uC3842AN SMPS idea for feedback - What do you think?

[WyverntekGameRepairs]

I was doing some thinking while reading over my thread I posted concerning my uC3842AN SMPS power supply questions regarding design choices of the PS1 SMPS, and I had a very interesting idea:
Because (in the way the PS1 was using it) the optocoupler is transmitting analog data, it can become dangerously unreliable as it ages and fades. It also could reduce the lifespan of the LED inside the opto due to it being in an ON state for so many hours. A great way to increase the lifespan of the LED is to allow it to turn off for a short period before turning back on again, instead of having it ON all the time. Since we are using analog in this circuit, it would make sense to use a A-D converter on the LED side and a D-A converter on the phototransistor side. Because it would become a pulse-width modulated signal at either 2.5V (or 3.5V depending on LED needs) or 0V, the brightness of the LED would not affect the output since the ON time of the LED would be measured instead of the brightness of the LED. In case you haven't caught on yet, this would be at a couple hundred Hz to even a couple KHz. To understand what I'm going on about, let's take a voltage of, say, 5V - and let's monitor it using an ADC. Let's assume this ADC can take a signal input ranging from -10V to +10V (meaning it can take AC as well as DC).Let's also assume that the ADC's median output (the output when 0V is being measured at the input) has a duty cycle of 50%. When the voltage is a stable 5V, the ADC outputs a square wave with a duty cycle of about 65%. However, as the voltage rises, the ADC measures that and increases the duty cycle depending on the voltage. But if the voltage begins to drop, the ADC measures that and decreases the duty cycle depending on the voltage. Basically, the ADC is measuring the voltage, and setting the duty cycle that coincides to the voltage measured on the input signal.
Going back to the optocoupler, if we connect the ADC output to the optocoupler, the LED will be turning on and off at a very fast rate. But as the duty cycle increases, the LED is on for a longer time, and vice versa. The LED is never purely on unless (and this is a worst-case scenario) the ADC receives a 10V input and thus outputs a square wave with a 100% duty cycle.
But what about the phototransistor side? Simple: The digital signal received would be converted to analog using a DAC that operates on the near same principle as the ADC, but in reverse. It measures the duty cycle of the input signal, and outputs an analog signal from -10V to +10V, depending on the duty cycle of the input.

This solution would greatly increase the lifespan of the LED, and would make it far more reliable in use, because instead of relying on the brightness of the LED (which changes over time as it ages and begins to fade), we are relying on the duty cycle of the high-frequency digital pulses transmitted by the LED.

I think that my own power supply will have a feature like this, because it is far more reliable and age friendly. As it grows old and begins to dim, the efficacy of the signal is not changed. The signal would not be damaged until the LED grows so dim that no signal is transmitted at all, thus causing the SMPS to stop working completely. I believe that this is also far safer - whereas in analog, the damaged analog signal due to the opto's aging and dimming LED would throw the SMPS controller IC's feedback measurements and calculations so out of whack that the SMPS would most likely suffer some form of damage or cause damage to the device it is powering.

What do you think to my solution? (I'm pretty sure a lot of people have used this solution in their projects and power supplies, but I thought of this just now and I want your opinions and feedback on it). Don't be afraid to correct me if I'm wrong about something, as I am very willing to learn something new.

Edit by gnif: Removed marquee, the information below isn't advertising any optos but rather providing technical data sheets on the science of optoisolators and best usage.

[SuzyC]

For technical questions, contact: optocoupler.answers@vishay.com ... In general, an optocoupler's life time is a period of 100000 hours.

Vishay says, "No Worries!"

Optocouplers are in use in every powered electrical instrument you can think of , from your smartphone charger to your TV to your PC and in every case, everything else in the P/S  is likely to fail first.

The nature of the feedback using an optocoupler in a SMPS is such that the brightness of the LED in the optocoupler is held constant over years by the LM431 on the cold side of the  P/S, so aging of the LED is compensated for.

What causes them to deteriorate is high temperatures, so keep things cool, man.

[SuzyC]

Lucky for you Opto's today cost only a few cents in large quantities, but you can get just one for $0.35 plus shipping from Vishay, but other cheaper Mfg's are probably just as reliable. Do a little research.

Toshiba says: https://toshiba.semicon-storage.com/info/docget.jsp?did=13438

The projected field lifetime for the LED = AF × stress hours = 184.7 × 1,000 = 184,767 hours (or 21 years). With the AF value
calculated, all data points of stress hours map to the expected field lifetime time.

