Problem with TL431 in a SMPS

Status
Not open for further replies.
I would avoid the planar types, unless you are forced to .... I have used Payton stuff... I had to use them for Military applications in fighter jets...270VDC down to 28VDC...at 1200Watts...
For transformers like in your application..I prefer EP cores from Ferroxcube...I had great succes with them....
I have designed cell phone chargers that use just triple coat on the secondary with no tape...It passed all worldwide regulatory requirements and has sold in the millions, no problems...
Leakage current was about 3uA....

Chris
 
TL431 is a op amp with internal reference voltage, high gain and low frequency pole, while zener is just a "constant voltage diode". Replace a zener with 431 will make a absolutely different loop responce, so you have to re-design the loop. It's difficult for flyback -- more difficult because you use a non-traditional controller.

So it's better use the zener.

Ultima Thule said:
My design is similar to "Engineering Prototype Report 14" except for the output voltage being 4,1 V/1,2 A instead of 9 V/0,3 mA and it also uses a TL431 shuntregulator instead of the zener used in their schematics to regulate output voltage seen in EPR 14 document on page 5.
Michael
 
Hi guys,

after having a long breake with my SMPS project I have continued with it, as it has to be done sooner or later! 🙁

Not much have hapened sice last time more than I replaced temporarly the TL431 with a chain of 5 x 1N4148 in the feedback circuitry which gives approx 4 Volts out at small loads.
The problem still exist that the output voltage is falling too much by load, so... I decided to put a 1k trimmer in parallel with the 500 Ohm already in parallel with the photo diode.

So I trimmed until I got my 4,1-4,1 Volt out @ 2 Ampere load.

Now after switching off the SMPS I meassured the total resistance in parallel with the photo diode and it showed 62 Ohm, further at power on the voltage fall over the photo diode was 0,92 Volt meaning I have a current of approx 15 mA, is this a normal value.. sounds very much to me. 🙄

Anyhow, the Tiny swith is sensitive with where the feedback is taken, if I connect it to the output side, meaning after the CLC filter it starts to weaving, but as soon feedback node is connected to the rectifier diode it stabilize and works very well.
So I will start to check and modify the feedback and see how I can use TL431, more later to come.

Michael 🙂
 
Hi Eva! 🙂

Yes, thats the approach I'm trying out right now for couple hours, but however I connect diffrent capacitors in diffrent ways, el-lytes, plastic, ceramic in diffrent values, it does just not help... 🙁

I have the photodiode connected before the inductor and tried to make the TL431 (I changed back fromthe diode chain to TL431 since my last post) "slow" and just maintain DC level, but I just don't get the output voltage stable by load.

I moved also the 817 optocoupler as close as posible to the feedback legg on the Tiny circuit as it is sensitive to picking up stary signals, now it's 2-3 mm, didn't help.

When I put a trimmer in parallel with the resistor connected between Anode and Ref of the TL431 I can readjust the output so I can keep 4,2 Volt upto 2,5 Ampere so there is capability, but this is max as it goes into hick-up mode when increasing further.

I still believe, as I wrote already in post #2, that the Tiny's special switching skipping sheme which increase switch pulses by load also increase the HF energy that is getting rectified in some way which adds up to the feedback signal in some way causing the decreasing output voltage by load.

BTW, as a note, when putting the TL431 back I did remove the trimmer in parallel with the photo diode.

Michael 🙄
 
Hahaha.. yes it MUST work, what is your adress so I can send it to you... btw, I like challenges too so maybe I don't want to leave away my baby anyway! 😀 😀

Anyhow, attached is a principle schematic of the feedback circuitry, it's not fully accurate as for instance C4 and C8 have not bean used at the same time (though tested also together), but it shows a bit my ideas.
The idea is to let HF information go through C5 and C6 while TL431 takes care of the DC level.

Is this similar (maybe not in the same order as my circuitry but to the idea) to how you have designed your feedback?

Michael

EDIT: Forgot to connect feedback earth and voltage earth together in my quick schematic I put up!
 

Attachments

  • tl431_feedback.gif
    tl431_feedback.gif
    7.5 KB · Views: 331
C6, C5 and C8 are not required, connecting R2 after the inductor and the optocoupler before the inductor is just enough. Note that C4 makes the TL431 to work as a short-circuit at HF, thus ignoring the feedback from R2 and acting as if the optocoupler was connected to ground through R1 in series with a voltage source.

Simple, isn't it? This technique obviously introduces some kind of crossover frequency because feedback is being taken from two different points and mixed, and L1 should make effect well above that frequency. Also note that R1 is actually setting the high frequency open-loop gain of the circuit, and C4 together with R2 and R3 are setting a open-loop zero above which that gain is effective.

The TL431 actually works as a "DC servo" 🙂 (speaking in audio terms).
 
Hi Eva,

thanks for your reply, yeah.. thats pretty much the basic setup I have worked with as you describe, so when it didn't solve the problem keeping DC level by load I added all these extra components.
In other aspects it has been stable except keeping the voltage level... 🙁

You say that C4 cap between A and Ref makes The TL431 work as an shortcircuit for HF, but here's the problem, it doesn't, it gives up and output impedance falls above 100 kHz while di/dt and dv/dt in SMPS circuits can have energy components easily upto the 100 MHz area.
The C4 cap doesn't help as CMRR is not enough, and further on can also be asymetrical depending if it's upgoing or downgoing signal, so that's why I did go even further to really let the HF energy bypass the TL431 by means of C5 and C6, at least this was the idea with these caps.

Well, tomorrow I will look into it further.
But once more the Tiny may skip pulses at ligth load as much as 9/10 pulses meaning the HF garbage decrease a lot and the so called rectifying phenomena decrease too, I think that with normal switching circuits that have a continous ongoing switching without advanced switch-skipping scheme like Tiny II doesn't show up this as clearly as the HF garbage is "constant".

Michael
 
Hi Eva,

look at the picture and think what... 😀

I have strugling with the powers*it so much that I was ready to trash that Tiny with a sledge until I "discovered" something! 😱

Well, those probes as seen in the picture was in the drama and I had them connected at the other end of my 65 cm pair of tiny 0,2 mm2 wires until I measured directly at the output of the PCB I discovered that the voltage is absolutely stable by load, ok.. so i thought that it's the resistance in my wires, checkd the diamater and calculated the area to be, as mentioned, 2,0 mm2.
As my multimeters can't measure mOhm accurately I calculated the theretical resistance to be 0,1 Ohm all together, but thats only 0,2 V fall at 2 A, what the heck...😕

So I measured at the other end of my wires, not much of voltage fall, as well theoretically checked, and thought what is this...

I started to connect/disconnect all banana plugs around the probes but with no change, the voltage fall was INSIDE these probes! :bigeyes:
And actually at 2A load I could feel a very small warmth, not that someone usualy registrate. On these probes are markings ">6A", I had almost 1,5V fall over these at 2,5 A, i wonder how many watts they will develope of heat when 6A goes through them!

Well, the case is closed, my circuit works very well and regulate within around 100 mV from 0,1-2,5 A!
What a frustrating story, and a great release! 😀😀

Cheers Michael
 

Attachments

  • test_probes_1.gif
    test_probes_1.gif
    33.2 KB · Views: 225
Status
Not open for further replies.