[471xx] Vref turn on time

Argail said:

The typical turn-on-time is 5ms, but no other informations.

No. What else information do you want/need?

If you switch it on with VCC=3V, it will settle to 1.14...1.26V (1.2V typ) within 5ms(typ)´at a load capacitance of 100nF. If you don't apply an external 100nF capacitor, it will settle significantly faster but will also suffer from much noise, so it isn't recommended.

It is likely that once it had settled for the first time, it may settle significantly faster if you re-enable it. How fast depends on the self-discharge of the 100nF capacitor and the off-time

You can enable external reference and apply a 470nF capacitor. This will amost triple your current consumption during VREF on, but will shorten the typical settle time to 100µs.

Argail said:

is it possible to know that Vref is ok

It's your reference. So what's your reference when judging the reference? Itself? Well, the SD16A will always tell you that the reference has reference voltage :)
If you have a voltage that is 100% stable (no matter how high the absoute value is), you can start sampling it and when the measured value does not change it means that your reference does not change anymore and has settled.
But this means doing excess conversions and therefore more power consumption.

The only way to do it with minimum power is to turn on reference when you need it, go to sleep for >5ms and then do the conversion. And spend some cents more for a higher quality capacitor.

Jens-Michael Gross said:

No. What else information do you want/need?

It isn't possible to know thar "vref" is good, so I must make a "wait loop" to wait Vref is settled. The wait time should be longer than the longest Vref power on, so 5ms is not ok. I have a look to the code example and see that they typically made a 13ms wait loop, the power consumption between a turn on time of 5ms and 13ms is not the same:p and perhaps the maximum power on for 471xx is 8ms, but this information is not available.

Jens-Michael Gross said:

You can enable external reference and apply a 470nF capacitor. This will amost triple your current consumption during VREF on, but will shorten the typical settle time to 100µs.

"I think there is something I don't understand in the SD16_A, on the MSP430F471xx functionnal block it's written that the converter doesn't have the buffer, it isn't this buffer who allows to power on the vref in 100us? If it is possible to set SD16VMIDON, 600uA*100uS <<< 260uA*5mS consumption wise (for vref power on time, and depends from the aplication use)."

Edit: my fault, output buffer and input buffer is not the same thing^^

Argail said:

the power consumption between a turn on time of 5ms and 13ms is not the same:p and perhaps the maximum power on for 471xx is 8ms, but this information is not available.

Indeed. And as I said, maybe the tiem is only a few microseconds if the external capacitor had been arelady charged and not discharged too much since last sample. Then it depends on VCC. The given 5ms are for VCC=3V only. And I guess for discharged capacitor.

if you have, however, an external stable voltage below Vref (such as a diode P/N voltage at low current), you can determine when sample taken with this voltage agianst Vref will no longer change due to changing VRef. But I don't know whether the additional power consumption of this external helper voltage will smaller than the power you save by shortening the waiting time. Maybe feedign it through a port pin only during the wait will help.

there's much room for experiments. And what is an embedded systems developer good for if he only uses preformatted code and pre-existing hardware? :)

Thank you for the time spent to answer me.

We have high dimensionnal restriction, the mechanical designer make funny bumps each time I ask him to add an IC:)  Moreover if I have to add an external Vref, I will select one with known parameters an directly use it.

Jens-Michael Gross said:

And what is an embedded systems developer good for if he only uses preformatted code and pre-existing hardware?

I'll ask my financial director to buy me 1000 IC and let me spend 1 work's week to characterize the maximum turn on time so I can add a single value in my uC comparaison worksheet, he'll probably be happy.

Argail said:

We have high dimensionnal restriction, the mechanical designer make funny bumps each time I ask him to add an IC:)

I know this. Luckily, I do both, hardware and software, so it's up to me to make both fit :)

Argail said:

if I have to add an external Vref, I will select one with known parameters an directly use it.

It adds to the costs and also adds to the power consumption. Also, if you switch the external reference on and off, you'll have to wait for it to settle too.

My proposal was a relatively cheap diode which will have an instantaneous forward voltage which is unknown but constant and fairly below Vref, and a resistor to limit the current to an unknown but constant and faitrly low value when you turn it on by an MSP output pin. All you need to do is to wait until the 'measured' diode voltage does not change anymore, which, since it's a constant value, means that the Vref does not change anymore.

Personally, I prefer using the TS431BCX as external 2.495V reference (+-0.5%), but then, I don't have any tighter power restrictions. And neither the few cents nor the SOT23-package do count.

Argail said:

I'll ask my financial director to buy me 1000 IC and let me spend 1 work's week to characterize the maximum turn on time so I can add a single value in my uC comparaison worksheet, he'll probably be happy.

That's a valid approach. After all, 1000 chips are much cheaper than single ones, and for production, you can reuse them, effectively lowering the production costs :)
We had a similar problem with the RF-Pics where sometimes the calibration values were lost due to overvoltage or soldering heat or whatever. So we needed to determine which static setting will most likely work if the calibration values were gone. Checked a few hundred chips. Saved a few hundred devices.