BQ25892: Stops ADC conversions on occasion

Hello Trevor,

Thanks for reaching out. I'll review this and follow up with you tomorrow, 3/13.

Best regards,

Alec

Hello Trevor,

Thanks for sharing your schematic and register dump.

You are correct that when the ADC is in continuous conversion mode, the CONV_START bit updates to indicate when a conversion is in progress. It sounds like most of the time you are able to use this behavior to log ADC readings from the device, but then sometimes the CONV_START bit stays 1, meaning you are unable to log the ADC readings. Is my understanding of the problem correct? This behavior seems unexpected.

You also noted that this issue occurs with different batteries. While I don't expect the battery itself to be the root cause, battery voltage can certainly impact behavior. As mentioned in section 8.2.8 of the datasheet, ADC measurements are only active when VBAT > SYS_MIN. From the provided register dump, it appears that your SYS_MIN setting is 3.5V and your battery voltage is read to be around 3.7V, but is this always the case?

How frequently are you polling CONV_START? I'm wondering if this could be software-related. It's possible that CONV_START is updating as expected, but you're polling at a time that makes it appear to be stuck at 1.

You also mentioned that removing the input source or battery resolves the issue. Could you clarify whether the issue only occurs when both an input source and battery are connected? Or does the issue persist until the input source is removed and reconnected? What is the input voltage and battery voltage when this issue occurs?

Best regards,

Alec

Hello Trevor,

I wanted to check if you still need help with this issue. Let me know if you have any updates or further questions.

Best regards,

Alec

Hi Alec,

Thank you for taking the time to review my issue.. and apologies for the delayed reply..

I'll try to answer your questions in order..

1. Yes your understanding is correct, we log ADC readings from the BQ25892 and yes sometimes the CONV_START bit remains '1' meaning we can't log readings anymore. This does not recover until either VBUS or the battery is removed.

2. The battery voltage at which I have seen this issue varies from ~3.7-4.0v.

Our application requires the battery to work from 3.5v to 4.2v..

I should also add, that when VBUS is present we set the SYSMIN to 3.5v and when VBUS is removed we set it to 3.0v.

3. We poll CONV_START every second.You make an interesting point with regard to polling, If I'm understanding you correctly you are saying that the CONV_START will set and clear itself on each ADC conversation cycle, regardless of whether the bit is read or not..If this is the case there certainly is a hole in our logic there.. 

So I will need to fix this in any case, even if it's not the root cause issue.. 

I will retest and let you know my findings 

4. The issue only occurs when both VBUS and the battery is connected, if either is disconnected the problem goes away..

When this happens VBUS is 4.8v and the battery voltage is 3.7-4.0V

I am out of office for the next 10days but as soon as I am back I will retest with a fix for point 3 and let you know my findings.

Regards

Trevor

Hi Trevor,

Thanks for the information. We can continue when you're back.

Best regards,

Alec

Hi Alec,

I'm back looking into this issue again and I implemented a fix for point 3 above.

I changed the code to perform one shot ADC conversions (as described in DS snippet below) i.e. the CONV_RATE remains cleared and the code sets the CONV_START bit to begin a conversion and then polls the CONV_START bit once a second until it clears indicating the conversion has completed.
The registers are read and parsed and then the process begins again.

Unfortunately this does not fix the issue I have been seeing.. 

Hi Trevor,

Thanks for reaching out.

The conversion time (tconv) is 8ms (typical), but I see that the datasheet description mentions the 1s maximum.

It's likely that polling every second is too slow to catch the CONV_START bit clearing if it's being set again.

Best regards,

Alec

Hi Alec, 

Thanks for your response 

I have changed the code to perform single shot ADC conversions it is my understanding that when using this method when I set the CONV_START bit to start the conversion it will clear once the conversion has completed no other conversions will be performed until I explicitly set the bit again in this case it shouldn't matter how often I poll I just need to wait for the bit too clear right? 

Regards 

Trevor 

Thanks for confirming, Alec.

Unfortunately, this doesn't seem to have any effect on the issue I'm experiencing.

To recap:

I currently have a battery connected to the system along with VBUS, and in this setup, ADC conversions appear to stop, with the CONV_START bit remaining set. However, if I remove the battery, the ADC conversions resume. Similarly, if I disconnect VBUS, the conversions start working again. But when both are connected, the issue persists.

Do you know of any other possible reasons why ADC conversions might stop with CONV_START remaining set under these conditions?

Again thanks for you help so far

Regards

Trevor

Hi Trevor,

The CONV_START bit stays high during input source detection, so it's possible the behavior that you're seeing is related to that.

It's helpful to know that this issue only occurs while VBUS and BAT are connected. I double-checked your register dump and schematic, but I don't see any obvious issues.

I'll investigate this further and try to recreate this behavior on an EVM.

Let me know if there are any other details that might be important to know. I appreciate everything you've shared so far.

Best regards,

Alec

Hi Alec,

I took some traces of VBUS, VSYS and VBAT when in the condition of the CONV_START bit being stuck..


So here is the trace when CONV_START is set and remains set. 
There's a dip on both VSYS and VBAT every ~43mS. I'm sure if this helps at all or is even related 

and for comparision here it is normally

Thanks again for your help so far.. Have a nice weekend!

Regards

Trevor

Hi Alec,

It seems the battery that was exhibiting the issue was making poor contact with the onboard connector. Once the connection was properly secured, it started working again.

From my previous experience with ADC measurements, the sampling should simply capture the input voltage at the moment of sampling — the shape of the input signal shouldn't matter. The behavior you're describing is unexpected and a bit concerning, as it suggests a battery could be connected but not making proper contact, with no easy way to detect this condition.

One possible approach to monitor this would be to implement a guard timer in the code. The timer could be reset at each ADC interval, and if the ADC stops converting for an extended period, a fault could be raised.

I am interested on your thoughts on this

Regards

Trevor

Hi Trevor,

The ADC in the BQ25892 is designed for high-accuracy measurements of slowly varying DC signals, as commonly found in battery charger applications. Because of this, there are inherent trade-offs when the input signal is rapidly changing or unstable.

This article explains this in more detail: How delta-sigma ADCs work, Part 1

Your idea sounds like a reasonable software solution for flagging potential loose battery connections. Are loose connections something you expect could occur in your application? And just to confirm, are you only seeing the ADC stop converting with this particular battery?

Best regards,

Alec

Thanks, Alec.

Loose connectors wouldn’t normally be expected—the connector linking the battery to the main PCB is typically a tight fit and isn't meant to be unplugged frequently in our application. However, since I used this battery for testing and was repeatedly plugging and unplugging it, it’s possible that I may have slightly loosened the terminals. As a result, the connection might not be as solid as usual.

So far, I’ve only observed this issue with this particular battery.

Regards
Trevor

Thanks Alec.. 

Regards