Strange phenomenon with TDA7294

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
If a psu's voltage is dropping but not appearing to reach the set current limit and is in otherwise good shape then it is reaching its limit but faster (peaks) than the display can report. If the indicated current goes up when you add capacitance to the psu's output then this is what is happening as the caps are able to source the current peaks and not trip your psu's current limit.
 
I've also observed that paralleling the onboard rail buffer 'lytics (which really aren't that big at 100 µF...) by 6.8 mF elcos increases the 4 ohms output power and the rail voltages under this condition significantly. So can I suspect the bench supply's current regulation/protection being a bit too fast?
Best fegards!
Yep. You are going to pull about 8Apk with 4Ohms @32V rails. Therefore, for lower frequencies your 3Apk supply isn't even capable of delivering 4Apk for an 8Ohms load without running into current limiting.
 
I've also observed that paralleling the onboard rail buffer 'lytics (which really aren't that big at 100 µF...) by 6.8 mF elcos increases the 4 ohms output power and the rail voltages under this condition significantly....
An experiment...
So, at the chip, parallel pairs 220u or 270u, then series diodes connect the dc cable. One diode per each rail, close to the amp. Preferably ordinary power diodes with high amp, low volt spec (datasheet voltage drop graph looks like minimal variance under load). At the point the DC power cable connects the diodes put 1 (one) 2u~4u7 250v cap from v+ to v- and optional ferrites at same locale. After a few inches of cable a minimum of 13,500u per each rail.
And, then measure again.
Carefully.

I had used an actual speaker during a cruder test. Although I got some bruises when my chair failed because of a little panic, I did find the output power most entertaining, and pointlessly too much for a house. :)

The functionality that I should claim for the little mod is reduced stereo crosstalk distortion. However, that's not why I liked it. Indeed, some clipping and similar useless signal, got lost, both in stereo and in mono as well.
Actually, why I liked it is subtract more than a volt of noise and get more headroom, from. . . where???

Anyway, get that far, and then undervolt for more generous tolerances that can be used for better quality. By many weirdly different ways to measure, capacity for hi-fi is just over 40 watts per chip, same as LM3886. But, you can multiply that wattage figure by the number of chips used. So, use the voltage corresponding to the ~40 watt target figure.
I don't know much about elegant measuring, but I imagine it could be informative to measure it working comfortably within normal capacity.

Edit: My idea there was arrange the boost, and then spend it for better quality.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.