I just want to list some polemic points about that GaN fashion for class D:
- That PE29102 driver thing does not have level shifter.
- That PE29102 driver thing does not have current limiter.
- That PE29102 driver thing is 60V only.
- That PE29102 driver thing has an under-voltage lockout way too low 3.2V for driving powerful GaN FETs for audio.
- That PE29102 driver thing has a joke datahseet with a block diagram likely not drawn by an electonics engineer, lacking a sample application circuit schematic with component values, etc. This is as if this being advertised for audio is not coming from Applied Science Experts.
- GaN FETs come in a package with minuscule thermal mass that has zero resilience against thermal cycling at lower end audio frequencies, like 60Hz for 30Hz class D reproduction.
- Switching losses are still the dominant factor in class-D audio amplifiers even if body diodes are ideal. Turn-on and turn-off are still limited to 500~1000A/us in order to maintain reasonable EMI generation for audio quality and channel-to-channel interference.
- Silicon FET are already pretty good measurably, their ceiling of operation speed is just the required speed for class-D audio (500~1000A/us) and are available in packages with high thermal mass at a fraction of the cost of GaN.
- Keep GaN for RF. GaN is superb for RF.
- The idea of trying to sell GaN for audio does not come from Applied Science Experts. This is a bubble.
- Investors shall instead just give money to audio gurus, just for exercising the metric system associated to the ear, as this is what keeps human class in place.
- That PE29102 driver thing does not have level shifter.
- That PE29102 driver thing does not have current limiter.
- That PE29102 driver thing is 60V only.
- That PE29102 driver thing has an under-voltage lockout way too low 3.2V for driving powerful GaN FETs for audio.
- That PE29102 driver thing has a joke datahseet with a block diagram likely not drawn by an electonics engineer, lacking a sample application circuit schematic with component values, etc. This is as if this being advertised for audio is not coming from Applied Science Experts.
- GaN FETs come in a package with minuscule thermal mass that has zero resilience against thermal cycling at lower end audio frequencies, like 60Hz for 30Hz class D reproduction.
- Switching losses are still the dominant factor in class-D audio amplifiers even if body diodes are ideal. Turn-on and turn-off are still limited to 500~1000A/us in order to maintain reasonable EMI generation for audio quality and channel-to-channel interference.
- Silicon FET are already pretty good measurably, their ceiling of operation speed is just the required speed for class-D audio (500~1000A/us) and are available in packages with high thermal mass at a fraction of the cost of GaN.
- Keep GaN for RF. GaN is superb for RF.
- The idea of trying to sell GaN for audio does not come from Applied Science Experts. This is a bubble.
- Investors shall instead just give money to audio gurus, just for exercising the metric system associated to the ear, as this is what keeps human class in place.
GaN stomps all over Si for switch mode performance I'm afraid, whatever you think, it will gradually take over a large percentage of power conversion including class-D amps as it permits smaller and more efficient power conversion, both of which count for a lot. For class-D faster switching means less intrinsic distortion as well as smaller filter components,
Ranting won't stop this trend!
Ranting won't stop this trend!
GaN is the future 😉
Regarding high voltage half bridge drive a multi chip solution can work using the Ti capacitive digital isolators and discrete fast logic with RC's for deadtime control. My memory is not great on this but I think this worked fine to a few MHz.
If anyone is looking for an easier route to GaN; Ti do fully integrated half bridges.
http://www.ti.com/power-management/gallium-nitride/gan-fet-power-stages/products.html
Regarding Evas concerns the heating of the die in a GaN device in a given application should be less due to increased efficiency. The switching node has less ringing that reduces EMI and allows for faster edge rates.
Regarding high voltage half bridge drive a multi chip solution can work using the Ti capacitive digital isolators and discrete fast logic with RC's for deadtime control. My memory is not great on this but I think this worked fine to a few MHz.
If anyone is looking for an easier route to GaN; Ti do fully integrated half bridges.
http://www.ti.com/power-management/gallium-nitride/gan-fet-power-stages/products.html
Regarding Evas concerns the heating of the die in a GaN device in a given application should be less due to increased efficiency. The switching node has less ringing that reduces EMI and allows for faster edge rates.
To NMOS, from your post #29:
You'll have to stick with the IR2110 to drive the TP65H035's. All of the big cascode units from Transphorm have a low-voltage, 'regular' MOSFET at the input -- it actually does the switching.
The LMG1210 provides 5V Gate drive specifically for 'straight' GaN devices -- but only those of the Enhancement mode flavor.
And that's only those discrete values -- no graph of variation over temp, at different Gate resistor values, or different Drain voltages, or capacitive vs inductive loads.
Cheers,
Rick
The problem with lead induction is the spikes they cause in Gate drive -- stability would be a separate issue.
You'll have to stick with the IR2110 to drive the TP65H035's. All of the big cascode units from Transphorm have a low-voltage, 'regular' MOSFET at the input -- it actually does the switching.
The LMG1210 provides 5V Gate drive specifically for 'straight' GaN devices -- but only those of the Enhancement mode flavor.
