Utveckling av en trådlös bioelektrisk signalinsamlingsmodul

Development of a wireless bioelectric

signal acquisition module

MARTIN ANDERSSON

Utveckling av en trådlös bioelektrisk

signalinsamlingsmodul

MARTIN ANDERSSON

2014-06-dag

Jan Wikander

Martin Edin Grimheden

Max Ortiz Catalan

Kontaktperson

Martin Edin Grimheden

Sammanfattning

Där finns många intressanta forskningsområden kring utvecklingen av mekatroniska

proteser, ett av dessa områden berör specifikt möjligheterna till trådlöskommunikation

mellan protesen och omvärlden.

Huvudmålet med detta arbete var att utveckla ett trådlöst bioelektrisk signal-insamlings

system för elektromyografiska signaler i real-tid hos patienter som har genomgått en

armamputation.

Som ett första delmål sammanställdes två designförslag för själva

hårdvarukonfigurationen och presenteras för selektion i samspråk med det

projektägande företaget Integrum. Denna hårdvarukonfiguration skulle bestå av följande

komponenter, en analog-front-end(AFE) samt en microkontroller(MCU) och slutligen en

wifi-enhet. Den främsta av de två designalternativen överensstämde 100% med den

konfigurationen som företaget hade arbetat fram som ett alternativ innan detta projektet

hade påbörjats.

Den resulterande produkten besitter kapaciteten att leverera data som överstiger

Integrums nuvarande trådlösa produkt med 400%, systemet är även väldigt stabilt och

robust. Dock är det osäker om denna kapacitet kommer kunna realiseras då en

multifasitets bug hittades under arbetets gång. Denna bug förhindrar att systemet körs

på maximal hastighet pga problem med den interna minneshanteringen samt att den

tillåter endast en vis kvantitet av dataflöde per tidsenhet.

Vid arbetets avslut arbetade Texas Instruments fortfarande med att lösa dessa problem.

Mjukvaran som finns tillhand för wifi-modulen besitter även ett antal begränsningar som

kan göra det svårt att utnyttja all överföringskraft i wifi-modulen, vilket kan leda till

flödesproblem mellan AFEn samt Wifi-modulen. Dock besitter produkten stora

möjligheter att uppfylla alla krav som ställs på den som en första generationens produkt,

vilket leder till en stark rekommendation för att den ska integreras i Integrums system.

Master of Science Thesis MMK 2014:13 MDA 456

Development of a wireless bioelectric signal

acquisition module

MARTIN ANDERSSON

2014-06-day

Jan Wikander

Martin Edin Grimheden

Max Ortiz Catalan

Contact person

Martin Edin Grimheden

Abstract

There are many interesting research areas surrounding the development of

mechatronical prosthetics, one of these areas is the possibility of wireless

communication between the prosthetic and the surrounding world.

The main goal of this thesis is to development a wireless bioelectrical signal acquisition

module for sampling of EMG signals from a arm amputee in real time.

As a first step two hardware design alternatives were put together, consistent of the

three following component areas. An analog-front-end (AFE), a micro controller (MCU)

and finally a wifi module. The main design alternative correlated 100% with the

proposition the company had put together as an alternative before the start of this

project.

The resulting product has the capacity to deliver data speeds that surpasses Integrums

current wireless product with 400%, and the system is very stable and robust. But it is

unsure if it will be possible to utilize the complete transfer speed, since a multifacility

bug was found during the project. The bug restricts the system from being run on

maximum speed because of problems with internal memory clearing and only a specific

amount of data is allowed to be transferred within a specific time window.

Texas Instruments where still working on these problems at the end of this project. The

provided wifi software has several limitations that may make it hard to fully utilize the

total transfer speed, and thereby may lead to data flow problems between the AFE and

wifi module. But the product has the possibility to fulfill all the stated requirements, and it

will serve well as a first generation product. Therefor it gets a strong recommendation

for integration into Integrums system.

Contents

1 Introduction, purpose and problem description 1

1.1 Introduction . . . 1

1.2 Purpose . . . 2

1.3 Problem description . . . 2

1.4 System requirements . . . 3

1.5 Terminology and definitions . . . 4

2 Method description 5 3 System overview 7 4 Related work on related fields of interest 9 4.1 Wireless biometric multi-channel system . . . 9

4.2 Programmable Neural Measurement System . . . 9

4.3 Wireless-Implanted Neural Recording System . . . 9

4.4 Anthropomorphic Robotic Arm . . . 10

4.5 Current Hand Exoskeleton Technologies . . . 10

4.6 Bio-inspired robotics hands implementation . . . 10

4.7 Improving Myoelectric Pattern Recognition Robustness . . . 11

4.8 Stress reduction at the skin-implant . . . 11

4.9 Targeted reinnervation . . . 11

4.10 Final thoughts . . . 11

5 The electromyogram characteristics 13 6 Analogue Front End 15 6.1 AFE theory . . . 15 6.1.1 AFE requirements . . . 15 6.1.2 AFE characteristics . . . 16 6.1.3 ADC types . . . 17 6.2 AFE alternatives . . . 18 6.2.1 Texas Instruments . . . 18 6.2.2 Intan technology . . . 21 6.3 AFE comparison . . . 24

7.2 MCU alternatives . . . 28 7.2.1 Texas instruments . . . 28 7.2.2 Atmel . . . 32 7.2.3 Other vendors . . . 33 7.3 MCU comparison . . . 34 8 Wireless connectivity 35 8.1 Decided connectivity . . . 35 8.2 Wifi theory . . . 35 8.2.1 Wifi standards . . . 36 8.2.2 Network protocols . . . 36 8.3 Wifi alternatives . . . 38 8.3.1 WL1271-TIWI . . . 38 8.3.2 CC3000 . . . 39 8.3.3 Compact-S-UART . . . 40 8.3.4 TiWi-SL . . . 41 8.3.5 WizFi630 . . . 42 8.3.6 TiWi5 . . . 43 8.4 Wifi comparison . . . 45 9 Design proposals 47 9.1 Alpha design . . . 47 9.2 Beta design . . . 47 9.3 Chosen design . . . 48 10 Resulting implementation 49 10.1 MCU . . . 49 10.2 AFE . . . 50 10.3 WIFI . . . 51 10.3.1 UDP vs TCP . . . 52 10.4 WBSAM software . . . 53 10.4.1 UART function . . . 53

10.4.2 Wifi start-up sequence . . . 54

10.4.3 Send function . . . 56

10.4.4 Receive function . . . 56

10.4.5 Network data check function . . . 57

10.4.6 Matlab commands function . . . 58

10.4.7 Timer workaround . . . 59

10.4.8 AFE sequence . . . 60

10.6 Client workstation . . . 62

10.6.1 Access point settings . . . 62

10.7 Matlab . . . 63

10.7.1 Matlab code examples . . . 64

10.7.2 Matlab limitations . . . 66

11 Recommendations and future work 67 11.1 Recommendations . . . 67 11.2 Future work . . . 68 12 Literature study 69 12.1 Literature study . . . 69 A AFE tabels 73 B MCU architecture 75 Bibliography 79

Chapter 1

Introduction, purpose and problem

description

1.1

Introduction

This thesis was suggested and supervised by Max Ortiz Catalan, research engineer at the orthopaedic company Integrum AB, located in Gothenburg.