Edit by gnif: de-googled the link

[gnif]

@WyverntekGameRepairs SuzyC is not advertising but referring you to some real information.

[gnif]

Anyway, let's get this thread back on track

This thread was always on track, the information @SuzyC provided did not advertise any brand make or model of opto, rather the science behind them and how they operate, how best to use them in a SMPS where the LED may fade (feedback is king) and information from a very reputable (industry leading) manufacturer and researcher of high quality optos.

For the record, the link provided is an application note titled "Basic Characteristics and Application Circuit Design of Transistor Couplers", with the outline "This document outlines the basic characteristics and application design of general-purpose transistor output photocouplers". It's an excellent read on the characteristics and usage of optoisolators.

[WyverntekGameRepairs]

 
Yeah, I screwed up.
Oopsie on my part for jumping to conclusions.
@SuzyC I have PM'ed you a personal apology.
Just a suggestion, you may want to change your format a tad... It looks like a blatant advertisement and could easily be taken as such at face value. Especially this one:

For technical questions, contact: optocoupler.answers@vishay.com ... In general, an optocoupler's life time is a period of 100000 hours.

Vishay says, "No Worries!"

Optocouplers are in use in every powered electrical instrument you can think of , from your smartphone charger to your TV to your PC and in every case, everything else in the P/S  is likely to fail first.

Now that I understand what you are trying to convey though...
I already know that optos last a very long time. However, I want to consider the worst case scenario in my power supply so I can prepare for the worst. What if the power supply is in a hot location? This can damage the opto, and make it dimmer a lot faster than normal. I want to be able to allow for my power supply to have a near-perfect feedback signal no matter how old or dim the opto is, especially because feedback is one of the most important parts of a SMPS. This is a design feature I call "quality." It must last for a long time, even in the worst case scenarios. I will be designing my power supply to be the best it can for it's application, but the fact is nothing is perfect. While it is rare that an opto fails or dims out quickly, it is still a possibility. I want to prepare for that possibility. That is why I am going the extra mile with these design choices - I need this thing to last no matter the scenario.

[MagicSmoker]

...I will be designing my power supply to be the best it can for it's application, but the fact is nothing is perfect. While it is rare that an opto fails or dims out quickly, it is still a possibility. I want to prepare for that possibility. That is why I am going the extra mile with these design choices - I need this thing to last no matter the scenario.

If you are worried about opto lifetime - and this is not an unreasonable concern for PSUs that are left plugged in 24/7 - then don't use an opto at all. You can often get good enough voltage regulation by adding another winding to the transformer and deriving the feedback from it. As a bonus, this winding can also supply power to the controller IC so that only a small trickle-charge from the rectified mains is needed for starting up. The attached LTSpice schematic illustrates the principle and gets you in the ballpark as far as voltages go (ie - I did not optimize frequency compensation or turns ratio).

[T3sl4co1l]

But by the time you're worried about opto lifetime or operating temp, you need to fix a lot more parts.

You'll be using UC1842 first of all.  Probably electrolytics --> fat fuck film caps.  Higher class insulation on everything, necessitating custom transformers.  Transient protection probably, which can still wear out (MOVs).  You can get TVSs or use other methods but they're more expensive, of course.

No idea who'd buy a brick-shithouse power supply that will more than outlast the already-aging things it's intended to power, and which costs several times its original retail price...

Tim

[WyverntekGameRepairs]

If you are worried about opto lifetime - and this is not an unreasonable concern for PSUs that are left plugged in 24/7 - then don't use an opto at all. You can often get good enough voltage regulation by adding another winding to the transformer and deriving the feedback from it. As a bonus, this winding can also supply power to the controller IC so that only a small trickle-charge from the rectified mains is needed for starting up. The attached LTSpice schematic illustrates the principle and gets you in the ballpark as far as voltages go (ie - I did not optimize frequency compensation or turns ratio).

Hm. You know what, you've got a point there. I think I've heard of this technique before, something like primary-side feedback? It omits the opto and replaces it instead with a coil on the transformer. And that also means less parts so less points of failure, so that could actually be a fantastic idea. Would there be any important aspects I would need to consider if I was to use this method?

But by the time you're worried about opto lifetime or operating temp, you need to fix a lot more parts.

You'll be using UC1842 first of all.  Probably electrolytics --> fat fuck film caps.  Higher class insulation on everything, necessitating custom transformers.  Transient protection probably, which can still wear out (MOVs).  You can get TVSs or use other methods but they're more expensive, of course.

No idea who'd buy a brick-shithouse power supply that will more than outlast the already-aging things it's intended to power, and which costs several times its original retail price...

