PAM8403 class D amp chip

[paulsherwin]

I've just bought 5 modules built around the PAM8403 chip from China for the princely sum of 99p. The modules are just a surface mount chip and a few other components on a tiny PCB. I've never used a class D chip before so thought I would buy some and have a play.

Does anybody have any comments regarding these chips? I wondered if there might be RFI problems - we shall see.

[Ed_Dinning]

Hi Paul, plenty of decoupling and very careful track layouts.

This particular beast does not use an output inductor, they could be a nightmare if the Q was too high/ low.

Ed

[ThePillenwerfer]

I've been using one to power a pair of PC speakers for a few years and never had any trouble with it.

It runs from a spare USB port as it's required voltage is low enough. I'd previously used a pair of LM386s but they were prone to pick up electrical notice from the computer, and weren't as powerful on 5V.

The only downside I'm aware of is that the only way to use them for mono is to only use one half; they can't be used in a bridged configuration.

[paulsherwin]

Thanks for the comments.

[MrElectronicman]

I have a couple of these as well, and also used them to replace an LM386. Not had any problem with them so far.
I was interested in the fact they could be run at 5 VDC from a cell phone type of charger, and am using one as a bluetooth remote speaker system, and the other as Raspberry Pi speakers.

[Argus25]

Originally Posted by paulsherwin

I wondered if there might be RFI problems - we shall see.

Paul,

Silicon Chip magazine in Australia did an excellent two part article on Class D amplifiers complete with a project to build one, it was based on the IRS2092 IC and used a pair of IRFB5615 output mosfets. It had an amazing 250W into 4 Ohms or 150W into 8 Ohms output. It ran off +/- 50V supply rails. This was in the Nov & Dec 2012 issues.

The switching frequency was 500kHz and the output filter consisted of a 22uH (5A rated) series inductor and a 470nF "X2" polypropylene filter cap (sorry to mention an X2 cap) , the -3dB point being about 49.5kHz, so there is about 43dB of attenuation of the nominal 50V rms switching waveform, which results in about 0.4V of the 500kHz signal remaining output.

So it is possible it could radiate RFI, but they took extra measures:

In their design they also added a snubber network after the filter described above, a 10R 1W resistor with a series 100nF cap....but also on the input to the filter, directly on the output (the lower mosfet's drain connection & the upper mosfet's source connection) they also added another snubber network consisting of a series 150pF capacitor and and a 10R 1W resistor to slow down the dV/dt of the switching waveform and cut back the EMI.

In addition they added clamp diodes (1N4004's) from the output (directly on the mosfets) to the + 50V rail and the -50V rail to prevent the output voltage excursions exceeding those values.

I hope that info is of some practical help,

best regards,
Hugo.

[Argus25]

..it is interesting on your PAM IC that they say its low EMI, probably because its all low power, but if it did make EMI you could always add a pair of small inductors in series with each speaker lead and a capacitor across them or the ferrite beads and the capacitors they suggest to ground in their data sheet.

[paulsherwin]

These are just tiddly little amps. Apart from general interest, I'm considering them as an alternative to an LM386 where a radio has an unrepairable amp stage, and as the basis of a signal tracer. This would usually mean leaving one of the amps unused so I need to investigate the implications of that for power consumption. The 5V supply rail is a useful feature in many applications.

I suspect small PCB modules like these may be used a lot by hobbyists in the future, given that through hole ICs are becoming increasingly difficult to find.

[Argus25]

That will be an interesting experiment to put them in a radio, especially an AM radio, because the AM radio is one of the most sensitive sniffer devices for RFI.

[dseymo1]

I've successfully used one of these modules to 'repair' a DAB radio, whose O/P amps were unidentifiable. I also have one driving a pair of bookshelf speakers for TV and casual Bluetooth listening.

At work we used one to power external 'speakers from one of our Mil Spec. laptops. The output went through the same 38999 (posh military connector) as the ethernet, ethernet stopped when the amplifier was in use. The solution was four, one per output line, bridged stereo, SMT inductors and capacitors. This clobbered the harmonics very well. The rise and fall times of these amazing devices is very short, in the order of nano seconds, to keep the efficiency up. The filter components ended up costing more than the amplifier chip.

[paulsherwin]

It is certainly a tiny chip to produce 2*3W without any heatsinking. It must generate almost no heat.

[julie_m]

A Class D amplifier delivers a pulse-width modulated output. It dissipates very little power because it is switching hard "on" or "off", so there is either no voltage across the switching transistors or no current through them.

[Argus25]

Probably then if the dV/dt at the output was reduced with snubber networks (directly on the output terminals) the efficiency would start to drop as it would create an intermediate state of conduction for the output devices and there would be more heat generated by the chip. But series L and parallel C would not do that as much and would at least get the RFI off the wires feeding the speaker, as these will act as antennas.

It was interesting that for the Silicon Chip project, they had to use two snubber networks and the L-C filter to tame the RFI, mind you its very high power, but still, the principles remain the same.

(One thing of interest here is that to get a "square looking" waveform, where the rise and fall times appear relatively rapid compared to the fundamental frequency, the rise/fall times need to correspond to at least the 21st harmonic. So lets say the switching frequency is 500kHz there will be RFI splattered all over the radio bands right up to 10.5 MHz).

[G8HQP Dave]

Although the datasheet announces it as "filterless" the small print advises adding a filter.

Be aware that "filterless" may be code for 'spread spectrum'. EMC regs were written under the assumption that interference would be mainly at discrete frequencies; a few sprogs can be tolerated because there is a good chance that they will not be on critical frequencies so can be ignored. Wideband noise ought to be at a much lower level because it affects all frequencies, but the regs do not require this. Hence some bright spark came up with the idea of getting below the threshold by spreading the muck everywhere.

[paulsherwin]

Originally Posted by paulsherwin

I've just bought 5 modules built around the PAM8403 chip from China for the princely sum of 99p.

They arrived today. The modules are absolutely tiny, about the size of a thumbnail. It's hard to believe such a thing can deliver 2*3W but we shall see. I'll post again if I discover anything interesting.