View Single Post
Old 11th May 2015
Here for the gear
studio96's Avatar

Even if the post is a bit old, some adds:

There is much more about making an amp than just using read built power devices. The mentioned ICEpower module are really well, that is right, but powering is an issue. I used them too when rebuilding my oonitors and did some tests.

Fortunately I had access to pretty precise analog network analyzers making it possible to meassure impedance and transformation curves directly. Also a very fine granular spectrum Analysis can be done with my eqipment.

First: Ripple damping is not ideal with class D amps in general (as always claimed) but ranges from something like 40dB to 80dB. So one can observe the introduced ripple very well in the output signal. If we want 24bit precision audio, care has to be taken to handle this. One can show easily why it is a difference to use linear power supplys or switched power supplies with low and high power head room.

Second: ICEpower modules do have a known issue. There will be a significant residual wave in the range of the switching frequency which is strongly dependent from the input stability. Which other class D's are more spread frequency can be observed. In theory this high frequency should not harm the speakers, but the debris wich can be meassured after the filters in the loud Speakers especially at the teater's is significant and again way from 24 bit precision true audio. The impact of this frequency can be limited in adding the regulator's output to the final signal by a special auto adaptive circuit which performes compensation. This is done in many high precision regulation Loops like sigma delta conversion. I managed to lower this residual power by ratio 3 by adding 4 passive devices.

Third: Regarding the above issues it is a difference if more than one module is powered from one or two supply paths. In my final system I am using an individual SPS-supply for all the modules. For my application this is ok but for an ideal supply a linear power supply was even a bit better.

Forth: Power Supplies emit EM-frequencies and so do nearly all the power signals in the device. Therefore shielding is a big task. It is pretty easy to meassure the left channel's signal with something around -100dB in the right signal's path and the same can be done with the dirt from the power supplies.


Having a closer look at all the devices I know, which use these ICEpower or also the Hypex or similar, they all look different, use other concepts and some devices clarify that the designer put the emphasis on cost saving than on quality.