Hacking Hardware GBA GBA VREF audio quality mod research

[Sono]

I've had great success smashing an axial 220nF ceramic capacitor in parallel across the VREF capacitor on my GBA SP and GBA Micro, so I want to ask others to try it out, if we can stop spamming useless snake oil "power clean" capacitors everywhere, and use science instead to just install one.

Board placement:

• AGB: C39, next to AMP-AGB
• AGS: C21, under the cartridge slot shield, needs lead length extension and maybe isolation so it doesn't short against the shield, and the shield can close
  • AGT: similarly C21, but on the other side going to the PMIC instead of the AMP-AGB, so there is seemingly not enough space there for this mod
• OXY: C41, just barely fits under the display, hitting both the back of the display, and the right side of the bezel

Chip pins:

• AMP-AGB: pin3, other side goes to AGND (which is tied to GND, excellent audio quality even with IPS mod, noise is borderline inaudible)
• PMIC: pin9, other side goes to AGND (on Micro it goes to GND lead of the headphone jack, which sounds like it causes some mild ground loop, or some other noise leaking in, but still a lot better with than without)

Please tell me if the install is successful or not!

Spoiler: Technical data (unfinished)

I'm still researching this, but it seems like there are multiple things in play.

Adding extra capacitance on VREF seems to improve stability of the bias voltage.

Also, the *shape* and size of the capacitor also plays a big role, as a big axial ceramic capacitor has much lower resonant frequency than the tiny SMD 0402 (USA units) one, meaning that it affects the signal differently, even if both have the same capacitance. Sadly I don't know the math behind it, but testing seems to confirm this somewhat.

I still have yet to find how VREF is generated, and from what basis (did not test if it's affected by the volume slider or not), but it seems to be the center bias voltage for turning single-ended input into differential, for amplification.

 

[impeeza]

From a completely noob here, Which are those snake oil "power clean" capacitors?

 

[Nephiel]

From a completely noob here, Which are those snake oil "power clean" capacitors?

Those are usually sold as kits for classic handhelds, claiming to get rid of audio hum, hiss or buzz on speakers/headphones.
They are little more than a big capacitor (or multiple small ones in parallel), connected across the output of the onboard power regulator. This smooths the voltage to the rest of the board, and can reduce power supply noise, but the results vary.
They are essentially just a few cents in parts, sold at a premium, so most of the time you'd be better off replacing all the old aging electrolytic caps on the board with new ones.

Instead, the method OP mentions is increasing the capacitance (and likely changing other characteristics) of a specific 100nF decoupling cap, used for the reference voltage of the audio amplifier IC. There's an old thread about it.

I'd expect the value, chemistry, and ESR of the capacitor to affect the results more than its shape, but I don't really know that much about the subject.

 

[impeeza]

Those are usually sold as kits for classic handhelds, claiming to get rid of audio hum, hiss or buzz on speakers/headphones.
They are little more than a big capacitor (or multiple small ones in parallel), connected across the output of the onboard power regulator. This smooths the voltage to the rest of the board, and can reduce power supply noise, but the results vary.
They are essentially just a few cents in parts, sold at a premium, so most of the time you'd be better off replacing all the old aging electrolytic caps on the board with new ones.

Instead, the method OP mentions is increasing the capacitance (and likely changing other characteristics) of a specific 100nF decoupling cap, used for the reference voltage of the audio amplifier IC. There's an old thread about it.

I'd expect the value, chemistry, and ESR of the capacitor to affect the results more than its shape, but I don't really know that much about the subject.

Thanks mate for the explanation.

 

[Sono]

Instead, the method OP mentions is increasing the capacitance (and likely changing other characteristics) of a specific 100nF decoupling cap, used for the reference voltage of the audio amplifier IC. There's an old thread about it.

I'd expect the value, chemistry, and ESR of the capacitor to affect the results more than its shape, but I don't really know that much about the subject.

Oh, the linked thread is interesting! They have it wrong way around, the arrow points to pin9 on AGB-AMP, which *is* VREF.
As for globbing pin9 and "pin10", don't! On some motherboard revisions I have seen it being GND, so it would short VREF to GND, which may not be so great for the chip in the long run...

And yeah, capacitors are weird. But size and shape also seems to matter. Electrolytic causes high-pass noise, whereas axial ceramic seems to low-pass the noise. The on-board SMD 0402 ceramic is clearly inadequate even from the start.