Is there a cheaper TLV431/LMV431?

[Fritz Schlunder]

Hello Folks,

Ok, the usual dilemma: A uC with ADC is around 40c more than one with
just a comparator. Too much. To obtain a nice VCC-independent reference
for things like poor-man's SD conversion I often employ a TLV431 or
LMV431, mostly pulsed to save power. However, those are around 10c. A
BAV99 is only 1.5c but not quite as precise. One could buy the TL431 for
a song but it wants too much juice to stay in regulation.

So, is there a cheaper 1.24V ref that can live on 100uA cathode current
or less and turns on and off within a few hundred nsec when modulating
the cathode? Or maybe a cheap part with another function that happens to
have an independent reference flying around in its trunk?

And Jim, no, I can't roll my own on silicon here...

I'm not sure what poor-man's "SD" conversion is/stands for, and I'm not sure
what your accuracy/speed/tolerance requirements are, but the PIC12F510
(http://ww1.microchip.com/downloads/en/DeviceDoc/41268B.pdf) may be of
interest. In volume it can get down to $0.56 each and it includes a built
in oscillator, a comparator, an 8 bit ADC, and a 0.6V absolute reference.
Presumably the reference has a tolerance of 0.550V - 0.650V absolute.

The PIC10F220/222 devices also include the 0.6V reference, but it might be a
little tricky to use depending upon what you are trying to do.

 

[James Waldby]

Joerg wrote:
....

It doesn't have to be 1.24V. It just has to be below 2V or so to
accomodate the typical two-cell design that must be allowed to function
down to less than 2.5V.

You might try adapting the following, eg substitute 22K in place of
R2=30K. Note, MMBD7000 is a dual Schottky diode a lot like the BAV99.
U2b doesn't do anything - it was merely simpler to breadboard as shown.

Vx U1a U1b
Vi------R2-----+----->|--->|-------+
| |
R1=10K | |
R2=30K R1 |
| U2a |
Vo----+---------->|-------+---Vg
+----------|<-------+
U1,U2=MMBD7000 U2b

Measured DC voltages, relative to Vg:
Vi Vo Vx
2.74 .456 .850
4.11 .464 .937
4.30 .465 .947
5.04 .468 .975
5.50 .469 .986
12.7 .477 1.099

-jiw

 

[Pooh Bear]

Hello Spehro,


Don't know but some of the SOIC seem to remain in production. Also the
LM431 in SOT23. Lots of brands of the 431 can be had for around 5c at
every street corner. But the guys using TO92 need to probably watch it.

For low power designs there is hardly any way around the TLV/LMV
versions and for some reasons those did not enjoy the bargain pricing
that they bigger brethren reached. Maybe because the non-economy version
is just fine for PC power supplies and thus sells more.

Are you saying the 'standard' 431 is on the way out ? Given the number used in
supplies that seems astonishing.

Graham

 

[Pooh Bear]

Hello Spehro,


Nice price, although with three resistors you'd be back above 5c. But
maybe less than a TLV.

How much do you pay for resistors ?

Graham

 

[Lostgallifreyan]

How much do you pay for resistors ?

Graham

I imagine that when the budget has tolerances like those, even their
mounting pad or hole costs are considered...

 

[Genome]

Hello Folks,

Ok, the usual dilemma: A uC with ADC is around 40c more than one with just
a comparator. Too much. To obtain a nice VCC-independent reference for
things like poor-man's SD conversion
Regards, Joerg

Arrrrrrrrrrrrrr, like wot someone else almost said..... Go ratiometric.

Arrrrrrrrrr. Measure the answer measure the answer, do the sum do the sum.

Arrrrrr

DNA

 

[Tim Shoppa]

Are you saying the 'standard' 431 is on the way out ? Given the number used in
supplies that seems astonishing.

They discontinued the CA3046 last year. As far as I'm concerned, now
ANYTHING is possible.

(Actually National sitll makes their variant in surface mount, but all
the major catalog suppliers had their shelf stock dry up within days of
the CA3046 part being discontinued).

Tim.

 

[Joerg]

Hello Graham,

Are you saying the 'standard' 431 is on the way out ? Given the number used in
supplies that seems astonishing.

No. Only some versions are obsolete. The 431 is here to stay. I'll
probably be in a nursing home before that one retires.

 

[Joerg]

Hello Graham,

How much do you pay for resistors ?

Usually around 0.8c, fully burdened.

 

[Joerg]

Hello James,

You might try adapting the following, eg substitute 22K in place of
R2=30K. Note, MMBD7000 is a dual Schottky diode a lot like the BAV99.
U2b doesn't do anything - it was merely simpler to breadboard as shown.

Vx U1a U1b
Vi------R2-----+----->|--->|-------+
| |
R1=10K | |
R2=30K R1 |
| U2a |
Vo----+---------->|-------+---Vg
+----------|<-------+
U1,U2=MMBD7000 U2b

Measured DC voltages, relative to Vg:
Vi Vo Vx
2.74 .456 .850
4.11 .464 .937
4.30 .465 .947
5.04 .468 .975
5.50 .469 .986
12.7 .477 1.099

Pretty cool solution! A diode pre-regulator. The MMBD7000 can be had for
around 2c a pop in qties (for a dual pack).

In this case it wouldn't work for me though since I have to clamp a uC
port. This in order to provide a rather precise high level for a PWM.
Besides tempco there is the job of muffling Vbat drift. Especially with
alkalines this voltage can range from 2.5V to 3.2V.