Tim

You gave me a right laugh indeed! And yeah, looking at the big picture these power supplies will probably outlive the consoles they are powering. And the people who own them.
Also, these will cost about $16. That's about what it should be, if not cheaper.

[T3sl4co1l]

You'll be lucky if China can make it for that price in quantity, let alone in small quantities, let alone in the US.  If you're selling hundreds/year I would guess $200 would be closer to profitable, assuming of course you get that many buyers at that price.

On the upside, you can afford a couple bucks for better parts at that sale price.

Relevant:

Hm. You know what, you've got a point there. I think I've heard of this technique before, something like primary-side feedback? It omits the opto and replaces it instead with a coil on the transformer. And that also means less parts so less points of failure, so that could actually be a fantastic idea. Would there be any important aspects I would need to consider if I was to use this method?

The transformer needs lower leakage inductance, which determines cross-regulation, and you need good cross-regulation to use aux supply for feedback.  Which, again, hints at a custom transformer.

And again, you can tolerate poorer regulation at the main supply if you post-reg (tack on some LDOs), at some cost to efficiency.

The other major way is to use an off-the-shelf module like a MeanWell something or other (ca. $10 in these ratings and quantities), and derive the other outputs with POL (point of load) DC-DC converters.

Tim

[MagicSmoker]

Hm. You know what, you've got a point there. I think I've heard of this technique before, something like primary-side feedback? It omits the opto and replaces it instead with a coil on the transformer. And that also means less parts so less points of failure, so that could actually be a fantastic idea. Would there be any important aspects I would need to consider if I was to use this method?

As T3sl4co1l already mentioned, the higher the leakage inductance the worse the cross-regulation. Still, if you can keep leakage under 3-4% (which is eminently reasonable, even with all the safety agency mandated crap) and the loading of the outputs doesn't vary too wildly this technique works well enough. Don't let perfect be the enemy of the good and all that. However, this technique should only be used on flyback converters, not forward types.

And yes, you are going to need a custom transformer here, but that was always the case because the only SMPS transformers you can buy off the shelf are overpriced jobs intended for dev kits and reference designs. The chance you will find one that fits your design specs is somewhere between nil and none.

EDIT - posted before including the other part of the reply:

You'll be lucky if China can make it for that price in quantity, let alone in small quantities, let alone in the US.  If you're selling hundreds/year I would guess $200 would be closer to profitable, assuming of course you get that many buyers at that price.

Aye, this is undoubtedly a questionable business venture, but that doesn't mean it isn't a worthwhile endeavor for the OP to undertake. You really don't learn the intricacies of product development until you've actually developed a product, after all.

[WyverntekGameRepairs]

Hm. You know what, you've got a point there. I think I've heard of this technique before, something like primary-side feedback? It omits the opto and replaces it instead with a coil on the transformer. And that also means less parts so less points of failure, so that could actually be a fantastic idea. Would there be any important aspects I would need to consider if I was to use this method?

As T3sl4co1l already mentioned, the higher the leakage inductance the worse the cross-regulation. Still, if you can keep leakage under 3-4% (which is eminently reasonable, even with all the safety agency mandated crap) and the loading of the outputs doesn't vary too wildly this technique works well enough. Don't let perfect be the enemy of the good and all that. However, this technique should only be used on flyback converters, not forward types.

And yes, you are going to need a custom transformer here, but that was always the case because the only SMPS transformers you can buy off the shelf are overpriced jobs intended for dev kits and reference designs. The chance you will find one that fits your design specs is somewhere between nil and none.

That's quite a bit to take in. You've got a point, too, perfection isn't everything. I just want it to be as functionally stable as possible is all. And yes, leakage inductance. That's going to be an assfest to deal with, especially if I tried using an off-the-shelf part. I think the reason I was leaning toward using an opto in the first place is because that is how the transformer that I was designing this around is built for.
The transformer had the right specs and everything, and it was designed specifically for the uC3842AN. It has the correct outputs, it has the correct safety regulatory specs, it has everything I need. It's when better concepts are introduced that I start getting frustrated, because the transformer I want probably doesn't have said feature.

Switch mode power supplies are not complex. They are big systems made of elementary circuits, and so they have many moving parts. Each moving part does something with precision. Then you have a design, but you have improvements but cannot make said improvements because certain parts cannot be changed, then one must compromise. I might have to forget about primary-side feedback. Unless... I might be able to stack transformers. Problem is, this might affect leakage inductance greatly.

I'm at a loss right now.