Noise elimination in Flyback SMPS

[techneo]

Hi,

 I have designed a switch-mode variable PSU which is a flyback topology based on UC1843 PWM chip. The PSU is working good with an output voltage adj from 15 to +24VDC. Schematics attached .





The problem is that i get a spike of +10V on my output every 400 to 500 msec. I have tried to use a pi filter but to no avail.

Kindly help.


-techNeo

[vern]

I have developed a SMPS that uses the LMV431 (but with a different controller, LM5020). I have no experience with the UC1843.
The circuitry around the LMV431 is a bit different, adds a slope to the compensation loop and a filter to the optocoupler supply.
See attached pdf.

[fpgaguy]

When you say 10V spike, do you mean it's Vout + 10V, or do you mean Vout - 10V ?

Without more data I would guess the pwm controller is sensing >1V on isense and skipping a pulse, or on comp

Does it change  this behavior if you vary the load by 50% or so ?

[witsanukai]

How long of spike period?
I am not so sure about your maximum current capability.
Try to take constand load for stability by add 240 Ohm 5W at out put and see the response and spike gone?
you may need to apply some small load for feed back stability.

[techneo]

When you say 10V spike, do you mean it's Vout + 10V, or do you mean Vout - 10V ?

Without more data I would guess the pwm controller is sensing >1V on isense and skipping a pulse, or on comp

Does it change  this behavior if you vary the load by 50% or so ?

It is Vout+10 V. The behavior stays the same if load is varied.


-techNeo

Posted on: January 22, 2019, 12:22:41 00:22 - Automerged

How long of spike period?
I am not so sure about your maximum current capability.
Try to take constand load for stability by add 240 Ohm 5W at out put and see the response and spike gone?
you may need to apply some small load for feed back stability.

The spike is 4-5 msec.
Full load capability of the PSU is 10A.

-techNeo

[PM3295]

You will need to add proper slope compensation to your current sense if your duty cycle is near or greater than 50% to prevent sub-harmonic oscillations. This can certainly produce the spikes you see.

[techneo]

You will need to add proper slope compensation to your current sense if your duty cycle is near or greater than 50% to prevent sub-harmonic oscillations. This can certainly produce the spikes you see.

So this slope compensation will be in the form of components as suggested in Reply # 1 above  or do i need to make a massive overhaul of the circuit ?
In any case i think i need to dig deep about you suggested .

[PM3295]

In the simplest  case, it is just adding a portion of your saw-tooth into the current sensing node via a resistor. To prevent loading on the saw-tooth it is recommended to use a buffer transistor as shown in my example. It is important to read up about it and calculating the correct value of the slope comp resistor according to your circuit parameters.

The scope plot shows the gate drive and the current feedback of a bad converter design with no slope comp that we took a look at for a client. The sub-harmonic oscillation is clearly visible producing wide sub-harmonic switching pulses. The blue trace shows the voltage spikes it produced on the +35 V bus on this boost converter.

Notice that this caused the core to start saturating (upward curve in current).

[vern]

didn't think of it, but I think PM3295 is right.
In the UC1843 datasheet it says for duty cycles > 50% you need slope compensation and there are examples how to do it. Looks similar to what PM3295 suggests.
That fits with the slow oscillation of 400 - 500ms, because it takes a while for the transformer core to get saturated.
I didn't think about it because the LM5020 that I use in my SMPS hat an integrated slope compensation!

[techneo]

didn't think of it, but I think PM3295 is right.
In the UC1843 datasheet it says for duty cycles > 50% you need slope compensation and there are examples how to do it. Looks similar to what PM3295 suggests.
That fits with the slow oscillation of 400 - 500ms, because it takes a while for the transformer core to get saturated.
I didn't think about it because the LM5020 that I use in my SMPS hat an integrated slope compensation!

I made the changes in my pcb as per your suggestions. The spike has reduced from +10V to +7V but has not been eliminated completely.

I will now look and try to implement the slope compensation examples...

-techNeo

[Sideshow Bob]

In switch-mode design it is quite important to selct the correct values but also the component specification. Like for caps low ESR and dielectric material. But also using the scope correct see public link below
http://www.interpoint.com/product_documents/DC_DC_Converters_Output_Noise.pdf
From the top my head I also suggest som Google search. The may give somewhat identical results. But I would recommend to sift throgh the first page at each and absorb relevant info
https://www.google.com/search?q=filter+switch+mode+noise
https://www.google.com/search?q=inductors+switch+mode+noise+reduction
https://www.google.com/search?q=switch+mode+noise+reduction

[PM3295]

I will now look and try to implement the slope compensation examples...

