Amplifier power stage : Silicon Carbide (SiC) and Gallium Nitride (GaN) ?

[jerome69]

Hi !
Just for my knowledge and perhaps will try to built something. Leave schools and electronics a long time ago and would like test these new devices.
New technologies for new power équipements, not for audio but interesting to see how it could work. Perhaps some improvement in linear audio and class D amplifier. See some few topic on SiC and GaN in this forum.
Have you try these new components in audio amplifier ?
They seem to be very expensive.
Some reference :
https://www.onsemi.com/solutions-ap...ehicle-electrification/wide-bandgap-solutions

https://www.digikey.in/Site/Global/Layouts/DownloadPdf.ashx?pdfUrl=4039FA9B108F42229FEF9E84EE758607

https://www.ti.com/lit/an/slyt801/slyt801.pdf?ts=1687040496468

Lot of fun !

 

[Mark Tillotson]

The problem with devices optimized for switching is that SOA is of very little concern as they are assumed to spend most time fully off or fully on, so often not very suitable for linear amplification.

For class D though the clear winner is GaN with the higher operating speed, although this mandates ultra-low inductance mounting techniques to take advantage off. I've played with 80V 20A EPC9203 modules for instance:
https://epc-co.com/epc/products/demo-boards/epc9203 (the module handles the daunting SMT device mounting) and yes they switch much faster and can be driven nicely from fast logic, but require 5V signal, 3.3V won't be enough.

 

[AnalogJoe]

GaN and SiC are very popular among RF/Microwave Power Amplifiers, because they allow for high voltages and have very high electron mobility. In audio, GaN has been recently getting a lot of attention, you can get several tens of Watts out of them and not even a heatsink might be needed. They are truly amazing pieces of technology.

 

[jerome69]

Thanks you Mark for the response.
I understand, for linear amplification things are tricky because it is not done for. I think too drive correctly these devices for linear operation is hard. The feedback loop circuit must work perfectly.
About class D, I didn't think about building the circuit and parasitic RLC, it is HF design ! Nice module (epc 9201/3) which can switch higher than 500 kHz. Virtually noiseless in the audio range with class d amplifier and smps !

 

[HAYK]

What makes the GaN good in class D, not the switching frequency that can reach several
MHZ, but it's low dead time. With SiC it is possible down to 20ns, with GaN, 4ns is at reach with recent drivers. The main distortion in class D is due to dead time. It has very low Ron.

 

[Mark Tillotson]

Dead time is the amount of slack the circuit designer gives to the circuit to avoid any risk of shoot-through, the device parameter is simply switching time (typically different for on and off, and usually quoted for 50% of max current). GaN is pretty asymmetric to drive and the gate is leaky, the drivers have to handle that and have an ultra-low impedance path to the FET(s) as overshoot on the gate can trash the whole device. Driver recovery limits the actual frequency, but of course you want efficiency and low distortion so the period of the switching waveform is much longer than the switching time of the devices ideally. For instance GaN FETs might be switching in around 5ns but the on and off times are more like 500ns.

 

[stocktrader200]

If the tpa series chips weren't so good , I would try a SIC part as the high side fet as that benefits the most from lower drive requirements.

 

[HAYK]

With Covid lock down I got interested with this subject, good and affordable GaN was GS61008, $12 from mouser, now it says obsolete and un available. It is 90A 7mohm. The problem to switch with low dead time was the drivers high to low delay mismatch. The LM5113 was 8ns but new Lmg 1210 is 3.5ns. TI has an integrated half bridge GaN 10A 80V, LMG5200, and provides EVM board. On YouTube a Russian fellow could pass the dead time to 6.5ns using it along FPGA delta sigma modulator.

Since I discovered the means to eliminate the dead time effect by ABD modulation, GaN became useless.

 

[NareshBrd]

The devices are mainly switching types or high power RF types, and tend to be expensive.
Most are designed for a much higher frequency than seen in audio, and there are many Class D chips available.
This discussion is veering down the Class T (Tripath) lane, read up on that, most of the claimed improvement was a marketing tool.

If you want to experiment, there are such chips in almost every mains operated LED lamp, controlling the switching transistors, and you get them in large quantities...in the trash at times, as the lamps have started to reach end of life.
Computer SMPS will give you higher power rated parts.

You can experiment with those parts, build such an amplifier, and see how it sounds in comparison to a normal Class D device.

 

[jerome69]

See this page about GS61008P very interesting application notes.
Interesting discussion, agree the design is very important.

 

[HAYK]

I have analyzed this class D module 2 years ago, they used old generation drivers with 10ns delay mismatch. I simulated one bridged using a pair of LMG1210, with open loop 4ns DT and 200khz carrier, I got 0.004%. With 50ns DT 400khz carrier for common MOSFETs it is 2.2%.