Dead Time and Crossover Distortion

[Genomerics]

As some of you may know I am putting together a website about switch mode power supplies and Class D amplifiers.

I've 'kind of' finished off the theory stuff about the switch mode power supply and moved on to the Class D amplifier part.

Whilst going through the motions I hit another 'thing'.

Cross over distortion due to dead time. I'd heard about it but never really thought about it..... then it 'appeared'. Now I might have to think about it.

Anyway, if you fancy a look then,

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

Home page is at

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

Sorry it was just one of those GOSH moments

Cheers

DNA

[flo]

Hi DNA,

Great website! Very informative and fun to read.

Can't wait for the next update

Florian

[Portlandmike]

Hi DNA,

I find the dead time induced distortion to be one, if not the most important root of distortion in class D's.

I also find it interesting in that no manufactures have jumped on the "class A~D" version of this. I'm sure they will come.

Its not nearly as bad as real class A vs class B, it just means you need to run the ripple current inductor very high, such that the peaks of the current are higher than the maximum expected output current. Yes the effeciency will be lower if all else stays the same, but it will still be better than any class AB on its best day! (Except maybe at idle.)

I'm planning to get around to doing some experiements with a UcD400 and smaller filter inductance, larger output caps....

Nice pictures on your site!

Mike

[Genomerics]

Originally posted by flo
Hi DNA,

Great website! Very informative and fun to read.

Can't wait for the next update

Florian

I had to take my head out of my bum to read this. Now I have to try and put it back in again.

Thanks, may all your breakfasts be seriously tasty.

Cheers

DNA

[analogspiceman]

Hello DNA,

It's been discussed on this forum before, but you are seeing the curious effect on deadtime of the peak-to-peak current of the output inductor current of a class-d amplifier.  As long as the output inductor current ripple is larger than the load current at the moment of switching, its current is in the right direction to force the output stage node voltage to begin slewing to the other rail immediately when both MOSFETs turn off at the beginning of deadtime (this is natural commutation - also known as soft switching).  For larger load currents, slewing doesn't commence until the end of deadtime when the other MOSFET forces the voltage to change.  Thus, unlike a linear class-b amplifier (where deadtime is at zero output) there are two deadtime zones at non zero load current (symmetrically spaced about zero).  Nice web site, by the way.

Regards -- analogspiceman

Here's a reference to a very good short technical letter on the subject of class-d deadtime distortion:

Effect of dead time on harmonic distortion in class-D audio power amplifiers

Mosely, I.D. Mellor, P.H. Bingham, C.M.
Dept. of Electr. & Electron. Eng., Machines & Drives Group, Sheffield

This paper appears in: Electronics Letters
Publication Date: 10 Jun 1999
Volume: 35 , Issue: 12
On page(s): 950 - 952
ISSN: 0013-5194
CODEN: ELLEAK
INSPEC Accession Number:6290250
Digital Object Identifier: 10.1049/el:19990705
Posted online: 2002-08-06 22:36:47.0

Abstract

A model is described for predicting the harmonic levels introduced by the use of dead time in class-D, PWM-driven audio power output stages.  The model demonstrates that the harmonic levels are a function of load impedance, modulation depth, dead time and switching frequency.  In addition, measurements show that, for audio applications, dead time is the dominant cause of power stage non-linearity.

Citing Documents

An integrated 200-W class-D audio amplifier, Berkhout, M.
Solid-State Circuits, IEEE Journal of
On page(s): 1198- 1206, Volume: 38, Issue: 7, July 2003

A 20-W stereo class-D audio output power stage in 0.6-/spl mu/m BCDMOS technology, Morrow, P.; Gaalaas, E.; McCarthy, O.
Solid-State Circuits, IEEE Journal of
On page(s): 1948- 1958, Volume: 39, Issue: 11, Nov. 2004

A current-mode power sigma-delta modulator for audio applications, Dallago, E.; De Leo, G.; Sassone, G.
Industrial Electronics, IEEE Transactions on
On page(s): 236- 242, Volume: 52, Issue: 1, Feb. 2005

[Genomerics]

Originally posted by Portlandmike
Hi DNA,

I find the dead time induced distortion to be one, if not the most important root of distortion in class D's.

I also find it interesting in that no manufactures have jumped on the "class A~D" version of this. I'm sure they will come.

Its not nearly as bad as real class A vs class B, it just means you need to run the ripple current inductor very high, such that the peaks of the current are higher than the maximum expected output current.

Nice pictures on your site!

Mike

Pictures are courtesy of LTspice and Xnview. LTspice lets me copy a 'bitmap' to the clipboard and then I can fiddle about with it in Xnview to crop it and turn it into a .gif file.

There is something going on in there. 200nS/2.5uS is 0.08. Also 50 + 50 is 100 and the 'blip' happens at 8V so, well I don't know. It's going to take some staring at but I'm sure someone else has explained it elsewhere.

Like I said it was a GOSH thing because, although I'd heard about it, I'd never seen it before. I mentioned it because it's a cute thing to see.

DNA

[Genomerics]

Originally posted by analogspiceman
Hello DNA,

It's been discussed on this forum before, but you are seeing the curious effect on deadtime of the peak-to-peak current of the output inductor current of a class-d amplifier.  As long as the output inductor current ripple is larger than the load current at the moment of switching, its current is in the right direction to force the output stage node voltage to begin slewing to the other rail immediately when both MOSFETs turn off at the beginning of deadtime (this is natural commutation - also known as soft switching).  For larger load currents, slewing doesn't commence until the end of deadtime when the other MOSFET forces the voltage to change.  Thus, unlike a linear class-b amplifier (where deadtime is at zero output) there are two deadtime zones at non zero load current (symmetrically spaced about zero).  Nice web site, by the way.

Regards -- analogspiceman

Thanks! That makes so much sense that I am feeling slightly stupid. I could have spent ages scratching my head over it...

DNA

[Genomerics]

Thank you Mr Analogspiceman, I have just quoted you on the website. I'll try to explain things better tomorrow.

If you want me to scrub that one then say so.

Cheers

DNA

[analogspiceman]

Originally posted by Genomerics
Thank you analogspiceman, I have just quoted you on the website. I'll try to explain things better tomorrow.

If you want me to scrub that one then say so.

If you think it promotes understanding, then quote away, by all means...

But perhaps, for the sake of better accuracy and clarity, I could ask you to make a slight correction to the text (replace "its current is in the right direction" so the text will read as follows):

It's been discussed on this forum before, but you are seeing the curious effect on deadtime of the peak-to-peak current of the output inductor current of a class-d amplifier.  As long as the output inductor current ripple is larger than the load current at the moment of switching, the sum of the two currents is in the right direction to force the output stage node voltage to begin slewing to the other rail immediately when both MOSFETs turn off at the beginning of deadtime (this is natural commutation - also known as soft switching).  For larger load currents, slewing doesn't commence until the end of deadtime when the other MOSFET forces the voltage to change.  Thus, unlike a linear class-b amplifier (where deadtime is at zero output) there are two deadtime zones at non zero load current (symmetrically spaced about zero).

Regards -- analogspiceman

[Genomerics]

Analogspiceman

Job done...

I will probably be using other words, assuming my head doesn't blow up in the process.

Cheers

DNA