-techNeo

Just keep in mind that the slope compensation injection will lower your intended peak current limit, so your sensing resistor should be reduced in value accordingly to compensate.

If you inject too much slope compensation, it may reduce stability margins, so the amount of needed compensation should be carefully calculated and evaluated.

[kreutz]

PCB layout and position of poles and zeroes on the control loop influence circuit's stability.  We don't have enough data to further help without random guessing. Please, post your complete  pcb layout as well as photos of the prototype.
R4C4 filter in the I sense signal is very important and its time constant must be << switching period. What is your switching frequency?  Also proper gap for the flyback transformer core avoid possible core saturation (There no info about your flyback transformer design). Also output Rectifier diode reverse recovery time is important because it influences switching noise on the I sense signal, so your choice of components is important (ie, 1N4007 is not remotely fast enough to be there). Review also your choice of 1N4148 on the snubber circuit. ESR of the output capacitors is very important for output voltage ripple, your choice of capacitor model is important. Physical layout of feedback signal wires is very important, they must be far from power signals and chip's vcc and Vref filter capacitors must be really close to respective IC pins and low ESR.

Could you, please, also post your circuit design calculations? Oscilloscope screen photos of the relevant signals will also help...

Also, for troubleshooting purposes, disconnect the output voltage feedback wire from the ic pin and use a constant bias (look for a way to fool the voltage feedback circuit), check if the overvoltge spike is still there, so we could get an idea of where the instability is coming from, I would recalculate the R4C4 filter, disconnect the 1N4007,  and check the core gap before doing extra troubleshooting. I didn't find the fr2007 diode datasheet on the web... could you post a link?

[dennis78]

I think diode is fr207 or fr1007 instead fr2007

[kreutz]

FR207 (fast recovery diode, 250 nS average recovery time) could be used instead of the 1N4148 in the snubber (clamp) circuit as well as a replacement for the 1N4007 in the secondary of the transformer if the switching frequency is < = 50 khz, if the switching frequency is much higher consider using ultra fast recovery diodes or super fast recovery diodes depending on the frequency. See "Selection of Ultra-Fast Recovery Diodes Used in Flyback Circuits": https://www.maximintegrated.com/en/app-notes/index.mvp/id/849. Review your selection of main secondary rectifier diodes if the switching frequency >> 30Khz, you will most probably need faster ones.

Another detail: can you, please, explain why you connected the zener diode D7 in series with the resistor R18 and connected the chip's VCC to the other side of that resistor (instead of connecting it to the junction R18-D7)? AFAIK there is no purpose for that zener. As it is, there is no regulation to Vcc whatsoever and that could influence circuit operation..

[PM3295]

We don't have enough data to further help without random guessing. Please, post your complete  pcb layout as well as photos of the prototype.
What is your switching frequency?  Also proper gap for the flyback transformer core avoid possible core saturation (There no info about your flyback transformer design).
Could you, please, also post your circuit design calculations? Oscilloscope screen photos of the relevant signals will also help...

We can only really help when we have all the details. I agree!

[techneo]

PCB layout and position of poles and zeroes on the control loop influence circuit's stability.  We don't have enough data to further help without random guessing. Please, post your complete  pcb layout as well as photos of the prototype.
R4C4 filter in the I sense signal is very important and its time constant must be << switching period. What is your switching frequency?  Also proper gap for the flyback transformer core avoid possible core saturation (There no info about your flyback transformer design). Also output Rectifier diode reverse recovery time is important because it influences switching noise on the I sense signal, so your choice of components is important (ie, 1N4007 is not remotely fast enough to be there). Review also your choice of 1N4148 on the snubber circuit. ESR of the output capacitors is very important for output voltage ripple, your choice of capacitor model is important. Physical layout of feedback signal wires is very important, they must be far from power signals and chip's vcc and Vref filter capacitors must be really close to respective IC pins and low ESR.

Could you, please, also post your circuit design calculations? Oscilloscope screen photos of the relevant signals will also help...

Also, for troubleshooting purposes, disconnect the output voltage feedback wire from the ic pin and use a constant bias (look for a way to fool the voltage feedback circuit), check if the overvoltge spike is still there, so we could get an idea of where the instability is coming from, I would recalculate the R4C4 filter, disconnect the 1N4007,  and check the core gap before doing extra troubleshooting. I didn't find the fr2007 diode datasheet on the web... could you post a link?

