Why is this opto-isolated feedback loop improvement method not more common in SMPS?

[treez]

Page 16 of the following shows a cascoded feedback loop optocoupler fed by a current source, this vastly improves SMPS transient response, and makes feedback loop compensation easier. Why is the technique not more used in offline, isolated SMPS’s?

Pg 16 shows the components in the schematic, and page 9 and 10 explain it..

https://www.onsemi.com/pub_link/Collateral/EVBUM2148-D.PDF

- - - Updated - - -

As such, can you see problems with cascading the opto feedback to the ICE2HSO1G LLC controller as attached?

ICE2HSO1G LC Controller datasheet:
https://www.infineon.com/dgdl/Infin...n.pdf?fileId=db3a30432a40a650012a458289712b4c

 

[D.A.(Tony)Stewart]

Naturally all the 2nd order lead-lag feedback networks improve transient response, if done right.

 

[treez]

thanks, Diode D5, on page 16 of the onsemi app note of post #1 is key.....this stops the error amplifier from railing, meaning it gets back into regulation quicker...less overshoot...this is such a brilliant improvement, that its amazing that every error amplifier in the world does not use this technique....ie, the diode d5.

 

[D.A.(Tony)Stewart]

I liked the high bandwidth of the opto-diode. Op Amp current source and the inclusion of a injection point with R19 for Bode Plots, something every SMPS regulator should have, yet so few realize it's importance.

Yes D5 clips large output errors during startup with a preset adjustable zener set to 2.94V with 3.3V feedback expected from output. But the AC gain is small and DC gain is 1+ so the effect is to compress the AC feedback. So without looking at it in more detail, I would say they spent a lot of time on fixing a transient problem. It reminds me of a Type II PLL where frequency error is the same amplitude as max phase error.

 

[treez]

thanks, this simulation shows it well, the cascade allows you to compensate the feedback loop without the dreaded low frequency opto pole.

 

[Easy peasy]

Its not always convenient to use a proper current drive to an opto, also many normal circuits act to lower the gain when driving an opto - which is usually good news for the loop, granted buffering the opto xtor o/p helps raise the pole freq of the opto a little, but a good quality opto with CM shield works pretty well too...

 

[treez]

Its not always convenient to use a proper current drive to an opto

...in the case above (page 9,10,16 of app note in post#1) you can see that the current drive to the opto is ultra-convenient, -just two opamps and some Rs and Cs.

good quality opto

I have been looking for a good quality opto for ages, for SMPS feedback use, ive found "poor" optos, and "less poor" optos, but never (what I would call) 'good quality'.
..you always get the dreaded low frequency pole, whose tolerance is enormous, and the CTR is so widely toleranced that it might just as well be a lottery number.

- - - Updated - - -

granted buffering the opto xtor o/p helps raise the pole freq of the opto a little

..a little?, most literature on this subject that ive perused is saying the pole is pushed up from about 10khz, to above 50khz

-its baffling to me that cascoding the opto is not more common in isolated smps feedback, -the Cascode BJT costs all of 0.02c.
............................................................................................
At the top of page 10 of the app note in post#1 (Left hand side near top), it says that

QUOTE:
As the (optocoupler) diode is driven with a current
instead of a voltage source, the error amplifier output does
not have to swing too far during a transient
UNQUOTE

..however, this is not entirely true, it’s the diode D5 (seen on page 16) that means that the opamp output doesn’t have to swing too far during a transient.

 

[Easy peasy]

most control loops are < unity gain below 1kHz, so having opto performance above 10kHz is of little benefit, remember the op-amp driving the opto-LED has to have a power bandwidth of ~50kHz if you are to get 50kHz thru the opto - so the cost is going up, and unless you overdrive the opto led you still need an opto with >10kHz BW - mainly determined by the turn off time of the xtor - so a better opto needed (better than a 618 for e.g.) as well - no free lunch sorry...

- - - Updated - - -

check out the likes of a 4503 for a quality opto for power converters...

 

[treez]

most control loops are < unity gain below 1kHz, so having opto performance above 10kHz is of little benefit, remember the op-amp driving the opto-LED has to have a power bandwidth of ~50kHz if you are to get 50kHz thru the opto

Thanks, though I’ll be more specific, if we cascode the optocoupler as above, then we are getting to the stage where it’s as if we are having to compensate a feedback loop where there is no optocoupler in it……and as most people would agree, that is a massive improvement, as compensating a straightforward non-isolated feedback loop with no optocoupler in it is particularly simple……this is the panacea that cascoding the optocoupler gives us. Surely this is a big advantage?

 

[Easy peasy]

it's only an advantage if you really need a high BW control loop, most converters do not, its only a very few DC/DC converters that require 5kHz or higher..... even getting that on a converter that has output C and series L in the power train is difficult....

 

[treez]

thanks, we need to iron out the undulating PFC output voltage into our 3.8KW LLC resonant converter, so we need a bandwidth some way above 100Hz. The load is a huge 360V battery.
I appreciate that 5khz feedback loop bandwidth is pushing it. However, its just simpler to compensate the feedback loop (whatever the bandwidth) when you don’t have the opto in the feedback loop, and cascoding makes it look like there's no optocoupler in the feedback loop.

- - - Updated - - -

The HCPL4503M optocoupler looks good..

HCPL4503
https://www.farnell.com/datasheets/1815146.pdf

..its low CTR would be worth it if really reduces the effect of the opto pole like it says in the datasheet.

I looked at the test circuits for it, and wonder if it can be connected to the feedback pin (FREQ) of an ICE2HSO1G LLC controller IC, like a regular optocoupler (eg CNY17 type)

…as on page 4 of the ICS2HSO1G datasheet…

https://www.infineon.com/dgdl/Infin...n.pdf?fileId=db3a30432a40a650012a458289712b4c

I am just wondering, if HCPL4503M really is that good, then why is it not commonly used in SMPS?

 

[treez]

Graph on top right of page 6 of this (CNY17) optocoupler datasheet shows how also the pole frequency gets moved higher by having more opto diode current…

https://www.vishay.com/docs/83606/cny17.pdf


..the problem is that control chips like ICE2SH01G have a maximum current of 330uA being pulled out of the regulation pin, so its simply not possible to have 10mA running in one’s opto-diode, as the CTR is too high at 10mA…shame..

Low CTR optocouplers needed….HCPL4503 is low CTR but looks like it is only for logic output, with the external resistor being needed for connection in pull up to 5V

 

[D.A.(Tony)Stewart]

ON Semi schema uses PS2703−1−M−A which has CTR of 100 +/-50 @10mA and CTR=10+ at 1mA and with 10us rise fall time into 1K, you should be able to modulate the output up to 50kHz. They boost it with a common base low input impedance to 2.5K out.

 

[Easy peasy]

HCPL4503 is low CTR but looks like it is only for logic output

in fact it is widely used in telecoms switchers which have high dynamic performance requirement, the CM shield reduces residual 100Hz coupling across the opto too, keeping 100Hz in the o//p to a minimum.

- - - Updated - - -

the small signal performance of the PS2703 might just scrape to 50kHz (very low signal) bit for larger dynamic signals, say 100mV rms on the o/p 50k seems very unlikely...

 

[treez]

The HCPL4503 optocoupler datasheet speaks of its "100's of times faster than normal optocouplers' performance, and massive reduction of base-collector capacitance compared to normal optocouplers, so why, I wonder, have onsemi.com, in the App Note of post#1, not mentioned using an optocoupler like HCPL4503?....I mean, its more convenient to just use a HCPL4503 than to put in the cascode stage.

 

[Easy peasy]

Did you perhaps, pause to consider, that if On Semi had made the 4503, the whole article would have been on that opto-coupler...? As it doesn't it chose to write an article that requires you to use more xtors and op-amps, coincidentally, a lot of which are made by On-semi....

- - - Updated - - -

p.s. you forgot to mention the CM screen, which is perhaps the most impressive bit about the opto....

 

[treez]

thanks...this opto (HCPL45030) has a datasheet "situation" about it..
https://www.edaboard.com/threads/344463/

p.s. you forgot to mention the CM screen, which is perhaps the most impressive bit about the opto....

..will this really matter for general analog feedback in SMPS?

 

[FvM]

"1 Mbit/s" optocouplers like Motorola originated 6N135 are around since at least 40 years and used also for high speed analog isolation since that time. I suppose an educated engineer should know about it.

The other point is that many SMPS designs are apparently fine with standard speed optocouplers.