EMI of a heater regulator

[Bernhard Kuemel]

Hi sed!

I made a variable duty cycle regulator to get a 170 W heater down to
about 30 W with a NAND that feeds back to its input capacitor via 2
resistors and another NAND drives a power MOSFET:

http://bksys.at/bernhard/img/44/regulator.jpg
(image + circuit diagram)

It did work, but produced EMI, so I added various inductors and a
diode as seen in the circuit diagram, but not in the photo. About 80
turns over a bundle of soft iron wire was not enough, so I tried a
ferrite insulation transformer, a big self wound coil on a EI core
of a former (silicon steel?) transformer and a microwave oven
transformer primary. The latter two eliminated the EMI but were too
noisy, acoustically.

I played a bit with the HV output of the MOT and when I arced to the
structure of my table the lamp went to full 100 W. The second NAND
had died appearently and went high permanently. My table top is wood
with plastic cover. The control circuit uses 0,3-0,6 mA, so I think
the NAND is not abused by driving the MOSFET.

The power of the lamp was fluctuating somewhat (other than beats
from the 50 Hz from the line and the regulating frequency) and when
I touch the fixed resistor I get bigger fluctuations and when I hold
it tighter the lamp may go out at all.

Any recommendations on how to eliminate the EMI (quietly) and why
the NAND died?

Parts list:

HEF4092BP 4 NAND Schmitt triggered
470 nF
22Kohm variable (actually only 18K)
47Kohm fixed
1N4148 2 diodes
STP3NC60FP PowerMesh II MOSFET, 600 V, 2 A
2 A bridge rectifier
10 V SMPS from a cell phone
various inductors
UF4007 diode
230 V rms main power

Thanks, Bernhard

 

[Bernhard Kuemel]

HEF4092BP 4 NAND Schmitt triggered

HEF4093BP

 

[Henry Kiefer]

Add a series resistor between driver gate and MOSFET. Maybe 100 to 1k.
Check if the power supply for the IC is correct and stabilize it with an
capacitor.

What is the value of the R-C ?

- Henry

 

[default]

Hi sed!

I made a variable duty cycle regulator to get a 170 W heater down to
about 30 W with a NAND that feeds back to its input capacitor via 2
resistors and another NAND drives a power MOSFET:

http://bksys.at/bernhard/img/44/regulator.jpg
(image + circuit diagram)

It did work, but produced EMI, so I added various inductors and a
diode as seen in the circuit diagram, but not in the photo. About 80
turns over a bundle of soft iron wire was not enough, so I tried a
ferrite insulation transformer, a big self wound coil on a EI core
of a former (silicon steel?) transformer and a microwave oven
transformer primary. The latter two eliminated the EMI but were too
noisy, acoustically.

I played a bit with the HV output of the MOT and when I arced to the
structure of my table the lamp went to full 100 W. The second NAND
had died appearently and went high permanently. My table top is wood
with plastic cover. The control circuit uses 0,3-0,6 mA, so I think
the NAND is not abused by driving the MOSFET.

The power of the lamp was fluctuating somewhat (other than beats
from the 50 Hz from the line and the regulating frequency) and when
I touch the fixed resistor I get bigger fluctuations and when I hold
it tighter the lamp may go out at all.

Any recommendations on how to eliminate the EMI (quietly) and why
the NAND died?

Parts list:

HEF4092BP 4 NAND Schmitt triggered
470 nF
22Kohm variable (actually only 18K)
47Kohm fixed
1N4148 2 diodes
STP3NC60FP PowerMesh II MOSFET, 600 V, 2 A
2 A bridge rectifier
10 V SMPS from a cell phone
various inductors
UF4007 diode
230 V rms main power

Thanks, Bernhard

Microwave Oven Transformer choke? That's gotta be gross overkill.

Too much inductance and too much stored energy - lucky to keep the
Mosfet alive when dealing with the spikes that could produce if
allowed to saturate.

anyhow . . . most EMI will respond to just a little bit of attention
to just the right point in the circuit - not MOT's - and you shorted
the HV winding? I can see the temptation - but would expect parts to
fry doing that.

The advice on using a gate resistor? Good. A small ceramic disk cap
from the gate ground may also help. Or a snubber around the mosfet
(.01 uf cap in series with a 100 ohm resistor).

If you have a scope - look at the signal to the load - it should be
on-off with little or no ringing or glitches in it.

If you are touching parts of the circuit and causing changes - that is
probably a switching glitch/ringing - I saw that a lot when designing
a 30 amp 24 volt pwm battery discharger - a snubber around the mostet
and cap on the gate made it smooth right out. Problem was the leads
out to the battery were radiating signal back to the gate of the
mostet.

You don't say what frequency the EMI is occurring. If it is affecting
FM radio, for instance, soft iron wire is probably not going to work.

Suspect layout - long leads carrying heavy current with fast rise
times will radiate EMI - likewise the power supply can carry noise
back into the power lines. Lots of inexpensive surplus IEC type power
line filters on the market - which will work as well on the load side
if the leads to heater are radiating. - but first try to get a smooth
on/off transition with no ringing - since ringing is bad for the
mostet wastes power and creates RFI.

