Convert 12-14 vdc to 9 vdc for a device only needing 0.3ma...

On Jan 2, 8:52=A0pm, Jon Kirwan wrote: ...

...

Minor correction - kT/q is about 26mV.

So for every doubling of current the voltage rises by this amount - each decade change will then result in the 60mV you mention (25mV*ln (10)).

kevin

Thanks, Kevin. I should have paid more attention to the d ln(I) in the divisor when writing. I knew this and if I'd bothered to actually calculate kT/q for a moment, I'd have remembered. My laziness. Appreciated.

Jon

The rule of thumb I teach is that for a 1N4148 or small signal bipolar eb junction, it is 0.5 volts at 1 mA rising 0.1 volts for each decade up to two decades up or down. 100 mA? 0.7 volts. 10 uA? 0.3 volts.

It isn't dead accurate, but it is easy for the students to remember.

Jim

the knee in the curve is in the 0.6-0.7 range, at very low currents

0.5V or less can be measured, and near the max working currents it can be more than 1V

bye.

There is no "knee" in a diode curve (until avalanche or failure). If you plot the diode curve on semi-log paper it's pretty straight.

I seem to remember closer to .6 than .5... maybe .57 or .58 at 1mA for the 1N4148. I rounded it up in my mind a while back when I was looking before. I suppose I could go to the bench and check, though. I agree about the 100mV per decade current change. For the bipolars I'm most familiar with as a hobbyist (2n2222 and 2n3904 and 2n3906), liking as I do one penny parts I can hold in my hand rather than tapping them out of a salt shaker by the dozens, is more like .65V at

100uA or so and going up by the usual 60-70mV per decade. So I'd estimate more like .71-.72V at 1mA for them. At a guess. If I weren't snowed in, I'd trapse over to the bench and check. But that's untrustworthy memory, for now.

Basically, our thoughts on the thrust of it are the same. Just that I don't lump the 2n2222 BE junction quantities at specific points to be the same quantities and quantified variations with current with the

1N4148. Mostly because it makes enough of a difference at times when figuring out how much remaining voltage I may have somewhere else to work with, I suppose.

Jon

So despite this curve, diode forward bias can still be a good voltage reference as long as the current is fixed/known right?

Is there a lot of variation between production runs or even within production runs?

I'm thinking now that even my concepts on zeners might need some fine tuning as well. Are zeners as rigidly plateu as I think? Do designers commonly get dripped up by the part of the curve inside the designated current range? Are zeners really as immune to current variation (within their range) as some of us get in the habit of thinking?

Is the plateu for high current zeners flatter or less flat than the plateu for lower current parts?

A resistor makes a pretty good voltage reference as long as the current is known/fixed.

There is variation, sure. Resistors can be easily had that are far more precise.

Depends. Low voltage (under, IIRC 6V, "zeners" are pretty bad. Above that they're actually avalance diodes and can have a pretty low dynamic impedance. If you want even lower, band-gap reference diodes are even flatter, though have limited voltage selections. It depends on your needs.

Commonly? Designers shouldn't. The curves are published. Some even have reasonable Spice models. ;-)

High voltage "zeners" tend to be flatter than low voltage. I've never looked at the threshold voltage vs. current for high vs. low current diodes. I've never used a zener at more than a few tens of milliamps. I tend to use them as references rather than shunt regulators, though I have one (albeit very low current) shunt regulator in my current design.

Zeners have pretty significant series resistances.

The crispness of their 'knees' varies with voltage rating--low voltage types are notoriously weak-kneed.

Cheers, James Arthur

I stand corrected.

I noticed the graphs on the IN914 datasheet agree completely with that, a nice long straight line from nanoamperes to the current limit.

If you go low enough in current will the drop go negative :)

'fair' perhaps I wouldn't go as far as 'good'

the main thing is you get a fair voltage reference with a lousy current source say you give a 1N914 12ma with 5% ripple (so it's varing from about 10mA to 14mA)

across the 1N914 you measure a voltage that varies from 716 to 723 mv so 719mv with 0.7% ripple

not a much as you get from things like temperature changes.

they vary with changing current too, but not to the same degree.

If the low power parts are built as scaled-down versions of the high power parts then performance is comparable, often they aren't and the lowe power zeners perfrom much worse.

[snip]

Except for a rather nasty -2mV/°C temperature coefficient.

Study up on "BandGap" references to get a good handle on semiconductor TC's.

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

If we knew what we were doing, it wouldn\'t be called research...
                    -- Albert Einstein

Yes. A negative current. ;-)