Self oscillating fun

[Lars Clausen]

EVA: you might get me to jump out of an aeroplane, with a parachute!

You can get around the body diode thing by either using a low volateg schottky in series with the MOSFET, and a fast 600V diode (not necessarily SiC, even if they are the best so far).
This is the way a 'certain dutch company' gets around it at their biggest module.
You can also switch actively, so the body diodes never go in the conductive quadrant. It's not impossible.

Right now i am working on a 25 kW grid-tie inverter for power storage (like UPS). It's actually very much like building a huge Class D amplifier.
I am guessing your application is something like this too. Maybe for solar energy?

[Eva]

My previous project was focused on 60Hz mains synthesis but my client deceived me and left me with a lot of unpaid work.

This new project is a very compact amplifier intended to drive the dodecahedron onmidirectional sound sources employed for soundproofing measuremens (usually in pubs and discos). Those sources are made up of 12 medium sized speakers (one per side, series-parallel connection) and are intended to produce in excess of 120dB(A) of pink noise. This requires over 3KW peak and 400W average, which is what a bulky and heavy linear amplifier is currently delivering, but this (mains powered) class D module is intended to produce 6KW peak and 800W average in just 17cm by 10cm by 8cm (excluding case and fan).

[Lars Clausen]

Pretty Cool Job Eva!

[Eva]

First more or less complete prototype assembled and under testing, there are still some design flaws to solve. At least it can produce some reasonably undistorted high power sound.

Now some pictures:









Main board. Preliminary version, I did not bother filling up high current paths with polygons in the bottom layer yet:


Bigh daughter board. The ground plane is full of cuts and does very little against picked up EMI, but a better SMD layout with a continuous ground plane will come later:


The small daughter board that sits on the two sockets on the right side is an auxiliary power supply. Due to my lazyness I have temporarily borrowed a similar aux. psu. module from another project, but I must design a custom one for this project and it must be quieter because the modulator is now clearly picking up its EMI.

Sorry, I won't publish shcematics or component values, but it's not that hard to reverse engineer having the PCB layouts...

[TOINO]

Originally posted by Eva


…, but it's not that hard to reverse engineer having the PCB layouts...

[RX5]

--->>>Sorry, I won't publish shcematics or component values, but it's not that hard to reverse engineer having the PCB layouts...



Eva,
Cool

[Tekko]

Yikes eva Thats prolly the most advanced class d i ever seen

[luka]

Hi

She tends to do stuff like that and as I've seen control board and main board from every angle, it is so populated as hell . But even that is not good enough, SMD's will be used... I mean damn...

[Eva]

Some bugs fixed:

- Added 4th order 15Khz input lowpass filter because the modulator mirrors all the input components above the switching frequency (30-80Khz) down to the audio frequency range. My laptop produces 150mV of 100Khz to 10Mhz trash per each volt of output signal at the line out jack and this was resulting in tons of white noise before the filter was added. This laptop is really a difficult sound source

- Due to a desperate lack of space, the unbuffered input signal trace was routed near 5V logic traces resulting in undesirable crosstalk. I got back a perfectly clean signal at the input of the self.osc. integrator by replacing this trace by a small piece of shielded wire.

- Due to a further desperate lack of space, the self.osc. integrator was placed near the floating section of the IR2113 gate driver resulting in too much capacitive crosstalk. Vertical copper foil shielding was added on both sides of the PCB resulting in a major removal of the unwanted componentes from the output of the integrator.

- A grounded flux band was added to the ETD29 magnetic snubber because it's also placed too close to the modulator resulting in crosstalk. This change also helped to get a cleaner signal from the integrator.

- My modulator contains an "idle dissipation reduction system" that introduces 800ns of turn-on delay during certain circuit conditions. This long dead time allows for full resonant operation when a low signal level or no signal at all is applied, and as a result, idle losses are reduced by up to 10W. However, due to a noisy current sense signal, the system was operating erratically and introducing dead time when it shouldn't resulting in audible distortion, so I temporarily disabled it. This has yet to be fixed.

BTW: I wonder if any similar idle loss reduction system is already patented.

[fredos]

How did you get more than 10 watts loss at idle? Even at 340V DC bus, at such low frequency you should have nearly nothing...
I miss something? Can you explaint me too the routing on your power section....It seem that you put back the DC bus to the aux PSU transformer...Except if that is a current sense transformer..

Fredos