Zener diode regulator question.

RDL2004

I'm working on a power supply project and I'm using an LM317 as the main regulator. I want the output to go down to almost zero, so I need a negative 1.25 v or so reference. I'm using a TL7660 and an LM385-1.2 and have that part working fine. The 7660 needs 10 volts maximum and the 385 needs a little over 6 volts minimum. Since the main input is 24 volts, I need to regulate the power to the 7660 to below 10 volts. I didn't have any small 3 terminal regulators that would work, but for some strange reason I had some 10 volt zener diodes, so I thought I would try using one of those. I never built one of these before, and in my opinion it doesn't work very well. Maybe I'm doing something wrong?




It doesn't appear to actually "regulate" until about 5 volts over its zener voltage spec, and even then it's not well regulated. I'm probably going to switch to a 78L09, but I'm curious if this is the way a zener diode normally works. The zener diode is a Philips NXP BZX79-B10, 113
For my testing I used a 12 volt LED (with internal resistor) as the load. The actual voltage reference circuit will draw about 15 mA, the 12 volt LED is probably under 10 mA.
The 1k resistor is maybe a bit too high in value, but I tried going down to 470 with no real difference. The power supply I'm using as the source won't go above 24 volts so I have no idea what happens up there, but it looks like the output voltage wants to continue to slowly rise.

Mike Green

Given the load of the LED and the value of the resistor, I'm not surprised at the curve you got. At 15V Vin, the Zener drops 10V and the 1K resistor drops 5V (without the LED). That's 5mA through the resistor. As the LED conducts, it will draw at least 5mA. That leaves no current for the Zener to use to regulate. The LED probably draws on the order of 10mA to 20mA at 12V. You'd want a Zener current of at least several mA. I'd use a resistor no more than 330 Ohms with a 1/4W to 1/2W rating.

RDL2004

Okay, I realize now how the forward voltage drop of the LED was affecting the point where the zener starts to conduct. Replacing the LED with a 1K resistor dropped the point where the zener starts to regulate to about 12 volts or so. I had some 100 ohm 1/2 watt resistors and a total of 300 ohms worked pretty well as you suggested. I have a much better idea of how these work now. I don't think I'll end up using one in this project, but I can see how they can be useful as long as you don't expect too much. It seems like the input voltage can't vary a whole lot and the current limit resistor needs to be carefully matched to the load, because when even I was getting 10-25 mA through the zener, the output voltage would still rise a volt or so when the input was varied from 15 to 24 volts.

Heater.

Don't expect components to be perfect.
A diode for example is said to conduct one way but not the other. Sounds simple enough.
Of course diodes can conduct a bit in the reverse direction, especially the one I made in lab at university:)
Of course diodes will eventually give up and pass current with too much reverse voltage.
Similarly, in the forward direction the humble diode is not a short circuit. It has an I/V characteristic.

When it comes to zeners, they are designed to give an emphasized "brick wall" characteristic. But of course it is not perfect.

If you really want to get down to basics you have to plot the I/V curves of these devices.

Tracy Allen

I don't know what you mean by, "the LM385-1.2 needs a little over 6 volts minimum." The LM385 is basically the same as a zener diode, except it is made with transistors and has a very sharp knee and tighter regulation than the 9V zener that you described in your experiment. It only takes about 10µA of current to go into regulation, and it stays flat up to about 20mA.

You could use a couple of LEDs in series forward bias to regulate the power voltage to the TL7660. LEDs have pretty good regulation, and the TL7660 works well at 3V. The resistors have to be chosen to supply current needed by all the consumers, the LED, the quiescent current of the TL7660, and the current drawn from the -V supply to bias the LM385 and the LM317. Just like the experiment, if any branch is starved for current it will give unwanted results.

RDL2004

"the LM385-1.2 needs a little over 6 volts minimum."

I admit that I don't understand why this happened, but when connected to the adjust pin of the LM317, if the voltage supplied to the 7660 dropped below 6.1 volts the output from the LM385 shifted from -1.25 to -0.6 volts. It's possible this was due to some flaw in the power supply to the negative voltage reference circuit while I was trying to make the zener diode work. When the problem first appeared, I did notice that the reference circuit in the LM317 data sheet showed a 10 volt supply to the LM385 so I assumed that may have been the problem. Since I'm going by notes I made almost a year ago and I now know that this part of the circuit works perfectly when supplied with 9 volts I will revisit this issue and see if I can reproduce the problem. I plan to rebuild the circuit today using an LM317 to supply power to the 7660/385.

