Axign/AX5689.... Fresh From the Bench: Improving GaN with Digital Control

[voodooless]

Hello everyone!
I started developing an amplifier on AX5689C, the output stages will be built on GaN ISG3208EA

Hi @GaNMaster, welcome to ASR.

Maybe start your own topic on this? Sadly, Google knows nothing about "GaN ISG3208EA". Can you tell us more about the concept? Will it have PFFB?

 

[GaNMaster]

Maybe start your own topic on this?

As soon as there is a result
https://www.innoscience.com/site/product_details/1021 - ISG3208EA

 

[GaNMaster]

Part of the schematic. But that's not the main thing.
The main thing in this work is the software.

 

[GaNMaster]

Measurements of the first test version.

 

[goryu]

That would answer what I've asked myself for a long time: why haven't Purifi or Hypex jumped on GaN, if it has so much potential? They probably did already.

Bruno has addressed this very question previously. His answer: he is able to implement current Fets just fine without the added costs. Looking at the performance of his Purifi designs, I think his point is well made.

 

[KL....]

Hello everyone!
I started developing an amplifier on AX5689C, the output stages will be built on GaN ISG3208EA

@GaNMaster, Thank you, fabulous, it will be fabulous to discuss/compare to the Axign Demonstration Board.... fabulous :=)
As I am, I am sure @pma will be very interested to discuss/compare, so here is (definitely) a good place :=)
The Title indicates, 'AX5689: Fresh From the Bench: Improving GaN with Digital Control'.... and your Implementation/version is exactly that :=)

Thank you for the first test version measurements, part schematic, and ISG3208EA information. 20-20K is looking very good, with parts that are (Truely) exceptional (7-900, 5k-30K), aren't they? Already very interesting and It will be very interesting to further discuss/compare these and other measurement/s, for this version, and (it sounds like) your next versions :=)
THD+N 0.012%
SNR 79.8 dB

 

[NTK]

Bruno has addressed this very question previously. His answer: he is able to implement current Fets just fine without the added costs. Looking at the performance of his Purifi designs, I think his point is well made.

Adding to what Bruno Putzeys had said on the subject of GaN, here are some excepts from his article "Life on the Edge – A Personal Perspective on the Past, Present, and Future of Class D Audio Amplifiers" at hometheaterhifi.com.

...​

The Revolution that is, and isn’t GaN​

...​

Let’s get this out of the way first: yes, GaN FETs are faster and easier to use than silicon. But at home audio power levels, the gap is narrow. 150V GaN FETs are about twice as fast as their much cheaper silicon counterparts. This means that if you drop GaN FETs into an existing class D amplifier, you can only hope to improve distortion by a factor of 2 or so. ...​

​

... By contrast, it’s been enthusiastically embraced as a band-aid and buzzword by parties who have little to bring to the table in terms of intelligent design or audio performance. The actual circuits in most commercial GaN-based amplifiers are nothing we wouldn’t have recognized 20 years ago. Often, they are basic open-loop schoolbook amplifiers, with lack of feedback ludicrously a claimed benefit.​

​

This is sad because GaN is a breakthrough, just not one that’s relevant for Hi-Fi and home cinema amps. This changes markedly, and spectacularly so, once you pass 200V. 200V silicon FETs are significantly slower than 150V ones. By contrast, there’s not much of a speed gap between 150V and 600V GaN FETs. While GaN isn’t about to change what audio performance is available, it’s certainly going to change what power level this performance is available at. I can assure you, anything north of a kilowatt, I’d use GaN. Not because it’s magic, but because it’s practical. That’s something to look forward to. Imagine a touring amplifier with audiophile performance…​

​

 

[KL....]

Part of the schematic. But that's not the main thing.
The main thing in this work is the software.

@GaNMaster, Thank you, yes and also, do you mind further explaining/expanding on the software?

 

[GaNMaster]

THD+N 0.012%

I made measurements of Sabaj A30 (I have it) it has distortion of 0.000xx

 

[GaNMaster]

By the way, I am absolutely sure that there are at least 4 versions of silicon for AX5689
Since I have collected quite a lot of information on this chip.

 

[KL....]

