Please take a look at the INN, INP, OUTP and OUTN (before LC filter) on the scope. For DC detect to trigger, the differential voltage between INN and INP should be at least 17mV for 420mS. This min voltage depends on the gain settings (DS pg 12). The outputs differential duty cycle should exceed 14%.
Also try using a smaller input caps such as 1uF or 0.1uF. This will give a smaller time constant and the inputs will be charged up to the 3V DC bias faster. I can review the schematic if you share.
Paul C. Chen
Audio Applications Engineer
Texas Instruments, Dallas TX.
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.