Dead Time and Crossover Distortion

[classd4sure]

You can turn the body diodes right off right with the use of a few diodes.

Relying on them too much will cause heavy EMI problems and likely failed mosfets.

I tend to think ZCS with sincle cycle, current mode control is the answer here.

 

[phase_accurate]

IMO better dead-time control like using adaptive schemes would be a more elegant approach than just increasing ripple current.
I always prefer solving problems where they arise and solving the deatime problem with better deadtime control would be just that.
If done in integrated form it would even be more cost efficient for large quantities than any workaround.

Regards

Charles

 

[Portlandmike]

IMO better dead-time control like using adaptive schemes would be a more elegant approach than just increasing ripple current.
I always prefer solving problems where they arise and solving the deatime problem with better deadtime control would be just that.
If done in integrated form it would even be more cost efficient for large quantities than any workaround.

Regards

Charles

Charles,

I studied this a year or two ago. Adaptive dead time does little to solve the problem it turns out. That's assuming you do a good job of adjusting it in the fixed dead time approach. At the time I had my head fully into it, but my recollection is faltering right now. I wish I could remember how I came to that conclusion, but right now it alludes me.

Bottom line is there is always a error caused by inductor current and it causes uncertainty. It takes longer to slew the current when it needs to be pulled down. Ideal dead time control is better, but an amp will always be lower distortion when its operated within the range of the ripple current. And, the distortion is not nice from a hearing standpoint in that its odd, and fairly high order.

Its very akin to saying good A/B bias is just a better way then Class A, and in many ways that's true if cost matters at all. I'm always more interested in best... not necessarly most economical. I"m not trying to make a living on it 🙂

Mike

 

[Genomerics]

Hi DNA,

Great website! Very informative and fun to read.

Can't wait for the next update 😀

Florian

Ahaaaaaa, try a bit of.......


http://www.genomerics.org/hbvmc.html

DNA🙂

 

[ChocoHolic]

Hello again

I've just uploaded the first part of the explanation as to what is going on. It looks like there are two things that need to be accounted for.

The first is a DC offset (just done that) and the second is the point at which this occurs (Too be Done).

Cheers

DNA

http://www.genomerics.org/outfilt.html

For some reason I am getting the message that this WEB site is forbidden for me. Do I need a special access allowance, or might it be forbidden in China?
But I am allowed to enter the main page....

 

[Genomerics]

Sorry, I have a tidy up so all of the links moved around.

The front page is at

http://www.genomerics.org

and the page where the part about crossover distortion begins is at

http://www.genomerics.org/classd/hbfl/outfilt.html

DNA

 

[ChocoHolic]

...THAT's quite nice !!

 

[sovadk]

Attached is a MATlAB script made by Flemming Nyboe from Texas Instruments. It can calculate the DC transfer function of a power stage and the resulting THD plot among others.

The script accounts for three types of errors:
1. Dead time distortion. The two deadtime zones in the TF are located as function of the ripple current. Also their witdh is a function of the ripple current. This script predicts that there's not gain in THD if you go for a large ripple current and achieve soft switching for all output currents. This has earlier been suggested her at diyaudio, that a large ripple current should solve all THD problems associated with dead time. But forget thant one guys!
2. Ron nonlinearities due to FET body diode conduction.
3. Finite turn of speed (effective threshold voltage varies with the drain current and distorts the modulation)

It's a very interesting piece of code and gives a very good insight to how different parameters affect the openloop distortion. Have a look at this screenshot:

An externally hosted image should be here but it was not working when we last tested it.

All my respect to the Class D guys at TI. They make cheap openloop chips (not much respect there), but the guys at the design team really know what they are talking about.

 

[sovadk]

There is also some slides associated with his MATLAB code. It might clarify how some of the distortion mechanisms are modelled.

In advance, sorry for the bad quality of the scan (I suck).
http://www.student.dtu.dk/~s042302/diy/Flemming-thd.pdf
Again it would be nice to be able to exceed the 100kb limit. If I could upload these things to this site, then it stays here forever and not just until I've finished at the university. Mods please!!!

 

[lumanauw]

IMO better dead-time control like using adaptive schemes would be a more elegant approach than just increasing ripple current.

Hi, Phase accurate,

How can I make an adaptive dead time system that automaticly adjust and ensures exact point between no dead time and no shoot through ?

 

[Eva]

I don't think that adaptive dead time is really required. If the circuit provides good switching precision, the delay from gate drive to the switching node becomes linearly dependent on output current (due to switching di/dt) and this does not produce substantial distortion, it only increases open loop output impedance slightly.

 

[ChocoHolic]

...also my impression is that you do not need an adaptive system, but high timing precision.
Basically this one is adaptive, but unfortunately designed for deadtimes which are to long for high quality class D. .... may be for subwoofer amps with low switching frequency...
http://www.national.com/pf/LM/LM5104.html