ADC Based Feedback

[NTK]

Another kit you may find interesting is the demo kit by GaN system (being acquired by Infineon), which Amir reviewed here.

It has an onboard DSP (called D2Audio) by Renesas (who acquired it from Intersil). Here are a few links for further info (which is as much as I know about this system). My impression is that its performance seems to be limited by the Renesas controller, and is significantly below that of the Axign. GaN systems is supposed to have a reference design with the Axign controller, but I don't think GaN systems made it into an eval kit. The software for the DSP was renamed to "D2Audio Customization GUI" from "Audio Canvas III".

GS-EVB-AUD-BUNDLE2-GS Evaluation Board | GaN Systems

gansystems.com

D2-92683 - Intelligent Digital Amplifier and Sound Processor

DAE-3™ and DAE-3HT™ Digital Audio Engine™ devices, System-on Chip (SoC) multi-channel digital sound processors and Class-D amplifier controllers.

www.renesas.com

https://www.renesas.com/us/en/document/apn/r32an0002-d2-audio-customization-gui-v3i-sound-processing-and-audio-algorithms-rev100?r=502821

500W Heatsinkless Audio Amplifier from Axign and GaN Systems Demonstrates a New World of Extraordinary Audio Performance | GaN Systems

Companies unveil revolutionary, high-efficiency Class-D amplifier at CES 2022 GaN Systems, the global leader in GaN power semiconductors and audio technology innovator Axign, debuted a new groundbreaking GaN-based 500W Class-D audio amplifier. This reference design merges best-in-class...

gansystems.com

 

[BKr0n]

Another kit you may find interesting is the demo kit by GaN system (being acquired by Infineon), which Amir reviewed here.

It has an onboard DSP (called D2Audio) by Renesas (who acquired it from Intersil). Here are a few links for further info (which is as much as I know about this system). My impression is that its performance seems to be limited by the Renesas controller, and is significantly below that of the Axign. GaN systems is supposed to have a reference design with the Axign controller, but I don't think GaN systems made it into an eval kit. The software for the DSP was renamed to "D2Audio Customization GUI" from "Audio Canvas III".

GS-EVB-AUD-BUNDLE2-GS Evaluation Board | GaN Systems

gansystems.com

D2-92683 - Intelligent Digital Amplifier and Sound Processor

DAE-3™ and DAE-3HT™ Digital Audio Engine™ devices, System-on Chip (SoC) multi-channel digital sound processors and Class-D amplifier controllers.

www.renesas.com

https://www.renesas.com/us/en/document/apn/r32an0002-d2-audio-customization-gui-v3i-sound-processing-and-audio-algorithms-rev100?r=502821

500W Heatsinkless Audio Amplifier from Axign and GaN Systems Demonstrates a New World of Extraordinary Audio Performance | GaN Systems

Companies unveil revolutionary, high-efficiency Class-D amplifier at CES 2022 GaN Systems, the global leader in GaN power semiconductors and audio technology innovator Axign, debuted a new groundbreaking GaN-based 500W Class-D audio amplifier. This reference design merges best-in-class...

gansystems.com

It seems that the software used isn't available for public use. Although, it does look eerily familiar.

It's actually funny you should post this specifically. I was just talking to Leo over at orchard audio and he uses those GaNs in his design. Actually speaking of him, he also recommended a company to get a DSP from (https://danvillesignal.com/) so if that works out then I won't even need to make a DSP. Apparently they can make them with compatibility for things like the software hypex uses. That would take out a ton of guess work.

 

[BKr0n]

So I've been doing a little reading and a little youtubing. This illustrates it best, but I think I'm still a little foggy on the particulars...

 

[Zapper]

So would that mean in a test scenario it would be better than just using a plain 1khz sine wave?

I'm very lucky this is a newby section... I'm a little ashamed to ask, but what exactly IS a pole? I'm fairly good with electrical and electronics theory, but I admit my math foundation is a little lacking

No problem. In this context the pole is simply the -3dB frequency of an RC low-pass filter: 1/(2*pi*r*c)

Unfortunately no other than financial of course, it has been my biggest hurdle in all of this. I've mostly been depending on software for testing and simulation. I figure programs like LTSpice are mature enough that they take real world factors into consideration when running simulations so that's what I've been using. I've also got things like sigmastudio, VituixCAD, pspice, fusion360, and kicad among others, but I do not have room for a home lab at the moment. For all intents and purposes, for reasons I would rather not get into on a public forum, I'm kind of in a shoebox lol

LTspice is very capable: I designed several ICs with it as a design engineer at Linear Technologies. But it isn't well suited for heavy-duty mixed signal (digital and analog) that you would get with a DSP based approach, because it is an analog simulator. It would work well for an analog Class-D approach. I've modeled switching power converter topologies using the behavioral models in LTspice. That approach would work well for Class-D amps too.

