Class D amp long term reliability

[antcollinet]

A couple of photos from my Hypex NC252MP (manufactured 49 week of 2020). So just more than a year younger than @restorer-john’s module. As mentioned earlier, there is some slight board flexure. Here is a view just after separating the heat plate…

View attachment 258721

As mentioned previously, it does appear that the thermal pad is pushing back on the output MOSFETs, but does not seem as bad as John’s module that overheated. Whether there is sufficient thermal conduction can be measured but I do not have proper equipment. Likely Hypex does though. In my QA tech days, it was a lot of intrusive thermal couple probes. Today, would use a good thermal imaging camera.

Here is a close-up of the circuit board in the area in question…

View attachment 258727

Overall, solder joints look good. Most likely done via wave soldering machine. The MOSFETs do look hand soldered, but this is not unusual in contemporary electronics manufacturing. They look like they are well done. I have been in plants all over the world and automation vs manual tradeoffs are commonly made depending on cost, time and quality concerns.

Just to be clear, am not an electronics designer and agree that board flexure does appear to be a design issue. However, this is just one more module in thousands that Hypex has sold. Not sure that the board bowing happens on NCore, SMPS or other MP series products despite sharing common integrated heat spreader design.

Hypex clearly designs their modules for mounting in real world speakers and so extra steps are taken to deal with shock and vibration. These are stresses beyond what most consumer electronics need to consider during operation. Interestingly, surviving shocks during shipping is part of design considerations that many lay folks take for granted. Any serious electronics vendor has a myriad of tradeoffs to consider. With this in mind, I think it is important to consider the overall design in comparison to others.

Regardless of cooling there is no way on earth - as a project manager for a global blue chip electronics manufacturer - that I would have been able to get a QA signoff for product release with even 1/4 of the flex in the PCB as shown in @restorer-john 's example (EDIT: or @pma s). In fact I wouldn't have even tried - I'd have sent the design team back to fix it without asking for one.

See that big R277 - that is a thin ceramic component along the line of flex. It can't bend, the solder joints can't move. Flexing the PCB puts it under enormous stress and there is significant risk of it (or any of the other ceramic coponents - pretty much all those SM resistors and capacitors) cracking and hence failing. And many of them are in the output transistor control circuit - so if they go the power device is likely to go as well.

 

[RickS]

Regardless of cooling there is no way on earth - as a project manager for a global blue chip electronics manufacturer - that I would have been able to get a QA signoff for product release with even 1/4 of the flex in the PCB as shown in @restorer-john 's example. In fact I wouldn't have even tried - I'd have sent the design team back to fix it without asking for one.

I understand and there is some flex in mine too. His may have been worsened by the actual meltdown.

See that big R277 - that is a thin ceramic component along the line of flex. It can't bend, the solder joints can't move. Flexing the PCB puts it under enormous stress and there is significant risk of it (or any of the other ceramic coponents - pretty much all those SM resistors and capacitors) cracking and hence failing. And many of them are in the output transistor control circuit - so if they go the power device is likely to go as well.

Not disagreeing but not going to speculate either. Hypex already acknowledged the bowing and said it is a non-issue for them. They have to manage their own business risk. It is one potential issue with one Hypex model. I have multichannel amp on my testbench with four NC502MP modules, they show no sign of the bowing despite sharing a common Hypex mechanical design. I worked for a very conservative, quality driven automation manufacturer. Our electronics had to operate under far worse environmental conditions than most consumer grade products do. Some of this involved special board spacing and other considerations like higher grade components. As stated, am not a design engineer, but from experience I know there are ways to make electronics more environmentally resilient. Many of these practices are not obvious to a casual observer. When properly designed, have seen these electronics continue to perform in some much uglier situations than a slightly bowed board.

 

[antcollinet]

They have to manage their own business risk.

Yes they do. And I can manage my own purchase risk accordingly.

 

[Davide]

I own NC252MP and NC502MP, and both came with flexed PCB (252 especially).
Also, even when unflexed, the PCB is under enormous load due to the compression of the thermal pads.
Therefore, premature failures do not surprise me.

 

[Rottmannash]

I own NC252MP and NC502MP, and both came with flexed PCB (252 especially).
Also, even when unflexed, the PCB is under enormous load due to the compression of the thermal pads.
Therefore, premature failures do not surprise me.

Who said there were premature failures?

 

[Davide]

Who said there were premature failures?

I said I wouldn't be surprised to see such cases.
And anyway, one case is reported in this thread (NC252MP, 18 months old).
Restorer-John's analysis most likely shows a link to PCB flex.

 

[offtrack]

I have a 2 x UcD180HG + HxR amp and SMPS180v4 power supply that was put together around 2010.

In late 2014 one channel of the original v1 modules was taken out by an early revision of the Soekris DAM1021 firmware*.
The UcD180HGv1's were replaced with v2 modules and the HxR regs swapped across.

The amp has been a daily driver and is still running the original SMPS.

https://imgur.com/a/vMdFI0V

* outputs didn't mute on power down resulting in a very nasty "bang".

 

[Philbo King]

I view reliability somewhat differently than perhaps a few here. In my mind, reliability is not incidental or accidental. It occurs when that quality is seen as a primary and desirable characteristic. In other words, you don't design for efficiency, or effectiveness, and have reliability just happen. You have to design for reliability at the same time as efficiency or effectiveness, and then see how efficient you can make the device, or how effective it can become without reduction of that necessary reliability.