 

[Ancient_Hacker]

You need both L and C to make a good filter. Try your coil again but
put 0.01uf from either side of the coil to ground.

 

[Bernhard Kuemel]

Add a series resistor between driver gate and MOSFET. Maybe 100 to 1k.

Hmm, that will slow down the switching. I was using an extra NAND to
minimize switching time to get mimimal switching losses/heat in the
MOSFET. I'm not sure, though, whether the on resistance or the
switching losses are the dominant heat source.

Check if the power supply for the IC is correct

It is rated 10-11 V and measures 10.3 V, well within the recommended
3-15 V for the HEF4093BP.

and stabilize it with an capacitor.

Ok. There must be a C in the SMPS, but I guess the additional C
shall stabilize against incoming EMI.

What is the value of the R-C ?

The oscillator C is 470 nF, the charging R is 47K, the discharging R
is 0-18K. The hysteresis voltage of the Schmitttrigger is 1 V
(5.2-4.2 V).

Bernhard

 

[Bernhard Kuemel]

Microwave Oven Transformer choke? That's gotta be gross overkill.

Sure, much too big, too. I just tried it because it was lying around.

Too much inductance and too much stored energy - lucky to keep the
Mosfet alive when dealing with the spikes that could produce if
allowed to saturate.

Hmm, I have a 100 W light bulb in series, and a diode that allows
the current to keep running when the transistor closes.

anyhow . . . most EMI will respond to just a little bit of attention
to just the right point in the circuit - not MOT's - and you shorted
the HV winding? I can see the temptation - but would expect parts to
fry doing that.

I thought, maybe when I arced some sort of puls made it through the
gate capacitance to the NAND. It's just strange that it happened
when I arced to the metal structure, while the top board is plastic
covered wood and survived quite a while of direct short circuit arcing.

The advice on using a gate resistor? Good. A small ceramic disk cap
from the gate ground may also help. Or a snubber around the mosfet
(.01 uf cap in series with a 100 ohm resistor).

Ok. Even slower switching then. Although I don't plan on playing
with the MOT any more and it worked quite long (at least about an
hour) before I did that.

If you have a scope - look at the signal to the load - it should be
on-off with little or no ringing or glitches in it.

No scope. It's on my whishlist for about a year now. I use a piezo
speaker from an old digital watch which showed me that the
oscillator is still working.

You don't say what frequency the EMI is occurring.

I first noticed it in the headphone connected to my sound card.

If it is affecting
FM radio, for instance, soft iron wire is probably not going to work.

I also get it in a small battery powered radio. So, is ferrite ring
better then? I'll try some.

Suspect layout - long leads carrying heavy current with fast rise
times will radiate EMI - likewise the power supply can carry noise
back into the power lines. Lots of inexpensive surplus IEC type power
line filters on the market - which will work as well on the load side
if the leads to heater are radiating. - but first try to get a smooth
on/off transition with no ringing - since ringing is bad for the
mostet wastes power and creates RFI.

Thanks, Bernhard

 

[default]

Ok. Even slower switching then. Although I don't plan on playing
with the MOT any more and it worked quite long (at least about an
hour) before I did that.

The cap is used in conjunction with a resistor and if the resistor is
small ~100 ohms and the cap in the 100 pf range it shouldn't slow the
mosfet switching substantially.

If you are encountering ringing or glitches in the transitions they
usually happen at relatively high frequencies 100 KHZ - MHZ range so
it doesn't take much to kill them..

The snubber works similar to the way the condenser across points in
old ignition systems used to. Without the cap you get a lot of arcing
at the points and a weak spark at the plug. With the cap the spark is
hotter and the points don't arc.

In the ignition you use a cap to increase the ringing (while saving
the points). The snubber absorbs the pulse of energy caused by
inductance in the heater and wiring to the heater when the mosfet
opens. The 100 ohm series resistance with the cap, keeps the ringing
down by lowering the Q of the LC circuit. Snubbers can work wonders.

How is the heater arranged? If it is coiled wire it may have a high
inductance - that and parasitic capacitance is a tuned circuit -
guaranteed to ring for a time after the power switches off diodes or
not.

Diodes? They have to be very high speed to deal with transients
effectively. 1N4000 series are useless in that application. An
ordinary silicon rectifier takes a long time to turn off once it
conducts.

Sorry about the scope - that would probably lead you right to it.

 

[Ban]

Hi sed!

I made a variable duty cycle regulator to get a 170 W heater down to
about 30 W with a NAND that feeds back to its input capacitor via 2
resistors and another NAND drives a power MOSFET:

http://bksys.at/bernhard/img/44/regulator.jpg
(image + circuit diagram)

It did work, but produced EMI, so I added various inductors and a
diode as seen in the circuit diagram, but not in the photo. About 80
turns over a bundle of soft iron wire was not enough, so I tried a
ferrite insulation transformer, a big self wound coil on a EI core
of a former (silicon steel?) transformer and a microwave oven
transformer primary. The latter two eliminated the EMI but were too
noisy, acoustically.

