Digital vs Analog Isolators in Smps Design

Green/yellow

...Jim Thompson

Maybe I'll go back and try again to model an opto.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

In my case, fast design time trumps expensive parts. My product idea may flop. All the time put into making the design cheap could be wasted. If the product doesn't flop, then the pcb can be cost optimized. Also, the product volume is low and each unit sells for about $700.00 ... A couple of extra bucks on expensive parts is ok at this time..

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

True. For my design, it only needs loop gain less than 10khz. Alls fine in that case if I use an optoisolator to feedback the output voltage. But if I change the design to use a digital isolator, I have to use something really fast since my PWM is at 1Mhz. Optos are sluggish in a digital role compared to the other isolation methods (GMR, galvanic etc..) I've never seen a digital isolator used in an smps design. Only linear.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

I just never saw a compelling reason to use any fancy isolators. One of my staples for very high isolation and where cost is warranted is the CNW137 widebody:

Rise/fall times in the 25nsec range, how much more would you want?

If you use a TL431 you can shave off another $0.15 :-)

:On Sat, 28 Feb 2009 08:14:08 GMT, Ross Herbert : wrote: : :>On Fri, 27 Feb 2009 19:58:46 -0800, D from BC wrote: :>

:>:On Sat, 28 Feb 2009 03:06:38 GMT, Ross Herbert :>: wrote: :>: :>:>

:>:>I pulled a psu from a defunct Epson inkjet printer and it uses a similar setup :>:>to your description. The mains input primary side switcher uses a mosfet :>:>controlled by bjt's and the secondary side uses a L4962EA 1.5A smps controller :>:>to produce a 42V output. :>:>

:>:>As for the OP's query, my initial reading is that the digital isolator cannot :>:>be applied to smps feedback control. The output control of an smps relies on :>:>

:>: :>:I don't think so. :>

:>:Once the analog error signal gets to a PWM (or other), the error :>:signal crosses into digital land. :>:This works when the PWM is on the secondary side. :>

:>Take a look at the data sheet for a typical controller IC commonly used in SMPS, :>eg. UC3842. :>

:>"Pin 2 - Voltage feedback - This is the inverting input of the Error Amplifier. :>It is normally connected to the switching power supply output through a resistor :>divider." :>

:>That says the feedback signal is analog, not digital. :>

:>In SMPS which don't use a PWM controller IC, the feedback signal is always :>analog. : :I think the UC3842 is meant to be used on the primary side. :I don't think a digital isolator can be used with a current mode :controller (UC3842) because it uses 2 feedback loops. One loop is the :current ramp from the primary side and other loop is the voltage :monitoring (via linear opto) from the secondary side. :To use a digital isolator, there can only be linear feedback loops on :secondary side. :(The UC3842 uses a primary (I mode) and secondary (V mode) feedback :loops.) :A digital isolator can work with a voltage mode controller used on the :secondary side. (The UC3842 is a current mode controller.) :With a digital isolator, it's the mosfet gate signal that is :isocoupled not a sample of the output voltage.

In this thread so far, you hadn't indicated that you were trying to implement a smps design with the PWM control on the secondary side. It was Tim who raised this possibility but you didn't indicate that you were going down this path. It is no wonder that I assumed that you were trying to implement digital feedback control from the secondary side to the primary side of a line powered smps.

Only now, when I pointed out that a typical widely used smps controller IC wouldn't work using digital feedback, did you specify that you were thinking of doing the pwm and feedback control on the secondary side. If I had known what you were trying to do then I probably wouldn't have commented at all.

Well..yeah.. My posts can be a bit buggy sometimes..

So I guess you haven't seen a smps design using a digital isolator instead of a linear isolator...

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

Let's see.. Smps clock: 1Mhz Min duty: 10% Tonmin = 100nS

ok...looks like the CNW137 is fast enough for my app.. :(

So maybe I was getting a little speed crazed.. :)

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

LMV431 ref voltage is 1.24V LM431 ref voltage is 2.5V

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

Sure, but do you need less than 2.5V Vout?

Yup.. 2V is required in my design.. It's not pretty but I got a LMV431 stuck on the output of an op amp to create a 2V offset. (Long story how that got decided.)

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

Ok, didn't know that. That'll be 15 cents extra :-)

Yeah, those legacy solutions. "Hey, that goes back to ol'Leroy, not purdy but it's always worked, why's it wrong now?"

I bought 2 coffees today for $4.50.. :O

I'll cheapen up the design later on.. No point in sharpening the pencils if they're being used for firewood. (uh...I just made that up..I'll think of a better expression later.)

I suppose it's subjective.. Maybe it's strange compared to the other ways I get an offset. Perhaps it's a sign I'm getting better at electronics..

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

And crack out the old percolator. $4.50 for a coffee is insane ;-)

That was two coffees, so it's half sane. ..or perhaps half-n-half.

I ordered a coffee for myself and another. It was a new coffee shop in the area and I didn't check the price. I was shocked at the cash.. For the volume, it was more expensive than Starbucks, Tim Hortons and McDonalds.. btw..I like McDonalds coffee the best.. Smooth and low caffeine. Over here, Tim Hortons coffee is like a narcotic. Drop a caffeine pill in McDonalds coffee and you get Tim Hortons. :P Tim Hortons coffee is more consistent than McDonalds.. Occasionally McDonalds coffee is too thin. Tim Hortons is very consistent. Starbucks is nice for that occasion funky taste.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

McDonalds is the *worst* coffee on the planet. I'd drink Starbucks first, and that sucks too.

Seen 'em, never been in one.

Don't do caffeine.

Dinkin' Donuts is quite good, though usually only lukewarm.

Ick. Green Mountain Coffee Roasters is the best I've had. We have it shipped in, though have seen it in grocery stores in the Midwest.

I hope you're not going back to the time when McDonalds coffee was really offensive. Like burnt solder rosin. I can't recall when McDonalds switched to better coffee.

I've tried coffee at a few of the truck stops around here and the coffee is surprising good. I suppose it takes good coffee to stop the truckers.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design