In prosthetic development leading manufactures are trying to integrate more and more technology into their products, both software and hardware. The main soft-ware research area in the prosthetic community lays in neural networks, meaning the implementation of artificial intelligence (AI) to interpret the bioelectrical signals perceived from the muscular mass. As the technology progresses more and more tools become available to improve the quality and quantity of the signal acquisition. Meaning the more data the system has access to the more exact interpretations can be provided and at the same time decreasing the learning time for the system. The demand for more data means the next generation AFE needs to have the capability to supply a higher number of channels to collect data from. The direct consequence of increasing the amount of channels will be the higher demands placed on the data delivery system. As the amount of channels increase it become more and more cumbersome to connect it to the patient, so to increase usability of the system the possibility of using a wireless solution for data transmission is a valid alternative. The prototype made during the thesis will be an wireless signal acquisition module, used for making a proof of concept of a rehabilitation system, that will be based on virtual reality on the platform BioPatRec with myoelectric control. The project will focus on increasing the amount of channels on the product to fulfil the demands on larger quantities of data sampling when it at the same time offer a wireless communication link.

1.2

Purpose

The purpose of the project is to develop a proof-of-concept prototype of a wireless signal acquisition module that will be able to send bioelectric signals by a wireless interface in real-time. The reasons for the project comes from the demand for a higher number of EMG-sensors in combination with better movability, connectivity and information gathering. And with the goal to one day allow a higher degree of autonomous prosthetics that will be able to be as fast, agile and sensitive as a normal human hand is.

This thesis will hopefully lead to increased possibilities for the research depart-ment to improve the research information gathering, and in the end improve the quality of life for amputees. And thereby leading to the possibility for people that has been born with defect hands, arms or amputee’s, to once again be able to fully interact with the surrounding world as a whole human.

Below are the areas of interest as well as the boundaries of the thesis, presented in combination with the EMG signals movement through the system with the possible usage of a Digital signal processor (DSP).

Figure 1.1. Signal paths.

1.3

Problem description

The center core of this thesis is to answer if it is possible for a signal acquisition model to move from cable bound to wireless communication, when at the same fulfilling all the stated requirements.

1.4. SYSTEM REQUIREMENTS

1.4

System requirements

In the following section are the requirements on the WBASM stated, some are re-quired to be able to supply the rere-quired level of data quantity and quality while others are needed to surpass their current system.

The system is required to be able to measure over 2000 samples per second(SPS) with a resolution of 16-bit or greater on at least 32 channels simultaneously, and at the same time be able to transfer the data wireless in real time.

These settings are required to be controlled through Matlab, as well as starting and stopping of the signal acquisition. The client workstation will acquire, store and present the data with the help of Matlab.

• A minimum of 32 channels per module.

• A minimum of 2kHz sampels per second per channel. • 16-bit or greater resolution on the ADCs.

• Real-time signal acquisition.

• The system must be able to simultaneously communicate with

sev-eral modules if necessary.

• 1,1 Mbps or greater communication link for a module handling 32

channels.

1.5

Terminology and definitions

Below are a few short definitions and clarifications of the terminology used.

The electromyogram signal: The electromyogram (EMG) signal is the representa-tive of the electrical activity produced by skeletal muscles during contraction and rest.

Myoelectric prosthesis: Myoelectric prosthesis are prosthesis that are controlled by the voluntarily contraction of muscles but aided by electrical power.

Bioelectric signal: In this thesis bioelectric signals are refered to as the electrical impulses of interest extracted from the human body, for example surface electromyo-graphy(sEMG).

Real-time: In this thesis real-time is referred to the computational time that a human being does not register as a delay from command to execution of movement. Analoge digital converter (ADC): ADCs are a electrical component used for con-verting analogue signals to a digital representable signals.

Analogue front end (AFE): An AFEs task is mainly to convert multiple analogue signals to digital signals. They contain amplifiers, multiplexers, control interfaces and they can have either a internal or an external ADC depending on design. Wireless network (Wifi): The wireless networks task will be to work as and commu-nication link between the wireless bioelectrical signal acquisition module (WBSAM) and the PC for transfer of the sampled data.

Micro control unit (MCU): The MCU main role will be to monitor and transfer the data generated by the AFE to the Wifi interface for further transfer.

Chapter 2

Method description

The first part of the thesis consisted of learning about EMG measuring, and the demands and limitations that could relate to the thesis. The accessible documen-tation in this area is quite extensive and overwhelming. Once the information had been striped down to its most basic fundamentals that related to the thesis, focus was changed to the technical aspect, meaning the specific components and their technical data that would be used in the project.

The first phase was to present two design options of a complete system, in re-gards to hardware options available on the market. These designs had to be redone twice because several MCU options went out of commission during this stage of the project. Selection of the different hardware options were based on purely tech-nical documentation and theoretical aspects of the information available. In some instances similar projects used a component or another, giving indications that the component would be a feasible alternative.

In discussion with the company the first design option was decided on, the pre-sented design correlated exactly with the hardware design that earlier had been planed for by the companies lead engineer. Giving a clear indication that the deci-sions made during the first phase where correctly made.

For a project of this size and organization complexity consisting of only one person, a macro cycle version of the V-model was used since it would provide a systematic and logical approach to the different areas of interest during the project. Mean-ing focus was applied to one area of interest at a time and each completed area was then integrated to the bigger system. The decided approach was to start with implementation of the MCU then the AFE and finally the Wifi.

