SystemD_2kW, any interest for an open design?

[ChocoHolic]

...finally... Fully assembled.
TO220 diodes for freewheeling on the backside. Required cooling depends on application.

[Reactance]

Originally Posted by ChocoHolic

...finally... Fully assembled.
TO220 diodes for freewheeling on the backside. Required cooling depends on application.

Firstly. Good job, looks professional a roadmap for a fully SMD design.

Did you manage to switch it on with a full volatge applied to it ? if not all the best.

[Baldin]

Originally Posted by ChocoHolic

...finally... Fully assembled.
TO220 diodes for freewheeling on the backside. Required cooling depends on application.

Looks real cool

[ChocoHolic]

Originally Posted by Reactance

Did you manage to switch it on with a full volatge applied to it ?

Of course it's up and running.
Pretty important thing is to change R410 and R411 to low values (50...100R), otherwise the clamping function of Q205 and Q206 is to soft and in combination with integrating gains the amp could fall into LF oscillation, when triggered with large rectangle signals.

After this, the intended fundamental behavior was in place.
Step response with rise times around 5us.
Clipping around 1.2kW into 2R, means with two amps in bridged mode we will get the 2.4kW into 4R as intended.
Noise is Ok, at least I cannot hear the noise on the speakers.
Dynamic hysteresis is also working and keeps output ripple small up to about 90% of clipping limit.

Attached a screen shot of pushing 1.2kW into 2R - slightly clipping - means 140Vpp.
And a second screen shot with 90% output voltage, showing still low ripple.
Probes were 10:1 in both cases.

...have to chase distortion mechanisms during the next days...
This is still a topic which my sims do not cover in a fully satisfying way.
So real life search and optimization is still necessary.

Will post further measurements in the next days.

For all who want to build the amp, please some patience.
As soon as everything is checked and optimized in an acceptable way, I will publish a builders package. Please do not start before that.

[phase_accurate]

Thats'r what I call dedication ! Really cool job. Congratulations

Regards

Charles

[slshellas]

That's what It needs a project. A very serious and deeply knowledge of the subject.
Congratulations !
Regards
Pan

[Workhorse]

Excellent build Choco

Cheers......

[ChocoHolic]

...Reactance asked for timing diagrams, so let's start with the measurements which are related to gate drive timing in idle situation.

The first two screen shots show Uds (upper trace) and Ugs (lower trace) of the lower MosFet.

Screen Shot 1 shows the half bridge sloping from negative rail to positive rail.
After the Ugs is being pulled below the treshhold voltage it still takes some tens of ns before the N-chanel becomes non conductive and the sloping of Vgs begins.
Despite being field controlled devices - MosFets have an turn off delay, which is caused by its internal gate resistance and it capacitances.
...Vds begins to slope, this is caused by the inductive current of the output filter.
About 20-30ns after the sloping of Vds has started it becomes more steep.
This is the point when the upper MosFet becomes conductive.
So with this dead time setting (25ns at the output of the IRS20957) and totem pole BJTs I am working with a forced sloping during idle.
This causes some more idle losses, but reduces dead time distortion.

Screen Shot 2 shows the half bridge sloping from positive rail to negative rail.

While screen shot 1 & 2 are showing only truly measured traces, the screen shot 3 & 4 additionally show a third constructed trace which shows the anticipated gate drive of the upper MosFet.

All measurements where done with 1:10 probes

[ChocoHolic]

Besides timing - the measurements of Vgs and Vds of the lower MosFet also can tell us whether the circuit behaves well at higher loads or if it is suffering badly from HF resonances.
For this I am typically doing a forced hard switching test.
Input of the amp shorted. From output to positive rail a load resistor.

The attached screen shots show the behavior with 6R load from the amplifier output to the positive rail.
While such tests in amps with 40V rails are easy going.. at 60V rails the arcing becomes already respectable.. and this amp with the 82V rails really works like a welding machine..
No uncontrolled ringing. Trouble free level of disturbances.
Compared to the earlier values in the schematics I adjusted the snubber to 10R and 330p.
A more sophisticated resonant snubbering with mini torroids across the legs of the snubber resistors showed further slight improvement, but is not worth the efforts.

[doctormord]

Great engineering work! Congrats!

Cheers doc