zener diode VI response

[cravenhaven]

Hi all,
I had designed a cct which included NXP BZT52H-B3V6 zener diodes as part of an input limiting filter. My inputs were to be in a range of 0-3.3 volts so I specified a 3.6v zener surrounded by 10K resistors.
However when I came to test the design I found that the zeners limited the input voltage to just 2V so I set about testing the response characteristic of the zener. I looked at the spec sheet but they didnt show a reverse VI characteristic for some reason.
I set up a test cct and measured the voltage vs current to understand what was happening and got this result.

To me this doesnt look like a 3V6 zener, I was expecting a fairly sharp cutoff at soemwhere around the nominal voltage.

Is this what you would expect from a zener ro is there something wrong with this product?.

Allan

[MatthewEveritt]

The data sheet specifies a working voltage of about 3.6v at 5ma so I'd say that data looks about right. I've not played much with zeners but I do know that they're not a sharp cut off.

[JohnnyBerg]

I think you mixed up the X and Y axis?

X axis = voltage
Y axis = current

Actually a nice curve then.

tip: turn your monitor 90° 

[macboy]

It looks about right for a low-voltage zener. Unfortunately, below about 6.3 V, zeners behave very poorly, having a very sloped VI curve (rather than sharply vertical). I say vertical because typically the V is on the X axis, and I on the Y.

Have a look at this document from OnSemi, page 20, Figure 7. You can see that the 6.3 and 7.2 V zeners have a good VI curve, and everything below looks just like what you have discovered.

[c4757p]

Quite typical.

Use a resistor to the power supply to pre-bias the diode with the correct current, then use a standard (non-Zener) diode to clamp into the prebiased Zener.

(Obviously you'll have to subtract some 0.6V from the Zener voltage.)

[SeanB]

Yes, low voltage zener diodes are pretty lax as to the actual voltage. They are sorted out of the pile during production into bins, the voltage across it at a specified current being the only factor. Run it at another current and it will have a different voltage drop. I wanted some zener regulators as rough voltage control, so went and bought a dozen diodes of 3V0, 3V3, 3V6 and 3V9 at 1W, and then simply ran them at 50ma and redid the selection, as I wanted 7V5 at 50mA, so it ended up with a chosen 3V0 and 3V6 diode in series. Perfect to limit the charge voltage in the cheap light so it did not cook the 6V SLA battery any more. Original used a current limit resistor off a supply of 12V open circuit, so it tended to kill the cells fast. I was not going to spend more than $5 on fixing the charger, including the cheap battery, so zener regulator it was. They work well now, at least 2 years per battery instead of 6 months.

[3roomlab]

Yes, low voltage zener diodes are pretty lax as to the actual voltage. They are sorted out of the pile during production into bins, the voltage across it at a specified current being the only factor. Run it at another current and it will have a different voltage drop. I wanted some zener regulators as rough voltage control, so went and bought a dozen diodes of 3V0, 3V3, 3V6 and 3V9 at 1W, and then simply ran them at 50ma and redid the selection, as I wanted 7V5 at 50mA, so it ended up with a chosen 3V0 and 3V6 diode in series. Perfect to limit the charge voltage in the cheap light so it did not cook the 6V SLA battery any more. Original used a current limit resistor off a supply of 12V open circuit, so it tended to kill the cells fast. I was not going to spend more than $5 on fixing the charger, including the cheap battery, so zener regulator it was. They work well now, at least 2 years per battery instead of 6 months.

this thread is good, i wanted to do an experiment similar as what you did ... but is on trying tempco actually (specifically NXP BZV55 )
X axis is total zener voltage ... height of boxes is tempco mV/^oC
each box represent a certain model of zener (eg 6.2v, tempco +2.3mV)
each group is how tempco suppose to null off (which i think will not in real depending on current passed)
(MJlorton have a video experiment on this)
so now i think, even total string of voltage will not be plausible LOL !

[macboy]

this thread is good, i wanted to do an experiment similar as what you did ... but is on trying tempco actually (specifically NXP BZV55 )
X axis is total zener voltage ... height of boxes is tempco mV/^oC
each box represent a certain model of zener (eg 6.2v, tempco +2.3mV)
each group is how tempco suppose to null off (which i think will not in real depending on current passed)
(MJlorton have a video experiment on this)
so now i think, even total string of voltage will not be plausible LOL !

The old-school way to null tempco in a zener is to use a 6.3 V zener in series with a forward biased diode. The tempco of each is approximately equal and opposite, so the overall tempco is "nulled" out (reduced). The tempco of each device varies with current, so the current can (and must) be adjusted to find the sweet spot. This is how older zener voltage reference diodes were made, with the forward biased diode built-in.

[SeanB]

If you want I still have a half box ( around 5k pieces each) of 12V and 24V 400mW ITT zener diodes. I use them either as zener diodes or as small signal diodes, depending on actual circuit voltage. They are starting to come off of the red and white tapes though.

[JohnnyBerg]

The old-school way to null tempco in a zener is to use a 6.3 V zener in series with a forward biased diode. The tempco of each is approximately equal and opposite, so the overall tempco is "nulled" out (reduced). The tempco of each device varies with current, so the current can (and must) be adjusted to find the sweet spot. This is how older zener voltage reference diodes were made, with the forward biased diode built-in.

Those where the days .. 

[SeanB]

When you bought one of those it came with a hand written label with the actual voltage ( to 2 digits) and the current ( to 2 digits after the decimal) of the actual value, and the room temperature at which this was done, and by whom. Impressive considering the readings were typically done with a moving coil meter, though this would only be a null detector in either a KV divider or similar, referenced to a copy of the standard volt.

3roomlab, I bought them at an auction for $2, in a lot. Nobody else wanted them, and the auctioneer started at $2. They do make good use once lights across a 12V lead acid battery.

[cravenhaven]

Thanks everyone for the replies.
I found it strange that the datasheet did not provide a VI graph for the reverse characteristics which seems to me to be the purpose of the device. Also the results I found would have suggested to me a nominal rating of 3-3.5 rather than 3.5-4 but perhaps that's splitting hairs.
The documents you linked to macboy was very indepth and the graph you showed seems to put a lie to the devices even being called zeners in the low voltage ratings.
The exercise has rammed home to me the need to test each part of a cct before commitment to any sort of production, and zeners need to be treated with suspicion.
I will experiment with the the bias cct C4757p, thanks for the suggestion.

[codeboy2k]

zeners are always grossly spec'ed.. they are not precision devices, and if you want any precision what-so-ever you need to age them and select them for voltage and tempco.

As the story is told, Fluke (HP, and others too) would age and select a zener and a resistor divider as a set, to get a precise 10V reference with an op-amp from a TC zener.

Use them directly from the manufacturer whenever you need an approximate voltage and can tolerate the lack of precision, but anything better than that requires time and dedication.  These days, it's far cheaper to buy a reference when a precise, absolute voltage is needed.

I have used zeners as voltage regulators for powering an op-amp, since the PSRR of the op-amp can usually take care of small mV deviations from the zener.