Chapter 3

System overview

To give a clearer view of the overall system a system overview is presented below. Signal acquisition from the amputees arm is conducted with the help of sEMG headers that the analogue part of the AFE samples. The AFE quantizes the reading to the digital domain for further transfer over the system represented by the MCU, Wifi and PC.

Figure 3.1. Overall signal path.

A mechatronic prosthetic arm is a complex construction, this project will solely focus on the sampling of EMG signals that are acquired from the user. But as will be discussed in the following chapter there are many research areas that directly involves the physical construction of the prosthetic arm.

Chapter 4

Related work on related fields of interest

The following chapter gives a small glim’s into related fields that may be of interested for Integrum, even if it is not directly connected with this project. This articles give an indication to how complex and exciting tomorrows prosthetics will become.

4.1

Wireless biometric multi-channel system

The article "A portable wireless biometric multi-channel system" [8] contains some very interesting parallels in comparison to what is hoped to be achieved with this thesis. The main difference from a hardware point-of-view is that it is a low rate signal acquisition data system that is constructed around the wireless interface ZigBee. The article has a very deep technical documentation that may be of use when developing similar systems.

4.2

Programmable Neural Measurement System

RHA2216 is one of the AFEs that has been investigated during this thesis and in the article "A Programmable Neural Measurement System for Spikes and Local Field Potentials" [15] it has been used in a very adaptable signal acquisition system close to the goal of this thesis. In that project a total of 8 RHA2216 was used to thereby allowing up to 128 channels for signal acquisition, thereby showing the potential to increase the possibility’s of future acquisition systems.

4.3

Wireless-Implanted Neural Recording System

A interesting area of research is the integration of electronics within the human body, the article "A Power and Data Link for a Wireless-Implanted Neural Recording System" [9] gives a insight to how in the future there may not be a need to penetrate the human skin for signal acquisition and power supplying. This could lead to some interesting aspects of the prosthetics development, in the way that once a solid anchor-technique has been acquired the internal electronics will still be able to

communicate with the outside world. Meaning for example that firmware updates will be possible and EMG data transfers to onboard computation in the prosthetic. But it will also lead to new demands on security from external hostile sources.

4.4

Anthropomorphic Robotic Arm

The goal for the development of a prosthetic arm is to give back the full interac-tion of the lost arm, and one area that consist of many complex problems is the development of mechanical hands that has a high level of dexterity that resembles the human hand and at the same time has low weight and high power efficiency. In the article "Development of an Anthropomorphic Robotic Arm and Hand for Interactive Humanoids" [17] a humanoid arm and hand is presented that gives a clear indication to the direction that the future development of hands and arms are proceeding towards.

4.5

Current Hand Exoskeleton Technologies

As the development of all related areas continue, the possibility to decrease the size and power consumption will only give further possibility’s for decreasing and created better and more adaptable prosthetic hands. The development of hand exoskele-ton is a complex and multidisciplinary research area and there are several ongoing project around the world and the paper "Current Hand Exoskeleton Technologies for Rehabilitation and Assistive Engineering" [11] goes through several interesting alternatives and challenges that needs to be overcome. The paper also contains some very interesting information related to the anatomy of the hand and the interaction and construction of the exoskeleton surrounding the hand.

4.6

Bio-inspired robotics hands implementation

Another paper of interest in a related area is "A survey of bio-inspired robotics hands implementation: New directions in dexterous manipulation" [13], this study goes into more sophisticated biomimetic robotic hands and arms. It also shows that the interest for more muscular inspired actuators are emerging in the development of movement in arms, in combination with greater usage of AI for grasping of objects. For a amputee to get used and comfortable with the usage of a prosthetic it has to interact to the commands issued by the amputee within a reasonable time, it has been found that a window between 150 and 250 ms is within acceptable limits when using pattern recognition based control, but also refers that for non pattern recognition a time delay of up to 100 ms is acceptable. This article can give a

4.7. IMPROVING MYOELECTRIC PATTERN RECOGNITION ROBUSTNESS

4.7

Improving Myoelectric Pattern Recognition

Robustness

To increase robustness for acquiring and recognizing myoelectric signals a study has been made to investigation the optimal placement, orientation and number of electrodes needed during electrode shift. The paper "Improving Myoelectric Pattern Recognition Robustness to Electrode Shift by Changing Interelectrode Distance and Electrode Configuration" [16] findings could there for be compared to the current standard used at Integrum and see if any benefits may be extracted for future use.

4.8

Stress reduction at the skin-implant

A article that may be of interest to the engineers of Integrum is the "A compu-tational model for stress reduction at the skin-implant interface of osseointegrated prostheses" [18] that shows a 90% decrease in stresses at the skin-implant interface. This area is outside my expertise but may be worth the time to investigate for a person who is familiar with the technical area.

4.9

Targeted reinnervation

Accessing neural information by reinnervation has shown some amazing results as presented in "Neural Interfaces for Control of Upper Limb Prostheses: The State-of-the-Art and Hopes for the Future." [19] and as well in "Robotic touch shifts per-ception of embodiment to a prosthesis in targeted reinnervation amputees." [21] and "Targeted muscle reinnervation for real-time myoelectric control of multifunc-tion artificial arms."[22]. By targeted reinnervation the available information used for prosthetic control increases and there by adds better control and increases the possibilities for the amputee to interact with the prostheses. This is now a clinically available treatment that has been performed on more then 40 patients.

4.10

Final thoughts

To summarize all the articles it is clear that with so many areas under development the future prosthetics will become more and more lifelike and offer a better movabil-ity and interaction for its users. But the complexmovabil-ity will also increase many times over, both in the aspect of complex algorithms as well as on the hardware side.

Chapter 5

The electromyogram characteristics

In this chapter a small summary of the EMG signals characteristics and electrical limits, that may be needed to be taken in consideration are presented.

The electromyography (EMG) signal represents the electrical impulse during con-traction and rest. The EMG signals is of a stochastic nature and can be represented by a Gausian distribution function. The EMG spectrum of useful information ranges from 0 to 500Hz [4] where the main information of interest is in the range of 50 to 150Hz [6]. The signals amplitude ranges from 0 to 10mV (Peak-to-peak) or 0 to 1.5mV (RMS), the EMG signal characteristics are the factors that set specific demands on the AFE, for example the frequency range and maximum frequency to determine the needed sampling frequency. Besides these factors the EMG signal will just represent a input signal for the system and this thesis, the reason for this is to limit the thesis areas to as a high degree as possible. If it is of interest to the reader there is a massive amount of information related to EMG measurement on the internet[2] [3] [6] [7] [30] [31] but this thesis will not go deeper into that topic since it is not of interest for the completion of the requirements. Below is the frequency spectrum of a EMG signal in a graphical presentation.

