Measured zener voltages

[paulossant]

I bought a pack of zener diodes from ebay a while back. Since it came from ebay, I didn't had any datasheet for those.
Now I just tested a bunch of zener diodes voltages vs current, and got a really spread of expected values. I expected the rated value around 1mA, but some of those are really far from that.
Are those really bad zeners, or this is expected?
The only one that behaved very well was the 5V1 zener, as expected since around 5V they are usually very well behaved.


All pictures and the .xls file are attached

[madires]

Look for datasheets of Zener diodes with the same wattage or package to get an idea about the specs. For example, in case of 500mW low voltage Zeners the Zener voltage is often stated for a current of 5mA.

[paulossant]

That's ok, but what surprised me, is that the IV curve hardly resembles a "knee" shape, as stated all over electronic books.

[markce]

The standard low voltage zener diodes are arround 400mW max, so mind the test current.
Different types of zener diodes have quite a different temp coefficient.
The internal resistance also influences the clamp voltage.

[tszaboo]

Low voltage zeners are terrible. They always have been.

[magic]

5V1 looks like the textbook curve, 4V7 is not far behind.

The high voltage ones could indeed have significant internal resistance. Connect a 47Ω resistor in series with 5V1 and you will probably see a similar curve above a few mA.

3V3

[schmitt trigger]

I believe that the combination of a log scale for the vertical axis, and a very compressed range for the horizontal axis, is giving you those weird curves.

Re-plot using a linear scale for the current, and start all the way from "0" in the voltage, and see how they look like.

[macboy]

Low voltage Zeners (< ~ 5V) use a true Zener effect, and this has a relatively soft knee. Above that, they are actually using avalanche effect which presents a much sharper knee.  In any case, plotting with a (less practical) linear I_F scale will illustrate the knee more clearly.

The 6.2 V one is perhaps the best behaved. Look at the first three data points at low current: nearly vertical plot. This is why these are nearly exclusively used in Zener voltage references. They are usually combined with a forward biased Si diode in series (either internally or externally) to cancel the temperature coefficient. This is why Zener references (LM399, LTZ1000, etc.) tend to be around 6.8 to 7 V range. 

Some of your lower voltage ones are clearly not what they claim to be. I'd expect the nominal voltage to appear somewhere around 1 mA (give or take an order of magnitude).

[magic]

On a related note, I bought a kit of ceramic disc capacitors from an auction site the other day and almost all were over 20% off

[paulossant]

The log scale was to get a better look at the supposedly knee. The 4V7 and 5V1 are as expected, but the 3V3 is absolutely awful.

[Vovk_Z]

I have tested now one of my 3V9 Zeners - it is different from TS (clearly better).
It has about 2.0 V ar 0.01 mA; 2.55 V (+-) at the 0.1 mA; then 3.3 V at 1.0 mA; and 4.1 V at 10 mA. It is ok.

(It is easy to test because I have "1 kOhm resistor at output" mode in my bench supply, so I can see Zener or LED voltage drop right onto device panel meter).

Tested one of my 3V3 Zeners:
1.60VDC at 0.01 mA; about 2.05VDC at 0.1 mA; 2.7VDC at 1.0 mA; and 3.56 V at 10.0 mA.

My Zeners are not from Ebay but from shops in my city (they usually sell normal, I mean good, parts).

[madires]

My personal rule of thumb is to drive low-voltage low-wattage Zeners with 3mA for best performance.

[schmitt trigger]

You may want to purchase a few devices from reputable vendors, and test under the exact same conditions and make A-B comparisons.

After all, these are Ebay components. There may be a good number of functional but borderline components.

[mikerj]

Low voltage zeners are terrible. They always have been.

Indeed, I really don't see much point in the very low voltage parts; they make terrible voltage references and are little use for clamping signals.

[TimFox]

Technically, the lower-voltage parts are true "Zener diodes", and have the "soft" V-I curve you measured.
The higher-voltage parts (above roughly 5 or 6 V) are "avalanche diodes" and have "harder" curves.
See:  https://circuitglobe.com/difference-between-avalanche-and-zener-breakdown.html

[Conrad Hoffman]

As suggested above, I'm sure low voltage zeners are good for something, but I'm not sure what it is. Further, friends don't let friends buy electronic parts from the big auction site because they tend to be of questionable pedigree and quality.

[TimFox]

At those low voltages, you have no choice but "true Zeners", unless you use band-gap devices such as the TL431 as shunt regulators.

[muvideo]

At those low voltages, you have no choice but "true Zeners", unless you use band-gap devices such as the TL431 as shunt regulators.

Some forward biased LEDs are not that bad, Red gives about 2V, Blue about 2.8-3V,
but their curve and internal resistance is not in datasheet usually and has to be tested.

[exe]

very interesting, I'd expect the best zenners to be 5.6V and 6.2V. May be 6.8V if including a diode that should do temperature compensation.

[srb1954]

Testing with 100mA Zener current is going to cause significant self-heating in the diodes, particularly if they are small packages. This will affect the shape of the curve due to the temperature coefficient of the Zener voltage - positive for voltages above 6V and negative for voltages below 5V, approximately zero between 5V and 6V.

If you really need to test the Zener voltage at high currents e.g. for clamping circuits you need to use a pulsed current technique to determine the true voltage vs current curve.

[TimFox]

Zeners around 6 V have the best tempco.  6.8 V reference Zeners are a 6.2 V Zener and a forward-biased PN junction in series, so that their tempcos cancel at a specified current.  There are similar devices with two or three PN junctions to get higher reference voltages.

[paulossant]

Testing with 100mA Zener current is going to cause significant self-heating in the diodes, particularly if they are small packages. This will affect the shape of the curve due to the temperature coefficient of the Zener voltage - positive for voltages above 6V and negative for voltages below 5V, approximately zero between 5V and 6V.

If you really need to test the Zener voltage at high currents e.g. for clamping circuits you need to use a pulsed current technique to determine the true voltage vs current curve.

I have taken the temperature effects into consideration. To more precise, I tested the zener diodes with mine diy constant current load.
I saw the temperature dependence affecting the readings, so for currents bigger than 25mA i have done that exact pulse measurements (by turning on and off for about 3-4 seconds, allowing me to take note of the values)

my diy constant load was not good for currents below 300uA, therefore the lack of measurements bellow that value, but I can test it with other setup.

[srb1954]

Applying test currents 3-4 seconds is still too long to avoid temperature effects. You really need to be looking at millisecond length test pulses to avoid excessive junction heating.

It is not the overall case temperature that matters but that of the small semiconductor junction in the centre of the diode. These have a very small thermal capacity and don't take long (<100ms) to heat up.

[TimFox]

I noticed how strong was the self-heating effect in 1 W Zeners when I tested 62 V diodes (axial lead) on a Grayhill test clip fixture (plugged directly into a DVM, fed from a current source), and saw how the voltage changed when I pinched the diode between thumb and forefinger.

[paulossant]

Applying test currents 3-4 seconds is still too long to avoid temperature effects. You really need to be looking at millisecond length test pulses to avoid excessive junction heating.

It is not the overall case temperature that matters but that of the small semiconductor junction in the centre of the diode. These have a very small thermal capacity and don't take long (<100ms) to heat up.

Testing both ways should give an interesting comparison. I will arrange a setup, maybe a micro with a one time shot pulse (25ms), acting on a transistor to turn on a voltage source momentarily, and just get a resistor and the zener, with a scope on the zener. Would that be a good way to test?