TL331: voltage divider resistor in V+/V

Hello Shu-Cheng,

The only thing to watch out for is the bias currents of the inputs creating a small error across high value resistors.

This is a bipolar device, so there will be 18 to 40nA (higher currents at lower temps) flowing OUT of the inputs. This is shown in Figure 2. This current will cause an extra voltage drop across the input resistance. Larger resistors (megoms) will have more error, and smaller resistors (kohms) will have less error, but at the cost of a higher divider string current.

In your case, the biggest error would be across the ~33k input impedance of the inverting input divider. 33k*40nA=1.3mV. So you may see about a 1.3mV change over temp in the reference voltage due to bias current. You can drop the resistor values by a decade to 10k and 4.99k to further reduce the error contribution, but you will then burn more current in the divider string.

You can attempt to "null" the bias effects by balancing the impedance at the inputs so that the bias currents cancel. To do this, you would make divider resistors similar to the values of the temp sens resistors *at* the threshold temperature. So if you want the trip point at 85C, then the divider resistors should be 6.2k and 1.45k. Doing this, you can scale-up the size of the resistors.

You may also wish to add hysteresis, as temperature is usually a "slow" signal, and you can get "chatter" as the temperature slowly approaches the threshold.

Hi Sirs,

Thanks for your reply.

If our NTC is divided voltage and it's for two comparator input, is there any problem with this? Is there a problem with insufficient current drive capability?

Here is our ideal schematic~~

[LE] 2/11/20 > Schematic clip removed

Hello Shu-Cheng,

You are just doubling the bias current as seen by the resistors, which would double the error. So now you have 36 to 80nA of current, or about 2.6mV max error. As long as you allows for this in the deign margin, you should be fine.

Otherwise, you would want to switch to a CMOS input device where the bias current is negligible. Unfortunately, we do not have a CMOS input device with the same pin out as the TL331. But if you can change your footprint, see the TLV7021.

Hi Sirs,

Thanks for your reply.

We have idea that use one LM2903 to replace two TLV7021, is this feasible? LM2903 Input  is CMOS architecture?

The circuit is drawn on the first page here, mainly the one as below

2/11/20 {LE]> schematic removed    

After the control signal from the NAND Gate, we want to turn off the Load Switch in the figure below.

Our schematic:

2/11/20 {LE]> schematic.pdf removed  

Shu-Cheng

I will review this more closely in the morning but I’m a little concerned about using the lm2903 as shown in the schematic. The biggest concern is that the lm2903 does not have a rail to rail input stage. It is a bipolar input stage. The tlv7021 that you mentioned is cmos and has a RRI. I suggest you look at the tlv7032 which is very close to being released. The datasheet is already available on the web and can be samples. This device is push pull output but you are not level shifting the output, so the push pull output may work for you and save you 2 pull up resistors. If you need open drain the tlv7042 would be what I would recommend. That part is also listed in the datasheet. I’ll review more tomorrow and send you another response. I just wanted to give you another option to consider while you were waiting.

Chuck

Hi Chuck,

Thank you very much.

Please let me know if have any update in morning.

Have a pleasant night!!