diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Class D (http://www.diyaudio.com/forums/class-d/)
-   -   My fully discrete ClassD (http://www.diyaudio.com/forums/class-d/54248-my-fully-discrete-classd.html)

MikeB 26th March 2005 11:29 AM

My fully discrete ClassD
 
Hi !

I've nearly sucessfully built my first ClassD, fully discrete using standard devices.
It plays like hell, but, i have some problems left..
First problem, without the 1nF in the VAS deadtime seems to be too short,
heating up the P-Channel...
The second problem, driver for opening the mosfets heats up, so i have to turn off after ~5 minutes.

The schematic i've constructed:

http://www.lf-pro.net/mbittner/ClassD/Classd_sch_3.GIF

The 0.05 ohms represent the fuses...

And the prototype: (not the latest)

http://www.lf-pro.net/mbittner/ClassD/IMG_0034.JPG

http://www.lf-pro.net/mbittner/ClassD/IMG_0035.JPG

Any ideas, suggestions ?

MikeB

destroyer X 26th March 2005 01:07 PM

Go ahead, discrete class D is great!...good luck!
 
Even with those little problems sounding good!...this is a promiss that will be very good when those details turned fixed.

regards,

Carlos

subwo1 26th March 2005 01:12 PM

Hi MikeB,
Have you tried increasing your final driver 33R emitter resistors? Maybe try just increasing the N-channel ones to something like 47 or 56R.

MikeB 26th March 2005 01:58 PM

Hi subwo1 !

No, have not tried yet, but you're right, the n-channel is faster !
The 1st version used 10ohms, then i increased to 33ohms, but did
not show any change... :( I will try !

What i do not understand, only the P-channel heats up, the N-channel
keeps cold, should crossconduction not heat up both ?

Mike

Giaime 26th March 2005 02:15 PM

Beautiful prototype!

About that P-channel heating, have you matched your mosfets? I suppose you did, but maybe they have drifted...

MikeB 26th March 2005 02:39 PM

Hi Giaime !

I did not really match the mosfets, i only matched them for identical
Vgs-threshold... I can't avoid having different Rds-ON, also the
p-channel has higher recoverytimes, but i have no idea how to
measure that !

Mike

analogspiceman 26th March 2005 03:05 PM

Re: My fully discrete ClassD
 
Quote:

Originally posted by MikeB [some snippage]
I've nearly successfully built my first ClassD, fully discrete using standard devices. It plays like hell, but, i have some problems left.. First problem, without the 1nF in the VAS deadtime seems to be too short, heating up the P-Channel... The second problem, driver for opening the mosfets heats up, so i have to turn off after ~5 minutes.

And the prototype: (not the latest)

http://www.lf-pro.net/mbittner/ClassD/IMG_0034.JPG

Any ideas, suggestions?

Since your output inductor is a coreless solenoid, it may be injecting enough stray flux back into your amp to cause problems. As a test, space it back on the end of a twisted pair of wires, 20 or 30 cm away from the rest of the circuitry.

I don't suggest it, but as a test to demonstrate how much trouble this inadvertent coupling can cause, you could then bend it back down very close over your circuit board, moving it and twisting it around until your circuit starts misbehaving (and probably breaks).

Regards -- analogspiceman

PS: What did you mean by "VAS"?

subwo1 26th March 2005 03:37 PM

Quote:

Originally posted by MikeB
What i do not understand, only the P-channel heats up, the N-channel
keeps cold, should crossconduction not heat up both ?

One thing that could heat the P-channel more is its greater rds-on. But like you were thinking, it would also seem to me that both should heat up, still. It may have something to do with the P-ch's body diode recovery being forced by the N-ch. Maybe something will occur to me later.

MikeB 26th March 2005 03:41 PM

Hi analogspiceman !

Hmm, i hoped the heatsinks would give some "shielding"...
A bit dumb, as the heatsinks are made of alloy...
Would it be enough for testing to put some steelplate between
coil and rest of amp ?

With "VAS" i meant the 2nd diffamp with currentmirror, loaned from
another ClassAB-design. With the 150ohm the voltageswing is
reduced to +/- 1.2volt, also giving low zout. The 1nF slows down this
signal, and with the 0.6v vbe-threshold of the following bjts i "cut"
this signal at it's half, giving ~100ns deadtime.

Mike

megajocke 26th March 2005 06:52 PM

I don't know if this is your problem, but 36 volts of gate drive is probably not good for those output transistors... At least not in the long run. But i guess puncturing the gate oxide would be more terminal than just heating up. Like Short and Kaboom!


All times are GMT. The time now is 02:37 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2