Class D amp long term reliability

[Inner Space]

... having spent my life in all facets of sales, from HiFi to houses, fashion to furniture ...

Off topic, but ... what? You? Fashion? The entire board wants to hear about this ...

 

[pma]

I wonder what is the life time of the capacitor in the output LC filter of almost any class D amp. If we consider usual values L = 20uH, C = 1uF and switching frequency Fosc = 420kHz, then the output filter capacitor is a subject of permanent high ripple current with high di/dt, this with non-zero ESR makes considerable power loss to be dissipated. The capacitor should be polypropylene impulse type and even then the ripple current is of concern. Any long-term experience with this here? Please see the explanation below:

L1 and C1 make the output filter

The plots show voltage before the filter, v(out1) with +/-30V amplitude, output voltage v(out) across the load R4, current through capacitor i(C1) and current through load i(R4). Please note the amplitude of capacitor current and the ripple current that will be there even at v(out) = 0.

 

[audio2design]

Probably one of the worst examples of this are companies selling led overhead light fixtures that require the customer to replace the whole fixture when the bulb no longer works. How about get a regular fixture and buy the LED bulb separate? The big box retailers also carry fixed 50 foot LED strings for Christmas. Defeats the purpose.

My fixed LED strings from 5+ years ago are still going strong. The ones from last year with replaceable bulbs are in the garbage.

 

[audio2design]

I wonder what is the life time of the capacitor in the output LC filter of almost any class D amp. If we consider usual values L = 20uH, C = 1uF and switching frequency Fosc = 420kHz, then the output filter capacitor is a subject of permanent high ripple current with high di/dt, this with non-zero ESR makes considerable power loss to be dissipated. The capacitor should be polypropylene impulse type and even then the ripple current is of concern. Any long-term experience with this here? Please see the explanation below:

View attachment 166908
L1 and C1 make the output filter

View attachment 166913
The plots show voltage before the filter, v(out1) with +/-30V amplitude, output voltage v(out) across the load R4, current through capacitor i(C1) and current through load i(R4). Please note the amplitude of capacitor current and the ripple current that will be there even at v(out) = 0.

At typical output power probably almost forever.

 

[audio2design]

LED bulbs are sold and they are replaceable like incandescent & florescent ones are. You don't toss the whole LED lamp fixture at replacement time.

Most thin downlight designs you so. But stick to better brands and they rarely fail.

 

[audio2design]

Another view is that manufacture of separate bulb and fixture may require substantially more resources than the single combined product. And assuming a 50,000 hour life, for a product like tree decoration it may be more sustainable to do the combined approach.

Your larger point is correct, though. Longer service life and repairability are generally characteristic of more-sustainable design, and therefore desirable.

I remember when Japanese cars became really popular and everyone complained about how hard they were to be worked on. And then the people stopped complaining about that because it was so rare that you had to actually work on them.

 

[pma]

At typical output power probably almost forever.

Based on???? The di/dt through the filter capacitor is almost independent of output power, it is defined by switching frequency, rail to rail swing, inductance and capacitance.

 

[audio2design]

But let's get back to the OP's question shall we? I have an 18 month old Hypex NC-252MP sitting here on my bench. Dead. Built 41st week of 2019. Sent to me by an ASR member to investigate why it failed as it had a very easy life in a multichannel amplifier. The brand and the member will stay anonymous as he was very happy with their customer service in rapidly rectifying the situation to his satisfaction.

Let's have a quick visual inspection to see if anything stands out.
View attachment 147422

I can immediately see some very dodgy hand soldering on the SMPS primary side switching transistors and some of the output MOSFETS, because these are the parts done by humans. Regardless, it should have never left the factory with soldering this poor. I think I am onto something:
View attachment 147423

Also, the board is clearly under a lot of strain (it bends) due to the screwing down of the PCB to the heat-spreader near the heat producing SMPS switchers:
View attachment 147425

The board bends enough to be 0.76mm out of alignment and the devices are no longer firmly pressed against the heat spreader:
View attachment 147426
The screws are also too far away from the devices and the impregnated heat transfer pad is too thick. (approx device package positions marked in red). They should not be using distantly mounted fixing/screw points and relying on the flexible PCB to maintain adequate device to heatsink contact. This is bad design 101.
View attachment 147424

First of all, let's dismantle the sucker and remove the heat-spreading aluminium plate and have a look-see:
View attachment 147427
Yes, you guessed it! It's gone fully Chernobyl!

Vaporised SMD components hanging off the primary side SMPS switching transistors. Basically, The board flexes under pressure (from the squishy "thermal" pad), so the transistors (FETs) are not adequately thermally coupled any longer to the aluminium spreader. They eventually fail from over heating (these particular devices are both internally shorted) and they take out some associated components around them:
View attachment 147428

View attachment 147429

A real mess. There's even a poor SMD transistor with its package cover blown right off. You can see the junction components! And a nice vaporized resistor or two, along with those poor MOSFETs that gave up due to their inability to get rid of heat.

All due to inadequately or properly mounting (you know, with screws, nuts and washers) and heatsinking the transistors. Hypex are simply squashing the devices against the aluminium plate, using plenty of PCBs screws and a foam heat pad, but not considering PCB material deforms under heat and pressure over a (short) period of time.

In my opinion, all the NC-252MPs will likely suffer the same fate in years to come, if this key section is not revised. Not remotely good engineering at all and certainly not likely to last the length of time customers deserve.

Apart from this glaring design fault, the rest of the board is of good quality. The components are adequate and very nicely laid out. I do not like TO220 heat producing devices squashed under pressure from PCBs. It doesn't work. It creates more problems than it saves money. And all the secondary rectifiers and output stage FETs are the same unfortunately.

