Problem with LPF on Output of Class D Amp

[MarkCSI]

Hi,

I am trying to use a Class D Amp to drive a non-audio application (360 - 12000Hz, load will vary), and am getting distortion on the output when no load or lighter load is present. The basic flow is:

Class D Amp -> LPF -> Step Up Xfmr -> Output

I am using the LPF in the Eval kit (Max9737). The LPF is designed for an 8Ω load.

What I am seeing is not the high freq switching distortion, but much lower. I have included a scope trace . BTW, it happens with or without the XFMR. The FAE said it is because ther LPF is tuned for 8Ω.

I have a few questions:

1. Does this make sense to anybody

2. He sais I should simulate the LPF to optimize. How does one create the class d signal for simulation?

3. Does anybody have a recommendation for a free/cheap simulator?

Thanks!

Mark

 

[chuckey]

What is F+ and F- in your CRO display? What is the wave form into the transformer? - is it a proper audio transformer or just a mains transformer?, does this effect show it self at low frequencies?. I see the load is given as 4K, what is the reflected load into the amplifier? does the amplifier work OK straight into a 8 ohm load?
Frank

 

[MarkCSI]

F+ and F- are the output of the MAX9737 Eval board. I have attached the schematic. This problem shows up only with light loads. If I put the 8Ω load on the output of the Eval board, it looks fine (other than a little 300kHz remnant of the class d switching). I have not checked with loads other than 8, 4k, no load, but I suspect it gradually goes away from 4k down to 8. It is not an audio XFRMR, but this problem shows up even without the XFRMR.

Do Class D amps only work with lower loads, or is this an artifact of their LPF? The Maxim FAE claims it is the LPF.

Thanks for your help!

Mark

 

[chuckey]

You could try damping sections of the filter with a 1K resistor to stop them ringing or putting a load of 80 ohms permanently to load the filter a bit.
Frank

 

[godfreyl]

3. Does anybody have a recommendation for a free/cheap simulator?

I recommend the free version of SYMetrix SIMPLIS. It's good and also easy to learn and use (at least compared to LTspice, imo).

 

[chuckey]

I have been pondering on your problem, and I don't think its anything to do with the filter. Consider a 300KHZ squre wave, this has got harmonics at X3, X5, X7 etc, now if you PWM it with a 10 KHZ sine wave, You get a sideband at 10 KHZ, 300 +-10 KHZ, 900 +-10 KHZ. Your LPF has to reject any thing higher then say 100 KHZ, only allowing the base band component through. So how can it generate 20 KHZ, 30 KHZ etc, i.e. base band components of distortion?
One normal attribute of a class D amp is to recover the energy in the frequencies rejected by the LPF (hence the inductive input impedance) and return it to the Vcc supply. The Vout(FOUT+&-) of the amplifier IC shows severe clipping and it should be a 300 KHZ PWM signal, so what is going on here? Has your CRO got a restricted bandwidth? There is also a little zener diode circuit, has this been used correctly?
Frank