class-d output filter indispensable?
At the output of all tripath's chips there is a 2nd or 3rd order filter to cut all the frequencies above 20KHz.
They say you need it to rebuild the audio signal from a PWM signal.
I would like to know if is really indispensable.
Inside a speaker, woofers and midrange already have a low pass filter (with a cutoff frequency significantly lower than 20KHz), so, for this frequencies, the
signal is correctly reconstructed from the speaker itself.
What about the tweeter?
The tweeter is an inductance of low value in series with 8ohm resistance. This means a first order low pass filter. I don't know the value of this
inductance, but, an hipotetic cut off frequency could be 50kHz (even lower, I think).
The frequencies of the output signal are around 650KHz, more than 3 octaves higher, so I think they could be eliminated even from the tweeter inductance.
Three years ago I builded a ta2024-based amplifier. I used a fenice board, and, after some listening tests, I completely eliminated the output filter.
Here some photos.http://img256.imageshack.us/img256/9418/dsc0478n.jpg
(the contruction is horrible, but, it was my first project, and it was very low budget).
Please tell me if I am wrong.
I studied electronic only recently and with online courses, so my Knowledge is still little.
I appreciate every suggestion.
I'm by no means an expert, but I believe the purpose of the output filter is to eliminate the switching frequency and in particular it's harmonics, which may have significant power at radio frequencies, if not filtered out. While the speaker no doubt acts as a filter, the speaker wires will carry the switching waveform and cause EMI. A shielded speaker cable might prevent this problem. I am also interested in using D-class without an output filter, since the output filter needs to be designed for a specific load. Without a filter there tends to be less distortion, at least in amplifiers that use no feedback. A shielded cable should have a shielded connector, however I've yet to find anything I'd really like using. XLR might be ok, but I'd prefer a connector that can't be confused with an interconnect.
In a fast switching class D amplifier there may be things happening up to 100Mhz at the switching node. Feding this signal to wires and a voice coil directly is very likely to cause trouble, not only due to strong radiation on each switching edge but because the output stage requires a controlled load at RF to prevent resonances that can hurt reliability and overall circuit performance.
Some form of output filter is always required for anything except driving tiny speakers with low supply voltages (like in a cell pone or a laptop).
However, sometimes a full output filter may be replaced by an EMI filter made with ferrite beads (the ones that appear more and more resistive above 10Mhz and as a low inductance value below 10Mhz) and 1nF capacitors. This may be a good solution for not very long wirings and not too big voice coils.
BTW: There are some control schemes with post filter feedback that achieve load-independent frequency response (no peaking at filter resonance).
In this case it's not about what you want, more about what other people don't want messing up their lives ;) The filter is there to prevent undue RF from inflicting your music on the local population via the radio. Or disrupting other peoples activities by interfering with them.
thank you Corben, and thank you Eva.
I'll try ferrite bead + 1nF cap.
Is economic and fast, so is easy to try.
Corben, actually my speakers are all shielded, and I also use shielded and twisted speaker cables (I leave the extremities of the shield disconnected).
I've also shielded everything in my ampli.
Maybe this is the reason why my ampli doesn't suffer the absence of the output filter.
Thank you event horizon,
thanks to your suggestions now I understand what is the problem.
May be I am you neighbour...:D:D:D
I'd like to revisit this topic for further comments. I measured a Class D amp and could get huge peaking (+3 dB @ 20 kHz, +17 dB worst at 32 kHz) in the amp's frequency response.
Most speakers should have a very high impedance at the switching frequency, although most impedance measuring systems for speakers can't measure ultra high frequencies. But that's another thread, so let's assume for the moment the speaker impedance is very high at the switching frequency.
Why then would RF radiate out? With a high impedance, seems like current should be low through the circuit... (Please be merciful, I'm an audio guy not an RF guy!)
And, I'd like to hear more about the output needing a "controlled load": could some high resistance/impedance be placed in PARALLEL with the speaker output to achieve this?
At low RF frequencies (like 1MHz), signals tend to stay in wires okay, because the wire is only a small fraction of a wavelength. Unbalanced noise will still couple capacitively to the surroundings, in the same way that residential mains (hot and ground) radiates 60Hz hum into any nearby open circuit.
A properly balanced or shielded cable has very little stray field at a distance.
Screened cables are useless unless you ground the screen to the circuit producing the RFI. A floating screen does absolutely nothing -- the electric field is still the average of the two wires, which means RFI for unbalanced signals.
First post! I stumbled across this thread and thought I would comment... I work as an applications engineer at Maxim Integrated Products in the audio department. We manufacture many audio chips including class D amplifiers. Though I generally do not work with the class D amps, I do know something about them. There are a few reasons for the filter on the output:
The first is eliminating unwanted EMI (radiation) coming from the cables that can interfere with nearby euipment (and violate FCC regulations). Some class D amplifiers such as our newer amps have slew rate limiting and spread spectrum modulation techniques to reduce the amount of radiated emissions. We call our class D amplifiers "filterless" because they can pass FCC limits with 1m cable and a ferrite bead filter. Sounds like the Tripath amp also has this kind of technology. A ferrite bead filter should be good enough. 1nF is a bit big, we usually recommend somthing in the 330pF range. Some amplifiers will take a performance hit with larger capacitance. Also, the more capacitance, the less efficient (more power consumption) the amplifier will be.
The second reason for the filter is for meaurement purposes. The switching frequency will confuse a lot of audio analyzers. All our evalution kits ship with these filters for that reason. They are not needed in the application unless long cables are used (>1m or so). I would not connect an unfiltered class D amplifier to a tweeter however as there is a whole lot of high frequency energy in the signal. A dynamic speaker will be just fine.
As far as the radiation is concerned (I am not an expert in this), I think the cable has enough capacitance to draw some current at the switching frequency (and it's harmonics more so). I know that we go to fancy EMI chambers with our amps an a meter of cable and you can for sure see the effect of the radiation (even with an inductive load).
An app note I found on our site talks about how to design cabling to help EMI:
I thought spread spectrum doesn't reduce the emissions but just spreads them around so they get under the required limits. My guess is that FCC limits were intended as a worst case for discrete emissions, reckoning that there is a fair chance that a particular frequency is not in use at a particular location. Wideband emissions undermine this so really ought to have a lower limit.
I must declare an interest: as well as audio I am into radio so I don't like the electronic smog/graffiti spread about by modern electronics such as filterless Class D and (mainly Chinese) cheap SMPS.
An unfiltered Class D amp is a radio transmitter, and should be regulated as such.
|All times are GMT. The time now is 02:26 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio