My First MAX9744 Audio Amplifier Board [Issue]

[mafmouf]

Hi all,

First post here. Please don't be gentle.

Working on an audio amplifier board using the Maxim MAX9744 chip. With the Adafruit 20W stereo amplifier board schematic and MAX9744 datasheet as reference I used Altium Designer 17/18 to do the schematic and layout. The board itself has some pass through stuff but that's not relevant to my issue. Got the board made at PCBZone and bought the stencil for it. Squeegeed the paste over, manually placed components and reflowed according to the profile.

OK. So. I checked the nets and they all seemed to be fine. Used a bench top dual power supply to apply the 5V and... the power supply started switching the current limiting and the chip started heating! Not what I hoped for. I later realised the 5v ->3.3V regulator did not have the correct resistors added, hence resulting in supplying 5V instead of 3.3V. Turns out the absolute maximum rating for Vdd is 4V

I removed the regulator and applied 3.3V directly from the power supply where the regulator would be. Same current limit switching from the supply. Conclusion is the board could be dead.

I then decided to make up a second board minus the regulator and still directly apply 3.3V. SAME RESULT

From here, it get's beyond my understanding. I feel I've done something fundamentally wrong which is annoying. I looked to resolve it myself and check all the details again, but can't seem to find anything.

Any help or suggestions would be super useful. If I'm missing anything, please let me know.

Thanks!
Mafmouf.

[SiliconWizard]

Have you thoroughly checked that there are no short-circuits? QFN packages with a large exposed pad are bitches.

[mafmouf]

Sure did! I checked every single adjacent pair of pins to make sure nothing shorted. On 2 boards. This is why I'm so stuck.

[bandgap]

OK. So. I checked the nets and they all seemed to be fine.

Does "I checked the nets" mean that you checked for shorts on the board before soldering components to it? If not, then I would look at your 3.3V rail to ground isolation with a bare board. Measure the resistance and see if it's low.

What happens if you desolder the QFN from the board? Does the short go away?

-Clayton

[kosthala1]

Are you sure that you have placed the qfn with the correct orientation?

[bandgap]

Are you sure that you have placed the qfn with the correct orientation?

Looks like he did according to the datasheet.

[vlimit]

Hey Mafmouf,

I wrote a long reply to this but it went MIA, so medium post this time:

It may not be a circuit fault. What is the current draw?

How hot did it get? Even with 0W output it will get a bit warm. Could you hold your finger comfortably on the device?

Power amps can draw a lot of current and this one has good protection, so set your current limit to 1.6A (max power draw is ~2 x (3 + 0.8)W -> 7.6W for 5V PVDD), connect 4R speakers & ground the inputs. Then you have conditions that match those in the graphs around page 8 for 0W output, so you can start to compare against the datasheet.

The data sheet says thermal recovery takes in the order of 3s, so get your scope out and look at the current draw to see if that's happening. It will be easy to spot. There are some other things worth reading in the Current Limit section.

Check the output load with a meter.

Power devices like this often rely on multiple power pins to get power to different parts of the die. The L&R channels are likely separately powered. Check this by putting a meter on a spare device. If it is the case, scalpal the PVDD track to the L&R channels (one at a time) to see if you can isolate a fault to a particular channel. It's an easy repair afterwards.

The layout is important with this device. Worth getting a scope on your rails and the output to check that things are looking clean (compare against the scope traces in the datasheet).

Will be easier to help if you post artwork with the polyfills intact (given that PGND is disconnected) and use higher res screenshots that show all of the net names.

If you are doing another rev then consider:
- Vias under the thermal pad
- Fatter PVDD racks. Should match the output track widths
- PVDD should be 12V if possible (v^2/R) to get the value out of the device
- Connect adjacent pins on the same net directly to eachother.
- Might as well ditch some of your redundant decoupling caps (eg C25 can go).
- No PCB loops (eg the PVDD loop at pins 4 & 5. The loop is right next to a high current track).
- Don't leave it to polyfills to connect your nets (PGND). Route everything, then pour your polys. This lets you control how the plane currents flow.
- Cleaning up your footprint silkscreens.

[mafmouf]

OK. So. I checked the nets and they all seemed to be fine.

Does "I checked the nets" mean that you checked for shorts on the board before soldering components to it? If not, then I would look at your 3.3V rail to ground isolation with a bare board. Measure the resistance and see if it's low.

What happens if you desolder the QFN from the board? Does the short go away?

-Clayton

Sorry this wasn't clear. I checked before and after soldering the components that each net was isolated and not touching shorting with any other net.

Are you sure that you have placed the qfn with the correct orientation?

Yep. I checked at least twice to make sure. I even compared it to the Adafruit 20W Stereo Amp board.

Hey Mafmouf,

I wrote a long reply to this but it went MIA, so medium post this time:

It may not be a circuit fault. What is the current draw?

How hot did it get? I can use a temp gun next time to recordEven with 0W output it will get a bit warm. Could you hold your finger comfortably on the device? Nope, got to a point where I couldn't hold it for more than a second.

Power amps can draw a lot of current and this one has good protection, so set your current limit to 1.6A (max power draw is ~2 x (3 + 0.8)W -> 7.6W for 5V PVDD), connect 4R speakers & ground the inputs. Then you have conditions that match those in the graphs around page 8 for 0W output, so you can start to compare against the datasheet. Thanks heaps, I will give this a go. WHat do you mean by "4R speakers", though?

The data sheet says thermal recovery takes in the order of 3s, so get your scope out and look at the current draw to see if that's happening. It will be easy to spot. There are some other things worth reading in the Current Limit section.

Check the output load with a meter.

Power devices like this often rely on multiple power pins to get power to different parts of the die. The L&R channels are likely separately powered. Check this by putting a meter on a spare device. If it is the case, scalpal the PVDD track to the L&R channels (one at a time) to see if you can isolate a fault to a particular channel. It's an easy repair afterwards.

The layout is important with this device. Worth getting a scope on your rails and the output to check that things are looking clean (compare against the scope traces in the datasheet).

Will be easier to help if you post artwork with the polyfills intact (given that PGND is disconnected) and use higher res screenshots that show all of the net names. This is good feedback. I will update when I can.

If you are doing another rev then consider:
- Vias under the thermal pad. Is this common/standard practice?
- Fatter PVDD racks. Should match the output track widths
- PVDD should be 12V if possible (v^2/R) to get the value out of the device
- Connect adjacent pins on the same net directly to eachother.
- Might as well ditch some of your redundant decoupling caps (eg C25 can go).
- No PCB loops (eg the PVDD loop at pins 4 & 5. The loop is right next to a high current track).
- Don't leave it to polyfills to connect your nets (PGND). Route everything, then pour your polys. This lets you control how the plane currents flow.
- Cleaning up your footprint silkscreens. What's a good example of bad silkscreens in this layout?

Thanks heaps for the lengthy reply. Direct responses are in the quote above in blue.

There is a lot I probably need to inspect before reporting back.