TL431 resistor selection

[(*steve*)]

A colleague asked me yesterday about resistor selection for the TL431.

His concern was that the datasheet specifies a ratio of resistances for setting the reference voltage, but does not comment on a suitable range of values.

My advice to him was that given the 1.5uA to 2uA Iref, the voltage divider should have a current 50 to 100 times this (0.1 to 0.2 mA).

I realise that as the chance it Iref is mostly determined by the chip temperature, higher current will result in less variation of reference voltage with ambient temperature is self heating.

Does anyone have any rules of thumb they employ with devices like this?

Alternatively, is there anything in the specs that I have failed to read, or any application notes on this device that deal with this issue?

 

[trobbins]

Steve - your advise seems fine, given nothing else comes in to play. The only general issue typically arising with TL431 is to keep regulation current above 1mA, but I think that is spelled out in the app notes.

 

[NickS]

After looking over the datasheet my impression was that the Iref range was 2uA to 4uA from pages 6&7 (I am not sure where you see a range 1.2 to 2uA). But that is not so important.

Regardless I too like your advice. I would add the obvious that the trade off for higher resistors would be that the threshold voltage shifts up. So if he needs the threshold tolerance to be tight then go with the 200uA but if it is a loose detection level than save power and go with the 100uA.

And of course the division factor matters here too if you are dividing too 75% of the input then your threshold will change less than if you are dividing to 25% of the input.

 

[(*steve*)]

After looking over the datasheet my impression was that the Iref range was 2uA to 4uA from pages 6&7 (I am not sure where you see a range 1.2 to 2uA). But that is not so important.

Different versions of the 431 have slightly different ranges of Iref. I was quoting from the specs for the device he has (which I couldn't quickly find on the web). The one I linked to is similar (but obviously not identical).

Regardless I too like your advice. I would add the obvious that the trade off for higher resistors would be that the threshold voltage shifts up. So if he needs the threshold tolerance to be tight then go with the 200uA but if it is a loose detection level than save power and go with the 100uA.

And of course the division factor matters here too if you are dividing too 75% of the input then your threshold will change less than if you are dividing to 25% of the input.

Thanks, looks like it was relatively safe advice. He plans on having a minimum current of 1mA for the device, plus what is required for the divider, plus a bit for luck

 

[(*steve*)]

Steve - your advise seems fine, given nothing else comes in to play. The only general issue typically arising with TL431 is to keep regulation current above 1mA, but I think that is spelled out in the app notes.

Curiously enough, he found some specs suggesting his device could get away with under 1mA. However he is the sort to test stuff out first, and found that the regulation was rubbish under 1mA.

I think it is a hazard of relying on spec sheets that are for similar, rather than the exact same part.

Unfortunately, if you purchase a small number of SMT parts, you may not actually be able to verify who manufactured them...

 

[trobbins]

1mA is the worst-case design level for normal regulation characteristic (in the Fairchild spec sheet I have). Going below 1mA may be acceptable - as you say, due to device variation (the typ level is 0.45mA) - but would also be dependant on application, as use in a closed loop system may cause unwanted problems due to significant gain change in the feedback leg.