I played a bit with the HV output of the MOT and when I arced to the
structure of my table the lamp went to full 100 W. The second NAND
had died appearently and went high permanently. My table top is wood
with plastic cover. The control circuit uses 0,3-0,6 mA, so I think
the NAND is not abused by driving the MOSFET.

The power of the lamp was fluctuating somewhat (other than beats
from the 50 Hz from the line and the regulating frequency) and when
I touch the fixed resistor I get bigger fluctuations and when I hold
it tighter the lamp may go out at all.

Any recommendations on how to eliminate the EMI (quietly) and why
the NAND died?

Parts list:

HEF4092BP 4 NAND Schmitt triggered
470 nF
22Kohm variable (actually only 18K)
47Kohm fixed
1N4148 2 diodes
STP3NC60FP PowerMesh II MOSFET, 600 V, 2 A
2 A bridge rectifier
10 V SMPS from a cell phone
various inductors
UF4007 diode
230 V rms main power

Thanks, Bernhard

Another candidate for a Darwin-Award, there doesn't seem to be even a GFI in
the distribution panel. The microwave transformer will do the rest. Bravo!

 

[GPG]

The snubber works similar to the way the condenser across points in
old ignition systems used to. Without the cap you get a lot of arcing
at the points and a weak spark at the plug. With the cap the spark is
hotter and the points don't arc.

In the ignition you use a cap to increase the ringing (while saving
the points).

The capacitor slows the rare of rise, allowing the points to develop
some clearance. With fast solid state switch the self capacitance of
the coil is enough.
But to the topic, an integral cycle controller will reduce the EMI

 

[default]

The capacitor slows the rare of rise, allowing the points to develop
some clearance. With fast solid state switch the self capacitance of
the coil is enough.

Huh? Slows the rate of rise? Current begins falling the instant the
points open.

The cap serves two functions. It appears as a dead short (that
instant only) across the points and keeps the current flowing while
the points open - suppressing the arc. It forms a tank circuit with
the primary of the ignition coil and gives a damped sine wave at its
resonant frequency - increasing the spark. That's the Kettering
ignition system from the first auto's that had points and distributors

The first transistor ignitions still had points but they only switched
the non-inductive current to the base of the transistor. The first
systems still had caps across the transistor to take advantage of the
resonant circuit.

Coil pickups replaced points and the ignition coils changed from
induction coils to pulse transformers. Entirely different systems
even if they use some of the same parts - the operation isn't the
same.

I don't understand where the "self capacity of the coil is enough"
comes in - the coil isn't across the points in any ignition system
(that I'm aware of) and inter winding capacity is generally minimized
in pulse transformers - slows the rise time.

But to the topic, an integral cycle controller will reduce the EMI

Integral Cycle Controller usually means letting a string of pulses
from a sine wave into the heater - gating them at the zero crossing -
no EMI. That's my understanding of ICC - did you mean something else?

The op has a 40 volt DC supply with a NAND gate pwm switching DC
through a heater modulating it for 30 watts average dissipation -
using a single mosfet.

That is perfectly reasonable and can be done with no EMI to speak of.
In that application he doesn't even need a particularly fast switch.
I didn't question his using DC - it can be done and it is what he has
chosen to do.

 

[GPG]

points open. No
The cap serves two functions. It appears as a dead short (that
instant only) across the points and keeps the current flowing while
the points open - suppressing the arc. It forms a tank circuit with
the primary of the ignition coil and gives a damped sine wave at its
resonant frequency - Yes

increasing the spark No
.. It saves enrgy being lost at the points
.. That's the Kettering

ignition system from the first auto's that had points and distributors

The first transistor ignitions still had points but they only switched
the non-inductive current to the base of the transistor. The first
systems still had caps across the transistor to take advantage of the
resonant circuit.

The self capacitance referred to above is sufficient to leave the cap
out

Coil pickups replaced points and the ignition coils changed from
induction coils to pulse transformers. Entirely different systems
even if they use some of the same parts - the operation isn't the
same.

I don't understand where the "self capacity of the coil is enough"
comes in - the coil isn't across the points in any ignition system
(that I'm aware of) and inter winding capacity is generally minimized
in pulse transformers - slows the rise time.

from a sine wave into the heater - gating them at the zero crossing -
no EMI. That's my understanding of ICC - did you mean something else? No
The op has a 40 volt DC supply with a NAND gate pwm switching DC
through a heater modulating it for 30 watts average dissipation -
using a single mosfet.

That is perfectly reasonable and can be done with no EMI to speak of.
In that application he doesn't even need a particularly fast switch.
I didn't question his using DC - it can be done and it is what he has
chosen to do.

Perhaps a triac and one of these ( or similar) will help
http://www.fairchildsemi.com/ds/MO/MOC3163-M.pdf