Originally Posted by irribeo View Post Bridge rectifiers in output Sure3116 will guide the oscillation voltage spikes that occur without speakers connected towards the PVCC capacitors. Post#1 tells us these could burn 50V filmcapacitors in open output instantly said:
Last edited:
Interesting PS that I am using on my DIY project.
HWB060S-15 Sanken | HWB060S-15-ND | DigiKey
Can be adjusted about -/+ 10%
Dwight
HWB060S-15 Sanken | HWB060S-15-ND | DigiKey
Can be adjusted about -/+ 10%
Dwight
You can calculate how much PVCC will rise in that scenario.
Energy in an inductor: 1/2*L*I^2. Energy in a capacitor: 1/2*c*V^2.
Lets set up a worst case scenario. PVCC is 18 volts, and there's 7.5A flowing in the 10uH output inductor (the output current limit of the TPA3116) - this won't happen in an unloaded output condition, but we'll pretend it's the case. And the output filter capacitor is at PVCC, and is extremely large in value, so that all of the inductor's energy is dumped back into PVCC and none into the output capacitor. Oh and there's two output inductors (one per channel) doing this.
Energy in the inductors = 0.5*10e-6*7.5^2 * 2 = 563uJ.
Now lets assume there's 100uF of decoupling capacitors on the PVCC rail, which is awfully low. Energy in those = 0.5*100e-6*18^2 = 16.2mJ. So adding our 563uJ of inductor to 16.24mJ, we get 16.76mJ of energy now stored in our PVCC caps. 16.76e-3 = 0.5*100uF*V^2, V = sqrt(16.76e-3/0.5/100e-6) = 18.31V.
To blow a capacitor, you need to put enough energy in it to raise it beyond it's voltage limit... and that takes energy.
Now if there's no rectifier present to clamp the outputs, and the amplifier's driving the undamped LC output filter with no load, certainly it's possible to blow the output capacitors. But your PVCC capacitors are safe.
Energy in an inductor: 1/2*L*I^2. Energy in a capacitor: 1/2*c*V^2.
Lets set up a worst case scenario. PVCC is 18 volts, and there's 7.5A flowing in the 10uH output inductor (the output current limit of the TPA3116) - this won't happen in an unloaded output condition, but we'll pretend it's the case. And the output filter capacitor is at PVCC, and is extremely large in value, so that all of the inductor's energy is dumped back into PVCC and none into the output capacitor. Oh and there's two output inductors (one per channel) doing this.
Energy in the inductors = 0.5*10e-6*7.5^2 * 2 = 563uJ.
Now lets assume there's 100uF of decoupling capacitors on the PVCC rail, which is awfully low. Energy in those = 0.5*100e-6*18^2 = 16.2mJ. So adding our 563uJ of inductor to 16.24mJ, we get 16.76mJ of energy now stored in our PVCC caps. 16.76e-3 = 0.5*100uF*V^2, V = sqrt(16.76e-3/0.5/100e-6) = 18.31V.
To blow a capacitor, you need to put enough energy in it to raise it beyond it's voltage limit... and that takes energy.
Now if there's no rectifier present to clamp the outputs, and the amplifier's driving the undamped LC output filter with no load, certainly it's possible to blow the output capacitors. But your PVCC capacitors are safe.
Agreed on both points.
Which is why I would use a fast rectifier to protect the output caps if I was in the habit of potentially operating without speakers.
🙂
At 18V nothing is dumped on PVCC capacitors, output capacitors are only protected above PVCC voltage by the clamp, are they not??? In your example outputcapacitors still get full output.
Last edited:
The output caps without diode protection could get 147 times more than the 100uF used in the calculations. 100/.68
45V
If someone assumed that only a 35V cap was needed for the filter...
Last edited:
even with simple dmm one can measure 132V and 114V etc etc, far above 45V with very slow meter
You can calculate that too. It's a series LC circuit. Peak current in the L is 7.5A limited by the TPA, 0.5*L*I^2 as before is 281uJ. At resonance, that 281uJ will swing between the L and the C.
281uJ into a 0.68uF cap... 281uJ = 0.5 * 680e-9 * V^2... solving for V, 28.77 volts peak. Now that's on top of 1/2 PVCC average voltage, so your output cap voltage at worst case 24V rails is 28.77+12 = 40.77 volts peak. Using 10A as the current limit to buy a bunch of headroom gives a delta-V of 38.36V, on top of 12V that's 50.36V volts peak across your output cap.
A DMM won't accurately tell you the peak voltage on a several hundred KHz waveform. Only tool that can really do the job is an oscilloscope.
281uJ into a 0.68uF cap... 281uJ = 0.5 * 680e-9 * V^2... solving for V, 28.77 volts peak. Now that's on top of 1/2 PVCC average voltage, so your output cap voltage at worst case 24V rails is 28.77+12 = 40.77 volts peak. Using 10A as the current limit to buy a bunch of headroom gives a delta-V of 38.36V, on top of 12V that's 50.36V volts peak across your output cap.
A DMM won't accurately tell you the peak voltage on a several hundred KHz waveform. Only tool that can really do the job is an oscilloscope.
Last edited:
I'm treating each output phase separately here.
If you have a film cap between two phases and they both do the exact same thing just in opppsite directions, voltage across it is 28.77V multiplied by two = 57.54V. Or 38.36*2 = 76.72V with the 10A current limit #. Yeah, use a 100V cap.
Edit: never mind, you're halving the cap value if you put it between the two rails. I'm posting this from my phone, I'll do a proper analysis later.
If you have a film cap between two phases and they both do the exact same thing just in opppsite directions, voltage across it is 28.77V multiplied by two = 57.54V. Or 38.36*2 = 76.72V with the 10A current limit #. Yeah, use a 100V cap.
Edit: never mind, you're halving the cap value if you put it between the two rails. I'm posting this from my phone, I'll do a proper analysis later.
Hi guys-
I've lost track- has anyone found a 3116 2.1 that they liked. I know the earlier discussions weren't too positive.
Happy Holidays 😀
I've lost track- has anyone found a 3116 2.1 that they liked. I know the earlier discussions weren't too positive.
Happy Holidays 😀
Thanks for posting that. Intro paragraph from the datasheet: "The HWB series employs proprietary Soft-switched Multiresonant Zero-cross (SMZ) type resonant-mode circuits to achieve large noise reduction in the converter unit. Moreover, this is a switching power supply which has realized ultra-low noise (ripple voltage, conducted emission, and noise electric field strength) like a dropper power supply, employing a proprietary resonant mode
hybrid IC and transformer."
Is that at all similar to the Connex power supplies, for example, the SMPS300RS? From the Connex description: "Half Bridge resonant SMPS [that] uses zero-voltage switching half bridge resonant topology... the transistors are soft-switched, and the current through the switches has sinusoidal shape, and there are no additional losses in output inductors. Also, the amount of EMI is much less than any conventional power supply, and could be compared with a linear well regulated power supply."
Does 10uH 680nF oscillate at 610khz or 61khz ? And with ceramic 680(470)nF this frequency drops to very low frequency, isn't that correct? Dug you might be able to see that with monoboards and slightly better than Sure' ceramics probably?? I don't think oscillation frequency is that high at all, yet dmm could be off by large margin.
Fr = 1/(2*pi*sqrt(L*C)) = 61KHz with 10uH/0.68uF.
Ceramic vs film won't make a difference, though the capacitance of X5R/X7R ceramics does drop with increasing voltage.
Ceramic vs film won't make a difference, though the capacitance of X5R/X7R ceramics does drop with increasing voltage.