Bandgap Buffer Amp in LL08

[JohnMusselman]

I'm using the LL08 and have some questions about the bandgap buffer amp.  The datasheet seems to hint that the buffer draws a bit of current, but does not specify how much.  My application runs at an average draw in the 10 uA range and I do see a lot more draw if I leave the buffer on all the time.

So I turn the buffer on before making an ADC measurement (once a second) and then turn it off afterwards.  But doing this,  the reading is extremely sensitive to temperature.  I added a small delay after turning on the buffer and before starting the ADC conversion and the problem seems to have gone away, but I'm still testing it.  I assume the buffer needs time to settle.

Any help on this?  What is the bandgap buffer current draw and settling time? I believe the datasheet is mute on this.

John

[bigmac]

Hello, and welcome to the forum.

I assume that your reason for measuring the bandgap voltage is that you do not have a sufficiently regulated ADC reference voltage, and you require the extra measurement to calibrate the other ADC readings.  I would therefore suggest that you measure the bandgap after you have done the other readings.  This will allow maximum settling time for the bandgap.

You might also determine whether there are significant bandgap measurement differences between long and short sample intervals - I would tend towards the long sample interval.  This will again give more settling time during the other measurements.

Yes, the adder current for the bandgap reference does appear to be missing from the datasheet.

Regards,

Mac

[JohnMusselman]

Mac,

Thanks for your reply.

No, the design is a solar powered device running from a supercapacitor.  I am measuring the bandgap voltage in order to determine the system voltage, which can vary.  My testing shows the reading is more stable with a small delay after turning on the bandgap buffer and before starting the ADC, but I still see a little problem.  So I'm concerned how much delay I need to put in to be SURE the buffer is stable.  It's important as we are going into production soon and the first run is 10K pcs. 

This settling time is also not mentioned in the datasheet.

Thanks,

John

[bigmac]

Hello John,

This would be how I would perceive the measurement process -

1. Since the result is needed at one second intervals, I assume you will have no interest in shorter term voltage fluctuations.
2. The ADC reading will be subject to noise fluctuations, especially if you are using 12-bit conversion mode.  To reduce the effect of these fluctuations, a number of ADC readings could be taken, and the results averaged.  Ideally, the readings should be spaced over a 16ms or 20ms interval, to reduce the effect of any induced hum from the mains.  You might consider a set of 8 or 16 readings to determine each result.
3. After you enable the bandgap reference and commence the individual readings, you could determine the point at which two successive readings give either the same result, or a difference of one, after the resolution of the result is reduced sufficiently for the normal noise fluctuations not to have an influence.  When this condition is met, it is likely safe to assume that the bandgap reference voltage is now stable.  The last two readings, and the ones that follow, may now contribute to the averaged result.
4. This process would also give the opportunity to report an abormal level of system noise should the readings not stabilise after a certain number of readings.
5. To minimize system noise during each ADC reading, a low power mode, such as wait or stop3, could be entered immediately following the start of each ADC conversion, and would remain until the next conversion within the batch was due.  This would also have the added benefit of further reducing the average current draw.

Regards,

Mac

[JohnMusselman]

Mac,

Thanks for your help.  Are you with Freescale?  I sent this in as a tech request.

The readings are perfectly stable at room temperature, also at any temperature if the bandgap buffer amp is left on all the time.  I don't have a system noise problem.  However, if I turn on the bandgap buffer and immediately take the reading, the reading is unstable at higher temperatures. 

If I add some delay before the reading, the readings are much better, but some units still show some instability.

I don't think it's a good idea to make multiple readings and assume that if they are close enough then they are right..  I just need to wait for the buffer to stabilize.  But how do I know if I've put in enough delay?  What I really need is for Freescale to characterize the settling time.

So I sent this to them.

If they can't help my only choice is to do a bunch of temperature chamber testing across many units, cross my fingers, and ship product.  And hope it works...

Anyway, thanks for your ideas.

John

[JohnMusselman · Accepted solution]

A follow-up

In case anyone follows this thread, here is the response I received from Freescale:

"I got the answer from the developer team but first please be aware that we have no characterization of the settling time after wakeup for the bandgap or the buffer. The bandgap is likely slower than the buffer due to a filter, but is enabled on wakeup (no code required), so it is probably enabled for a while.

"The recommendation from the developer team is that you use a delay of 20 uS

"On the power side of things the regulator and bandgap draw about 120uA – this is the big draw. The buffer is around 15uA in low power mode. The buffer may start faster in high  power mode, but that mode is around 50uA or so. The bigger consumer of power is clocking and what speed the part is running for the core and adc – the core run power for some number of micro seconds will have a larger impact."