In a recent press release, Analog Devices featured a new 3D ToF camera so now seems a good time to cover the IEC 61496 series of human presence detection standards as part of my 2022 focus on functional safety of robots. For a review of the ADI 3D ToF camera listed to the July 2nd episode of the robotics report podcast at time 1:14, the camera specs can be found here.
Figure 1 - ADTF3175 3D ToF camera from Analog Devices
While 3D ToF cameras have lots of use in non-safety applications for robots, cobots and mobots, such as for box dimensioning they could also be very useful for safety and especially to implement SSM (speed and separation monitoring). Traditionally SSM was implemented using laser scanners and the relevant functional safety standard for laser scanners are given in international consensus standard IEC 61496-3. However, since 2018 IEC 61496-3 also includes 3D ToF opening the possibility to us this new and interesting technology in robot and mobot safety systems. In a future blog I will cover functional safety for 3D ToF systems but today I will concentrate on the IEC 61496 series of standards which was developed by IEC TC 44 which also has responsibility for the IEC 62061 machine safety standard.
The IEC 61496 series is a multi-part standard covering ESPE (electro sensitive protective equipment). Specifically, such sensors are for human presence detection as opposed to detecting objects. Parts of the IEC 61496 series include
- IEC 61496-1 – covers the common requirements for all ESPE and was revised in 2020
- IEC 61496-2 – covers AOPD (Active optoelectronic protective devices - light curtains)
- IEC 61496-3 – covers AOPDDR (Active optoelectronic protective devices responsive to diffuse reflection e.g. laser scanners)
- IEC TS 61496-4-2 – covers reference pattern techniques – project a pattern of light and check for distortions in the field to detect if people are present
- IEC TS 61496-4 -3 – covers stereo cameras
- IEC 61496-5 – in development to cover radar devices
The application of the IEC 61496 series is covered by IEC TS 62046.
As far as I can make out only part 1 is a harmonized EU standard but from the list available here it may longer be so since 2020.
Standards which refer to IEC 61496 for their human presence detection requirements include:
- ISO 3691-4 for industrial trucks (see 4.1.10)
- ANSI R15.08 – Industrial mobile robots standard (see 5.1.8)
- ISO 10218-2 – industrial robot safety standard (see 5.10.5.1)
- ISO 12100 – basic safety standard for machinery (multiple references)
- ISO 13482 - Service robot standard (see 6.5.2)
Typically call outs of the IEC 61496 series look something like the below.
Figure 2 - excerpt from ISO 10218-2:2011 5.10.5.1
Like all standards there are some things in the IEC 61496 series I like and there are some things I don’t like. Let’s start with some things I don’t like.
Firstly, it includes the idea of “types”. Sensors compliant to this standard can be type 1, type 2, type 3 or type 4 with the higher types corresponding to a higher level of safety. While types include some aspects of systematic capability, they also include performance and fault tolerance requirements.
The standard gives rules on which type is suitable for a safety function of a given SIL or PL and conversely what SIL or PL is required for the design of the control system of an ESPE.
For instance, if you are implementing a SIL 2 and/or PL d safety function table 2 of IEC 61496 tells you that you need a type 3 safety sensor. Type 2 is good enough for SIL 1 and you need type 4 for SIL 3/PL e. Conversely if you are building a type 3 human presence sensor table 1 of IEC 61496 tells you that your control systems need to be developed to SIL 1/PL c for type 1, SIL 2/PL d for type 2 as shown below. However as we will see later the strict fault tolerance requirements mean that PL d must be implemented with at least a CAT 3 architecture.
Figure 3 types and required safety performance according to IEC 61496-1 table 1
Not every type of sensor is available in every type. So that sensors according to IEC 61496-3 are only available in type 3 meaning they can only be used up to PL d, SIL 2. If you risk assessment says you need PL e, SIL 3 then you need a light curtain which is available in type 4, if you continue to implement your safety according to the IEC 61496 series.
The use of types seems unnecessary to be, could the standard not just have used SIL or PL. We have enough terminology already.
However, my bigger objection is that each type also comes with fault tolerance requirements. The wording is very similar to the category descriptions from ISO 13849 but just enough different to be troublesome.
Figure 4 - fault tolerance requirements by type from IEC 62046
Below is a comparison of the type 3 description and a CAT 3 description from ISO 13849. I invite you to compare and contrast and draw your own conclusions. For me it is effectively insisting on ISO 13849 categories at a time when most other standards are dropping them as being design restrictive. In fact the last line below from the type 3 description seems more like CAT 4 than CAT 3.
Figure 5 - comparing type 3 with CAT 3
Moving to things from the IEC 61496 series I like. One of the nice things about the IEC 61496 series is that it gives you a very nice list of performance tests for your chosen sensor.
Figure 6 - an extract from table 1 of IEC 61496-3
The above extract shows that testing must be done with a black test piece, a white test piece and a retro-reflective test piece and the positions at which the testing must be done. Other tests specified in IEC 61496 include background interference tests, multi-path reflection tests, light interference tests. This test list seems comprehensive and offers a good way to compare sensors and ensure your sensor meets a minimum capability.
Looking at IEC 61496-3 more closely – its scope includes indoor use only and it is limited to sensors using radiation in the range 820nm to 950nm which is interesting given that many automotive sensors I believe are up around 1400nm. Its black test piece is specified as being for a 1.8% coefficient of diffuse reflection. This value harks back to the roots of IEC 61496-3 being to cover laser scanners which measure only on a single plane. When measuring in a single plane you may have to deal with black wellington boots, black corduroy pants or black shoes. No other part of the person may be visible to the laser scanner. This value is unhelpful for 3D ToF cameras which feature whole body detection with a FOV (field of view) often exceeding 100 degrees in both dimensions. It is unlikely that someone is on your factory floor dressed all in black from which they have managed to keep all dust and other higher reflectivity contaminants. Other standards such as ISO 3691-4 allow the use of a 5% coefficient of diffuse reflection for whole body detection, and this seems more reasonable for 3D ToF.
If one of the parts of IEC 61496 doesn’t cover your technology or you need to do something out of scope of IEC 61496-3 (e.g. outdoor use) then the newly developed IEC 62998 series offers an alternative route to using the IEC 61496 series and I will try to cover the IEC 62998 series in a future blog.
I started this blog with a mention of the new ADI solution for 3D ToF cameras, the ADTF3175 datasheet can be seen here. This specific instantiation of the design, including the new sensor chip does not meet the IEC 61496 requirements and would probably need more VCSEL to increase the illumination to detect very black objects at more than a couple of meters. Who knows what the future will bring!
All blogs in this safety matters series can be found here.
