Hi Gentlemen,
I'm dipping my toes into Class-D amplifiers, and I've done a bit of testing on several of them, since I have access to an older Audio Precision machine at work (I work mainly with speakers).
I've noticed that the LC output filter of a Class-D amplifier can have a major impact on the frequency response of the amp. The ones I've seen so far (mostly on low-cost amps) have had a corner frequency near 20kHz, and they can be optimized either for 8 ohm operation or 4-ohm operation. When designed for 8 ohm operation, they usually roll off prematurely when connected to a 4-ohm load. This may be good for most speakers with an inductive HF driver, but might cause HF loss if connected to a non-inductive HF driver like a planar, AMT, or PZT driver. When designed for flattest response into a 4-ohm load, then the frequency response into an 8 ohm load begins to rise into a broad, underdamped "hump" just before the corner. I can only wonder what these output filters do to the amp's response curve into a very inductive load, like a full-range driver.
All of this has got me thinking - why not design these passive output filters for a higher corner frequency, like 40 kHz? That way, when these impedance-related effects happen, it's well outside the range of our hearing, and most tweeters are already beginning to mechanically roll off. I think it's possible that a lot of output filters are being designed by EE's that are just assuming that the top end of human hearing is 20 kHz, and then thinking, "Well, let's just put the corner frequency of the output filter there," but I don't know - maybe there's some good reason why a higher corner frequency would be bad.
Just to give my above question some more context, I've been playing lately with a TPA3110 board from Sure Electronics powering my desktop audio system, and I really like its sound for being such a low-power amp, but I found it created interference with my wireless mouse and keyboard. This chip is billed by TI as a "FilterFree" amp, through use of a proprietary switching method. I tried adding the recommended ferrite bead filter to the output wires (they were omitted from the board itself), and it helped, but the same RFI problem just manifests at higher playback levels now, so I'm thinking that the typical LC output filter is inescapable for very RFI-sensitive applications like a computer desk with wireless peripherals, and now I'm interested in ways to minimize the impact of the LC filter on the sound of the amp.
Anyone care to comment on the above? Thanks!
I'm dipping my toes into Class-D amplifiers, and I've done a bit of testing on several of them, since I have access to an older Audio Precision machine at work (I work mainly with speakers).
I've noticed that the LC output filter of a Class-D amplifier can have a major impact on the frequency response of the amp. The ones I've seen so far (mostly on low-cost amps) have had a corner frequency near 20kHz, and they can be optimized either for 8 ohm operation or 4-ohm operation. When designed for 8 ohm operation, they usually roll off prematurely when connected to a 4-ohm load. This may be good for most speakers with an inductive HF driver, but might cause HF loss if connected to a non-inductive HF driver like a planar, AMT, or PZT driver. When designed for flattest response into a 4-ohm load, then the frequency response into an 8 ohm load begins to rise into a broad, underdamped "hump" just before the corner. I can only wonder what these output filters do to the amp's response curve into a very inductive load, like a full-range driver.
All of this has got me thinking - why not design these passive output filters for a higher corner frequency, like 40 kHz? That way, when these impedance-related effects happen, it's well outside the range of our hearing, and most tweeters are already beginning to mechanically roll off. I think it's possible that a lot of output filters are being designed by EE's that are just assuming that the top end of human hearing is 20 kHz, and then thinking, "Well, let's just put the corner frequency of the output filter there," but I don't know - maybe there's some good reason why a higher corner frequency would be bad.
Just to give my above question some more context, I've been playing lately with a TPA3110 board from Sure Electronics powering my desktop audio system, and I really like its sound for being such a low-power amp, but I found it created interference with my wireless mouse and keyboard. This chip is billed by TI as a "FilterFree" amp, through use of a proprietary switching method. I tried adding the recommended ferrite bead filter to the output wires (they were omitted from the board itself), and it helped, but the same RFI problem just manifests at higher playback levels now, so I'm thinking that the typical LC output filter is inescapable for very RFI-sensitive applications like a computer desk with wireless peripherals, and now I'm interested in ways to minimize the impact of the LC filter on the sound of the amp.
Anyone care to comment on the above? Thanks!
Last edited:
Top end of hearing IS 20kHz at best, although for me it is more like 17kHz ;-(
To get rid of the impedance dependent filter behavior, post filter feedback is used (e. g. UCD or the like).
To get rid of the impedance dependent filter behavior, post filter feedback is used (e. g. UCD or the like).
The reason the -3dB point of the filter is not set higher is because it would reduce the amount of carrier wave attenuation.
Filter-free designs simply use the inductance of the speakers voice coil as part of the output filter, they are a cheap and nasty solution.
Filter-free designs simply use the inductance of the speakers voice coil as part of the output filter, they are a cheap and nasty solution.
Not quite. They use a modified modulation technique that reduces the impact of RF noise. Naturally it also have an impact on sound quality but refinements in recent years have all but eliminated that. The main thing with filter-free designs is to keep speaker leads as short as at all possible.
Thanks to all for the responses.
I notice that most output filters are of the second-order (12dB/oct) LC type. Could a third-order or fourth-order filter be applied at a higher corner frequency to achieve more switch noise attenuation while facilitating placing the filter corner further out of the audioband?
I notice that most output filters are of the second-order (12dB/oct) LC type. Could a third-order or fourth-order filter be applied at a higher corner frequency to achieve more switch noise attenuation while facilitating placing the filter corner further out of the audioband?
Yes it could, but such filters are bigger and more costly, and more sensitive to component tolerance and driver being close to ideal 8 ohms.
- Status
- Not open for further replies.