Hello everyone, I am a university student working within a group of three on the design of a class D amp as a graduation project, with the end of goal of automobile application. At the moment, we are in the early pre-design stage which entails considering relevant design options for different aspects of the amplifier.
For the transistors we are using LMG3422R030 GaN transistors capable of operating at a switching frequency of over 1 MHz (copy of the datasheet attached below). We are aiming for a 500W output power on an 8 ohm loudspeaker with as minimal THD as possible without compromising efficiency. After evaluating bridge designs, we have settled on the full bridge design.
These are our main unresolved design concerns as of yet:
1. Modulation scheme: PWM compared to Sigma-Delta
2. Open Loop design compared to Closed Loop design
If possible, references and resources would be much appreciated.
For the transistors we are using LMG3422R030 GaN transistors capable of operating at a switching frequency of over 1 MHz (copy of the datasheet attached below). We are aiming for a 500W output power on an 8 ohm loudspeaker with as minimal THD as possible without compromising efficiency. After evaluating bridge designs, we have settled on the full bridge design.
These are our main unresolved design concerns as of yet:
1. Modulation scheme: PWM compared to Sigma-Delta
2. Open Loop design compared to Closed Loop design
If possible, references and resources would be much appreciated.
Sounds like an ambitious project 😉 ... good 🙂 ..... look forward to follow, and hope you'll post your findings and results.
A few thoughts / my 2 cent:
The chosen GaN device is though not the easiest to use in bridge mode as you'll need to isolate the input part:
https://www.ti.com/lit/ug/snou176b/...ps%3A%2F%2Fwww.ti.com%2Ftool%2FLMG342X-BB-EVM
If you want to showcase GaN technology, maybe use a GaN fet which can work directly with already well known drivers like IRS2092.
GAN039-650NTB could maybe be a candidate?
Another alternative could maybe be the LMG5200 half bridge. Only drawback is the limited 80V max voltage .... which will of course not give the desired 500W @ 8 ohm ... look at https://orchardaudio.com/starkrimson_audio_power_amp_module/
Orchad audio makes a 500W module which seems to be based on what looks like Si8244 and GS61008P in full bridge.
https://compoundsemiconductor.net/article/114332/Audiophile_amplification_gains_from_GaN
Closed loop means better THD, and is easiest implemented in a self oscillating PWM design.
Purify and Hypex designs have probably some of the lowest distorting designs available, so think this is a good sign that it's the right way to go.
A clocked design could potentially have some stability benefits at e.g. past clipping, and can still be found in many proff amps.
Also look up AUDAMP24 which uses IRS20957SPBF + IGT40R070D1
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d462712ef9b701713070a36c08c4
A few thoughts / my 2 cent:
The chosen GaN device is though not the easiest to use in bridge mode as you'll need to isolate the input part:
https://www.ti.com/lit/ug/snou176b/...ps%3A%2F%2Fwww.ti.com%2Ftool%2FLMG342X-BB-EVM
If you want to showcase GaN technology, maybe use a GaN fet which can work directly with already well known drivers like IRS2092.
GAN039-650NTB could maybe be a candidate?
Another alternative could maybe be the LMG5200 half bridge. Only drawback is the limited 80V max voltage .... which will of course not give the desired 500W @ 8 ohm ... look at https://orchardaudio.com/starkrimson_audio_power_amp_module/
Orchad audio makes a 500W module which seems to be based on what looks like Si8244 and GS61008P in full bridge.
https://compoundsemiconductor.net/article/114332/Audiophile_amplification_gains_from_GaN
Closed loop means better THD, and is easiest implemented in a self oscillating PWM design.
Purify and Hypex designs have probably some of the lowest distorting designs available, so think this is a good sign that it's the right way to go.
A clocked design could potentially have some stability benefits at e.g. past clipping, and can still be found in many proff amps.
Also look up AUDAMP24 which uses IRS20957SPBF + IGT40R070D1
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d462712ef9b701713070a36c08c4
The most performant designs are self oscillating PWM with post filter feedback due to the high open loop gain they exhibit (E.G Hypex) and load insensitivity. They also support good audio performance at low switching frequency (200-300kHz) which is good as you still have turn on and turn off losses in any hard switched application like a class-d amplifier even using GaN. Using such a topology you should optimize the full bridge for losses at the switching frequency which will result in lower Rdson parts, a good starting point is when your switching losses are equal to your conduction losses. If you are developing a GaN full bridge from scratch (rather than integrated drivers) you need to take into acount the switching transients are exceptionaly high speed and this can result in misleading waveforms on test equipment without specialist probe techniques (transmision line probes), GaN transistors are also very sensitive to gate voltage overshoot, this is a combination that can result in a lot of confusion and failed transistors.
I just got done reading up on GAN FETS and listening to an hour long webinar on the subject. They say that these GAN FETS are the way to go. Supposedly they come much closer to achieving the prefect square wave and are a marked improvement over silicon FETS. I don't know how true that is but they do offer a reference evaluation platform consisting of a two channel amp and switching power supply.
Gan says this about the bundled platform:
Over at Mouser there is the video of the webinar if anyone is interested in watching it.
https://www.mouser.com/new/gan-systems/gan-gs-evb-aud-bundle2-gs-eval-platform/
Their reference design is pretty pricy at $1449.99 USD.
https://www.mouser.com/c/?marcom=198888653
Has any one here tried this amp at all and if so how good is it? Is there any benefit over say ICEPOWER?