By the way, not related to to LM385 voltage problem, but here is a chart where I was using 300 ohms for the current limit resistor which also shows the current through the zener diode (in mA), in case anyone is interested. You can see that the zener is working a bit better, but there is still more Vout rise vs. Vin than I would like.

PJAllen

Here's an approach to consider, it's an unconventional implementation of a 7805.
If you don't otherwise require any current from the 7805 then use a 78L05.

Tracy Allen

Rick,
I don't have a clear picture in mind of the circuit you're using for this, so it's hard to analyze the fishy behavior. I was imagining that the '7660 provides a negative supply referenced to ground, and the '385 is fed by a resistor from that supply and provides a regulated -1.2V, which feeds the bottom of the adj divider on the LM317. That negative supply would have to sink on the order of -5 to -10 mA. And as above there would have to be some current through the LM385, so overall the '7660 would have to supply -10 to -15mA.

For the zener diode, the voltage is not constant due to a parameter that you can find in a zener data sheet, the "dynamic resistance". A real zener diode is modeled as an ideal zener diode (absolutely flat constant voltage independent of current) internally in series with a small resistor. Example, from its data sheet, the TMZ5221B has a dynamic resistance of around 30Ω around a test current of 20mA. Dynamic resistance of zeners is usually lowest for zener voltages of near 7 volts.

Your test diode voltage changed by about 1 volt as the input current went from 0 mA to 35 mA. 1V / 0.035 A = 28Ω. That is the slope of the curve, the dynamic resistance.

RDL2004

The circuit I'm using for testing purposes is pretty much out of the data sheets. I'm using a BK Precision 1651 as my power source. I did a quick check this morning powering the LM385 from the BK 1651 and not the 7660 and I was able reproduce the problem where I got negative 1.24 volts out from the LM385 at less than negative 6 volts in, until it was connected to the resistor divider on the LM317 adjust pin when it rose to about -0.9 or so. I still need to investigate that more closely though.



Here are the results I got this morning using a purely resistive load on the zener.
Current was measured between the zener diode and ground (schematic in the upper right corner).





It's working pretty much as expected. I still don't think it makes a very good regulator though. I guess in a commercial environment where cost is a driving force, the difference in volume pricing between 4 pennies for a zener diode and 25 pennies for a 78xx regulator is significant, but for hobbyist and experimenter type use, I say forget about using zeners.

When I was taking the LM317 circuit apart this morning to rebuild it on a different breadboard, I had one of those /facepalm moments when I noticed this:



Yes folks, that is a 1992 vintage Teledyne Components TC7660 chip. Not the TI device I thought I was using (though it doesn't seem to have be the cause of any of my difficulties).

I know I ordered some TI 7660s last spring, I just have to find them now.

edit:

One more thing, I was reading the data sheet for the LM385 and it seems its current limiting resistor (R4 in my schematic above) is way too low. It should probably be something more like 50K to 100K ohms, not 680 ohms, but I pretty much copied that part of the circuit directly from the LM350 data sheet. Go figure.

Tracy Allen

You can reduce the current in the LM317 adj circuit by increasing R1 from 120Ω to 240Ω, and use a 10k pot for the adjustment instead of 5k. That halves the current sink requirement of the negative supply. Note that the current in the adj circuit is 1.2V / R1, so with R1=240Ω, the adj current is 5mA. There is also a bias current from the adj terminal, ~50µA. For stability you want 1.2 / R1 >> 50E-6. The negative supply then should provide 10mA, 5mA for the LM317 plus 5mA to keep the LM385 in regulation.

If you have an oscilloscope, look at the ripple on the '7660 pin 5 negative output. The oscillator frequency of a TL7660 is ~10kHz. One period is 0.0001 second. Ballpark, the ripple with 10mA sink current and a 10µF capacitor is
dV = dt * I / C
..... = 0.0001 second * 0.01 A / 10E-6 F = 0.1V
so that should be okay.
The LM385 should provide good regulation on the negative side, despite the ripple or lack of regulation going into the '7660.

jmg

RDL2004 wrote: »

.. here is a chart where I was using 300 ohms for the current limit resistor which also shows the current through the zener diode (in mA), in case anyone is interested.

Or, you could replace that 330R with an OnSemi NSI50010YT1G, which is a Constant Current Regulator & LED Driver 50 V, 10 mA.
(they also have 20mA models)

This would run the 7660 in effectively constant current mode : the voltage would be whatever was needed for 10mA constant current, and the 680R on the -ve side sets the V/I transfer.