Modern ABS diagnostics are evolving far beyond simple pulse counting. While we’ve long mastered Inductive and Hall Effect sensors, the AK Protocol wheel speed sensor presents a new frontier in automotive serial data. These 2-wire MRE sensors don’t just transmit speed; they deliver 9 bits of critical data, including direction, air gap reserve and internal error flags.
In this guide, we will master AK protocol decoding with the PicoScope oscilloscope to bridge the gap between raw voltage and meaningful data. We’ll cover how to avoid common shark fin measurement errors, apply precise Math channels for frequency graphing and master Manchester decoding to see exactly what the ABS ECU sees.
Looking back over the years, we have the following wheel speed sensors covered in our Guided Test menu here https://www.picoauto.com/library/automotive-guided-tests/sensors/
- Inductive (Passive)
- Hall effect (Active)
- Magneto-resistive element [MRE] (Active, including direction of rotation)
- MRE (Active, including PWM to indicate direction of rotation & magnetic field strength)
Following on from the above, it is hard to believe the “latest and greatest” active MRE / Hall effect speed sensor capable of transmitting even more data (utilising the AK Protocol) has been around since 2008!
Analyzing AK Protocol Signals: Real-World PicoScope Oscilloscope Captures #
Above is a capture taken from a Mercedes A Class with the road wheel initially stationary and then rotated by hand.
- Starting with the Scope 1 viewport, even though the road wheel is stationary, our wheel speed sensor is still transmitting data every 147 ms. (See zoomed view in Scope 2 viewport). Here we can see the appearance of 9 data bits (0-8) after the initial (wider) speed pulse, which are delivering information to the ABS ECU and a clear message to say I’m live, online, healthy and no errors
- Moving now to the Scope 3 viewport, we can see how the initial speed pulse has dramatically increased in amplitude to denote wheel rotation and speed (See zoomed view in the Scope 4 viewport). Note how the frequency of the speed pulses increases (becomes closer together) as the wheel speed increases.
Can we Still Graph the Speed of These Sensors with a Math Channel? #
Yes, we can with a little modification to the Math channel formula, as we only need to focus on the frequency of the first speed pulse and not the data bits.
In the image above, we can see our data bits peak at 60 mV, yet our speed pulse midpoint equates to 110 mV. Given PicoScope graphs frequency based on the rate at which the signal crosses 0 V, we need to lower the waveform by 110 mV such that only the speed pulse crosses the 0 V line (Refer to the pink reference waveform above, which has been lowered by 110 mV).
Using the Math channel freq(A-0.11), the signal on channel A is lowered by 110 mV (0.11 V), allowing the speed pulse alone to cross the 0 V line, enabling PicoScope to graph just the frequency of the speed pulse and so the road wheel.
AK Protocol Explained: Analyzing the 9-Bit Serial Data Packet #
This is dependent on several factors, ranging from the manufacturer of the sensor (Infineon or Allegro) to proprietary bits assigned by the vehicle manufacturer.
If you look at the serial data list for a wheel speed sensor utilising the AK protocol, you may not only see wheel speed, but direction of rotation, validity of direction, sensor air gap value, validity of air gap, magnetic field strength, mechanical error flag and calibration data. (All this from a 2-wire wheel speed sensor)
The more you research the AK Protocol, the more concise data you will find about the assignment of each bit for the sensor under test (the image below serves as an example only)
Summary of the Standstill Bits (0–8) #
Can We Decode AK Protocol? Mastering Serial Data in 2-Wire Sensors #
The answer is yes, but with an understanding of Manchester encoding/decoding, as all is not what it seems when we look at the relationship between voltage and logic levels.
Typically, with a differential bus (like CAN), we can say that when the voltage is high, it is dominant, and when it is low, it is recessive. With Manchester encoding, an AK protocol data packet is divided into bit periods (50 µs) wide. Midway through this bit period (25 µs), we are guaranteed a voltage transition, and it is at this point that we measure the voltage to determine the logic level (high voltage = logic 1, low voltage logic 0)
Let’s put this into action with the Mercedes screen capture below, using a crossing point through the data packet of 20 mV (See decoder settings lower right of image below)
Note the speed pulse width (Approx: 50 µs) defines the baud rate at 19.2 kBd. The decoder is set to ignore the speed pulse (Start bit 1) and decode only the 9 data bits (0-8) into binary values. As we can see, the decoder feature has not worked due to an error on my behalf (Remember, if you think you have found something wrong, blame your measurement technique and configuration first!)
Notice how the data bits resemble a shark fin and appear rounded rather than clearly defined, as can be seen with other networks. This is due to a measurement error, as I have our bandwidth limit (20 kHz) set to on, which cannot be turned off post capture to reveal the raw data (unlike our DSP/Lowpass filter).
Given our baud rate is 19.2 kHz, we need a bandwidth of 5 times this value to faithfully reproduce the signal in the scope viewport (5 x 19.2 = 96 kHz).
Not to worry, I have another screen capture below from a Volvo XC60, where I have applied a 100 kHz low-pass filter, which has the effect of removing noise without excessive rounding of the data bits.
A combination of low-pass filtering and the ability to fine-tune the serial decoder threshold (bit crossing point) is what makes PicoScope an invaluable tool when it comes to serial decoding, as we can accommodate most signals regardless of environmental noise and voltage levels.
Now, to delve deeper into the correlation between voltage and logic levels, let’s look closely at the decoded data field (0-8 bits)
Note in the image above how we have a speed pulse of 50 µs followed by a rest period of 25 µs. From this point on, our data field is divided into 50 µs bit periods. Below I have added 18 partition rulers to help visualise a bit period, the voltage transition point (which occurs at 25 µs) and the location where the voltage is measured to determine the logic level (high voltage = logic 1, low voltage logic 0)
Another technique that can help, where edges can be hard to define, is to apply our sign Math channel. Here, we use a similar technique to the frequency Math channel, where we lower the waveform such that the data field passes through 0 V (-164 mV).
Our sign Math channel will then return any value above 0 V as +1 V and any value below 0 V as -1 V (See image below)
Above we have the sign Math channel sign(A-0.164) in pink, where each bit is clearly defined thanks to the logic of the Math channel defining clear values (+1, -1) where our waveform crosses the 0 V.
Note that AK protocol wheel speed sensors can be found on all types of vehicles, as everyone will benefit from the additional data provided over the sensor wires. Below we have an Iveco S-Way tractor unit; based on the above information, I think we can all guess what sort of wheel speed sensor this is.
I chose to capture the above sensor in a slightly different fashion to the Mercedes & Volvo cars, which were measured across the sensor AC Coupled.
Here we have DC-coupled the sensor to reveal the DC offset (1.206 V), which should not be ignored (See this case study here https://www.youtube.com/watch?v=xr4dicP8mBA)
Note: I have included current, which is the true signalling method used for these sensors. Given that the voltage and current are proportional to one another, any change in current (load) will also be captured within the voltage (which we have proven with the Mercedes & Volvo). With that said, if you can measure current, we capture work done and qualify the system integrity in the process
The one caveat we have to manage is the low current nature of these sensors (Specification: 7 mA, 14 mA and 28 mA).
As in the video link above, using a current multiplier increases the measured current by a factor of 10, such that the clamp can capture and display these values with ease. With the help of a math channel (B/10) we then reveal the original current values.
Above, we have 7 and 17 mA, respectively, between our data bit peaks and speed pulse, where we must allow for errors with clamp drift and offset.