Hi All,
I thought, rather than pull he CD650 mods thread even further off topic, I'd start this one.
I've been having loads of fun recently with making my TDA1541A DAC a dual TDA1541A DAC. Its lovely. I was surprised, it might be partly because its output is louder (I suppose doubled) and the SNR is improved, but there is tons of detail, the sound does appear different.
Anyway, while doing this, I thought I'd finally tackle another issue I've had. It's subtle. I couldn't hear it through my speakers, but it was really obvious on quiet tracks through headphones. There was this kind of hashy (as if digitally induced) background noise. Not pleasant at all.
Now I've done several things to improve grounding, and its nearly gone, but one of the 'errors' I had made was that I thought it would be good to seperate out the agnd and dgnd an tie them together at the star ground. Reading about (I think it was Guido on ecdesigns' thread, and/or the TDA1541A info thread) it was mentioned that standard practice was to tie these two together as soon as possible -i.e. as close to the pins as possible.
I did this, well as close to the pins as possible, and the noise issue was drastically reduced! It was better (because it was the shortest route due to obstacles caused by components) to go via a decoupling ground point rather than a wire from pin to pin.
I still have a slight noise, but its all but gone and I suspect I'd need a ground plane to resolve this completely.
BUT, why is the TDA1541A so fussy about this, and the pins need to be tied together? In what circumstances would it be possible to have these two grounds seperate (i.e. why did Philips make it two pins in the first place?!).
Cheers,
Phil
I thought, rather than pull he CD650 mods thread even further off topic, I'd start this one.
I've been having loads of fun recently with making my TDA1541A DAC a dual TDA1541A DAC. Its lovely. I was surprised, it might be partly because its output is louder (I suppose doubled) and the SNR is improved, but there is tons of detail, the sound does appear different.
Anyway, while doing this, I thought I'd finally tackle another issue I've had. It's subtle. I couldn't hear it through my speakers, but it was really obvious on quiet tracks through headphones. There was this kind of hashy (as if digitally induced) background noise. Not pleasant at all.
Now I've done several things to improve grounding, and its nearly gone, but one of the 'errors' I had made was that I thought it would be good to seperate out the agnd and dgnd an tie them together at the star ground. Reading about (I think it was Guido on ecdesigns' thread, and/or the TDA1541A info thread) it was mentioned that standard practice was to tie these two together as soon as possible -i.e. as close to the pins as possible.
I did this, well as close to the pins as possible, and the noise issue was drastically reduced! It was better (because it was the shortest route due to obstacles caused by components) to go via a decoupling ground point rather than a wire from pin to pin.
I still have a slight noise, but its all but gone and I suspect I'd need a ground plane to resolve this completely.
BUT, why is the TDA1541A so fussy about this, and the pins need to be tied together? In what circumstances would it be possible to have these two grounds seperate (i.e. why did Philips make it two pins in the first place?!).
Cheers,
Phil
Because if there was one pin and the leads tied inside the package, the analogue and digital current loops would share a common impedance (a common bond wire inside) and so you'd get noise again.
Try to tie them together as close as you can to the package, consistent with minimising the bypassing loop areas.
Try to tie them together as close as you can to the package, consistent with minimising the bypassing loop areas.
Hi Martin,
I sort of see your point, but I still imagine the same would be the same with the join just outside the chip. Not sure what the difference is, unless perhaps its related to decoupling and the ground current loops related to that?
I never could get the hang of grounding issues!
Cheers,
Phil
I sort of see your point, but I still imagine the same would be the same with the join just outside the chip. Not sure what the difference is, unless perhaps its related to decoupling and the ground current loops related to that?
I never could get the hang of grounding issues!
Cheers,
Phil
yes, exactly that - imagine you want the bypass cap from anlog / digi supply returning as close to the body to the relevant ground pin as poss, then these pins are joined at or after that point. The currents flowing in each loop are kept separated as far as possible.
IIRC there's an AnalogDevices or Linear Tech application note which describes all this really well... why separate analog/digi ground plins, and why splitting teh ground plane here is a bad idea.
IIRC there's an AnalogDevices or Linear Tech application note which describes all this really well... why separate analog/digi ground plins, and why splitting teh ground plane here is a bad idea.
> why did Philips make it two pins in the first place
You got something like 5-10 nH of inductance in the pin and bonding wire etc (not sure of the value but should be in the ballpark). So that looks small but at 10 MHz that's about half an ohm of impedance... so the little fast bursts of current from the digital part of the chip over that small impedance would probably be larger that 1 16-bit LSB and ruin the chip's performance... so that's why 2 ground pins.
Look at the ESS sabre it has ground pins all over the place, which is excellent.
Anyway, there is always coupling inside the chip between digital and analog ground... so you should try to connect them with low impedance, the ideal thing being a ground plane...
You got something like 5-10 nH of inductance in the pin and bonding wire etc (not sure of the value but should be in the ballpark). So that looks small but at 10 MHz that's about half an ohm of impedance... so the little fast bursts of current from the digital part of the chip over that small impedance would probably be larger that 1 16-bit LSB and ruin the chip's performance... so that's why 2 ground pins.
Look at the ESS sabre it has ground pins all over the place, which is excellent.
Anyway, there is always coupling inside the chip between digital and analog ground... so you should try to connect them with low impedance, the ideal thing being a ground plane...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.