Long term dependable digital power supplies?

Thread Starter

Elerion

Joined Sep 11, 2017
85
Hi everyone.

There're now lots of comercial power supplies in the market with a high degree of digital features.
Some of them even store all changes in voltage/current in memory. If you just unplug the unit at any time, then power again after a couple of hours, the same voltage/current settings are there.

If even the slightest change in voltage is stored in flash memory inside a microcontroller, it seems not very dependable gear in the long run. Flash memory wears and fails after a heavy use. And I'm not talking about supply where you store the settings whenever you need to. I'm talking about supplies that are continuously writting to memory any minor change.

This seems to be a trend. I'm not saying to go back to just pure analog and potentiometer driven units (which, by the way, seem the most dependable ones), but I wander if this new fancy supplies are made to stay, or to be replaced in a couple of years of heavy use.

What do you think about this?
 

crutschow

Joined Mar 14, 2008
24,296
EEPROM or Flash memory are typically good for about 100,000 erase cycles, so that would be about 28 cycles a day for 10 years.
I would think it is not written to the memory until the power is shut off, so that's probably more than sufficient for most bench supply applications.
 

MaxHeadRoom

Joined Jul 18, 2013
19,968
A lot depends on your application/end use etc, power rating etc.
In some areas, SMPS do not stand up well. Which makes a rugged linear supply more of an asset.
Max.
 

OBW0549

Joined Mar 2, 2015
3,203
What do you think about this?
Writing to flash continuously with every minor change would be a pretty dumb way to do it, and I seriously doubt any supply manufacturers are clueless enough to do that.

Far more likely they store on power-down, as @crutschow suggested or, when settings are changing they wait until a few seconds after the settings have stopped changing and only then do the write to flash.
 

MrChips

Joined Oct 2, 2009
20,167
We used to have a Data General Nova 2/10 computer with the most robust power supply.
There were no electronics involved, just one constant voltage transformer and a capacitor.
 

MaxHeadRoom

Joined Jul 18, 2013
19,968
In what sense?
In my area of expertise, namely Industrial control systems they were often used in shop floor systems, due to the number of failures that I came across I always spec'd in linear versions, either commercial or custom built to suit.
Max.
 

Analog Ground

Joined Apr 24, 2019
305
It sounds like you are asking about existing commercial power supplies. The endurance of any Non-volatile memory, NVM, (Usually Flash memory) is a reasonable question to ask the manufacturer. If you are asking about including NVM in a new design, there are ways to use blocks of NVM to get a very large number of write cycles. Which are you asking about, existing or a new design? If existing, do have an example model?
 

Thread Starter

Elerion

Joined Sep 11, 2017
85
Which are you asking about, existing or a new design? If existing, do have an example model?
I just tried with a new cheap power supply from an unknown manufacturer. HANMATEK. HM305P.

Far more likely they store on power-down, as @crutschow suggested or, when settings are changing they wait until a few seconds after the settings have stopped changing and only then do the write to flash.
The settings are definitely not saved upon shutdown. I tried powering down using the front panel switch and unplugging the power cord, and that makes no difference. The settings are restored either way.
They seem to be stored as OBW0549 said. When changing voltage or current, there're a few seconds before the panel locks again and no digit flashes on the display. That is the moment where the settings are saved. If I power off before the panel locks, the new settings are lost.

I would think any reputable manufacturer would minimize the number of writes to non-volatile memory, and only do it upon power down.
I agree. 100%. Unfortunately, seems not the case with this model. And I find it quite a bad thing. Because it is quite easy to make LOTS of changes to voltage (or current) in many applications, especially if using the supply as a test voltage/current source.

I suspect this is not true in all cases. Does it make sense? In what cases? Why?
Sure, not true in all cases. But this may happend in many other designs, and you don't know until you buy it. Lots of power supplies nowadays seem just clones, except quite expensive ones. Even known manufacturers are making "low cost" units (under 300 $ or €), which are almost exactly as other clones, but still much more expensive.

Don't you think many electronic devices are just failing more often that ever (designed to be replaced in a near future), when we are needing the exact opposite? In the sense that the world is getting full of technological trash, and many things are just thrown away with a minor fault that the user doesn't know how to repair. I see that all the time.
 

nsaspook

Joined Aug 27, 2009
6,886
The digital components are likely to be the most reliable parts in a modern power supply. The vast majority of repairs I've seen are cooling related (once the infancy failure period is done). Bad fans and overheated/failed power components are common. After 20 years of operation the 'bad fan' problem becomes endemic in large industrial operations. We have life extension protocols specifically designed to replace the original high quality fans with new high quality fans before failure.
 

Thread Starter

Elerion

Joined Sep 11, 2017
85
The worst thing would be, if the non-volatile memory fails to retain the setting, the supply would become unusable.
Well, that's quite a bad thing, don't you think?

The digital components are likely to be the most reliable parts in a modern power supply. The vast majority of repairs I've seen are cooling related (once the infancy failure period is done)
That's true. But it is something I (and many people) could repair (including replacement of burnt transistors or passive components, and even replacing a transformer). But if a microcontroller fails (or a complex and undocumented digital circuitry board), that's definitely much harder to fix. Don't you think?
 

nsaspook

Joined Aug 27, 2009
6,886
Well, that's quite a bad thing, don't you think?



That's true. But it is something I (and many people) could repair (including replacement of burnt transistors or passive components, and even replacing a transformer). But if a microcontroller fails (or a complex and undocumented digital circuitry board), that's definitely much harder to fix. Don't you think?
It's statistically unlikely the digital components will fail just from normal operation on a well designed board. 99% of the repair root causes I see are heat/environmental excesses, aging capacitors, bad internal/external wire connections and bad power (surges, cycles) related. I've got equipment that's been running so long 24/7, the photo-optical systems are failing because the leds are fading long after they have exceeded the lifetime hours limits. Long term dependability IMO is unlikely be limited by properly used digital components on high quality substrates as they were built under the most perfect/cleanest conditions on this planet, actually tested for proper operation and sealed to isolate them from the world.
 
Last edited:

MrChips

Joined Oct 2, 2009
20,167
I worked for many years with electronic equipment and computers. My anecdotal evidence has been that equipment failed more often after a power off/on cycle. For this reason, we left all computers running 24/7.

My most recent evidence was two years ago when I went away on vacation for four weeks. I turned off the power on my desktop PC before leaving work. When I returned, the PC would not boot. Memory was bad. HD was corrupted. Had to scrap the entire system and replace it with a new one.
 

SamR

Joined Mar 19, 2019
1,787
Long ago I was given the concept of reliability of electrical devices as being 25 years MBTF (mean time before failure). In standard distribution curve that tells me that even after quality control release of the product a very low percentile will fail out of the box and an equal low percentile will never fail but the Mean Value of them will last 25 years. Some less, some more.
 
Top