This will be difficult to achieve with TP65H035's. Note in the partial page 4 -- there's a 30 nsec longer turn-off delay, than turn-on -- surely a tough hill to climb.
And that's only those discrete values -- no graph of variation over temp, at different Gate resistor values, or different Drain voltages, or capacitive vs inductive loads.
Cheers,
Rick
Last edited:
Oh, and I'd highly recommend that you design out the CD4049, and any other 4000-series parts that are in the signal path -- if you're going to try some GaN parts. The 4000-series is plenty useful for lots of things, but only barely fast enough (50 nsec @ 15V/50pF) for regular MOSFETS, even if you're running them at 12 or 15V.
I understand the temptation since we want to end up with 12V drive, but practically all 'modern' gate drivers expect only TTL/LVCMOS levels, and are going to do their own level shifting anyway.
Rick
I understand the temptation since we want to end up with 12V drive, but practically all 'modern' gate drivers expect only TTL/LVCMOS levels, and are going to do their own level shifting anyway.
Rick
I went to AXPONA 2019 hi-fi show in April , Chicago and experienced this Class D - GaN for myself the very first time. I never heard of GaN device before auditioning to this new GaN class D amp. My jaw was dropped on the floor of how 3D of the sound stage and how wide dynamic range it gives out.
I have always been a die-hard fan of pure class A and SET tube amp. Never like the idea of digital class D. However, this real experience with Class D- GaN FET shook and challenged all of my doubts and beliefs.
This is the amp that I have listened at the Axpona show this year AGD PRODUCTION
Regards,
Tom
I have always been a die-hard fan of pure class A and SET tube amp. Never like the idea of digital class D. However, this real experience with Class D- GaN FET shook and challenged all of my doubts and beliefs.
This is the amp that I have listened at the Axpona show this year AGD PRODUCTION
Regards,
Tom
have look website,company have attend at fair 2019 in Germany
long time no have visit Munich Hi-FI fair, because Im love PA
It will be interested which Kind of Gan Fets and driver this company use, I think its not Transphoform
Can you recomend me device what is fast enough to change CD4049 ?
I not want redesign everything new, is there device in independnet mode like CD4049 available ?
Dunno how the CD4049 is being applied, but one big difference is that the CDxxxx parts can handle up to (and switch) 15VDC. 74HCxxxx parts get pretty unhappy an anything higher than 5VDC. If the application needs more than 5V, then the 74HC part isn't going to do it.
Great, but pretty sure you'll have to take the "CD" out of the p/n --
74HC4049 will do it.
Regards,
Rick
74HC4049 will do it.
Regards,
Rick
CoolGaN from infineon.
There is this EVAL_AUDAMP24 here:
Gallium' 'Nitride' ''('GaN')' - Infineon Technologies
Does anyone have any experience with this? Not clear if the board is available for ordering.
There is this EVAL_AUDAMP24 here:
Gallium' 'Nitride' ''('GaN')' - Infineon Technologies
Does anyone have any experience with this? Not clear if the board is available for ordering.
I will try 400V Class D GAN FET in my next project
after finished discrete Kartino UCD Xlite AMP, I will replace IRFP4227 with IGT40R070D1 GAN FET
optimized for Class-D Audio amplifier applications.
but I have to make new special PCB to assemble/soldering GAN FET
400 V Class D Gan FET works and is compatible with existing control Mosfet Class D Driver (IRS20957 / IRS 2092)
In theory, it should also work with UCD Xlite
Its DIY lets try and fail to learn,
Last edited:
The switching speed of 2092 is limited at 800kHz, which i think GAN can operate from 1Mhz to 10Mhz, which will reducing the LC filter components size dramatically.
Hi, I would be interested to test this one :
EPC9106
Can you send me a private message in order to know how to get it.
Hi!
I have some experience with cascode TPH3208 and SI823x drv.
In my design, the optimum THD for these FET turned out to around 5.5 ns, and the switching was too noisy in the FM range.
When using IRFB5615, there is significantly less switching interference, and all parameters are better in the same design. Optimum Dt 3.3-3.5ns. 850kHz
While working with this project an interesting thought came up.
The SI8233 driver may well work with GAN transistors with a little refinement of the gate control circuits. You just need to add the negative gate bias circuitry. In addition, the Vgsth is additionally balanced for free.
I'll try it sometime.
I have some experience with cascode TPH3208 and SI823x drv.
In my design, the optimum THD for these FET turned out to around 5.5 ns, and the switching was too noisy in the FM range.
When using IRFB5615, there is significantly less switching interference, and all parameters are better in the same design. Optimum Dt 3.3-3.5ns. 850kHz
While working with this project an interesting thought came up.
The SI8233 driver may well work with GAN transistors with a little refinement of the gate control circuits. You just need to add the negative gate bias circuitry. In addition, the Vgsth is additionally balanced for free.
I'll try it sometime.
The price of this board is nearly 500$... -> EPC9106, EPC EPC9106 в наличност. Купете EPC9106 с най-добра цена на Components-Store.com