Which Class D Amplifier? PS Audio, Ghent, Nord, Merrill or other???

^^ The reason we aren't using GaNs yet is we are in a prototype stage. We developed a means of eliminating the need for dead time circuitry in class D amps (for which we have a patent pending).

The main reason to use GaNs is to minimize dead time (dead time increases distortion). So you are seeing GaN-based class D amps right now that have no dead time circuits simply because the transistor is so fast that with conventional switching speeds no dead time circuits are needed.


We're able to switch at the same speeds with conventional devices.


What this means is that for any given semiconductor tech that comes along (GaN and whatever **will** follow) we'll always be able to switch at higher or equivalent speeds with older technology (which is less expensive, although GaNs are already pretty inexpensive, being only about $3.00 each). Of course the flip side is that we can use the latest tech and switch faster. But in a prototype that sort of thing isn't necessary.

   I'm also curious about how your development of a class D amp is progressing and whether it will utilize GaN or any other newer technology.  

@noble100
Right now we're not using GaN transistors but they are certainly on the radar. So far we've been building our modules by hand. GaN devices are much smaller than what we are using, which makes the assembly a lot more difficult (or easier, if we farm it out). Otherwise progress seems to be pretty good.

 The EPC eGaN FETs switch 10x faster than silicon MOSFETs and have zero stored charge. I'm guessing that probably improves sound quality when used in a class D amp.

While they are faster its incorrect to characterize them as having 'zero stored charge'. The gate (input) capacitance is the big issue; on traditional MOSFETs it can easily be around 2000pf, which requires a bit of current from the driver circuit to switch the device at speed. GaNs tend to have less gate capacitance (about an order of magnitude less, depending on the device) and that is part of why they can switch faster. But that gate capacitance still plays a role in the driver design.
Much moreso than actual switching speed, getting dead time down is what reduces the distortion in a class D amp. Many of the GaN devices switch fast enough that amps with zero dead time can be built at pretty close to current switching speeds (400KHz or so).

And from Merrill Audio who use the higher switching speed

" Internally the Element 118 Power Amplifier Monoblock PCB boards are manufactured to the highest excellence available and further engineered for sustaining a most stable temperature distribution. Using 8 pounds of pure copper per monoblock, the highly refined heat distribution system maintains a steady temperature environment on the PCB boards, a condition absolutely essential for linear operation."

George, looks like you made my point.

I tend to believe the manufacture of the board, EPC and it’s designer Steve Colino, who stated only heatsinking is required on these GaN devices if the switching frequency is raised. Not by what you say.

Well you're not likely  to be designing a class D amp anytime soon, so its no worries for you. For those that **do** design such amps, the idea of no heatsinks is scary to say the least, and the prudent designer that knows the amp is going to see some thermal abuse in the home will add heatsinks. To not do so is to court disaster- If it turns out that a model is overheating, the boards in the unit will have to be redesigned and that sort of thing can put a company under really fast.

You can buy your own EPC GaN boards from Digikey, and yes they operate with no heatsink.
Because EPC’s Steve Colino said to me, (and I posted up his email), they are at 600khz switching speed, you can increase that to 1 or 1.5mhz as Steve said, but you will have to use a heatsink on them.

@georgehifi
Yes, the board itself is the heatsink, but is an evaluation board. If you ran this thing under heavy conditions, it would fail unless cooled by a fan or the output devices were mounted on a proper heatsink.

The Class-D is just going to be a money spinner for him, he buys the modules does a small mod so it can be said to be better, throws in in a box, and there’s his beer money.

@georgehifi
Just for the record, this statement of yours from the beginning of this thread is entirely false. Our design is from the ground up entirely our own so we don't use anyone's modules. We have a patent pending on how we eliminate the need for dead time.

By eliminating dead time we can use much slower transistors than Gan devices and yet switch at the same speeds. Of course we can use Gan devices too.

BTW looking at the heatsinks in the SU-G30 above, that section is the GaN amp, it may have higher than todays 600khz switching because it’s using these heatsinks, as EPC said to me if left at 600khz there is no need for any heat sinking on any of the GaN boards, so it maybe higher.

Heatsinks are required, even for Gan devices switching at lower speeds, regardless of what EPC allegedly said. The reason is that all output devices have a finite 'on' resistance which is exactly zero. It might be 0.05 ohms, but its not zero. This means that it will make some heat when full on and that heat builds up over time.