LM4132-Q1: LM4132-3.3V reference produces up to 4.0V at the output during power-up!

Hi Kai,

This is an interesting behavior. I need to test this on the bench after the holidays to see if I can replicate it. The only thing I can recommend at the moment is to have a larger Cin cap and to make sure Cin > Cload as per section 8.2.1.2.2.
Can you tell me more about your application?
Do you have a schematic?
Is this an automotive application?

-Marcoo

Thanks for the fast response, dear Marcoo!

Right now the circuit exists only on the bench. The output of a simple LM7805, powered by a mains transformer, rectifier and storage cap is connected to the "Vin" input of LM4132-Q1-3.3V. The "EN" input is directly connected to "Vin". At the "Vref" output sits a pure resistive load which pulls no more than 20mA. I tested this scheme with all loads from 0...20mA and all kinds of Cin/Cload combinations, strictly following the datasheet recommendations. I even checked several different LM4132. But always the same behaviour: Randomly occuring overshoots at the output of LM4132, up to 4,0V.

There is a datasheet of the competing reference ADR3525 from Analog Devices. On page 18 of

www.analog.com/.../ADR3525_3530_3533_3540_3550.pdf

there's an interesting note about "Vin slew rate considerations" reporting on overshoots and transient anomalies during slow rising input voltages. I guess I experience something very similar...

Below you see a scope plot I have prepared for you. The upper line shows the input voltage of LM4132, the lower line the "Vref" output. As I told before, the output voltage of LM4132 seems to be glued to the input voltage and rises and rises and rises, but never exceeds 4.0V. There seems to be an internal zener clamp... 

The reason why I took the Q1-version of LM4132 is because we intended to pull up to 20mA out of the reference. So, we wanted to have a bit headroom. Today I know, that our load is drawing only 15mA maximum so that we could take the REF1933, if it's better suited...

This application with the LM4132 is not automotive but industrial sensoring.

Kai

Or see section 10.2.2.3 in the datasheet of TI's REF3433:

10.2.2.3 VIN Slew Rate Considerations
In applications with slow-rising input voltage signals, the reference exhibits overshoot or other transient
anomalies that appear on the output. These phenomena also appear during shutdown as the internal circuitry
loses power.
To avoid such conditions, ensure that the input voltage wave-form has both a rising and falling slew rate close to
6 V/ms.

Kai

In the meantime I have made some meaurements on the REF3033. With slow-rising input voltages there's no overshoot at output, but with fast-rising input voltages in combination with little load capacitance there IS, as is mentioned in the datasheet. Even with 100nF_X7R an overshoot is produced:

But the overshoot disappears (<50mV) with 2µ2_X7R_0805:

In the following plots you can see the behaviour with slow-rising input voltages, again with 2µ2_X7R_0805:

No overshoot! No sticking of output signal to the input signal as it can be seen with the LM4132!

(In order to check the behaviour with as many power-ons as possible, I didn't wait for the input voltage to settle to 0V before each power-on. But this makes no difference.)

Hi Kai,

I haven't go to test these parts in the lab yet for overshoot but it sounds like you found a good replacement. Most families of devices such as the LM4132 and the REF30xx are different architecture so it makes sense they have different overshoots. Is there a particular accuracy/temp co requirement you want in addition to 15mA?

-Marcoo

Hi Marcoo,

thanks for your answer!

Initial accuracy is not so much an issue, but temp co is. 60ppm/°C maximum for the REF3033 is a bit high. So I want to check the REF3133 and REF1933 too.

In the following plot you can see the behaviour of REF3133 with fast-rising input voltage. At output sits a load capacitance of  2µ2_X7R_0805:

There's also an overshoot of 3,8V...3,9V! And there's a turn-on delay too, as mentioned in the datasheet. Interesting to note, that the overshoot disappears when the load capacitance becomes zero. This is also mentioned in the datasheet, though.

Now I want to check the REF1933...

Kai

So, here are the measurements on the REF1933. The settling behaviour is ideal, as can be seen in the following plots:

Fast-rising input voltage:

And slow-rising input voltage:

All plots were made with a load capacitance of 2µ2_X7R_0805.

Interesting to note: There's no output voltage up to 1.8V input voltage! And, at 2.0V there's an extreme peak in the current consumption of REF1933 of 2.8mA, which is seven times higher than the normal supply current at standard input voltage! This behaviour is ultra sharp and abrupt and can cause instability in certain applications (e.g. supply with high source impedance). This anomaly isn't mentioned in the datasheet.

REF3033 and REF3133 do also have this current consumption anomaly, but in a much softer form because the increase is only about 100µA. Datasheet of REF3133 shows a plot of this behaviour.

Kai

Hi Kai,

My testing agrees with yours, the LM4132 does have an overshoot that can be minimized with capacitors but it will still exists. I would say the REF1933 and REF2033 do not have the overshoot just like you tested.
The 2.8mA of the REF1933 could be from the output load if the output load is on and trying to drain all the power.

Is it an option to put a RC filter + buffer at the output to minimize the overshoot?

-Marcoo