Electronics > Projects, Designs, and Technical Stuff
Larger output choke for cheap class D amplifier
Circlotron:
I've got several of these class D amplifier modules.
https://www.ebay.com.au/itm/IRS2092S-500W-Mono-Channel-Digital-Amplifier-Class-D-HIFI-Power-Amp-Board-FAN/311844643896?hash=item489b63a438:g:bfcAAOSwU8hY63c-
I haven't actually driven a speaker with them yet, only a resistive dummy load. I've got them powered with +/- 67 volts and they switch at IIRC 377kHz or thereabouts. At the output there is about 2V p/p ripple at the switching frequency. If you have a 20kHz sinewave coming out the amplifier at only 10V p/p it looks pretty ordinary, with lots of shaky wiggles. Not only that, the output inductor runs quite warm to the touch, even with no signal. I know for a fact that I would never be able to hear this switching frequency ripple, but I like the idea of reducing it a whole lot so that the signal looks cleaner. With the existing inductor the frequency response starts to roll off at about 45kHz, but I can only hear to 7.9kHz so we can get a bit aggressive with the inductor here. What would be a good starting point for toroid material, size and inductance? Size is not an issue - it doesn't have to fit on the board. The output filter cap looks to be 470nF in the pic.
BrianHG:
That output inductor supposed to be a tuned circuit to remove the 377Khz. I'm assuming that the 2vpp leakage is at 377Khz. That tuned circuit should have a really negative db notch at that frequency. Changing the inductor/capacitor value for a larger one may simultaneously cut some of your audio band, but, you chance the new circuit will no longer have quite as good 377Khz notch. (This is assuming the design on that board is properly tuned in the first place.)
Let's take a look at this from a slightly different angle. +/-67v supply is a 134v p/p 377Khz signal feeding that inductor/cap circuit. The output is only 2v p/p. That's 67:1. Instead of modifying that circuit, what if you were to externally add a second identical filter. That remaining 2v p/p will now be down to 0.03v p/p. If not, the oscillations you measure may be ground bounce through you scope probe and the only way to get a good reliable measurement here is to use a differential probe, or double check the 2v p/p reading with an external true-rms voltmeter.
Changing the current inductor with one with a higher Q and or choosing a lower series resistance cap on the output will also lower that 2v p/p. If the heat your module is generating is not because of the poor power mosfet switching timing, but because that the current inductor doesn't have a high enough impedance at 377Khz, then changing that inductor will make your module run cooler. A test to find out may be removing the module's filter cap. If this doesn't affect the amp's drive of the mosfets and the module runs a LOT cooler, then you know that a better inductor may be the way to go.
Your inductor choice probably needs to be the same inductance and at minimum capable of handling the same amount of raw current, however, you want an inductor with the really high Q as close as possible to the 377Khz, especially paired with the module's filter cap.
As for the inductor selection, or you winding your own, there are others here better suited to answer that question for you.
Circlotron:
--- Quote from: BrianHG on August 06, 2020, 05:02:17 am ---That output inductor supposed to be a tuned circuit to remove the 377Khz. I'm assuming that the 2vpp leakage is at 377Khz. That tuned circuit should have a really negative db notch at that frequency. Changing the inductor/capacitor value for a larger one may simultaneously cut some of your audio band, but, you chance the new circuit will no longer have quite as good 377Khz notch. (This is assuming the design on that board is properly tuned in the first place.)
--- End quote ---
Thanks for your reply, BrianHG. I have to disagree with you somewhat. If the cap and inductor resonated at the switching frequency, seeing they are in series they would be practically a short circuit at that frequency. Minimum impedance anyway. BIG current would be flowing. What's more, the voltage across the cap (speaker terminals) would be off the planet. I'm guessing the inductor would be about 100uH 33uH and in conjunction with 470nF they would resonate at 23.2kHz. Having said that, I wonder why things don't go pear shaped when the amp outputs a 23kHz signal?
Also, it's definitely the inductor that gets hot. The rest of the amp is quite okay heatwise. I'm expecting it is skin effect as well as core losses. Litz wire might smarten things up.
Edit -> simulation says inductor is 33uH.
BrianHG:
--- Quote from: Circlotron on August 06, 2020, 06:01:00 am ---
--- Quote from: BrianHG on August 06, 2020, 05:02:17 am ---That output inductor supposed to be a tuned circuit to remove the 377Khz. I'm assuming that the 2vpp leakage is at 377Khz. That tuned circuit should have a really negative db notch at that frequency. Changing the inductor/capacitor value for a larger one may simultaneously cut some of your audio band, but, you chance the new circuit will no longer have quite as good 377Khz notch. (This is assuming the design on that board is properly tuned in the first place.)
--- End quote ---
Thanks for your reply, BrianHG. I have to disagree with you somewhat. If the cap and inductor resonated at the switching frequency, seeing they are in series they would be practically a short circuit at that frequency. Minimum impedance anyway. BIG current would be flowing. What's more, the voltage across the cap (speaker terminals) would be off the planet. I'm guessing the inductor would be about 100uH and in conjunction with 470nF they would resonate at 23.2kHz. Having said that, I wonder why things don't go pear shaped when the amp outputs a 23kHz signal?
Also, it's definitely the inductor that gets hot. The rest of the amp is quite okay heatwise. I'm expecting it is skin effect as well as core losses. Litz wire might smarten things up.
--- End quote ---
Opps, did I say resonate. The inductor should have the highest possible series resistance at 377Khz, IE, not conduct the signal through. As for the cap, it should be in parallel across the speaker's +/- output to ensure no residue 377KHz makes it through. I must of made a mistake, but in the past, when we wanted to erase such a tone, we would create a notch filter with a tuned inductor whose impedance would have highest decibel cut/highest resistance at the frequency we wanted to eliminate, the 377KHz.
BTW, for the IRS2092, the cap is parallel to GND across the speaker output. The inductor is is acting like a series choke/notch filter as seen in the datasheet example schematic:
Circlotron:
20 kHz output, 4V p/p into 8 ohms.
Navigation
[0] Message Index
[#] Next page
Go to full version