Explain this transformer phenomenon..............

[ian]

I'm using a microprocessor to control the brightness of a lamp
by detecting the "zero" voltage point on the A.C., then delaying
a pulse to a solid state relay. Everything works fine.................
(famous last words)..............until..........Murphy's law kicks in.

I've figured out the problem but I'd like an understanding of the
phenomenon.

The circuit is very simple, A 24V step down transformer is fed into
a full wave Schottky bridge rectifier. No capacitor is used on the
rectified positive side, instead a voltage divider of a 30K resistor
on the + side and a 4.3k resistor to -. The mid point is fed into
the base of a 2n3904 transistor so when the + voltage drops below
4.8V the divided point drops below .6V and the transistor stops
conducting. The positive side of the full wave rectifier is fed into
another diode to a filter capacitor. The diode allows the "raw"
rectified voltage (before the filter cap) to drop to 0.

The problem...........
The pulse train from the transistor stops after a while...........

What's happening?
After investigation and scopeing I've found the "raw" positive voltage
from the bridge rectifier doesn't drop all the way to 0. It is actually
"chopped" at the bottom for about 1.1mS. In fact, it
stays above 4V. Then, as the transformer? heats up under load,
the minimum voltage climbs higher. Eventually it climbs over 8V
keeping the transistor always conducting, hence, no pulse train.
Now I *think* this is caused by transformer saturation but I don't
understand the theory. Even a saturated transformer must reverse
polarity so the voltage must drop to 0. And what's causing the
"chop" at low voltage?

The transformer is a Hammond BE2G class2 and the rectifiers are SB560.

[sghioto]

Ian,
Does the voltage go to zero if the filter capacitor is removed? If so then you need to isolate the filtered dc voltage more. Use another bridge rectifier across the 24 volt winding for this purpose. Suspect some leakage in the diode connected to the + rectifier output causing this problem

Steve G

[MrAl]

I'm using a microprocessor to control the brightness of a lamp
by detecting the "zero" voltage point on the A.C., then delaying
a pulse to a solid state relay.
<snip>
The positive side of the full wave rectifier is fed into
another diode to a filter capacitor. The diode allows the "raw"
rectified voltage (before the filter cap) to drop to 0.

Hi ian,

The problem can be that you are using another Schottky for this
(fifth) diode, and the leakage is working with your 30k+ resistor
divider impedance to keep the divider biased from the cap supply,
or that the diode being used has higher than normal leakage.

The solution would be to either use a regular diode like 1N4148
or possibly even 1N4002, and/or lower the impedance of your
divider to less than 10k. I would think the bridge rectifier section
could remain as Schottky's, but these type of diodes are not really
necessary even for a zero crossing detector, as 1N4002 or similar
diodes would work just fine and have less leakage.

Schottky diodes have rather high reverse leakage, in many cases
it looks similar to a regular diode (with lower forward drop) but with a 20k
resistor in parallel ! Diodes like 1N4148 can have an equivalent
reverse leakage that looks like a 500 megohm resistor in parallel...
quite a difference: 500 megohms compared to 20k.

[sghioto]

I wired this circuit up yesterday, basically trying to duplicate Ian's design and it's functioning fine after 12 hours. Didn't have a 24 volt transformer but used a 16 volt ac alarm transformer instead.


Steve G

[MrAl]

Hi again,

That's great. I suspected that a 1N4148 diode would work better.
You can verify ian's problem by connecting a 20k resistor in parallel
with the 1N4148 to see if the circuit stops functioning normally
(no or reduced pulses out).

[ian]

I applied 20V to the diode I'm using but there was no apparent leakage
through the SB560.

I did reduce the 30k resistor to 15K and the +voltage drops to 0 but
the input to the transistor base became "fuzzy" and did not drop
enough to put the transistor into conduction. Strange.

I have now put the 30K resistor back and reduced the 4.3k resistor
to 2K for more of a low voltage drive. It seem to be working now.
The low voltage on the A.C. drops to 2V and does not seem to climb
even when the transformer heats up. The divided voltage into the
transistor is better defined.

[ian]

The odd thing about this circuit is that the bottom part of the
rectified waveform is distinctly "clipped". It is a nice sine wave
until about 4V then it is suddenly flat. I don't know what would
cause that.

[MrAl]

Hi again ian,

Well, i looked up the spec on that diode SB560 and it says it has
pretty high leakage current... 0.5ma to 25ma, and it only takes
100ua to make the two resistors 33k and 4.7k produce 4vdc, which
is what you are seeing.
Do you have a 1N4148 or even some other 'standard' si diode
laying around you can try?

Yes it is possible that the problem is caused by something else,
but what you are seeing is exactly what you would see with a
diode that is leaking according to it's very own data sheet.
For this reason alone please try another diode, of a different
type (standard si like 1N4148 or even 1N4001 or similar).

[ian]

I did have two circuits doing the same thing when I was scoping the
waveforms.

I also do not understand this "leakage". I put 40V across the the
diode and measured the leakage. I tested 20 and only a few
leaked anything measurable, like a few microamps.

[MrAl]

Hi ian,

It could be that something else is wrong, but it must be something
simple or the wiring is wrong. Is the transistor ok? Is the cap
connected in the right place?

Here's why it looks like the diode. Note how sometimes the waveform
tips dont reach all the way down to zero volts as you reported.
This seems to only happen when there is leakage in the diode.



If you swap the diode for 1N4148 and it still doesnt work, then
send the circuit to me and i'll fix it for you. Swap that diode first
though.

[MrAl]

Hi again,

In addtion to the waveforms in my previous post, i'd like to add that
the 1N4148 diode isnt really made for rectification into a cap filter,
so maybe a 1N4001 diode would be better. A 1N4148 diode would
still be good to do a quick test with though.