 

[Joerg]

Hello Fritz,

I'm not sure what poor-man's "SD" conversion is/stands for, and I'm not sure
what your accuracy/speed/tolerance requirements are, but the PIC12F510
(http://ww1.microchip.com/downloads/en/DeviceDoc/41268B.pdf) may be of
interest. In volume it can get down to $0.56 each and it includes a built
in oscillator, a comparator, an 8 bit ADC, and a 0.6V absolute reference.
Presumably the reference has a tolerance of 0.550V - 0.650V absolute.

One application is a poor man's sigma-delta converter (SD). There you'd
have to compare an input signal to a feedback signal and the only way to
generate that on a cheap uC is PWM. You regulate that PWM signal to
become equal to the input signal. Then you count how many high pulses
you've got in a given window, say 1024 for a 10-bit conversion. This
tells you the magnitude of the input signal. But it only works if you
can be sure about the amplitude of those pulses and that's where the
TLV431 comes in as a clamp.

The PIC10F220/222 devices also include the 0.6V reference, but it might be a
little tricky to use depending upon what you are trying to do.

In many cases I'd need the math horsepower of the 16-bit MSP430.

 

[Joerg]

Hello Genome,

Arrrrrrrrrrrrrr, like wot someone else almost said..... Go ratiometric.

Arrrrrrrrrr. Measure the answer measure the answer, do the sum do the sum.

Arrrrrr

Huh? You still need a reference somewhere in the game. Else the system
cannot know whether something is 1.24V or not.

 

[Fred Bloggs]

The TLV431 can pull that off. Look at figures 15 and 16:

http://focus.ti.com/lit/ds/symlink/tlv431a.pdf

You've got pretty good eyes if you can read to within 12.4mV on that
graph- and that's just 1%.

At low currents you'll have to give it a little "capacitive english" to
get results similar to figure 16.

You will? That's where it turns in its best performance. The 1mA pulsed
waveform is visibly worse- plan on 10's microseconds for 1/2LSB error at
any reasonable resolution.

 

[Mike Monett]

Huh? You still need a reference somewhere in the game. Else the system
cannot know whether something is 1.24V or not.

Regards, Joerg

Hi Joerg,

Genome may have a point. Is there a reason you need the absolute value of
the voltage? Or is there a way to convert it to a ratio of something else?

Say you have a voltage divider. You can generate a PWM voltage and compare
it to the output of the divider. When they are equal, the duty cycle of the
PWM generator is equal to the resistor ratio. You know the value of one
resistor, so a simple calc gives the other. It seems almost anything else
can be done in a similar fashion.

For example, I have several inexpensive digital thermometers that run off a
single watch battery of ~1.5V. Somehow they give accurate readings, and I'm
pretty sure they don't have any voltage references buried inside

Regards,

Mike Monett

 

[Lostgallifreyan]

Hi Joerg,

Genome may have a point. Is there a reason you need the absolute value
of the voltage? Or is there a way to convert it to a ratio of
something else?

Say you have a voltage divider. You can generate a PWM voltage and
compare it to the output of the divider. When they are equal, the duty
cycle of the PWM generator is equal to the resistor ratio. You know
the value of one resistor, so a simple calc gives the other. It seems
almost anything else can be done in a similar fashion.

For example, I have several inexpensive digital thermometers that run
off a single watch battery of ~1.5V. Somehow they give accurate
readings, and I'm pretty sure they don't have any voltage references
buried inside

Regards,

Mike Monett

Interesting. It reminded me af a zener problem I had, using two, and trying
to get accurate relations between them. I solved my problem with one, and
it;s wasn't a reference, merely a means to get some approximate constant,
and I used a divider across it to get my other value, which remained locked
to the zener voltage by proportion, solving all my problems at once.

The point being that any datum will do, so long as you choose it well, and
use it as your sole reference. It certainly works for navigation on tides,
and those are both complex, and with high risk stakes.

 

[Joerg]

Hello Fred,

You've got pretty good eyes if you can read to within 12.4mV on that
graph- and that's just 1%.

That's not how it works in a PWM application. On a micro the typical
pulse widths are several ten usec. The TLV is pretty much hooked up like
in the schematics next to figs 15 and 16, sans 50ohm of course. The
absolute value doesn't have to be within a percent but it needs to get
there really fast, and it needs to be remain the same.

What I was saying here is that the TLV431 ramps surprisingly fast for
such a jelly-bean part.

You will? That's where it turns in its best performance. The 1mA pulsed
waveform is visibly worse- plan on 10's microseconds for 1/2LSB error at
any reasonable resolution.

You just need to push it up the ramp and then ease back.

 

[Joerg]

Hello Mike,

Genome may have a point. Is there a reason you need the absolute value of
the voltage? Or is there a way to convert it to a ratio of something else?

Say you have a voltage divider. You can generate a PWM voltage and compare
it to the output of the divider. When they are equal, the duty cycle of the
PWM generator is equal to the resistor ratio. You know the value of one
resistor, so a simple calc gives the other. It seems almost anything else
can be done in a similar fashion.

Mostly I need the PWM circuit to provide a current or voltage of known
magnitude, or measure it. The resistor ratio won't tell me that.

For example, I have several inexpensive digital thermometers that run off a
single watch battery of ~1.5V. Somehow they give accurate readings, and I'm
pretty sure they don't have any voltage references buried inside

The ones with a diode as a sensor need some kind of reference and the
versions I had seen actually had one. Thermometers with a temp-variable
resistor don't, those can compare with a known resistor in the usual
dual-slope arrangement. In the same way that an ohmmeter does not
necessarily need a reference whereas a voltmeter does.