Filterless Class D

[Dave Z]

By the way sacrificing chickens also helps...

kidding...kidding

Dave

[soongsc]

Originally posted by Dave Z
soongsc

The output voltage is determined by the duty cycle, not the frequency. If you are at 50% duty cycle, then your output voltage will be 1/2 your supply voltage regardless if you are switching at 250kHz or 1MHz. Do you mean ripple current in those output caps?

Dave

That's what I was thinking as well. But doesn't the audio signal change the duty cycle during operation? Would the speaker driver act like output bead?

[Ipanema]

Dave,

Thanks for the explanation. The problematic I mean is how would they affect my audio system e.g. burn tweeter, bad sound, interfere with my TV etc. I would like to know whether the amount of EMI present in the filterless class D chip offered by TI or Maxim will cause any problems mentioned above.

As for as I know, the output LC filter only filter off EMI at the output connected to speaker. With or without LC filter, the EMI issue still exist on the amp board. Hence, good PCB layout and shelding requirement are required for both filter or filterless design.

Regards.

[Dave Z]

The speaker will act like a filter which is part of the premise why filterless class D works. But, the inductance is not the same as a true inductor. The output inductors can run quite warm. This heat will now have to be dissipated from the voice coil so make sure you derate the driver. Also since the inductance is less, not as much high frequency will be attenuated so it will start to become harsh which is why I don't use filterless above 5W -10W or so. Unfortunaltey even though the speaker will attenuate some high frequency, it will only attenute differential mode high frequency so it will not reduce EMI like a ferrite bead (plus it is also at the end of the speaker wire in stead of getting rid of it at the source). Speaker drivers are also not low impedance at high frequencies.

You will see output frequency response graphs for class D ICs based on filter values for 4, 8 or whatever load impedance. This is theoretical and not real world. As pointed out a speaker really is an inductor. So as frequency rises so does the impedance. You must do an impedance plot vs frequency of the speaker cabinet to see what the impedance of the speaker is where the LC filer starts to roll off. This should be the entire network. If there are crossovers, zobels, and multiple drivers in the speaker cab, they must ALL be connected while doing the impedance plot. The point is to see the true impedance of what the amp sees. You may find an 8ohm cabinet may really be 12-20 ohms at 20kHz. This will effect the filter response. If Z is higher than what the filter is designed for, you could get an unmatched load which may cause the lowpass LC filter to have a high Q. You could actually get a rise in output power at that freq which may cause damage or sound harsh. If real Z is lower than what the filter is designed for you will be attenuating some high frequencies. It's a shame IC manufacturers don't explain that better.

Dave

[Ipanema]

THanks for the explanation. Class D is very tricky to work with.

[phase_accurate]

I would even dare to say that the parasitic capacitance of a voice coild could have a significant effect.

IMO filterless class-d is just something for handheld devices and the like.

Regards

Charles

[Esperado]

Originally Posted by Dave Z

Speaker drivers are also not low impedance at high frequencies.

Shame on the enclosure manufacturers to not compensate their loudspeakers for a flat impedance curve, adding compensation network both for inductance and motional impedance of their drivers.
Shame on the reviews to not put accent about the importance of a near flat impedance curve when their review speaker enclosures.