Spikes on DC supply

[premkumar9]

Hi,
I have developed a unit based on 89S52 uC. It is used for pumping control and installed in a noisy (electrical noise) environment. In the last two months 4 OR 5 times it hanged with all pins high. It seems that it require some methods to removes spikes in Vcc supply. Since the unit is already installed in the site, there are some limitations on the possible modifications.
Can anybody please suggest some practical methods? Is there any device to arrest spikes(special diodes etc.) which I can add to my circuit as a temporary solution?

 

[Electronworks]

Most ways to improve the spikes on the supply will involve reducing the input voltage to the microcontroller. A linear regulator with a high power supply rejection ratio (PSRR) will do the trick with plenty of capacitive decoupling before and after the regulator - read the datasheet carefully to see if there is a maximum value of capacitance you can put on the output - it may go unstable.

I would suggest a 100uF tantalum in parallel with a 100nF ceramic on both the input and the output.

If the noise still gets through, a series inductor on the input will produce a second order filter when used with the capacitor.

If you cannot live with the voltage drop a linear regulator gives you, then you need to use series inductors and parallel capacitors to try to reduce the noise.

You only need to add this in series with the microcontrolller, not the rest of the circuit, so this will help to keep the power requirements of your linear reg low.

Finally, you might want to try putting the microcontroller circuit in an earthed metal box to stop any radiated noise getting in to your circuit.

Let me know if any of the above is feasible

 

[pasanlaksiri]

Thanks for the nice tips.

 

[Hero999]

Have you try adding ferrite beads to the power supply?

 

[premkumar9]

Supervisory Circuits

Have you try adding ferrite beads to the power supply?

I am thinking of adding supervisory Circuit. Could locate max ICs like MA814 etc. Can you please suggest similar ICs which may be more common?

 

[Nigel Goodwin]

For a start post the circuit you are using, we don't know what precautions you have already used.

 

[premkumar9]

For a start post the circuit you are using, we don't know what precautions you have already used.

I am using a simple RC network (1µF and 10K) to reset the µC on switch ON. But seems that in a noisy environment spikes will affect the uC through this RESET circuit. The solution is found to be using a supervisory circuit. So I want to try one. MAX 814 is one such IC. But would like to know some other ICs commonly used for this purpose so that I can check the availability locally.

 

[Nigel Goodwin]

No, you need to show the entire circuit, or if not that, at least all the power supply section and decoupling capacitors.

 

[premkumar9]

No, you need to show the entire circuit, or if not that, at least all the power supply section and decoupling capacitors.

Improving power supply is one side which will be looked in to. Right now I am looking for other ICs for supervisory purpose. Anyway I think that the supervisory IC is needed for my application.

 

[Nigel Goodwin]

Improving power supply is one side which will be looked in to. Right now I am looking for other ICs for supervisory purpose. Anyway I think that the supervisory IC is needed for my application.

Possibly, but it's probably been caused by poor design and poor layout, and you're only assuming the problem is entering via the reset pin, which may not be the case.

 

[Oznog]

I would suggest a 100uF tantalum in parallel with a 100nF ceramic on both the input and the output.

These days, a high-capacitance, ultra-low ESR MLCC cap is QUITE cheap for low voltage.
You can get like 10uF-22uF pretty cheap and tiny. There's generally no need for 100uF in combatting electrical noise. Noise will not be sustained that long.

In some cases you can cause latchup by inducing >Vdd on any input pin, so don't neglect this possibility. And you can do it while it's an OUTPUT, too. If the pin's low and you drive current in from an external source, the pin drive tends to ground it out. If the pin's high and you drive current into it, it jumps to the Vdd rail and that can cause latchup.

Leading to the next point not everyone understands: I/O pin protection against overvoltage consists of shunting current to the Vdd rail, and the Vdd input pin, which will actually raise Vdd on the board trace. A regulator mainly has the drive to INCREASE voltage on the output. It often has very little ability to shunt the excess voltage and LOWER the voltage on the output rail. Anything on the board which shunts excess voltage onto the Vdd rail has the potential to cause latchup in anything which uses the Vdd rail.