 

[DonH56]

Seems like an interesting project for a person very keen on learning. However, I do not see how someone can claim they are good with electrical and electronic theory, but do not understand pole/zero analysis and basic control theory. A course or two at a local tech school might be really helpful if you do not want to go the BSEE route. OTOH, I have seen a lot of EE's the past few years that have very little analog knowledge from college.

A sine wave is only a single frequency. A square wave has a range of frequencies that are odd-order harmonics (multiples) of the fundamental frequency. Look up a square wave on Wikipedia or someplace. A band-limited square wave excites a broad range of frequencies and is thus a good test of stability. A swept (in frequency) sine wave can provide similar information.

A simulator is only as good as the models you give it, and real-world parasitic elements in the design and layout are difficult to accurately predict without a more powerful simulator and/or lots of experience and analysis. There are also details of device models, such as self-heating and saturation or overload effects, not always well-modeled (if at all). It is not that the simulator cannot handle these effects, it is rather that the models provided may not include them. Some, like layout parasitics and coupling among signals from the wiring or PCB, require a different tool to extract or a lot of hand analysis. At some point the circuit may become too large to be practical; I have simulated analog IC designs that exceeded one million components after parasitic layout extraction. Getting that simulation to run was a huge effort.

I agree with @Zapper that using a behavioral model for the amplifier is a good approach for a class-D amplifier design using a digital/DSP-based control loop. But I am nervous about you tackling such a project without knowing the fundamentals of feedback first.

The class-D audio amplifiers I have seen oscillate at less than 1 MHz, typically in the 500 kHz range or so. Getting high-power devices to switch at 6-12 MHz is challenging and I am not sure any audio amplifiers run that high. That said, I have not kept up, so maybe some are running that fast? Not any I recall Amir testing, but I do not read all the test threads either.

A class G or H amplifier uses switching (G) or tracking (H) power supply rails typically wrapped around a class A or AB amplifier core. Feedback is usually purely analog in such a design; using an ADC and DSP, you'd need a DAC as well (additional latency, which may or may not be a problem) to supply the appropriate feedback signal injection (may be voltage or current). Keeping the circuit operating smoothly when the supply rail changes can be tricky, but using a digital circuit to sense the input signal and change the rails could be interesting. If the input was digital, you could synchronize switching or rail changes to the input signal using digital delays, though aligning the signal path delays may still be tricky. Without test equipment, there's a high probability of producing a smokebox IMO.

Starting from a kit, or evaluation board, you can gain some knowledge with a proven design before moving on to customize it. Perhaps that would be a good option?

 

[BKr0n]

No problem. In this context the pole is simply the -3dB frequency of an RC low-pass filter: 1/(2*pi*r*c)

Ah OK so it's just a step attenuation. I guess that would make a zero +3db?

But it isn't well suited for heavy-duty mixed signal (digital and analog) that you would get with a DSP based approach, because it is an analog simulator.

Yes I'm beginning to notice that. Though I would imagine that in this context a PWM signal should still be fine with it.

Seems like an interesting project for a person very keen on learning.

Just because it's hard doesn't mean it's not worth doing . Never thought this was going to be easy lol. Just from the learning aspect it's been quite a journey.

However, I do not see how someone can claim they are good with electrical and electronic theory, but do not understand pole/zero analysis and basic control theory. A course or two at a local tech school might be really helpful if you do not want to go the BSEE route. OTOH, I have seen a lot of EE's the past few years that have very little analog knowledge from college.

My background is more in test and troubleshooting than design. I've worked on all kinds of stuff from consumer to aerospace, but never really had the opportunity to go back to school and finish my degree. Life kinda happens sometimes ya know?

