Class D amp long term reliability

[witwald]

There are also some aspects the audio test equipment we have today does not accurately measure (e.g., imaging specificity), and which must be evaluated in actual operation.

@antennaguru In your work, for which elements do you rely on subjective assessments to help define the design?

"Imaging specificity" sounds like it is a very specific term, pardon the pun. How is it defined? Do you have at your disposal a number of (calibrated?) golden ears that agree on the amplitude, for want of a distinct measure, of that imaging specificity? Or is "imaging specificity" purely a subjective impression as perceived by a listener, some who care about it and some who don't? Would "imaging specificity" be a feature of loudspeaker performance that is best measured in an anechoic chamber?

 

[restorer-john]

Are those MR16 LEDs color/wavelength correct?

Standard Restorer-John test is a big pile of resistors. If I'm questioning any colours, especially on slightly discoloured resistors, I need a better light. Right now, there's about four different colours of 'white' from three identical lamps plus the ceiling LED. I even changed over 2 lots of 130x5mm LEDs to try to match colours in each lamp after batch sorting 260 matching from 2000. Gave up and decided to go all out with my project. Wish me luck. A skylight would be good, but I am loathe to have a giant hole cut through our roof and a light-tube installed.

 

[Doodski]

Standard Restorer-John test is a big pile of resistors. If I'm questioning any colours, especially on slightly discoloured resistors, I need a better light. Right now, there's about four different colours of 'white' from three identical lamps plus the ceiling LED. I even changed over 2 lots of 130x5mm LEDs to try to match colours in each lamp after batch sorting 260 matching from 2000. Gave up and decided to go all out with my project. Wish me luck. A skylight would be good, but I am loathe to have a giant hole cut through our roof and a light-tube installed.

I swear by good lighting. Nice bright moveable lighting with a solid background illumination for the room with good magnification optics and many faults can be found. Here I like a nice warm bulb to be near because it's cozy. Don't have the extra design considerations that you do.

 

[audio2design]

Standard Restorer-John test is a big pile of resistors. If I'm questioning any colours, especially on slightly discoloured resistors, I need a better light. Right now, there's about four different colours of 'white' from three identical lamps plus the ceiling LED. I even changed over 2 lots of 130x5mm LEDs to try to match colours in each lamp after batch sorting 260 matching from 2000. Gave up and decided to go all out with my project. Wish me luck. A skylight would be good, but I am loathe to have a giant hole cut through our roof and a light-tube installed.

Just get a proper LED driver and some 90+CRI COBs in the CCT of your choice from Digikey. 5,000-10,000 lumens plus is cheap

 

[storing]

And that's where the entire EU regulations get absolutely innane. The filtering/suppression in and around the power inlets, physical switch etc consume some power. The poor owner can't even have an indicator to know if his product is even plugged in- too much standby power.

The regulations might be silly but on the other hand what they regulate is perhaps also not free of silliness: nothing wrong with technological advancement, but is having to toggle a power switch really that big of a deal that it becomes ok to waste energy on? I'm not saying we should ditch it completely, but it doesn't hurt questioning this imo, it's like standby is mandatory for everything these days but is humanity really becoming so lazy that flipping a switch is becoming too much?

Reminds me of that Simpsons episode years ago where fun was being made of Apple: 'the light is on to confirm it's powered off'. A rather sharp observation actually.

 

[pma]

@antennaguru is correct. Imaging can only be assessed with a pair of speakers, not a single speaker test. You need a human, not a piece of test gear, because there isn't one. It is a glaring omission in ASR speaker reviews.



5 Watts at 4 ohms FR and you think that categorizes how an amplifier will behave at elevated powers, across the audible bandwidth into wildly varying loads? Why do you think manufacturers and Amir do a frequency response plot at 1W (or 5W in his case)? Because that's about as good as it gets. A best case freqency response. But representative of the real world? Nope.

I run my frequency response plots at full rated power and 1W (and often 1/2 rated power). Many an amplifier falls flat in a heap when asked to deliver a flat response at high power. And that is just fixed dummy loads. Throw a simulated speaker at an amplifier and sweep it at full power...

Notice how few manufacturers are prepared to quote their power bandwidth these days? (-3dB)

For example, at full power, where does the amplifier commence rolling off and at what frequency is the amplifier -0.1dB down?

I really do not think that there is a behaviour of class D as a general rule to change FR with power and to have power BW strongly dependent on output power. Below please find a measurement on my UcD180 sample amplifier

Hypex UcD180HG HxR amplifier module analysis and review

Hypex UcD180HG HxR module analysis and review After buying and testing AIYIMA A07 amplifier (based on TPA3255 class D chip) in January 2021 I was quite disappointed with the frequency response modulated by speaker complex impedance and its audible consequences. This is, unfortunately...

www.audiosciencereview.com

of frequency response at various output power into 4ohm load. This amplifier has a usable power somewhere at or below 130W/4ohm, defined solely by the linear PSU with 100VA transformer used. Physics cannot be violated and in fact 100W is a reasonable maximum, then the PSU output voltage falls too low and the amplifier starts clipping. However, there is no affect to the FR and you can see that the plots are spaced at distance of exactly 1dB even at 20 kHz and even where the amplifier is visually clipping. So I see no rule and no room for general statements here. I is all about the specific circuit solution used, regardless the class of operation. And you can read your requested -0.1dB decay, again at exactly same frequency for every plot.

