schottky tempco

Depends.

A 74S, 74LS or 74ALS series chips may be different from each other and likely different from a Schottky diode. There are also differences between SiC or GaAs or...Schottky diodes.

If for a specific component, look up the datasheet. If you want a general range of temperature coefficients, look up representative components within the families, create a table, and find the mean, median, or mode, as desired (or see if Google really is your friend and if someone has already done this).

Happy hunting.

Richard

I need a diode with a low voltage drop and a known temperature coefficient. I'm going to put it in series with the adjust terminal of a lm317 in constant-current configuration, so the voltage with respect to ground at the output pin of the 317 will be 1.2 volts plus the voltage drop across the schottky (which will have a constant current through it). This way I can tune the circuit to give me the exact dV/C I need. It will be for temperature compensation in a lead-acid battery charging circuit. Since I need about 2000 ppm/C, or maybe a little more, a "regular" silicon diode won't work. It would have 2.2 mV/C out of nearly 2 volts (1.2 + Vf), giving me in the 1000 ppm range. So I need to consider using a schottky.

Apparently temperature coefficient varies with the log of the current according to the shockley equation. If I can determine the parameters (like the ideality factor) for a particular schottky I can get the math worked out and fiddle with the current setting to get the temperature characteristic of the circuit right.

I'm going to set the current at 5 or 10 mA to make sure the 317 works right.

Schottkies tend to run less, -1.5 maybe, except that the very small signal-level diodes have decreasing tc's as the current increases. Some go to zero tc at 10-20 mA. I think the exact tempco depends on the metal used.

How about an LM35? It outputs 10 mV per degree C, so you could scale that as needed. But it won't sink much current, so you'd have to buffer it with an opamp or something.

Don't run an LM35 from over +5 volts! And don't pull the output below ground!

John

Whatever kind of temperature sensing element used, be it a diode, LM35 or whatever, needs to be screwed on to the battery terminal. The LM35 probably comes in some fragile package like a TO-92. On the other hand, I can solder a ring terminal right onto the lead of a diode and it will be rugged enough to screw to the battery terminal, and the diode's lead will provide a good thermal path from the battery into the diode. Ambient temperature sensing really isn't feasible here. The battery enclosure's temperature will undergo temperature variations to extremes, from around freezing to at least 85 C (outdoor machinery). The temp sensing must be through a direct connection to the battery terminal. I think I'm kind of stuck with a diode just for physical reasons. Thanks for the advice, John.

Well, then use a diode, or maybe a TO-220 transistor, which would be better thermally. The tempco of a silicon diode is around -2.5/650 =

-3800 PPM, so mix that with some constant voltage to get the result you need.

John

Oh: National makes a TO-220 LM35!

John

why not NTC or PTC resistors? you can get those with face mountable tabs.

Sure. NTC's have a huge change with temperature, but are pretty nonlinear. There are some simple ways to linearize them over a reasonable temp range.

RTD type ptc's usually run around +4000 ppm. The ceramic ptc's are pretty much too nonlinear to be useful for temp sensing... thy most;y are used for current limiting.

John

John Larkin wrote: (snip)

...or temperature limiting.

Any diode will have about .003 * Vf tempco; Shottky diodes have lower Vf, so lower tempco. If you need something higher than a silicon diode, use two silicon diodes in series, OR make a difference with some fixed voltage, as

(A+ B*T ) - C = result

so as to subtract from the non-temp-dependent A term; B can be any fraction of the result that you want, just chose C correctly...

Schottky diodes aren't specified for their operating conditions under modest currents, so you'll get nothing from their data sheets. Use a transistor instead as a planar diode (connect B and C together) for temperature sensing.

I think if you make the LM317 current source with a fixed resistor, and feed the output current into an R + diode series pair, you'll find the tempco you want is achievable that way.

Yes, that's the way to do it of course.

I have been simulating simple transistor circuits in an attempt to make a lead acid battery float voltage reference that has the proper temperature coefficient.

One way, using a 1 mA current source looks very good. It uses a low current transistor, like a 2N5089, with a resistor divider collector to negative rail, base to the divider node, and a small emitter resistor to provide the negative feedback needed to reduce the tempco by the right amount. Values that simulate a good result are: emitter resistor 508 ohms Base to negative rail resistor 1.3k Collector to base resistor 2.28k

Another version, fed from a +12 volt regulated supply through a 10k resistor (instead of a regulated 1mA source) simulated well with the following values: emitter resistor 411 ohms Base to negative rail resistor 1.3k Collector to base resistor 2.45k

I was trying to match a reference curve of

Brute force, nothing fiddly:

A sensor (either analog out such as LM34/LM35 or digital out) tosses a fairly reliable temperature to a microprocessor, and the microprocessor tosses a fairly reliable reference out a DAC (perhaps needing to be multiplied up, depending on the DAC's abilities). Given all the other stuff that a good battery charger should be doing (switching modes and voltages depending on time of charging, etc), and the present price of microprocessors, it does not appear to be overkill to use one - the job gets done, and variations in batches of parts are expected to be well controlled in dedicated temperature sensing parts.