ADS8028 EXT REF

Hi Robert,

Welcome to the TI E2E forum! Great question!

The internal reference can safely supply 2mA to any external circuitry and remain stable for A-to-D conversions. Above 6mA, you will see a noticable voltage droop as seen in Figure 8. The 2mA value is a safe margain to prevent additional gain error in the ADC measurement resulting from an inaccurate reference. The DC leakage current is the amount of current the ADS8028 reference input will sink while not sampling (or standby).

For an external reference, you will not be sourcing 2mA to the ADS8028. The ADS8028 reference input is a dynamic load that will have an average current that scales with the sampling rate. The average input reference current vs the sampling rate is shown in the following graph:

Now, since the reference input of the ADS8028 is a dynamic load, there will be moments where the reference input current will be larger than the average current. I have measured the largest peak current to be around 0.32mA. Your external reference needs to remain stable even with this dynamic load. Typically, the reference will drive a 10uF cap since the voltage on that cap will not droop more than 1/2 LSB during a conversion.

Hint: While testing, if you can reduce the SCLK (and therefore the sampling rate) and see a performance inprovement in conversion results, then most likely the reference is not remaining stable during the faster conversion. If instead you keep SCLK the same speed and reduce the application of chip select (/CS)  (also reducing the sampling rate)and see an improvement in measurement results, then the input driver to the analog input (AIN0, AIN1, etc) is not charging the ADS8028 input sampling capacitor fast enough at the higher sampling rate.

Let me know if that helps answer your question.

Regards,
Chris

Hi Chris, Thanks for your quick response and comprehensive answer.

Great to see a forum that works so well with TI employees checking it and responding.

This does answer the question very well. The data sheet could benefit from clarifying these points and the addition of the test results and your second graph. The first graph is in the datasheet. This would help other customers.

Thank you very much - Robert

Hi Chris, I need a bit more detail...

I have to supply the 5V reverence to external sensors, this supply can be shorted to GND or +32V. The supply itself is protected.

This supply therefore also has to be the reference for the various ADC devices in the design.

Our external signal inputs are all protected from shorts to +32V so the ADC is safe.

To protect the ADC REF supply I can put (lets say) a 1K resistor in series with the reference to the ADS8028 and a 5V1 or 5V6 zener to protect the REF and a 10uF capacitor with very low ESR (ceramic and / or tant). From your reply, it is clear the REF pin does not take a DC current (other than leakage) but a current related to the internal charge/discharge of the SAR conversion capacitors. You state the max 'peak' observed is in the order of 320uA. I assume this is a pulse of very short duration?

Please could you confirm:

1) The current pulse(s) are very short (how long typically would be helpful).

2) I assume these 'current pulses' are co-incident with the SCLK +ve and / or -ve edges?

Using the suggested resistor&capacitor network is simple and effective solution provided the ripple at REF is under say 0.5bit to maintain accuracy?

Tracking regulators are a complication as would be a unity gain buffer for the ADS8028 REF, we also have to consider errors introduced by offset voltages, drift etc and therefore dont  necessarily provide a 'better' solution without considerable cost.

Many thanks again.

Robert

Hi Robert,

Yes, you are correct. These current pulses are 100-200ns long and they correspond to one of the SCLK edges. I am not sure which edge they correspond to at the moment, but the pulses occur for each bit decision during a conversion.

Regarding your filtering circuit: I recommend removing the 1k resistor in series with the reference input to the ADC. Even though the reference current is dynamic, there will be a DC current to recharge the capacitor as it depletes. I did a simulation to show what happens to the capacitor voltage after many conversions at a 1 MSPS sampling rate:

Notice that the capacitor does not recharge fast enough during continuous conversions, therefore the reference voltage will droop and result, in a gain error in the ADC's measurement results.If you are sampling slower and non-continuously, you may be able to use a small resistor in series with the reference input (which does help to filter noise). By removing this resistor, I am assuming that your reference is able to remain stable with the large (10uF) capacitive load. If your reference does not remain stable with the capacitive load, then we would need to consider different buffering schemes.

Regards,
Chris

P.S. The E2E will be undergoing some maintenance for the next hour and there may be some issues posting to the forum during this time.

Hi Chris, thanks very much. All done for now.

Robert