I have attached my pcb file.

One good news is that with slope compensation i have been able to reduce the spike to +3V.

-techNeo

[M@X77]

viewing your pcb my first think is the big copper areas under transformer. May be they can catch magnetic field and gives these spikes, tray to do a big ground area on top layer under the trnasformer

[vern]

techneo, you should check kreutz post 119:  he is of course right that VCC of U1 and + of C13 are connected to the wrong side of R18!
That could destroy your chip and it can also cause these spikes since U1 VCC is not stabilized at all in your circuit!
And using a 1N4007 in a SMPS is also a big NONO. It has a reverse recovery time of 30us, which is way to slow for this SMPS.
In your circuit it is only to supply the control LED, you can use the 1N4148 instead, it's a lot faster and delivers enough current for your purpose.

I overlooked those two things when I first looked at your schematic.

[techneo]

FR207 (fast recovery diode, 250 nS average recovery time) could be used instead of the 1N4148 in the snubber (clamp) circuit as well as a replacement for the 1N4007 in the secondary of the transformer if the switching frequency is < = 50 khz, if the switching frequency is much higher consider using ultra fast recovery diodes or super fast recovery diodes depending on the frequency. See "Selection of Ultra-Fast Recovery Diodes Used in Flyback Circuits": https://www.maximintegrated.com/en/app-notes/index.mvp/id/849. Review your selection of main secondary rectifier diodes if the switching frequency >> 30Khz, you will most probably need faster ones.

Another detail: can you, please, explain why you connected the zener diode D7 in series with the resistor R18 and connected the chip's VCC to the other side of that resistor (instead of connecting it to the junction R18-D7)? AFAIK there is no purpose for that zener. As it is, there is no regulation to Vcc whatsoever and that could influence circuit operation..

Ok i will order the FR207 to use in place of 1N4007.

I have, however, a question ... Since I am using 1N4007 to feed the network so as to trigger an LED as soon as output voltage is available. Will it also affect the output voltage because of its slow response time?


-techNeo


Posted on: February 12, 2019, 12:52:16 00:52 - Automerged

techneo, you should check kreutz post 119:  he is of course right that VCC of U1 and + of C13 are connected to the wrong side of R18!
That could destroy your chip and it can also cause these spikes since U1 VCC is not stabilized at all in your circuit!
And using a 1N4007 in a SMPS is also a big NONO. It has a reverse recovery time of 30us, which is way to slow for this SMPS.
In your circuit it is only to supply the control LED, you can use the 1N4148 instead, it's a lot faster and delivers enough current for your purpose.

I overlooked those two things when I first looked at your schematic.

I am using R18 to limit the current to the zener diode because it had been my observation that in case of any overvoltage spike , the zener gets damaged. So to limit the inrush current for an over voltage condition, i used R18.


-techNeo

[witsanukai]

series inductor at output for 1.5uH can help?
just try to filter out spike.

[kreutz]

Posted on: February 12, 2019, 12:52:16 00:52 - Automerged

I am using R18 to limit the current to the zener diode because it had been my observation that in case of any overvoltage spike , the zener gets damaged. So to limit the inrush current for an over voltage condition, i used R18.


-techNeo

You should connect the chip's VCC pin to the junction between the zener diode and R18, that way The Chip's VCC will be regulated by the zener diode. You might have to recalculate R18 to take into account the Vcc pin current at rated voltage and  zener diode maximum current (derated). For calculation see example #1 at https://www.electronics-tutorials.ws/diode/diode_7.html

[vern]

I have, however, a question ... Since I am using 1N4007 to feed the network so as to trigger an LED as soon as output voltage is available. Will it also affect the output voltage because of its slow response time?

no, it will not affect your output voltage, but it might get hot and reduce the overall efficiency of your power supply.
The 1N4007 has a reverse recovery time of about 30us, this is the time it needs to change from the conducting (forward) state to the blocking state if voltage is reversed.
During that time current travels in the wrong direction through the diode.
That means that over frequencies of 33kHz the diode is not rectifying at all, it acts more like a resistor.
A good rectifier diode for SMPS is more in the nanosecond range. The faster the better. Since you don't need much current for your LED you can use the 1N4148, which has a reverse recovery time of about 8ns and a peakk forward current of 500mA, which is more than enough.