1-2 cm for the high-frequency decoupling is way too far. Put two X7R dielectric ceramic capacitors right by the IC. I generally use 1 uF, 50 V for this. Then next to those add two 10-33 uF electrolytic caps. I often use 22 uF but have used 33 uF as well. These need to be close to the IC as well. Then you want two bulk capacitors (min. 470 uF but 1000 uF is better and often lower cost than 470 uF as it's a more popular value). Put these were the power enters the PCB.
You can read about my rationale for this decoupling here: https://neurochrome.com/pages/supply-decoupling
Also see the Application section of the LM3886 data sheet.
You can find an example of this decoupling scheme and see component placement here: https://neurochrome.com/products/lm3886-done-right
Tom
You can read about my rationale for this decoupling here: https://neurochrome.com/pages/supply-decoupling
Also see the Application section of the LM3886 data sheet.
You can find an example of this decoupling scheme and see component placement here: https://neurochrome.com/products/lm3886-done-right
Tom
it depends on too many factors to give a single number, but:
- The closer to the IC the better;
- The smaller (physical size) the ceramic decoupling cap the better because of lower package inductance, e.g. 0603 instead of 1206 and SMT instead of TH (inverse packages like 0306 and 0508 are even better in this regard, but slightly pricier);
- The larger the capacity the better, for example, 2.2 uF X7R 0805 will perform better here than a 0.1 uF 0805 X7R if they are otherwise completely the same.
Edit: also, ceramic caps with higher voltage ratings suffer less from a reduction in effective capacity when biased with a DC voltage as is obviously the case here. X7R dielectric ceramics are the way to go, none of that X5R rubbish.
Last edited:
It's usually not about degradation, but oscillation. High speed/sensitive ics can suffer even if 5mm away from pin. Usually manufacturers write not to go above 2mm in datasheet for such products. There has been a lot of problems of this kind with swapping for modern opamps in older gear where decoupling was not that well done..
In any case, neuro has a lot of exp with LM3886, i'd take his word for that info
Amp chips can oscillate at RF and still amplify audio fairly well - but with raised noise and distortion, so oscillation (if it doesn't overheat the device) can show up as degradation of the signal quality. Also if instability is borderline the oscillation may only be happening in some parts of the audio waveform, so for instance might only affect loud transients and sections.
Sure. And maybe your tweeters will survive that. Maybe.
There's no doubt that lower inductance between the decoupling network and the IC will be better, so why not just scoot the components up to the LM3886? Your question is akin to "have you ever measured how much sugar you can add to gasoline before your car's performance suffers?" Why add sugar in the first place?
Be a bit careful with the physically small packages. You're right about the inductance. Smaller package -> lower package inductance. But often the voltage coefficient of the physically smaller capacitors is worse than that of a physically larger capacitor, even for the same dielectric, so you need to make sure you pick a capacitor that gives you enough capacitance. There's definitely a tradeoff there.
Tom
At RF frequencies the impedance of the speaker tends to RISE, limiting the amount of current it draws. More likely to kill tweeters oscillating at 100 kHZ than in the MHz. Also more likely to actually happen. If it’s enough power to damage a tweeter the zobel resistor will be getting HOT. If it’s not, you are below damage level. But possibly enough to get the amp to audibly misbehave with the valid audio signal.
If you have a zobel getting hot, do NOT connect the speaker until the situation is corrected. It’s a very simple test - put your finger on it. If you see smoke obviously you can skip that step.
If you have a zobel getting hot, do NOT connect the speaker until the situation is corrected. It’s a very simple test - put your finger on it. If you see smoke obviously you can skip that step.
Or if you are lucky enough to have a thermal camera you can avoid burning your finger(!) Often a hot part will smell hot, so I'd say look for smoke, check for hot smells, then touch tentatively.
You used to be able to get tee-shirts and mugs that changed colour when hot - ideally you'd be able to get resistors and other components with the same pigment that would allow quick identification of hot components - but that's a big ask I suppose.
You used to be able to get tee-shirts and mugs that changed colour when hot - ideally you'd be able to get resistors and other components with the same pigment that would allow quick identification of hot components - but that's a big ask I suppose.
Hello,
I have little knowledge about the parts used on the fifopi Q7 reclocker but there should be some chips there i guess.
The designer and users keep writing great stories about improving power supplies. There is a so called dirty side needing 5 volts and there is a 3,3 volt where people " go mad " using big supercaps.
These coke can seize caps are not that easy to put close to the circuit. BUT theoretically one should not use a long cable between the caps and the board. Will try reducing it from let us 8 inches to 2 inches be beneficial?
Most people will use a big regulated supply board while current will be around 60 mA but big supply board will make it harder to put it closer to where the 5 volt is needed.
The board has a terminal to attach the supply wires. My idea is to use a tiny 5 volt shunt supply made by a local technician. Am i right to think that a properly designed shunt supply should work both ways so also keeping " dirt " produced by the digital board from other elements in your advice? I will remove the connector from the digital board and connect the shunt circuit right there where i think it should be. That shunt circuit has a small cap at its output and if it to far from the " consumer" one should add a similar cap right at the chip.
So many people make this device with several " signal" boards and several supplies and then having cables running all around. It looks so " ready for improvement" to me.
These dc supply cables should they be routed close to chassis, twisted?
All suggestions are welcome.
Greetings Eduard
I have little knowledge about the parts used on the fifopi Q7 reclocker but there should be some chips there i guess.
The designer and users keep writing great stories about improving power supplies. There is a so called dirty side needing 5 volts and there is a 3,3 volt where people " go mad " using big supercaps.
These coke can seize caps are not that easy to put close to the circuit. BUT theoretically one should not use a long cable between the caps and the board. Will try reducing it from let us 8 inches to 2 inches be beneficial?
Most people will use a big regulated supply board while current will be around 60 mA but big supply board will make it harder to put it closer to where the 5 volt is needed.
The board has a terminal to attach the supply wires. My idea is to use a tiny 5 volt shunt supply made by a local technician. Am i right to think that a properly designed shunt supply should work both ways so also keeping " dirt " produced by the digital board from other elements in your advice? I will remove the connector from the digital board and connect the shunt circuit right there where i think it should be. That shunt circuit has a small cap at its output and if it to far from the " consumer" one should add a similar cap right at the chip.
So many people make this device with several " signal" boards and several supplies and then having cables running all around. It looks so " ready for improvement" to me.
These dc supply cables should they be routed close to chassis, twisted?
All suggestions are welcome.
Greetings Eduard