Chapter 6

Analogue Front End

In the following chapter the theoretical aspects of the AFE module and its areas of relevance to this thesis are briefly discussed, with the follow-up of selected AFEs that are of interest for implementation. And in the end a comparison of which AFE that will be recommended for implementation.

6.1

AFE theory

A AFE is used to acquire and transform the analogue EMG signal to a digital representation for further transfer through the system. The AFEs that exist on today’s market comes mainly in two forms, those with ADCs installed and those without.

6.1.1 AFE requirements

To decide which AFE that is most suitable for acquiring EMG signals some require-ments must be established to be able to minimize the available alternatives. In the search for suitable AFE only two companies have been found that offers alternatives that fulfil the following requirements in relative terms:

• 16-bit ADC resolution or higher. • Low Input Referred Noise.

• High Common Mode Rejection Ratio. • Very high input impedance.

• Acceptable gain.

6.1.2 AFE characteristics

Below are different AFE characteristics of importance presented in short terms. Number of channels:

The demand of a minimum of 32 channels per module is related to the need to be able to increase the amount of monitored areas of muscular tissue. No system at Integrum today has such large amount of channels on a wireless interface.

Samples per second (SPS):

Because of the Nyquist theorem, the lower limit of SPS needed to be able to monitor the surface electromyography sEMG signals is around 1000 SPS, the given require-ments is that the system should be able to acquire a minimum of 2000 SPS. This requirement is not a problem since there are several options available from vendors such as Texas Instrument and Intan technology that can cope with the requirements. ADC-Resolution:

The decision on the amount of bits, 16-bit or 24-bits will be decided in combina-tion with SPS as well as the number of active channels. The benefits of increasing the resolution from 16-bit to 24-bit will be the possibility to easier distinguish the baseline noise from the EMG signal [31]. But this will increase the demand on the data delivery system because of the increase amount of sampled data.

Amplification (Gain):

Because of the low voltage levels that the EMG signal contains it needs to be in-creased to workable levels for the following electronic components [7]. Low gain can to some extent be compensated by higher resolution [31].

Common mode rejection rate (CMRR):

A high CMRR is equivalent to a higher degree of rejection of the common input signal from the bipole electrode. A CMRR of 90dB is generally sufficient to suppress external noise [6] [30]. Since the goal is to try and guarantee the maximum quality of the EMG signal, a CMRR at the level mentioned above will be considered a goal to try and achieve.

Input referred noise (IRN):

Both the AFE and the internal ADCs have a certain amount of input referred noise that should not be confused with quantization noise. IRN is one of the factors that will be used to separate the different AFEs and ADCs from each other. The lower IRN the better for the final signal fidelity.

6.1. AFE THEORY

6.1.3 ADC types

Below are the two most common used ADCs that are used in association with AFEs that also fulfil the stated requirements before. Each type has its benefits and drawbacks that will be briefly highlighted since both types will be present in the AFE options presented later on.

Delta-sigma

A delta-sigma ADC averages multiple samples from the input for a specific set of time before outputting the digital correspondent value. The digital filter easily han-dles the noise and act as an anti-aliasing filter so thereby the delta-sigma converter only needs an external second-order lowpass filter on the analogue input. These kind of ADCs are most commonly used in applications that require high resolution for example audio systems, one trade-off compared to Successive Approximation Reg-ister (SAR) ADCs is the increased consumption of power due to the higher clock rates[29]. Since compensation of low gain can be accomplished by increasing the resolution in the ADC, the use of a 24-bit delta-sigma ADC may be an alternative if the gain is too low with a 16-bit SAR. A delta-sigma ADC has a larger output-data delay because of its architecture, this may prove a problem for information gather-ing at an sgather-ingle point of time unless several ADCs are used in parallel on over all channels.

Successive approximation register

In contrast to a delta-sigma ADC a SAR ADC has the benefit of taking a "snap-shot" of the signal and serial retrieval of the generated data on all active channels. SAR ADCs are suited for applications that require fast response and low latency, such as multichannel data acquisition systems. The serial retrieval will also simplify the data collection in comparison to the delta-sigma ADC. The use of a SAR ADC seams to be a good match for the requirements of this thesis[29]. One limitation is that the SAR ADC is currently limited to 16-bit resolution, if higher resolution us required then a delta-sigma ADC has to be utilized.

It can there by be summarized that each ADC type has it´s benefits as well as its downsides, but in the end most of the downsides can be compensated by smart architectural implementation, as shown in each AFEs schematics later on. Instead the overall picture has to be taken into account, meaning weighing each hardware component to be able to judge if the stated downsides of individual components are acceptable.

6.2

AFE alternatives

Texas Instruments and Intan Technology are the two companies that showed the greatest potential on the AFE market according to my research. In the following sections their different AFE alternatives of interest are presented and discuses.

6.2.1 Texas Instruments

Texas Instruments has several alternatives in the AFE department, to be able to decide on the most appropriate alternative for this thesis the posted requirements will limit the available alternatives.

ADS1299

There is only one option from TI that is of real interest since the ADS1299 char-acteristics has so much more to offer then the other options, as can be seen in the table on the next page. The ADS1299 is specifically manufactured to be used within medical systems and ECG/EEG measuring applications, but will work just as well with EMG measuring [31]. As can be seen no the next page it far surpasses the other options TI can offer when it comes to giving the best IRN, Signal-to-noise ratio (SNR), gain and CMRR. The downside is the power consumption, if compared to the closest competitor, the ADS1298, it is seven times higher. The ADS1299 can be used in daisy-chain mode and there by allowing multiple ADS1299 devices to be used in a high count channel systems.

The ADS1299 is a new version compared to the one presented in [31] where the downside of low gain has been rectified. The gain has been increased to give it further headway against it competitors specifically the alternative from Intan Tech-nology. One interesting option that is offered with the ADS1299 is the right leg drive, that can be used to reduce common mode interference. The importance of the right leg drive will be discussed with the Integrum in a later stage.