What made you guys decide to use the GAN FETS over traditional silicon MOSFETS anyways?
Gan says this about the bundled platform:
Over at Mouser there is the video of the webinar if anyone is interested in watching it.
https://www.mouser.com/new/gan-systems/gan-gs-evb-aud-bundle2-gs-eval-platform/
Their reference design is pretty pricy at $1449.99 USD.
https://www.mouser.com/c/?marcom=198888653
Has any one here tried this amp at all and if so how good is it? Is there any benefit over say ICEPOWER?
What made you guys decide to use the GAN FETS over traditional silicon MOSFETS anyways?
Overall GaN fets are probably the way forward, and will over time replace normal mos fets in a lot of places improving efficiency.
For amplifiers I'm not sure this will be soon though. Devices are pricy, and the circuit needed is more complex, which will also lead to higher price.
And for amplifiers you have some really good Mos Fets which is not that easy to beat.
And at the end of the day it's the total amps which should be evaluated, on performance, most importantly power, THD and achievable EMI/EMC.
Again looking at e.g. Hypex/Purify you have THD numbers like for a very good opamp!
But it's a very good university project to do, to find out exactly what the benefits and draw backs are 😉
For amplifiers I'm not sure this will be soon though. Devices are pricy, and the circuit needed is more complex, which will also lead to higher price.
And for amplifiers you have some really good Mos Fets which is not that easy to beat.
And at the end of the day it's the total amps which should be evaluated, on performance, most importantly power, THD and achievable EMI/EMC.
Again looking at e.g. Hypex/Purify you have THD numbers like for a very good opamp!
But it's a very good university project to do, to find out exactly what the benefits and draw backs are 😉
Do a scientific investigation. This means, dont ask a typical audiofil or gear head on your typical audio forum about subjective matters like sound etc.
//
//
GaN is cute and all, it doesn't change the bulk of the footprint of the PCB.
Which still consists of the same capacitors, gate driver/controller, preamp section and inductors etc. With higher frequencies, the inductors can be made a little smaller.
That being said, 600-1000kHz switching is not a problem for most MOSFETs .
So at the end of the day you didn't win anything on the practical side.
Improving efficiency is kinda of a non-argument. With regular MOSFETs an efficiency well above 90% can be expected. Some can do 95%. A couple more percent isn't really gonna change much in the overall picture.
Circuit won't be more complex, since many controllers are compatible with driving GaN.
I see benefits in GaN, but absolutely not in audio. In a dedicated ASIC at best otherwise.
Or maybe in extreme high power applications.
If you want to improve, look for higher order feedback control loops. But good luck avoiding patents.
Which still consists of the same capacitors, gate driver/controller, preamp section and inductors etc. With higher frequencies, the inductors can be made a little smaller.
That being said, 600-1000kHz switching is not a problem for most MOSFETs .
So at the end of the day you didn't win anything on the practical side.
Improving efficiency is kinda of a non-argument. With regular MOSFETs an efficiency well above 90% can be expected. Some can do 95%. A couple more percent isn't really gonna change much in the overall picture.
Circuit won't be more complex, since many controllers are compatible with driving GaN.
I see benefits in GaN, but absolutely not in audio. In a dedicated ASIC at best otherwise.
Or maybe in extreme high power applications.
If you want to improve, look for higher order feedback control loops. But good luck avoiding patents.
Last edited:
Have you checked out their evaluation bundle, the reference amp and power supply? What's your opinion on it?
When I was looking at it I couldn't help but notice it looks nothing like my Icepower 400A modules. It is much larger with a lot more components and they look like two completely different topologies. Class D seems like it can be done in so many different ways as there are so many different class D amps out there that it is hard to make that connection between say my Orion XTR PRO 2400 class D mono block from around 2008 or so and my Icepower multi channel amp that I put together recently to power a set of 3 way active towers I'm building. Granted one is a 12 volt car audio and the other is a mains powered home HiFi but they are both class D and they don't look anything alike. Class D has come a long way over the last 15 years or so. According to what I'm reading the GAN reference amp apparently has onboard DSP. I haven't seen DSP implemented on any of the class D evaluation amps yet. Interesting.
I totally agree. You can't really get any enjoyment out of listening to a MOSFET or a GAN FET for that matter.
its an evaluation board not a commercial amp its designed to demonstrate technologies: https://audioxpress.com/news/gan-systems-updates-reference-design-for-class-d-audio-amplifiers
That's a good example, they even mention in the article "smaller".
I don't see the smaller part..........
This is a super cool project. Two partners and I had a very similar senior project back in 2004.
It was to build a class D amplifier for a car, as well. We utilized a 64V power supply with the assumption vehicles would be moving to larger batteries.
There were lots of lessons learned. Best of luck with your project. I can't wait to see your results.
Which university do you go to? We went to Worcester Polytechnic in Massachusetts.
It was to build a class D amplifier for a car, as well. We utilized a 64V power supply with the assumption vehicles would be moving to larger batteries.
There were lots of lessons learned. Best of luck with your project. I can't wait to see your results.
Which university do you go to? We went to Worcester Polytechnic in Massachusetts.
Not really.
Can't find the dimensions of the one in the article.
But they have another similar version, where the power amp section is about 83*60mm give or take.
That's nothing special at all.
Last edited: