Practical Accuracy with Tiny III
Posted by: split63
on
The basic design is 115V AC input, fully rectified, switched by Tiny III, then passed through a 9:118 transformer to ultimately create a 1.5A 5V output. The final output voltage is compared to a reference voltage and then a comparator turns the feedback opto on or off.
On the secondary of the xfrm, the output voltage passes through a Schottky rectifier, then filtered by 1000uf, then through a 10uf inductor to form the output. The output is further filtered with 100uf cap to ground.
My question is, what sort of accuracy is practical with this setup. The typical Zener diode approach is poor at accuracy as the Zener and the Opto performance vary widely.
댓글
With a zener diode used as a reference the accuracy you can expect is about 8%. With a 1% TL-431 shunt regulator as an error amplifier you can get about 3% regulation
Also, I think your 10uH inductor value is too large. With 10uH inductor and 100uF cap the LC corner frequency is about 5 kHz. Thats too aggresive. I reccomend that you reduce the inductor to 2.2 uH to shift the LC corner frequency to around 10 kHz.
I'm curious why you think that the LC is too aggressive? From what perspective?
Anyway, where does your 8% and 3% accuracy estimates come from? In my version, I have replaced the Zener and TL431, with a comparator and 1% reference. This takes the slop out introduced by the LED and on the current transfer ratio of the opto, both of which have large variation. With the comparator, the LED of the opto is driven hard on, thus negating the softness of the opto.
If the version with the reference provides 3%, then this approach should be much better. But its unclear to me how I would determine the actual accuracy considering lot to lot and device tolerances.
The LC post filter is very much a part of the feedback path. It can therefore introduce phase lags at higher frequencies. Beyond the corner frequency the phase can dip sharply (by -180 degrees) causing instability in the power supply.
This is true for any power supply. Just try to increase the LC post filter value and you will introduce instability.
From a practical standpoint the corner frequency should be above 10 kHz for TinySwitch based designs to avoid this kind of an issue.
As far as the tolerance of output is concerned - For Tinyswitch based applications the output tolerance is really dictated by 3 things
1) The tolerance of the reference (TL431, zener, comparator)
2) The ESR of the output capacitor
3) The quality of the optocoupler (CTR and delay time associated with it)
The TinySwitch Enable current threshold variation does not play a crucial role other than setting the Q point for the feedback loop.
Assuming that you are using a low ESR output capacitor, the biggest factor then is the tolerance of the reference. Zeners are generally 3% to 5% tolerance which is why over temperature you can expect about 8% - 10% overall regulation. Additionally zener feedback provides no additional low frequency gain. TL431 or other error amps can have tighter reference (1% to 2%) and also provide for substantial low frequency gain resulting in much better regulation as compared to zeners.