6.2. AFE ALTERNATIVES

Below the ADS1299 and the ADS1298 that are the best option available from TI are presented:

Characteristics ADS1299 ADS1298

Resolution (Bits) 24 24

Number Input Channels 8 8

Input-referred Noise (uVpp) 1 3

PGA Gain (V/V) 1,2,4,6,8,12,24 1,2,4,6,8,12

Sample Rate (Max) (kSPS) 16 32

Power Consumption (mW) 5.125 0.75

INL (ppm) 8 8

Interface SPI SPI

Approx. Price (US) 61 42

Reference Mode Int, Ext Int, Ext

Offset Error (uV) 60 500

Offset error drift (nV/C) 80 200

Gain Error (typ) 0.1 0.2

Digital Supply (Min) (V) 1.8 1.65

Digital Supply (Max) (V) 3.6 3.6

SNR (dB) 129.3 112

Total harmonic distortion THD (dB) -99 -98

Analog Voltage (Min) (V) 4.75 2.7

Analog Voltage (Max) (V) 5.25 5.25

Architecture Delta-Sigma Delta-Sigma

Table 6.1. Specifications of the ADS1299 and ADS1298.

Each ADS1299 channel has a 24-bit, delta-sigma ADC, which is a cleaver way to give the possibility to take snapshots on each channel at the same instant just like a SAR ADC would permit, but now with the usage of a ADC with 24-bit resolution instead of 16-bit. The construction can be seen on the next page. The ADC uses a second-order modulator optimized for low noise applications. The digital decimation filters in the ADS1299 can be used to filter out the noise at higher frequencies, they can also supply decimation filters to provide anti-alias filtering. Thereby reduces the complexity of the analogue anti-aliasing filters typically required. Interesting tables of Input-referred Noise as well as Gain vs Bandwidth for more specific comparison are presented in appendix A.

6.2. AFE ALTERNATIVES

6.2.2 Intan technology

Intan Technology is the other company of interest for the AFE. Intan is offering several interesting options with both 16- and 32-channels AFE, with the option of the RHD and RHA series that represents a Digital and Analogue output interface. The RHD communicates by the standard SPI interface that will work with all MCU presented in this thesis. The architecture used by the RHD/RHA-series allow sam-pling to be done at the exact same instant which is optimal for the usage in a signal acquisition system such as is under development in this thesis. The RHD-series can also perform the operation to automatically calculation of the absolute value of the raw EMG waveform and there by lessen the burden on the controller in an EMG acquisition system, and to some extent seclude the need/demand of a on chip DSP on the MCU. Note worthy is the new release of the open-source USB/FPGA inter-face that Intan offers for the usage with their AFEs, main reason that this does not add any benefits is my lack of experience of FPGAs and C++ that would seriously hamper the thesis.

RHA2116/RHA2216/RHA2232

The analogue alternative of Intan technology AFE is identical to the RHD series except that it is not interfaced by SPI and it does not include a ADC, thereby allowing the end-customer to decide on the level of resolution to use. If the RHA series will be used a decision must be made on both what level of resolution the ADC should have as well as what type of ADC that should be utilized. Below are a few important points from Intans own homepage regarding the RHA series [53].

• Fully integrated amplifier arrays; no off-chip filter capacitors

re-quired.

• Low input-referred noise: 2 microvolts RMS.

• Low power operation: less than 500 microwatts per channel.

• On-chip high speed analog multiplexer allows many amplifiers to

share one external A/D converter.

• Lower cutoff frequency of all amplifiers set by external resistor;

adjustable from 0.02 Hz to 1.0 kHz.

• Upper cutoff frequency of all amplifiers set by external resistors;

adjustable from 10 Hz to 20 kHz.

• True zero gain at DC rejects electrode offset voltages. • In situ electrode impedance measurement capability.

RHD2216/RHD2132

The RHD series contains amplifiers, analogue and digital filters and multi-frequency electrode impedance measurement modules on one chip with the size of 8mm * 8mm. The RHD series has a custom built 16-bit ADC built into the chip that has the capa-bility of 1.05 MSPS or sampling at 30 kSPS on each channel on a 32 channel system. The RHD-series has the option of using Complementary metal–oxide–semiconductor (CMOS) as well as Low-voltage differential signaling (LVDS), if CMOS is used it is recommended to restrict sampling to 10kSPS on the RHD2216 and 5kSPS for the RHD2132 on each channel to keep noise gathering from increasing above acceptable levels, however by not using LVDS the risk of introducing more noise in the system is increased. The schematics of the RHD2132 is presented on the next page to clarify the difference between the use of a SAR and Delta-Sigma ADC implemented architecture. Below are a few important points from Intans own homepage regard-ing the RHD series[54].

• Fully integrated electrophysiology amplifier array with on-chip

16-bit ADC and industry-standard SPI.

• ADC supports sampling 32 amplifier channels at 30 kSamples/s

each.

• Low input-referred noise: 2.4 microvolts rms typical.

• Standard four-wire 16-bit SPI interface with CMOS or low-voltage

differential signaling (LVDS) I/O pins.

• Lower/Upper cutoff frequency of all amplifiers set by on-chip

reg-isters; adjustable from 0.1 Hz to 500 Hz and 100 Hz to 20 kHz.

• Integrated multi-frequency in situ electrode impedance

measure-ment capability.

• Optional on-chip DSP high-pass filters for amplifier offset removal. • Auxiliary ADC inputs for interfacing additional off-chip sensors. • Individual amplifier power up/down for power minimization.

6.2. AFE ALTERNATIVES

6.3

AFE comparison

Comparasion between the ADS1299, RHD2216 and RHD2132:

Characteristics: ADS1299 RHD2216 RHD2132

Resolution (Bits): 24 16 16

Number Input Channels: 8 16 32

Input-referred Noise (uVrms): 0.1 2.4 2.4

PGA Gain (V/V): 1,2,4,6,8,12,24 200 200

Sample Rate (Max/Channel) (kSPS): 16 30 30 Power Consumption (mW): 5.125(CMOS) 3.5(LVDS) 7(LVDS)

Interface: SPI SPI SPI

Approx. Price (US): 61 195 290

Reference Mode: Int, Ext Int Int

SNR at 2kSPS, (dB): 46 60 60

Analog Voltage (Min) (V): 4.75 3.2 3.2

Analog Voltage (Max) (V): 5.25 3.6 3.6

Right leg drive: Yes Yes No

Architecture: Delta-Sigma SAR SAR

Table 6.2. AFE comparison.

