[SOLVED] Isolated Flyback SMPS- High ripple on the output voltage?

Hi,

your attachments don´t work.
Mind that there is a limited time: from uploading the attachments to sending the post.

abdo.medo1000 said:

voltmeter shows exactly 15.1 V (RMS).

Click to expand...

"RMS" should be used for AC voltages.
For DC voltages use the non_RMS mode --> average.

Many multimeters when switched to "RMS or AC" filter out the DC and show the AC only.
In your case it would show the ripple voltage in RMS only.
In doubt you should read the manual of your voltmeter.

****
Many SMPS (also many liner regulators) don´t work properly on no load. They need a minimum load.
--> If you want to check performance of your power supply, always connect a suitable load.

****
Besides the schematic, the PCB layout has big influence on performance. Especially, stability, ripple, noise.
Thus I recommend to post pictures of the PCB layout.

Klaus

KlausST said:

your attachments don´t work.

Click to expand...

scope's shots

KlausST said:

"RMS" should be used for AC voltages.
For DC voltages use the non_RMS mode --> average.

Many multimeters when switched to "RMS or AC" filter out the DC and show the AC only.
In your case it would show the ripple voltage in RMS only.
In doubt you should read the manual of your voltmeter.

Click to expand...

You are correct, I use DC voltmeter to measure the voltage, and it shows me steady voltage value (15v) without any fluctuating.
But when I plug the oscilloscope prop to the output of the circuit as it shown above, the minimum voltage is 13.60V and the maximum voltage is 16.80V. And the RMS voltage value is 14.78(many times it shows the same voltage shown on the multimeter).

KlausST said:

Many SMPS (also many liner regulators) don´t work properly on no load. They need a minimum load.
--> If you want to check performance of your power supply, always connect a suitable load.

Click to expand...

I added resistance load (10 ohm) and the circuit did not change and the same fluctuating is happening when adding load or without load.
Also, there is continually load on the output (2 of 510 ohm series resistors are connected on the output of the circuit)

as shown on the schematic below


Thanks in advance

Hi,

the circuit does what you told it to do: It uses ON/OFF method instead of PWM regulation.

May I ask where your schematic is from?

--> My recommendation: use the feedback as recommended in the datasheet.

Klaus

KlausST said:

May I ask where your schematic is from?

Click to expand...

I got it from an engineer works on the power electronic field but nowadays, he is not available for that I could not ask him about this problem.

So you say that the problem on the feedback part, right ?

This is the datasheet I found

but there is no any part or example about the Feedback !

how should I do the feedback ?

Hi,

I'm unfamiliar with UC3845. That is why I ask: Is it right that Vref is connected to the opto feedback - which is a varying signal (in schematic in post #1)? Is that a 'typical application' from the datasheet? It would explain fluctuating, non 5.0V voltage.

Hi

abdo.medo1000 said:

So you say that the problem on the feedback part, right ?

Click to expand...

Your feedback is not used. FB pin is connected to GND.

abdo.medo1000 said:

but there is no any part or example about the Feedback !

how should I do the feedback ?

Click to expand...

I think in every of the design examples of the datasheet the FB pin is used properly.
Read and follow the datasheet.
Also goto the IC manufacturer´s web site and look for application notes / design notes for additional assistance.
They provide the documents exactly for you. They are for free, so use it.

I recommend this, because I´d do the same.

Klaus

KlausST said:

Your feedback is not used. FB pin is connected to GND.

Click to expand...

I checked this application note from Onsemi

"an1327-d application note from onsemi "

on figure.8 they connected the FB to ground directly, and also I couldn't see the Vref pin of the IC on this circuit schematic.

d123 said:

Hi,

I'm unfamiliar with UC3845. That is why I ask: Is it right that Vref is connected to the opto feedback - which is a varying signal (in schematic in post #1)? Is that a 'typical application' from the datasheet? It would explain fluctuating, non 5.0V voltage.

Click to expand...

HI there,

No, this circuit not from the datasheet.
Because the datasheet did not provide any example circuits

Hi,

Okay, without reading the datasheet: a Vref pin is usually tied to a stiff/reliably steady supply voltage, not a feedback node. OnSemi may/should have technical support documents like app notes or maybe an evaluation module with its corresponding user manual that help with correct wiring, if that's part of the issue, on the IC's page of their website.

d123 said:

Hi,

Okay, without reading the datasheet: a Vref pin is usually tied to a stiff/reliably steady supply voltage, not a feedback node. OnSemi may/should have technical support documents like app notes or maybe an evaluation module with its corresponding user manual that help with correct wiring, if that's part of the issue, on the IC's page of their website.

Click to expand...

Yes, I totally agree with you

I checked this application note from Onsemi

"an1327-d application note from onsemi "

On figure.8 I couldn't see the Vref pin of the IC on this circuit schematic.

But the used IC has Vref pin !

