Class D Design Issues

@nitrate
EVERY CLASS D I BUILD HOWEVER GOOD IT IS ALWAYS ENDS UP DEAD EITHER BY STARTUP PROBLEMS OR BAD LOADS!

You should try IRS20955 split supply version or IRS20124 single supply version. They are mosfet drivers for high and low side, designed specific for class d amplifiers and have integrated protections for power mosfets.I never had start up problems with my amps.
maybe fets get their supply before oscillator section has started!(this can short out output mosfets to load or to supply rails...)
What topology of mosfet drive do you use,
is there soft start implementation?
What is the nature of start up failure?
...
 
Variac said:
richie00boy

is this the ClassD4sure thread you were referring to?

http://www.diyaudio.com/forums/showthread.php?s=&threadid=42146

No it wasn't that, it was a specific thread about designing and building class-d amp from scratch. Maybe I will try to find it when I have more time.

PS I can see classd4sure is "account disabled" was he found to be a sock puppet or something?
 
Long time no 'D' lol

Wapajk,

My driver topology consists of MAX262 drivers driving the fet directly and magnetic isolators providing the drive signal to the max i.cs. The high side is powered off a bootstrap pump. Dead time is set elsewere, not relient on gate capacitance/resistance. It is not shoot through currents destoying the fets, i'm pretty sure it is the low frequency of startup causing too much current to be passed through the output filter. Also when dodgey loads are put on the output i think this is altering the filters responce causing the same thing. After all speaker loads are reactive and the filters parameters are somwhat dependent on the output load. All these factors combining gives rise to the failures. I will do more research on this and sort things later when i have more time, i'm really busy with other things at the mo. I have succsessfully created electric motor controllers with the same topology that can throw lorry starter motors around the bench so i'm convinced the problems are to do with frequency/filter interaction. I even once tried using dc-dc isolated converters instead of bootstrap, and implemented the old resistance in the gate deadtime instead of external control, it worked exactly like all my other output stages ( i must admit much simpler to implement but more expensive ) but the same old thing. Too much stress and bang!.. or just failure on startup. I must admit however the breadboard designs prob arnt helping things and could possibly be the root of the problem :cannotbe:

^^^^^^^^^ I know, i know, breadboard.. pls don't shout at me lol

Anyways, nice to visit the D class.. i'm a great fan of this topology..

LaterZ
Leigh
 
Sorry, maybe I've missed something. I looked at MAX262 datasheet - "MAX262 : uP Programmable, Universal Switched Capacitor Filter"
Did you mean MAX232 or something else.
I'm a bit confused here.

MAX232 is by no mean good driver for mosfets. Note that it has short-circuit current of only 60mA. that is to low for class D and also typical output swing of +-7.5V isn't to good(note that swing is rated under normal operation into load of 3kohm)
Also magnetic coupling isn't to good for gate driving in Class D(it can be ok if done correctly but will never be better than hi side and lo side mosfet drivers like IRS2110 or irs20955,irs20124,irs2011...)
maybe you could post schematic. It will help greatly!!
 
Opp's, sorry i meant MAX626, and the magnetic isolator in question is NVE's IL711-2. Both are pretty quick and allow me great flexibility in design, i also use them for other stuff too so that sort of influences my component choice :)

Some of my work is here

http://www.diyaudio.com/forums/showthread.php?postid=729138#post729138

... not sure if the schematics were up to date though. Should give you an idea of whats going on or rarther was lol

Leigh
 
My impression is that more than 50% of the unsolved DIY headaches in class D are caused by te difficulties of proper timing, proper gate drive control and parasitic resonances.

For a first rough idea about dead time you can watch both gate signals simultaniously while applying
zero volts rail at the power stage. As long as the rail is zero you do not need a differential probe to watch the high side gate and you will also not burn the power stage.
Only if these signals are OK, then it makes sense to power up the rails for the power stage.

To get a feeling about the power stage ringing in any class D amp it is be helpful to watch the gate drives and halfbridge sloping in a load condition slightly above filter ripple current. Most simple method to generate such a load condition is to connect a resistor between positive rail and amp output. The resistor should have a value that the resulting DC is slightly above the filter ripple.
In troublesome designs it is a good idea to start with reduced rail voltage. Already watching the lower gate drive and the half bridge sloping in this condition is usually telling you a big story.
To find undesired resonances you should have a scope with 100MHz bandwidth or higher. Depending on components and layout most designs show heavy resonances between 30MHz and 100MHz.
Curing the resonances is possible by ceramic smd caps at the rail, RCL snubbers at the rail and most effective by RCL snubbers between drain&source. The L of the snubber is the parasitic L of the snubber geometry, which you must not ignore. Unfortunately.
 
i found an early (1965) PWM in a IEEE circuit book. i built one of these about 20 years ago, and couldn't get it to work. i tried modeling it in LTSpice recently and found out why it didn't work when i built it. it needs DC bias on the input signal.
 

Attachments

  • class-d-1965.jpg
    class-d-1965.jpg
    16.7 KB · Views: 5,975
I'm tendando to repair the amp SPL Dinamics ext-3000D, I had to reconstruct the entire power by replacing all IRFP064N is pilotaggi 2sa1275 transistors and other components, rebuilt part because board went to fire! wings all ok, a problem also found in secondary 4 final IRFP360LC short only that I can not finding is replacing with IRFP360 but nothing goes into protection led green then red, tolgo line end where I added the final 4 audio is always the same I realize that warms the pre-pilot that these integrated smd installed 7 +7 pin (symbol carteggiata how do I go back to its value) on bassetta the tolgo is finamlmnete only part that makes a sound type frying is I realize that the integrated pre 4556D warms a lot but I do not know where to buy it, the problem is to know the code dell'integrato sanded welded on basseta you know the code! then there are also two integrated covered by black resin but I suppose they are good!

modul DWM3640nhv_v30 , I serve the value of U2, U1 production for what it says IRF

His clones with the same chassis:
hifonics Goliath GX3000d
Ground zero GZPA 1.5000 SPL
E-Audio 3000D
American Bass VFL 200.1