Hypex can do a whole lot better if they are selling supposedly state of the art equipment. 18 months lifetime for an entire 250WPC module including PSU is just not remotely good enough.

0.75mm seems like a lot but that heat sink pad is thick and would still maintain good contact. Typically failure like this in the SMPS is either surge event causing avalanche breakdown beyond the device rating or glitch on the controller. A bad connection would of course cause it. If the gate does not shut off on the control current peak, poof goes the FETS and the current sense resistor. Ditto for the gate drive and resistors. SOT23 could be part of gate drive.

The layout looks a bit amateur in the power section. That was not designed by someone who does high volume low cost.

 

[Colonel7]

My fixed LED strings from 5+ years ago are still going strong. The ones from last year with replaceable bulbs are in the garbage.

I've had the opposite experience with outdoor Christmas lights strings. Longest lasting was a bit less than 3 years. 2 failed in less than one month

 

[audio2design]

Based on???? The di/dt through the filter capacitor is almost independent of output power, it is defined by switching frequency, rail to rail swing, inductance and capacitance.

Di/dt is not average ripple current. Now apply time to di/dt.

 

[audio2design]

I've had the opposite experience with outdoor Christmas lights strings. Longest lasting was a bit less than 3 years. 2 failed in less than one month

The newer commercial ones do not have removable bulbs. Everything soldered. Socketed bulbs are big failure points. Look at the FIT of a single LED in the billions of hours versus the mechanical connection. Mech connection far more failure prone than a soldered LED

 

[Ken1951]

I've had the opposite experience with outdoor Christmas lights strings. Longest lasting was a bit less than 3 years. 2 failed in less than one month

My LED outdoor light strings, 5 or 6 I think, are going strong after at least 5 years I think. So much less hassle than the old ones, and no broken bulbs to deal with.

 

[antennaguru]

Just went to turn on a little Class D amp using an IcePower 50ASX2 module for a small room, and heard only a loud click over the small loudspeakers. I swapped the amp out for a chip amp with a pair of LM1875 devices and there was music again in that room.

I took the IcePower 50ASX2 amp to the workbench and connected it to larger speakers and heard a loud hum when I turned it on. It seems like it failed and I need to open it up and take a closer look inside the casework to see if the failure is visible. It only saw very light occasional service for a few years, so I'm quite surprised for it to fail. I'll report back after I have look-see.

 

[antennaguru]

OK - so another Class D amplifier module bites the dust! There was no reason why it should have failed as it worked fine the last time I used it and it was never abused and only saw light service. The other day it wouldn't turn on.

This one is an IcePower 50ASX2 that was in service for maybe 5 years. It's just making a humming noise now, and not passing any signals. I opened it up and nothing visually seems wrong. I did a make/break with every one of the professionally made (Ghentaudio) jumpers and there was no change. Again the only option is to remove and replace the bad module. I have had 2 class D plate amps fail already so am getting weary of the disposable nature of Class D modules.

I am attaching pictures that show how I installed this module in the case, how there was convection cooling across the module as well as ample room inside.

This is the front/top view of the case:
ICE 50ASX2 a by antennaguru uda, on Flickr

This is the rear/bottom view of the case:
ICE 50ASX2 b by antennaguru uda, on Flickr

This is the inside view of the bad module:
ICE 50ASX2 c by antennaguru uda, on Flickr

You can click on the pictures to enlarge them.

 

[pseudoid]

forever

Who would be the manufacturer of a "forever" capacitor?
Must be like the marriage vows:

"For every 1,000 marriages in the last year, only 14.9 ended in divorce, according to the newly released American Community Survey data from the Census Bureau. This is the lowest rate we have seen in 50 years. It is even slightly lower than 1970, when 15 marriages ended in divorce per 1,000 marriages."

 

[Count Arthur]

OK - so another Class D amplifier module bites the dust!

Have you tried bybassing the rotary power switch?

Unless they are pretty beefy, most rotary switches aren't designed to deal with much current, 3A perhaps, I wonder if the switch has failed.

For something like a power amp, for longevity I would typically use a switch rated at 16A @ 250v AC for the mains: https://cpc.farnell.com/marquardt/1...-off-16a/dp/SW05357?st=16a 250v rocker switch

 

[antennaguru]

Have you tried bybassing the rotary power switch?

Unless they are pretty beefy, most rotary switches aren't designed to deal with much current, 3A perhaps, I wonder if the switch has failed.

For something like a power amp, for longevity I would typically use a switch rated at 16A @ 250v AC for the mains: https://cpc.farnell.com/marquardt/1551-3102/rocker-switch-spst-on-off-16a/dp/SW05357?st=16a 250v rocker switch

Yes, I checked that. It's turning on but just humming loudly - and pushing the speaker cones out to do it. It does not pass signals at all - just hums and makes you want to make it stop.

There were extra contacts so I doubled up the contacts for each lead. Switch is fine.

 

[valerianf]

I would look first at the 2+4 big electro chemical capacitors. Usually it is low quality one.
You need to desolder them and test them with a LCR meter.
It is not easy to desolder because they are glued.

 

[antennaguru]

I would look first at the 2+4 big electro chemical capacitors. Usually it is low quality one.
You need to desolder them and test them with a LCR meter.
It is not easy to desolder because they are glued.

Why bother? It's a $100 module. Class D modules were destined for the landfill when they were designed/built. Just remove and replace...

 

[Doodski]

Why bother? It's a $100 module. Class D modules were destined for the landfill when they were designed/built. Just remove and replace...

Which makes me come to think about the future stuff. With globalism contact(s) list(s) and futuristically thinking very progressive companies. The sky really is the limit.