The RHD-series have built-in LVDS drivers and receivers for LVDS communi-cation which will help signal fidelity in comparison to the ADS1299 CMOS archi-tecture. The ADS1299 is dc coupled, whereas the RHD-series has programmable high-pass filters to remove electrode-tissue offsets and low-frequency drift and the RHD-series also has software-programmable upper and lower bandwidths.

ADS1299s main advantages are in the area of extremely well documented data sheets almost to the level that it contains too much information, higher resolution, much lower IRN, lower power consumption and it’s price. With the downsides of low gain and the usage of the CMOS architecture, and with the use of the ADS1299 it will most likely be required to use a daisy-chain to fulfil the requirement of at least 32 channels, but it also leaves headway to go beyond that amount of channels. With the downside of longer development time compared to the RHD-series. It is not a easy or straight forward decision when it comes to deciding which AFE to use when main reliance lies on the theoretical aspects of the components. Their previous generations have been researched before [31] and they both where declared to be more then adequate for the intended applications. In many aspects the sig-nal fidelity seams to be higher with the usage of the ADS1299 as can be seen in the above table, but Intans ease of implementation with the supplied software and

6.3. AFE COMPARISON

By using the ADCs that already are optimized for their separate AFE, develop-ment time will be reduced and another factor that may present a future problem is prevented since integrated ADCs are less prone to external interference. And there by the RHD series is the series to focus on.

In the RHD-series there is both a RHD2216 and a RHD2132 channel option, the only difference between these two option is the cost and the RHD2216 possibility of using the right leg drive. Since this thesis goal is to increase the amount of channels in the current system the RHD2132 seams to be the logical choice, but the benefits of the right leg drive has to be discussed with the company before a final decision is made.

Chapter 7

Micro Control Unit

7.1

MCU theory

There is no universal standard on how to decide which MCU to utilized for a specific application. It is a quite complex task to balance the need of operational power and power consumption in combination with selecting a MCU with the required peripheral units. The possible selection of MCUs in todays industry can also be quite overwhelming, for this thesis a great deal of the different peripheral options that are included in the MCU are of no interest. The few peripheral options that are of interest does force one to look at the more expensive segments on the market. What needs to be kept in mind when it comes to the decision on which MCU that should be used are for example:

• Bus architecture.

• Connectivity (SPI, Ethernet, Wifi). • The possible use of internal DSP. • Power management options. • Technical support.

The MCU needs to be able to handle a large amount of data flows, that will put some demands on the bus architecture but also how the input and output connections communicate internally on the MCU. By aiming high and acquire a powerful MCU with a solid bus architecture and thereby take the downside of high power consumption, one can strive to "guarantee" that the MCU will have the capability to handle the stated demands.

7.2

MCU alternatives

The two biggest vendors on the market is TI and Atmel, in the following sections the best alternatives from these two vendors.

7.2.1 Texas instruments

Texas Instruments has a great amount of MCUs in their portfolio ranging from sim-ple 8-bit MCUs up to ARM based MCUs. The main criterias when going through the alternatives are for example the system architecture and connectivity such as SPI, UART and Ethernet. The computational power is not of great importance since no calculations are planed at this stage. If needed a DSP can be connected to the MCU at a later product development cycle.

7.2. MCU ALTERNATIVES

PICCOLO F2806x

The piccolo has been used by Integrum before, the MCU is however not optimal since the Piccolo only has two SPI. There by limiting the system to the usage of only one SPI, it could be used to fulfil the channel requirements if a 32 channel AFE is used. This will limit the AFE choice to the RHD2132/RHA2132 or the ADS1299 in a daisy-chain mode. Both of the SPI can be made available if the wifi interface will use UART instead, this would allow the usage of two AFEs but the UART interface may limit the transfer speed. There is also a 64-channel system in the pipeline from Intan so in the future this system could use 64 or a maximum of 128 channels. Simplicity is a goal to aim for but one should not limit oneself by invest time in a system that can not be further developed, even though there a clear benefits of using a system that is known.

No reference for the usage of any kind of wifi-interface has been found on TI home-page. But according to the documentation for TI SimpleLink it should not be a problem to interface the F2806x with a wifi-module, but this can not be verified. Below are some MCU specifications from the manufacturers homepage [38] and the internal architecture is presented in appendix B.

Specification F2806x: CPU C28x Frequency (MHz) 90 FPU Yes CLA 1 VCU Yes RAM (KB) 100 Flash (KB) 256 PWM (Ch) 17 CAP/QEP 4/1 ADC 12-bit

ADC Conversion Time (ns) 289 12-bit A/D (Channels) 12

I2C 1 UART (SCI) 2 SPI 2 CAN 1 Timers 3 32-Bit GP,1 WD GPIO 40 USB 1 IO Supply (V) 3.3

ARM AM3715

The AM3715s computational power will not be fully utilized for this thesis but it’s rich connectivity both when it comes to wifi by TIs own wifi module as well as the possibility to use Ethernet, SDIO and SPI makes it a very interesting alternative. The SDIO host controllers interface allows high-speed transfers and also deals with SDIO protocol at transmission level, packing data, adding CRC, start/end bit, and checking for syntactical correctness [40]. By using a wifi module that uses SDIO all SPI can be used for the AFE, depending on which AFE module used the amount of channels be maximized up to 128 channels without implementing daisy-chain mode. By the usage of the AM3715 the Ethernet can be used for transfer error code if problems arise with the wifi module, adaptability by changing wifi interface if prob-lems occur is also a positive aspect of this MCU. The main downside of this MCU is its computational power, but by using scheduling it may be circumvented and used as a way to guarantee that the system may always have access to the computational power when need. TI also supplies the AM3715 with support of their SYS/BIOS Real-Time Kernel and AM3715 now also works with SMX RTOS if it is decided that a RTOS is of use. Below are some MCU specifications from the manufacturers homepage [56] and the internal architecture is presented in appendix B.

