Which is –in your opinion- the ‘digital’ and which is the analog part of a PWM class D amplifier?
George
I've done very careful listening tests with ferrite beads in the inputs of a preamp.
The ferrites are designed to attenuate RF ingress above 20 MHz. Below 1 MHz, the insertion loss is low, and at audio (i.e. under 100kHz) it's in low single ohms.
There is NO audible difference with the ferrites in place, other than a complete absence of RFI.
Better still, when I look at the response with a signal generator and scope, I see no anomalies that could possibly point to anything detrimental happening with the sound
The ferrites are designed to attenuate RF ingress above 20 MHz. Below 1 MHz, the insertion loss is low, and at audio (i.e. under 100kHz) it's in low single ohms.
There is NO audible difference with the ferrites in place, other than a complete absence of RFI.
Better still, when I look at the response with a signal generator and scope, I see no anomalies that could possibly point to anything detrimental happening with the sound
Last edited:
Warning, trap here!
If the input is analog (and there is no DSP section), I consider the whole class D PWM amp as an analog design.
If the input is a digital stream (say I2S), the input section is of a digital design.
George
Last edited:
Ferrite is ok, the problem is how it's used. I've heard awful effects nearly every time I've applied certain types. However I have no objections to transformers and CMC's, etc... there's a reason that I've noted.
I could probably tell you numerous things that make a big difference... You're welcome to PM me.
In general the models I've seen correctly predict complex impedances and their implications, the problem in usually the user.
A good place to look might be the difference between the impulse response measured with an impulse vs. any of the techniques that spread the energy over a broad range of frequencies.
Why would anyone volunteer to support your business endeavours for free, only to be insulted three pages later?
When a designer is not competent enough for a particular task, and has to ask for free help at a diy forum, the crazy one isn't the designer, it's the person who's willing to pay him and the person who teaches him how to do the design properly and does so without sharing in the paycheck.
You could do a lot worse than study the writings of a couple of classD designers on here - Chocoholic and Eva. Also study the implementation details of the IRS2092 from the various evaluation circuits available from IR (now part of Infineon).
Class D suggestion: If you're going to use MOSFETs in common source switching configuration to produce the high current square waves, spend a bit of time considering the idea of inserting a low-value noninductive resistor in series with each drain. External resistances on the order of RdsONmax and/or on the order of 2-3 times RdsONtyp. Why? So that your risetime, and your final flat-top voltage level, is repeatable from one MOSFET to the next, from one batch of MOSFETs to the next, and from one finished Class-D amplifier to the next.
Another suggestion: when designing your nonoverlap generators, focus on fast output risetime more than you focus on getting a complete and total nonoverlap where the waveforms cross literally at the rail. You've got Vth of safety margin to play with: exploit it!
Another suggestion: when designing your nonoverlap generators, focus on fast output risetime more than you focus on getting a complete and total nonoverlap where the waveforms cross literally at the rail. You've got Vth of safety margin to play with: exploit it!
Last edited:
Just Class D problems, not how to design Class D. Others will design the digital part, and they have done very well, so far as I know. I doubt that any here could or would tell me anything useful knowingly. I am not personally going into Class D design, I will leave that to Bruno and others.
- Status
- Not open for further replies.