As soon as there is a result
https://www.innoscience.com/site/product_details/1021 - ISG3208EA

@GaNMaster, Thank you :=)

Innoscience ISG3208EA
100V Half-Bridge SolidGaN with Integrated Gate Driver
The ISG3208 is a 100V, 45A, half-bridge SolidGaN in a compact 6.5mm x 6.5mm LGA package.
1.1 Features

• 3.1mΩ(HS)/2.7mΩ(LS) Half-Bridge GaN FETs with Gate Driver
• 80V Continuous, 100V Transient Voltage Rating
• Capable of up to 5MHz Switching
• Independent High-Side and Low-Side PWM Inputs
• High-Side and Low-Side Driver Interlocking
• Internal Strong and Smart Bootstrap Switch
• Adaptive Shoot-Through Protection
• Fast Propagation Delay (20ns Typical)
• Excellent Delay Matching (1ns Typical)
• Built-In UVLO, OVLO, OTP Protections
• 35μA Low VCC Quiescent Current
• Integrated VCC/BST capacitors
• High dv/dt Immunity up to 50V/ns
• Adjustable Turn-on Speeds
• Optimized for Easy and Low-EMI PCB Layout
• En-LGA 6.5mmx6.5mm with Top Side Cooling

1.2 Applications

• Half Bridge, Full Bridge, SR in LLC Converters
• High Frequency Buck and Boost Converter
• Datacenter and Automotive Power Supply
• Class-D Audio, Motor Drive Applications

 

[KL....]

Maybe a little later. This is the most secret part.
By the way, I am absolutely sure that there are at least 4 versions of silicone for AX5689.

@GaNMaster, Thank you, yes, the software is very important and I have observed/thought, more than 5 Iterations/versions where the required voltage has lowered with each Iteration/version, is that also your observation? I notice with the AX5689GRK that the Voltages are now 1.2V and 3.3V :=)

 

[GaNMaster]

@GaNMaster, Thank you :=)

Innoscience produces a lot of good GaN, I used LSG3202, LSG3206.

 

[GaNMaster]

is that also your observation?

I haven't seen datasheets for earlier versions. But the appearance of the chips changed.

 

[KL....]

@GaNMaster, is this a mono or stereo version/implementation?
If stereo, do you happen to know the crosstalk measurement?
What PS are you using?
What PWM frequency are you utiliseing?

Re Graph.... what Load and Wattage did you use?

Many questions to ask, hope that you do not mind :=)

 

[GaNMaster]

Of course, stereo.
I didn't take big measurements, but they are decent.
Power supply 24 volts.
PWM frequency 1.2 MHz.
Load - my monitors, 8 ohms. On the graph 5 watts.
ADC Cosmos.

 

[KL....]

@GaNMaster, Thank you

• I didn't take big measurements, but they are decent....
  • Your measurement is appreciated and look forward to future version/s measurement/s :=)
• Power supply 24 volts....
  • The Innoscience ISG3208EA can utilise 80V Continuous, so 24V is small but sufficient, isn't it?
• PWM frequency 1.2 MHz....
  • Your Implementation must be the first to (successfully) utilise 1.2 MHz, well done :=) This implys that your implementation/combination of AX5689GRK/ISG3208EA/PFFB is very compatable, doesn't it? Do you intend to try an even higher PWM frequency? If yes, what PWM frequency/s might you be considering?
• Load - my monitors, 8 ohms. On the graph 5 watts....
  • This means that the Load is varyable, doesn't it. What are the monitors and/or do you have an Impedance Graph for the monitors?
• ADC Cosmos....
  • is that the E1DA Cosmos ADC? How are you useing this?

 

[GaNMaster]

1) The project I measured uses ISG3202, which is essentially the same thing. I initially wrote that the new project uses ISG3208
2) I don't see any point in making the PWM frequency higher than 2 MHz.
By the way, I don't use PWM in this project, I use ZCM. It seems that this is the result of not very beautiful numbers on the graphs. But, it sounds better than Sabai A30.
3) The monitors are made on a ceramic speaker from SB acoustic. Unfortunately I don't have any Impedance Graph for the monitors.
4) I use a high-voltage (42 V) input of the Cosmos ADC.

 

[GaNMaster]

can utilise 80V Continuous

This is a test sample, I'll take measurements at 50 a little later.

 

[KL....]

@GaNMaster, Thank you
PWM 2MHz frequency.... definitely have not seen this done, so with your implementation, let us see what measurements result :=)
ZCM?.... do you mean PCM as input?
I use a high-voltage (42 V) input of the Cosmos ADC.... ok, so 42V not 24V, is that correct?

Is this the SB acoustic Drivers/Speakers? If yes then Load Impedance appears to be >=5.7Ω

SB acoustic 6″ SB17CAC35-8 / Ceramic​