I don't see that happening today. Not in audio, and not in anything else. Jim Taylor

Excellent explanation. I put in 9 years in safety-critical and flight-critical avionics design. They had a team of EEs whose only job was to assess reliabilty of both components, circuit assemblies and entire boxes. Their failure criterion was less than 1 failure per 10^9 flight hours (about 10 times stricter than nuclear power plants), to be put into air transport jets expected to be in service for more than 50 years..
Any box or circuit that didn't meet muster was duplicated or triplicated, as needed.
The science of electronics reliabilty is well established.
The questions are 'what are you willing to pay?' and 'what are the consequences of a failure?'. Nobody wants to pay $500k for a stereo component...

 

[pseudoid]

Excellent explanation. I put in 9 years in safety-critical and flight-critical avionics design. They had a team of EEs whose only job was to assess reliabilty of both components, circuit assemblies and entire boxes. Their failure criterion was less than 1 failure per 10^9 flight hours (about 10 times stricter than nuclear power plants), to be put into air transport jets expected to be in service for more than 50 years..
Any box or circuit that didn't meet muster was duplicated or triplicated, as needed.
The science of electronics reliabilty is well established.
The questions are 'what are you willing to pay?' and 'what are the consequences of a failure?'. Nobody wants to pay $500k for a stereo component...

MIL-HDBK-217 became the bible of MTBF but it never said: "An elephant is a mouse built to government specifications!"
Forget about the toilet seat and the pencil (eventhough some may consider them to be mission critical)!

 

[Davide]

On this forum one guy report a smoked NC122MP.
One user also says he had 6 modules gone up in smoke, but I sincerely believe this guy less. Either way, that would be interesting analyze the case.

 

[fpitas]

On this forum one guy report a smoked NC122MP.
One user also says he had 6 modules gone up in smoke, but I sincerely believe this guy less. Either way, that would be interesting analyze the case.

Hypex said they cannot do anything and I have to go through the manufacturer, which happens to be a chinese no brand assembler selling on taobao who just put the Hypex card in a box with cables and a switch, is unlikely to provide any post sale assistance or warranty.

Sounds...sketchy.

 

[antcollinet]

Hypex said they cannot do anything and I have to go through the manufacturer, which happens to be a chinese no brand assembler selling on taobao who just put the Hypex card in a box with cables and a switch, is unlikely to provide any post sale assistance or warranty.

Sounds...sketchy.

Moral of that story is to purchase from a reputable manufacturer/supplier. The contract is with them, not with hypex.

 

[jbattman1016]

I've got one of the older pioneer SC-05 units running still. I do not believe we can generalize that every class D design is going to fail any sooner, but still not enough data points I think.

 

[fpitas]

Moral of that story is to purchase from a reputable manufacturer/supplier. The contract is with them, not with hypex.

That; and I have some question if it was even a real Hypex module, vs a knock-off.

 

[GXAlan]

I've got one of the older pioneer SC-05 units running still. I do not believe we can generalize that every class D design is going to fail any sooner, but still not enough data points I think.

That generation of Pioneer was excellent (for its time). Those used official ICEamp products and the good thing about ICE power at the time was how much it was being used by B&O products so there was a clear track record of in-field use.

 

[EJ3]

My gear has mostly been through a repair/replace scheme after 40 years (the length of time my NAD 2200's, ADVENT 300 receiver with pre-outs, Technics SL-M3 TT & DUAL 1229 TT) went without failure (as well as some other gear). Now, every 5 years I get them checked out, if there are some updated electronics, I usually have it done then.
Maybe I was lucky (with all of this gear it is a longshot).
But I have an expectation of any gear I buy to last at least 10 years before needing a refurbishment or repair. To me that is the bare minimum of longevity.
All my cars (and their electronics) have lasted either that long or 300,000 miles (sometimes both) without catastrophic failure. And I do NOT baby them.
Why would I not expect that (10 years, at least) of an item in my home (a protected, climate controlled environment).

 

[levimax]

Why would I not expect that (10 years, at least) of an item in my home (a protected, climate controlled environment).

Seems like some around here want SOTA performance for $100 with long life. Maybe too much to ask for.

 

[pablolie]

Reliability has zero to do with something being class anything... it's down to implementation. In fact the lower temperatures of Class D *should* correlate with higher longevity, because a lot of component failure is driven by operating at higher temps and (unfortunately) overheating or suffering to spikes in up-down cycles.

 

[dlaloum]

Reliability has zero to do with something being class anything... it's down to implementation. In fact the lower temperatures of Class D *should* correlate with higher longevity, because a lot of component failure is driven by operating at higher temps and (unfortunately) overheating or suffering to spikes in up-down cycles.

There are two sides to that coin - Class A /B/ AB amps have always been built with decent heatsinks, they generate heat and need the heatsinks to ensure long term reliability.

Because the Class D amps generate so much less heat, they are often sold with minimal heatsinking, or just bolted to the case - for the case to act as heatsink....

Without a means of disposing of the excess heat, the fact that they run cooler and more efficient won't eliminate the risk of overheating... and therefore abbreviated lifetime.

All things being equal - ie: using similar levels of heat management / heatsinks, a class D amp should be more reliable and have a longer lifetime due to much lower temperatures.... But manufacturers leverage the reduced heat emission, to remove expensive heatsinks or heatsink cases.... and the end result is..... variable

Plenty of thermal images of class D amps showing that some of their components are getting seriously hot.... with good heatsinking given class D's efficiency, this should not happen.

Such is life in a capitalist society.

 

[levimax]

Such is life in a capitalist society.

How much do you think a class D amp would cost and how many choices do you think you would have and how reliable would they be in a "command economy"?