TMAG5170-Q1: TMAG5170 Queries

Prathmesh,

DRV5055 and TMAG5170 can both be used in angle encoder applications.  DRV5055 is only sensitive to a single field component, so angle measurements require two devices.  One the other hand, TMAG5170 is a 3D sensor and can capture the field of all three vector components.  There should always be at least two which may be used for angle calculations, but amplitude correction may be necessary to provide normalized inputs for the angle calculation.  

DRV5055 is an analog output device with a sensing bandwidth of 20 kHz.  TMAG5170 is a digital output device and the maximum sample rate with two channels is 13.3 kSPS.  

DRV5055 (propogation delay is 10 us)

TMAG5170 - (2 Ch. sample time is approx. 75 us.)

Using a 10 kRPM rotation, we would see 167 rotations per second.  It is plausible that either device might be used.  To improve angle calculations with TMAG5170, it may be desired to increase the sample averaging, but this will increase total sample duration and add extra latency.

Thanks,

Scott

To capture angle data using a linear Hall-effect device, it is necessary to have two outputs which are 90 degrees out of phase.  Since DRV5055 is only sensitive to the Z-axis component, it will require two devices mechanically separated by 90 degrees.  Placed by a rotating diametric magnet, this will produce a sine/cosine relationship between the two outputs.  This provides the necessary inputs required for an arctangent calculation.  TMAG5170 has 3 mutually orthogonal Hall elements integrated in the same package.  This allows a single device to be used to capture the needed field data since the X, Y, and Z magnetic field components are naturally 90 degrees separated in phase.  

The referenced plots are missing in your reply, but propagation delay will result in some fixed phase delay between the actual angle of the magnet and the result of the angle calculation.  For the highest degree of accuracy in position control at such speeds, it is necessary to use the propagation delay with the total communication/calculation time to account for the change in the magnet position.  Otherwise the resulting angle calculation will always lag behind the real position of the magnet.  This is discussed in the following app note

https://www.ti.com/lit/an/sbaa539/sbaa539.pdf

The linked document SLYA059 is specific towards a magnet which is traveling in a linear path near the sensor.  The plots shown in this document are using an axial polarized magnet with no rotation.  If this type of magnet were placed a the outer circumference of a rotating wheel/gear then the resulting response would be somewhat similar to the plots in this document.  This is not typically the approach used for angle encoding, however.  Rather the magnet is centered on the axis of rotation and the sensor is placed nearby such that the magnet does not leave the vicinity of the sensor. 

It is possible for the sensor to "miss" the passing magnet if the field changes fast enough.  All Hall-effect devices must integrate the input signal for a short period.  The resulting conversion will be an average value of the field during the entire input integration portion of the sampling time.  For TMAG5170 this is represented in the following image from the datasheet:

There is a tool which you might find helpful here:

https://www.ti.com/tool/HALL-PROXIMITY-DESIGN

Thanks,

Scott

The maximum speed on TMAG5170 depends on how much averaging you need to achieve your accuracy goals and the total "look ahead" you can manage.  

The document I linked previously has a plot which you can use as a reference:

This plot only considers the total time required to complete one conversion and the respective change in angle that would occur based on the rotation speed.  

In high speed systems, the controller must be able to forward predict the position based on known speed information.  The maximum speed will be set by the accuracy requirements in your system, and the ability of your algorithm to predict current position based on the sensor data (which has propagation delay).  Ideally, however, the magnet should rotate less than 180 degrees during the integration period of the device, and I would probably try to target a delay less than 45 degrees as a starting point.

DRV5055 is a linear analog device, however.  The operating bandwidth is 20 kHz, which should easily accommodate 10 kRPM.

Prathmesh,

Below is an image of an on-axis rotation, where the magnet is centered on the axis of rotation:

With the sensor placed below the magnet, also on this axis of rotation, the field from the magnet will be parallel to the face of the magnet and directed towards the S pole of the magnet.  As the magnet rotates, and ideal sine/cosine will emerge for the X and Y axes.  Since the vector is parallel to the magnet in this position there is no Z-component. 

In practice, your system tolerances and outside influencers, such as the earth's magnetic field, may result with some small amount of field observed in the Z direction.  This should remain very low, but it is hard to quantify for every magnet/air gap/system.  This portion of the field may be used for predictive maintenance.  However, this location is not always ideal for all designs, and some systems require that the sensor be placed off center.  This will still produce X and Y fields, but their amplitudes may be mismatched.  In this second case, there will also an increase on the Z component.  This will change depending on the location of the sensor and the shape and material of the magnet.

Thanks,

Scott

Prathmesh,

There appears to be a bug in the latest revision, and I have notified those responsible for the tool.  I believe you are attempting to find the ideal range for your specified magnet.  I have tried to simulate your settings using another simulation tool I have access to.  

In the image on the left it shows the 20 mm diameter magnet with a 10 mm gap from the sensor location. The distance sweep starts in this position and then moves the magnet closer to the sensor location.  (This sensor marker does not account for package sizes).  Since the magnet is not rotating in this simulation, we see that the field is entirely in the Y direction.  The final position is 10 um above the sensor to avoid the sensor entering the magnet space.  

This N52 magnet is quite large, so it produces a fairly strong field which would saturate the inputs of TMAG5170A1 when the magnet is closer than 7.5 mm to the sensing element.  

Thanks,

Scott

Prathmesh,

We anticipate having an updated tool available by mid Feb.

Here is the simulation with 10mm diameter, and 10 mm height, N52 Magnet with a 3 mm X offset

I also re-ran the simulation from yesterday with no offset since it was run with a radius of 10 mm (Diameter = 20 mm)

Please notice that with this particular sweep, that the offset was in X and the polarization is in Y, so at this particular rotation we only see a vector component in the Y direction. If we run the offset position with a 90 degree rotation we get the following result

Prathmesh,

The 75 us is taken from these items on the data sheet:

For a single sample on a single axis, the tmeasure time is 75 us.  Continuing down to the footnotes, the additional time per conversion is explained

The conversion process is pipelined, where the ADC conversion can take place at the same time as the input integration:

By comparison, DRV5055 is purely analog output, so it does not include any ADC conversion timing, and this would need to be done external to the device.  To continue functioning with high speed rotation, it would be necessary to use a high speed ADC for this.

This conversion timing does mean that TMAG5170 will be limited at high speeds, but the upper speed limit depends on the required accuracy during high speed rotation.  As the motor slows, the latency will become less significant.

We don't currently have any faster devices released which would provide you an linear analog output with 2 or 3 axes.  TMAG5110 and TMAG5111 are 2D latches, which may be used for speed and direction encoding.  A latch device will detect 4 unique positions per pole pair in the rotating magnet, so resolution is limited.

Thanks,

Scott