Specification AM3517:

ARM CPU ARM Cortex-A8

ARM MHz (Max.) 600

On-Chip L1 Cache 32 KB (ARM Cortex-A8) On-Chip L2 Cache 256 KB (ARM Cortex-A8) Other On-Chip Memory 64 KB

General Purpose Memory: 16-bit GPMC32-bit (SDRC, Async SRAM, NAND flash, NOR flash, OneNAND flash) USB 3 EMAC 10/100 MMC/SD 3 CAN 1 UART (SCI) 4 I2C 3 SPI 4 DMA (Ch) 32 Ch SDMA I/O Supply (V) 1.8,3.3

7.2. MCU ALTERNATIVES

ARM Cortex-M4

The Cortex-M3 might have been the perfect option for this thesis with it’s four SPI and Ethernet connection since this would give the extra benefit of future error searching of the wifi by using the Ethernet. But since it has been discontinued during this thesis and succeeded by the TI Cortex-M4 that lacks the Ethernet connectivity, the Cortex-M4 is the only valid option of the Cortex series with four SPI. One downside contra the Cortex-M3 is the bus architecture shows that the SPI is connected by the Advanced Peripheral Bus (APB) instead of the Advanced High-performance Bus (AHB). This could present a problem if the TI engineers has not allowed enough overhead for all SPI to be active at maximum speed at the same time, which is something to keep in mind. Below are some MCU specifications from the manufacturers homepage [57] and the internal architecture is presented in appendix B.

Specification LM4F230:

I2C 4

CPU ARM Cortex M4F

Flash (KB) 256

SRAM (kB) 32

RTC Yes

Max Speed (MHz) 80

QEI 2

Battery-Backed Hibernation Module Yes Internal LDO Voltage Regulator Yes

USB D, H/D, or OTG OTG

Memory Protection Unit (MPU) Yes

Motion PWM 16

Boot Loader in ROM Yes

Watchdog Timers 2

Analog Comparators 2

SSI/SPI 4

Digital Comparators 16

Maximum 5-V Tolerant GPIOs 43

ADC Units 2

Internal Temp Sensor 1

CAN MAC 2

16 MHz Precision Oscillators Yes

UART 8

GPIOs 43

ROM Software Libraries Yes

7.2.2 Atmel

Atmel also has a great amount of alternatives of MCUs, but for some reason At-mel has chosen not to invest in a great amounts of SPI ports on their MCUs. It also seams that Atmel has documentational problems since the information on the number of SPIs on their MCUs does not correspond with their data sheets.

ARM SAM3X4C

Atmels ATM-based solution SAM3X series is a Cortex-M3 MCU with focus on connectivity with the options of Ethernet, CAN and high-speed USB. The SAM3X is the MCU that fitted the thesis specifications perfectly. Below are some MCU specifications from the manufacturers homepage [58] and the internal architecture is presented in appendix B. Unfortunately it has also went out of stock during the thesis so this option has been disqualified.

Specification SAM3X:

Flash (Kbytes): 256 Kbytes Max. Operating Frequency: 84 MHz

CPU: Cortex-M3

Max I/O Pins: 63

USB Transceiver: 1

USB Speed: Hi-Speed

USB Interface: Host, Device

SPI: 4 TWI (I2C): 2 UART: 4 CAN: 2 LIN: 3 SSC: 1 Ethernet: 1 SD / eMMC: 1 SRAM (Kbytes): 64

NAND Interface: Yes

I/O Supply Class: 1.62/3.6 Operating Voltage (Vcc): 1.62 to 3.6

MPU / MMU: Yes / No

Timers: 9

32kHz RTC: Yes

7.2. MCU ALTERNATIVES

7.2.3 Other vendors

There are several other vendors but to be able to limit the time spend on finding a system that will be able to fulfil the requirements, focus was applied to the two of the biggest vendors on the market. The benefit of this decision is the level of support and documentation as well as possible help on forums around the world is greater then with smaller vendors [43].

7.3

MCU comparison

It is not possible to decide on which MCU that should be used for this thesis without taking into account the limitations the wifi module will add to the system. But below is short summary of the presented MCUs and the main areas of interest.

MCU Comparison:

Flash (Kbytes): AM3517 F2806x LM4F

Flash (Kbytes): 256 256 256

Max. Operating Frequency (Mhz): 600 90 80

CPU type: Cortex-A8 C28 Cortex-M4

FPU: No Yes Yes

Bus Architecture: ARMv7 Harvard Harvard

Max I/O Pins: 186 54 43

SPI: 4 2 4

UART: 4 2 8

Ethernet: 1 0 0

SD / eMMC: 3 0 0

Temp. Sensor: No Yes Yes

SRAM (Kbytes): 64 No 32

NAND Interface: Yes No No

I/O Supply Class: 1.8, 3.3 1.8, 3.3 3.3

Operating Voltage (Vcc): 1.8 3.3 3.3

MPU / MMU: Yes/No No/No Yes/No

Timers: 12 3 6

32kHz RTC: No No Yes

Calibrated RC Oscillator: Yes Yes Yes

Table 7.5. MCU comparison.

The MCUs that are presented here are all of interest, except the F2806x that has too strong limitations for further development even though it might be able to fulfil the requirements in this thesis with the usage of the RHD2132 or ADS1299 in daisy-chain mode. Each of the other presented MCUs are platforms that the project may grow on and to one day even allow a even higher number of AFEs as well as a adaptability to future wifi modules.

Chapter 8

Wireless connectivity

There are several wireless options available that does not perceive as a big obstacle to incorporate with the system that is under construction. Some of the more in-teresting alternatives are Wifi, Zigbee and Bluetooth. Both Atmel as well as Texas Instruments has available solutions, but of those solutions only Wifi is able to handle the amount of data transfer speed that is required. Zigbee also called IEEE802.15.4 has a maximum transfer speed of 1 Mbps [45], and as for Bluetooth 2.0 + EDR, it only has the possibility of transfer speeds up to 3 Mbps with a though rate of 2.1 Mbps that will limit the system for future development[46].

8.1

Decided connectivity

The current wifi modules in the industry today are small and compact which is a fantastic achievement when those aspects are combined with the amount of data that they are able to send. The next stops in the evolution in prostheses research and development is the possibility of using wireless systems with high capability to send and receive data to help read and interpret the EMG signals. The bottleneck in the data speed transfer today is the wifi interface, the fastest wifi modules available today run on the 802.11n protocol.

8.2

Wifi theory

The main theoretical aspects of interest surround the different available wifi stan-dards and protocol for data transfer, in the following sections these aspects are briefly discuses and their direct impact on this thesis.

8.2.1 Wifi standards

IEEE802.11 is a set of standards for implementing wireless local area network (WLAN), below are the standards that fulfils the requirements of the wireless net-work within this thesis [49].