A simulator is only as good as the models you give it, and real-world parasitic elements in the design and layout are difficult to accurately predict without a more powerful simulator and/or lots of experience and analysis. There are also details of device models, such as self-heating and saturation or overload effects, not always well-modeled (if at all). It is not that the simulator cannot handle these effects, it is rather that the models provided may not include them. Some, like layout parasitics and coupling among signals from the wiring or PCB, require a different tool to extract or a lot of hand analysis. At some point the circuit may become too large to be practical; I have simulated analog IC designs that exceeded one million components after parasitic layout extraction. Getting that simulation to run was a huge effort.

I noticed that when I tried simulating circuits with the INA851. Some of the results I was getting with it didn't make sense. I've been using simulation just enough to where I can get a stable signal. From there I'll just bread board and test. I just want to make sure that I have a functioning circuit.

The class-D audio amplifiers I have seen oscillate at less than 1 MHz, typically in the 500 kHz range or so. Getting high-power devices to switch at 6-12 MHz is challenging and I am not sure any audio amplifiers run that high. That said, I have not kept up, so maybe some are running that fast? Not any I recall Amir testing, but I do not read all the test threads either.

I may be mistaken myself. From what @Zapper was saying, the clock signal for the amplifier is supposed to match that of the DAC the signal is coming out of. I know things like DSD can get even into the MHz range. Though if we're only going by the audible spectrum, the 300khz described before should be more than sufficient.

A class G or H amplifier uses switching (G) or tracking (H) power supply rails typically wrapped around a class A or AB amplifier core. Feedback is usually purely analog in such a design; using an ADC and DSP, you'd need a DAC as well (additional latency, which may or may not be a problem) to supply the appropriate feedback signal injection (may be voltage or current). Keeping the circuit operating smoothly when the supply rail changes can be tricky, but using a digital circuit to sense the input signal and change the rails could be interesting. If the input was digital, you could synchronize switching or rail changes to the input signal using digital delays, though aligning the signal path delays may still be tricky. Without test equipment, there's a high probability of producing a smokebox IMO.

If I was going G or H, I would probably stick to the analog domain for the most part. Digital in class D to me makes a little more sense because, while not technically a digital signal, PWM needed to synthesize the output can be interpreted by ICs that process them, and as such, can be used to manipulate the signal in the digital domain. I did have a neat idea I was playing around with before I went with class D involving using a PWM signal to time a tracking envelope for class H, but I figured class D would just be more compact and efficient while achieving similar fidelity.

Starting from a kit, or evaluation board, you can gain some knowledge with a proven design before moving on to customize it. Perhaps that would be a good option?

I've definitely been considering this. If I could just get a hypex board to do the DSP rather than get the whole amp, that would be awesome. But they only sell them WITH the amplifiers. Still not a bad option, but I don't have any experience with them and don't know all the caveats of using them. My biggest concern is simply managing the power outputs. If they're adjustable to the drivers they're attached to (as to not overload them), then sure I may go that route. I have also been looking around at amplifier designs from all sorts of manufacturers. The 1ET400 has been my prime candidate since its noise figures are awesome for its price. As for DSP, the ADAU1467 is also a good plug and play option and there are tons of boards out there with them. Or I could just make a board with one, but one step at a time lol

 

[DonH56]

A pole or zero is an inflection point. A single pole will cause the transfer curve to deflect from a straight line to one that drops (slopes) at a rate of 6 dB/octave, with the -3 dB frequency being the pole frequency. Substitute "zero" for pole and then the curve increases (slopes upward) at 6 dB/octave. Add a pole, and the curve slopes down at 12 dB/octave. A single pole or zero introduces -90 or +90 degrees of phase shift (in the limit, a decade or so from the pole/zero frequency) so as you add poles/zeros, you increase phase shift through the device or filter. If the phase shift is 180 degrees the output is the inverse of the input. If that happens in a negative feedback loop, which inverts the output and applies (subtracts) a portion back to the input to compensate distortion, then 180 degrees is inverted and adds to the input, causing the output to grow further, which in turn increases the feedback signal, and the amp will oscillate (and shut down or self-destruct).

I suggest spending some time searching online articles, Wikipedia, and such for more background on poles, zeros, feedback, stability, and all that jazz.

Transfer function - Wikipedia

en.wikipedia.org

Zeros and poles - Wikipedia

en.wikipedia.org

Pole–zero plot - Wikipedia

en.wikipedia.org

https://web.mit.edu/2.14/www/Handouts/PoleZero.pdf

HTH - Don