P.S.: the power limit where the amplifier and power FR starts to behave violently is much more easily readable from the appropriate THD vs. power plot at well chosen frequencies

At least I can get more info from such plot than from the first graph in this post. We can see that the 126W power is above reasonable power working space of the DUT amplifier.

 

[restorer-john]

I really do not think that there is a behaviour of class D as a general rule to change FR with power and to have power BW strongly dependent on output power. Below please find a measurement on my UcD180 sample amplifier

Hypex UcD180HG HxR amplifier module analysis and review

Hypex UcD180HG HxR module analysis and review After buying and testing AIYIMA A07 amplifier (based on TPA3255 class D chip) in January 2021 I was quite disappointed with the frequency response modulated by speaker complex impedance and its audible consequences. This is, unfortunately...

www.audiosciencereview.com

of frequency response at various output power into 4ohm load. This amplifier has a usable power somewhere at or below 130W/4ohm, defined solely by the linear PSU with 100VA transformer used. Physics cannot be violated and in fact 100W is a reasonable maximum, then the PSU output voltage falls too low and the amplifier starts clipping. However, there is no affect to the FR and you can see that the plots are spaced at distance of exactly 1dB even at 20 kHz and even where the amplifier is visually clipping. So I see no rule and no room for general statements here. I is all about the specific circuit solution used, regardless the class of operation. And you can read your requested -0.1dB decay, again at exactly same frequency for every plot.

View attachment 176104

Nicely done Pavel. There is essentially no change in FR with this particular Class D module at various powers.

I'll see what I can come up with tomorrow as a comparison.

 

[restorer-john]

Just get a proper LED driver and some 90+CRI COBs in the CCT of your choice from Digikey. 5,000-10,000 lumens plus is cheap

Good idea.

 

[Madjalapeno]

I took some FLIR pictures of one of my NC400's last night as they were feeling warm.

More here: https://www.audiosciencereview.com/forum/index.php?threads/flir-camera-shots-of-nc400.29554/

 

[pma]

I took some FLIR pictures of one of my NC400's last night as they were feeling

More here: https://www.audiosciencereview.com/forum/index.php?threads/flir-camera-shots-of-nc400.29554/

They need a proper heatsink, as everything, if we ask for long life span. Most assemblers make it wrong, small and cheap cases.

 

[Madjalapeno]

They need a proper heatsink, as everything, if we ask for long life span. Most assemblers make it wrong, small and cheap cases.

I agree, but the SMPS is harder as it is not directly attached to the casework and relies on ventilation. Those bent aluminum heatsinks over the caps might be better if they were straightened.

 

[restorer-john]

Isn't it the transformer running the hottest? Why is the crosshair/target on the cooler parts?

 

[Madjalapeno]

Isn't it the transformer running the hottest? Why is the crosshair/target on the cooler parts?

It's been a while since I used the camera and I had forgotten how to move the crosshair. (and I was two margaritas in)

 

[DonH56]

Thinking out loud... Given their method of operation I would expect a decent class D amp to exhibit flatter full-power bandwidth than a comparable class AB amplifier just because the bandwidth of the class D's driver and output stages must be much greater to support the high switching rate. The class D output devices are always switching rail-to-rail at a high rate, so as current in them increases, it won't decrease the audio bandwidth much despite the heavier load on the driver stage. In contrast, as current (power) goes up in conventional amplifier outputs, the 3 dB bandwidth is much closer to the top of the audio band, and so decreasing bandwidth due to the heavier load (mainly on the driver stage due to higher current in the output devices, which increases their input capacitance) causes more significant roll-off.

 

[audio2design]

Isn't it the transformer running the hottest? Why is the crosshair/target on the cooler parts?

We don't know the heat sink temperatures as they were not treated to read properly on the IR camera.

 

[pma]

Thinking out loud... Given their method of operation I would expect a decent class D amp to exhibit flatter full-power bandwidth than a comparable class AB amplifier just because the bandwidth of the class D's driver and output stages must be much greater to support the high switching rate.

I am sure you remember troubles of many many class AB amplifiers from seventies, eighties and maybe even later with cross-conduction current when the output stage was asked to operate at high frequency and high power. 20kHz test at large output swing was able to disclose such behaviour. I was thinking about the similar thing that you have described ...

 

[Shazb0t]

Nicely done Pavel. There is essentially no change in FR with this particular Class D module at various powers.

I'll see what I can come up with tomorrow as a comparison.

I'm glad that you've acknowledged this. Interested to see what you'll be able to come up with.

 

[DonH56]

I am sure you remember troubles of many many class AB amplifiers from seventies, eighties and maybe even later with cross-conduction current when the output stage was asked to operate at high frequency and high power. 20kHz test at large output swing was able to disclose such behaviour. I was thinking about the similar thing that you have described ...

Yes indeed, that was one of the big problems they had! Along with some insanely nasty clipping behavior when the designers (raises hand, been there, done that) neglected to consider what happened when clipping effectively opened the feedback loop. I made a few bat-killers back then...

 

[Rottmannash]

Isn't it the transformer running the hottest? Why is the crosshair/target on the cooler parts?

So the power supply gets hotter than the amplifier module?? Is this typical? If so is it just when idling?

 

[restorer-john]

So the power supply gets hotter than the amplifier module?? Is this typical? If so is it just when idling?

The small transformers in SMPSs can run extremely hot, especially when delivering high power.