[SOLVED] TL431 fan-out with multiple op amps?

[d123]

Hi,

Is it okay to do this: Use one TL431 as the reference voltage for each of the non-inverting inputs in a quad op amp, if all the LEDs are meant to have a constant current of 8 - 10mA?

(The simulator shows the quad op amp in question as individual parts, reason for each needing connection to V+ for simulation to work.)



I ask as it looks a nice, easy less-parts solution, that works in a simulation of 10 seconds, or across a wide temp range, but something makes me think that in practice it may not be a good idea to have four op amps pulling on one reference.

As a sidenote, the simulation works, but erroneously tells me that the whole circuit would use 126mA rather than the ~50mA - 60mA I'd expect, and that even just one quarter of the LMC6464 + one NMOS + 1 LED drawing 8mA would use 23mA on the op amp supply pin, I know it isn't so as I've checked that out with real components and at most it draws ~1mA (15mA = 7805, 2mA; LED, 10mA; TL431, not even 1mA; LMC6464 supply pin, not even 1mA).

Actual circuit is 1 x LM7805, 4 x logic level ZVN4206 (~1.5 Ohms RDSon for what I'm doing), that quad op amp, those resistor and capacitor values - but for blue LEDs will more likely be R: 150 - 180 Ohms, and standard 2 x 10mm clear red LEDS, 2 x 10mm clear blue LEDS.

 

[betwixt]

I see no problem. The TL431 is rated to sink up to 100mA although I would never run one at anything like that current myself. The current it can supply (it isn't fan-out, that refers to digital circuits) is controlled by R11, not the TL431 but as the op-amps input current will be less than 0.2uA each, 11.4K should be fine as long as Vs is high enough for about 1mA to flow through R11.


Ignoring the current drawn by the ICs themselves, which is very small, each stage should try to maintain ~2.5V across the 220 Ohm source resistor, that means ~11mA should flow. The total should therefore be a little higher than (4 * 11mA) + TL431 current or somewhere between 45 and 50 mA.

Brian.

 

[asdf44]

I don't see any problems. The input pins arn't going to draw much and if its a quad part all the opamps are obviously close to one another so layout shouldn't be too bad.

It's a good flexible circuit but there are a lot of ways to do this. For example the TL431 is quite cheap and small so there wouldn't have been much cost or size in having four of them - one each doing the job as both reference and feedback (google 'TL431 current sink'). This might be beneficial if you wanted the LED's spaced farther apart.

Also at 11 mA you could consider ditching the mosfets and let the opamp directly source current into the LED's assuming the voltage isn't excessively high (and if it is, change so the LED's are in series).

I see you commented on the mosfet Rdson but in this application it will be in the linear region so Rdson won't be a factor. If you do keep the mosfet note that you can probably get away with a smaller package like SOT-23.

Finally there are 1.25V variants of the TL431 which would reduce the shunt voltage which would be particularly helpful if it meant two LED's could run in series.

 

[KlausST]

Hi,

I agree with the above.
But I recommend to add a resistor between shunt and feedback capacitor.

In your circuit the Opamp drives a 100nF load in series with the 220 Ohms shunt.
I've not checked Opamp datasheet, but it is a bit low impedance at high frequencies.

10k and the feedback capacitor 2n2 gives the same time constant of 22us.

Klaus

 

[betwixt]

Klaus is right, I didn't consider the feedback issues.
The supply voltage isn't specified but there is no advantage to using MOSFETS in this application and if bipolar NPN transistors were used instead, the circuit would work on lower voltages. Essentially, the op-amp output would only have to lift by Vbe instead of the higher Vgs needed by FETs.

Incidentally, the 2N6755 is a particularly bad choice in this circuit, it has as much as 4mA leakage and 4V gate threshold. It is a big metal can power device rated at 60V/14A and 75W dissipation. You could probably use a small signal bipolar transistor (2N2222 for example) to cut costs and size considerably.

Brian.

 

[KlausST]

Hi,

maybe BSS138 or FDV303N... about 1.5V V_GS_th.

I usually prefer mosfets to bipolar, because of accuracy (gate current vs base current error). But maybe accuracy is not that important here.

Klaus

 

[d123]

Hi,

Good to know that at least it's a usable circuit, thanks.

Hi asdf44,
I simulated with the TL431 as the comparator, as you mention, but being fussy, the temperature performance wasn't shown to be as good as with the op amp. Thanks for all the other helpful insights, too.

Hi, Klaus,
:roll: ...still having problems with feedback component values so I put the 100nF there, have to admit 100nF looks/looked questionable, I'll try out your recommendation, thanks.

Hi Brian,
Thanks for clarifying that's it's okay to use. The simulator doesn't have the ZVN4206 (VGSth 1 - 3V max), so I use the first NMOS in the drop-down list of parts which is the 2N6755, I mentioned which transistor I am using in the description in the first thread, interesting to know about the leakage.
Thanks for clarifying the total current for the circuit, the DMM tells me the same for all four LEDS, etc, but the simulation was playing head games with me and wanted to check it wasn't trying to tell me something I wasn't understanding. Full circuit draws about 54mA, 10mA per LED, and another 5 of which are used by the 7805, and the missing 9mA would be bits and pieces in resistors, the TL431, etc., that are too low for this DMM to measure.

This is only a "mucky" circuit to see if a couple or so of red and blue LEDs (which I'm betting are not the right wavelength, anyway!) do influence seedling growth, so precision is not so much of an issue, more a degree of constancy in the LED current across a normal temp range and a few hours a day.

Thanks very much for all the helpful advice and recommendations.

 

[d123]

Hi,

I've finished the slackly designed LED circuit, I just wanted a quickie to see if my cat's Catnip seedlings show any improvement in growth as they have to be indoors at this time of year, and this place is gloomy and cold, but whilst putting it together, in the process it's brought up a couple of questions about transformer-based linear supplies, that I'd prefer to post in a new thread...

I know there are mistakes in the final design, next time would probably be an adjustable 1.22V 4040/4041 dedicated to each LED colours Vf to get the right feedback signal to match what's left over after the LED's Vf (the blue ones use about 3V), so probably need to bump up the supply to 6V.

Anyway, here's the schematic, and some photos of "bungling Dan's" handiwork, pretty woeful effort, as can be seen 4 LEDs maketh not the growlight, but a makeshift prototype is that, I suppose. Thanks for the help, and Merry Christmas if you celebrate it

 

[betwixt]

That looks very nice. :smile:

Do bear in mind that a constant current generator is, within limits, immune to voltage variations so you can probably leave the voltage regulator out altogether and that will let you use the other LEDs without further modification. Worth noting that most 'super bright' LEDs are based on phosphor technology and their Vf tends to be much more uniform across the color spectrum than with older LEDs.

Have a great Christmas!

Brian.