 

[Electronworks]

Personally I would want to get rid of the spikes before adding a reset circuit. A reset circuit will take the part out of a hung state, but will reset every time a spike appears on the rail. Your processor will not hang every 3 days, but might reset every 10ms. Which is worse? A MAX814 is good, or failing that, the MAX809 is more popular

 

[premkumar9]

Thank you for the informative reply.

These days, a high-capacitance, ultra-low ESR MLCC cap is QUITE cheap for low voltage.
You can get like 10uF-22uF pretty cheap and tiny. There's generally no need for 100uF in combatting electrical noise. Noise will not be sustained that long.

In some cases you can cause latchup by inducing >Vdd on any input pin, so don't neglect this possibility. And you can do it while it's an OUTPUT, too. If the pin's low and you drive current in from an external source, the pin drive tends to ground it out. If the pin's high and you drive current into it, it jumps to the Vdd rail and that can cause latchup.


Leading to the next point not everyone understands: I/O pin protection against overvoltage consists of shunting current to the Vdd rail, and the Vdd input pin, which will actually raise Vdd on the board trace. A regulator mainly has the drive to INCREASE voltage on the output. It often has very little ability to shunt the excess voltage and LOWER the voltage on the output rail. Anything on the board which shunts excess voltage onto the Vdd rail has the potential to cause latchup in anything which uses the Vdd rail.

What devices we can use for clamping the spikes in the Vdd rail? Is there any device with high surge current capacity so that it can be connected to the Vdd rail without a current limiting resistor?

 

[premkumar9]

Possibly, but it's probably been caused by poor design and poor layout, and you're only assuming the problem is entering via the reset pin, which may not be the case.

I am not assuming like that. I understand that the first priority must be to reduce spikes in the supply and ground lines. but still I think I need to have a supervisor circuit to handle a situation where spikes cause problems in spite of all precautions to avoid it.

 

[premkumar9]

Have you try adding ferrite beads to the power supply?

I use 7805 three pin regulator to provide power to my uC. Also use an EL and CD capacitor at the O/P pin and Input. Where do you suggest to add a ferrite bead?

 

[premkumar9]

Most ways to improve the spikes on the supply will involve reducing the input voltage to the microcontroller. A linear regulator with a high power supply rejection ratio (PSRR) will do the trick with plenty of capacitive decoupling before and after the regulator - read the datasheet carefully to see if there is a maximum value of capacitance you can put on the output - it may go unstable.

I would suggest a 100uF tantalum in parallel with a 100nF ceramic on both the input and the output.

If the noise still gets through, a series inductor on the input will produce a second order filter when used with the capacitor.

If you cannot live with the voltage drop a linear regulator gives you, then you need to use series inductors and parallel capacitors to try to reduce the noise.

You only need to add this in series with the microcontrolller, not the rest of the circuit, so this will help to keep the power requirements of your linear reg low.

Finally, you might want to try putting the microcontroller circuit in an earthed metal box to stop any radiated noise getting in to your circuit.

Let me know if any of the above is feasible

I added a 6.8uF tantulum cap at Vcc line (near to the Vcc supply to the RESET circuit). Also added a 0.1uF CD at Vcc pin of IC to ground. Previously I used 10uF, 10K combination to RESET during swith ON. Now I changed it to 0.1uF CD from Vcc line to RESET pin (hoping that it will make the RESET less responsive to spikes.

 

[Hero999]

I use 7805 three pin regulator to provide power to my uC. Also use an EL and CD capacitor at the O/P pin and Input. Where do you suggest to add a ferrite bead?

Put the ferrite beads before the LM7805.

 

[Electronworks]

I added a 6.8uF tantulum cap at Vcc line (near to the Vcc supply to the RESET circuit). Also added a 0.1uF CD at Vcc pin of IC to ground. Previously I used 10uF, 10K combination to RESET during swith ON. Now I changed it to 0.1uF CD from Vcc line to RESET pin (hoping that it will make the RESET less responsive to spikes.

What was the outcome? Have you actually put a scope on various parts of the circuit to see where the noise actually is? Is it being cleaned up by the 7805 or is it appearing after the 7805?