Hi,

The voltage reference is an output, page 1 of attached datasheet. A basic resource - in you - is when a datasheet lacks information, look at a few others for the same part by other manufacturers. Sometimes e.g. NXP datasheet fills in information gaps that those scanned versions of old CMOS parts don'always or clearly provide, or ST, and vice versa.

OnSemi datasheet for UC3845: page 1, 6, 8, 9, 10, and subsequent pages have typical applications, etc.

Hi,

I recommend to find the source of your schematic There should be an explanation.

(indeed I wonder why people rather rely on "no name " schematics than on the original manufacturer recommendations. )

abdo.medo1000 said:

I added resistance load (10 ohm) and the circuit did not change and the same fluctuating is happening when adding load or without load.

Click to expand...

This makes no sense to me. I expect the behaviour to change a lot.

***
For my taste C13 value is too low. But a bigger size makes overall behaviour more slow.

--> I´d try to reduce C20 to 10n to make the feedback more fast. Let´s see what happens.

If you want me to continue, then please post the PCB layout.

Klaus

Did you try to short L1 and see if this are better?
Having an additional LC filter after the flyback can make compensation very hard.

See if shorting L1 fixes the issue.

Your circuit can run with 15V@10R mean the power delivery well.
Problem is your feedback loop. It is not operate in linear condition, it works in saturation mode. So that, if you see duty will rise to much then down to min. Duty will not stable when fixed load -> control loop is not stable.
When control loop stable, output ripple voltage frequency = switching frequency. I saw your ripple with very low frequency.
With using directly Vcomp short to ground, you change PWM mode from voltage mode to current mode. It means, when output rise, you imediately reduce duty (or peak switching current). Instead of using FB pin, this has delayed through its PID controller. This will help circuit adapt better to load respond. You can still use FB pin for the second protection loop, you devide voltage from auxilry coil. If opto spoiled, the output voltage will not rise higher than which you set. Or pull up with NTC to Vref, if temperature rise too high in some condition, It will make the output voltage reduce and lower output power.
Now you have to take care: maximum output current of Vcomp pin, opto have to sink higher little bit this current.
Opto has its CRT - Current ratio. It mean if Ic = If x CRT. But with opto CRT value is not fix, its is a range, like 200-400%. So, when design, have to select good range to ensur opto operate in linear mode.
Your PID control loop is : R15, C15, R22, R16, C20 and opto U6. I will not use C20 = 1uF, often <10nF as it use to reduce EMI noise to control pin Vcomp. It make delayed to control loop and make it hard to stable.
15V output, opto drop A-K ~1V, TL431 well work when V(A-K) > 1V, current always consumpt 1mA. Select If in linear range of opto If = Ic (max current source from Vcomp pin) / CRT_min. With 1mA bias for Tl431, use 1k resitance parallel with AK pin of opto 1k = 1V/1mA. Example If = 5mA, R in seria with opto from 15V: Rf = (15V - 1V_opto - 1V_Tl431)/ 5mA = ~2.6k Ohm. You are using 1k Ohm, that make opto conducted saturation, it work as on and off.
Normally, I often use reistor in serial wit C15 to easy change the gain of PID.
You can make circuit properly working by adding resistor devider from VDC_in which after retifier. This make current sense pin will be offset higher if input voltage higher, reduce I_peak through transformer. This make sure transformer will not enter saturation mode, or circuit operate with higher defined power rating. Higher power make product hotter, saturation point of transformer will go down, product maybe boom... or reduce life time.
Snuber RCD also effect to control loop when light load. Make sure leakage inductance of transformer is smallest as posible.
You don't have inrushed current limitter like resistor or better NTC. This make spike current when plug device to grid power, sometime will spoil diode rectifier, process day by day.
You don't have separate sink & source parth for controlling mosfet, to optimize EMI. High EMI make noisy to control loop.
Sensing output after L1 mean delayed control loop, because LC circuit act as low pass filter. But small L1 is acceptable if fsw > f(LC) >> fc of control loop (often select = 1/5fsw).
Control loop stable will make ripple become lowest, spike on mosfet will be lower, efficiency higher, product cooler, lower ripple current for both input & output capacitor -> longer life time, lower I_peak -> smaller transformer, lower EMI, ... Many importance things arround control loop.

Magnethicc said:

Did you try to short L1 and see if this are better?
Having an additional LC filter after the flyback can make compensation very hard.

See if shorting L1 fixes the issue.

Click to expand...

Also try to get ride of R3 (the resistor connecting COMP and VREF he is unnecessary.

Hi,

The design is using the opto in the common-emitter configuration, which is fine.

Taihung has made a lot of valid points. However, let me add one important observation.

The UC3845 is for DC input flyback due to its UVLO hysteresis. The UC3844 is for AC input. I assume that is the reason you are using as much as 660 uF of bulk capacitance at the input. I'd suggest you replace the chip.

Let us know the outcome.

Thank you for your kindly answers, I redesigned my circuit using the typical application design on the datasheets, and now it works fine.


Here is the datasheet that I used, and also it has all components calculations.
(UC3845A datasheet from TI)

thanks all again