VL53L1X Calibration Effect

[will3]

I am observing an unexpected reading issue. 

I am using Calibration and I use a physical distance to Calibrate.  However,  an object right in front of the sensor (1-20mm'ish away) reads a value of 0 (zero).  If I change the Calibration to physical distance + 20 (rough) I get a better measurement of the object when its in front -- I get in fact more of a measurement I would expect.  However,  I would need to calibrate at a distance greater than the physical distance -- which is not what I thought one is supposed to do? 

[John E KVAM]

There are 2 calibrations - offset and crosstalk.

If you do not have a coverglass, you do not need to do crosstalk.

The user manual that describes this is hidden in our software release. (But if you don't use ST's software you would not have found it. Not your fault.) It's UM2931. Find it on ST.com

if all your distances are 2cm short, then you have an offset of 2cm. 

You can simply add 2cm, or you can set the offset register by making call in our API. (Or just copy that one function into your code.)

if you have a coverglass, you are going to need to do the crosstalk cal. after doing the offset on a near, bright target, find a dull far target that is 10 or 20% short. 

Then call our crosstalk correction function. 

it estimates how many photons are reflected by your coverglass.

[will3]

John,

Thank you.  I understand.  We do not have a cover glass.

What I am struggling with is the sensor's ability to measure close objects.  I am seeing odd readings.  

Is the vl53l1x able to read objects in let's say less than 10mm?

Thanks again,

Will

[John E KVAM · Accepted solution]

If you want a sensor to see 4M like the VL53L1X can, you have to give up a little in the very short distances.

If you switch sensors to the VL53L4CD you can do a bit better. I can get good results down to 5mm, but below that one only knows 'very near'. It's linear, but not really accurate at the very close distances. 

The best compromise between the two sensors is the VL53L4CX. Distance to 6m, linear down to 1cm. 

The L4 line has a different VCSEL (laser). Narrower FoV at 18 degrees. But it's single mode makes it a touch better at the very near distances.

Both are pin compatible with the L1 line - but slightly smaller and less tall. (Cheaper too.)

I'd invest in the P-Nucleo-53L4A2. (This is the eval kit for the VL53L4CX) Cost is $56. 

Try the L4CX. See if it works using the software from ST.com. You will know in a few minutes if it works for you.

Then download and run the software for the VL53L4CD on the same board. It will work. See if the much easier software works. If it does, then go with the L4CD - slightly less expensive - but the max distance is 1.3M

- john

[will3]

John,

As usual, thank you so much for your quick response :)

We are somewhat invested already in the vl53l1x.  What would you say is the most trustworthy shortest distance for this sensor?

Thanks again,

Will

[John E KVAM]

the problem in linearity. below about 2cm the Numbers start going up! So if you get a distance of 3cm you don't know if it's a real 3cm or 1cm. 

If you can guarantee that nothing will go below 2cm, then you can trust any number greater than 2cm.

I generally suggest that if you can recess your sensor 2cm deep into your structure it works well for you. 

But consider the VL53L4CX. 6M max, good linearity down to 10mm. Same pin out as the L1 you have. 

Software almost identical. Differences are 18 degree FoV and 0.5mm shorter. 

- john