802.11b/g:

This wireless standard operates at a bit rate of 11 Mbps with a 5 Mbps average through output when TCP is used in the 2.4Ghz band [42] [49].

802.11g:

This wireless standard operates at a bit rate of 54 Mbps with a 22 Mbps average through output and offers a forward error correction code in the 2.4Ghz band [42] [49].

802.11n:

This wireless standard operates at bit rates from 54 Mbps to 600 Mbps on both the 2.4 Ghz as well as the 5 Ghz band. The wifi modules in this thesis with the classification of 802.11n can operate at a maximum bit rate of 54, 65 and 150 Mbps [49].

8.2.2 Network protocols

A decision on which protocol to utilize must come from the minimum speed that is required to be able to supply the needed transfer rate.

Generated data by the WBSAM depending on amount of channels, AFE resolu-tion and the amount of SPS. A "worse case scenario" (Channels*Resoluresolu-tion*SPS) is shown below with the maximum resolution on the ADCs with the related number of channels and SPS:

• 32 channels with 16-bits resolution ADC with 2kSPS will generate

1.1 Mbps of RAW data.

• 64 channels with 16-bits resolution ADC with 2kSPS will generate

2.1 Mbps of RAW data.

• 32 channels with 24-bits resolution ADC with 2kSPS will generate

1.6 Mbps of RAW data.

• 64 channels with 24-bits resolution ADC with 2kSPS will generate

8.2. WIFI THEORY

User Datagram Protocol (UDP)

UDP is designed for several connectionless (Protocol that does not require session initiation or maintenance) data streams and does not guarantee delivery at all or in any order. The benefits of this protocol is that it has low overhead and can thereby be used for a system that does not need a guarantee of delivery [42].

Internet Control Message Protocol (TCP/IP)

TCP/IP is a set of communication protocols used for network communication, Transmission Control Protocol (TCP) and Internet Protocol(IP). The main ben-efit of using TCP/IP it the guarantee that the sent information will be received, the direct consequence of the guarantee is the need for a large overhead. For exam-ple a 11 Mbps TCP/IP system can only push a maximum of 5 Mbps of RAW data because of the overhead demanded by the TCP/IP [51].

8.3

Wifi alternatives

Below are the different alternatives of wifi modules for MCU connectivity presented that has been found to be of interest for this project.

8.3.1 WL1271-TIWI

The perfect wifi module for the AM3715 MCU is the TI WL1271-TIWI 65 Mbps wifi module from SL Research (SLR) that is constructed to seamlessly interact with that specifc MCU. This wifi module fulfils all demands for the usage in this thesis. Below are a few specifications from the manufacturers homepage [59].

• Best-In-Class WLAN and Bluetooth Coexistence Technology on a

Single-Chip.

• Enhanced Low Power (ELP) Technology for Extended Battery Life. • On Board TCXO, Power Regulation and U.FL Antenna Connector. • Hardware and Software Pre-integration with TI AM/DM37x (ARM

Cortex-A8), AM18xx (ARM9), and OMAP4 (ARM Cortex-A9) Platforms.

• Software Upgradable for ANT and Bluetooth Low Energy. • FCC/IC/CE Certified.

• Operating Temperature Range: -40 to 85. Wireless specification:

Standards Support: Bluetooth v2.1 + (EDR)IEEE 802.11 b/g/n Compliant Typical WLAN Transmit power: +20dBm, 11Mbps, CCK (b)+14.5dBm, 54Mbps, OFDM (g)

+12.5dBm, 65Mbps, OFDM (n) Typical WLAN Receiver sensitivity: -89dBm, 8% PER, 11Mbps-76dBm, 10% PER, 54Mbps

-73dBm, 10% PER, 65Mbps

Bluetooth v2.1 + (EDR) Bluetooth Transmit Power: +9.5dBm Typical_{-92dBm typical Bluetooth Receiver Sensitivity}

8.3. WIFI ALTERNATIVES

8.3.2 CC3000

Texas Instruments CC3000 is a 802.11b/g wifi module that is specified to work off-the-shelf with the Cortex-M4 option that was presented in the MCU section. Since the CC3000 only is a 802.11b/g system it will only be able to supply 11 Mbps and at the same time require one of the SPI ports. One downside of the combination of the Cortex-M4 and CC3000 is that the connecting data bus is a APB bridge, meaning that the maximum level of speed will not be acquired, it can not be verified that the engineers at Texas Instruments have made a correct balance between the data flows from the cortex-M4 and the CC3000 by the usage of the APB. But this also gives a clear indication that the CC3000 will not be the solution that will supply the maximum data transfer, but that may be weighed against that the two parts are constructed for each other and should work immediately off-the-shelf. Below are a few specifications from the manufacturers homepage [60].

• IEEE 802.11 b/g.

• Embedded IPv4 TCP/IP stack.

• TX power: +18.0 dBm at 11 Mbps, CCK. • RX sensitivity: -88 dBm, 8% PER, 11 Mbps.

• Works with low MIPS and low-cost MCUs with compact memory

footprint.

• FCC, IC, and CE certified with a chip antenna. • Integrated crystal and power management.

• Complete platform solution including user and porting guides,

Ap-plication Programming Interface (API) guide, sample apAp-plications, and support community.

8.3.3 Compact-S-UART

The company Wiicom has a wifi module by the name of Compact-S-UART, this modules main downside is the UART based connectivity meaning the fastes bus type will not be utilized throughout the connected MCU. Unless the MCU has the RS485 connection options that will allow greater internal data speed. But as stated in the data sheet the module is specified to a interface with the maximum speed of 2.5 Mbps that will limit the data transfer and in the end disqualify the wifi module as a possible candidate, even though it should be a easy task to connect it to any of the MCUs. Below are a few specifications from the manufacturers homepage [61] of interest.

• Compliant to 802.11b/g and single stream 802.11n. • Host interface through USDB.

• Ad-hoc and infrastructure operation modes support. • WPA/WPA2-PSK and WEP security modes support. • WEB Server on port 80.

• TCP and UDP protocol support. • Power-saving management.

Wireless specification:

Data Interface USDB support up to 2.5 Mbps (without overhead) RF Frequency Band 2.412 - 2.484 GHz (integrated antenna)

Data Memory On Board FLASH up to 32 Mbit Security Protocol WEP, WPA and WPA2-PSK WLAN Functions Ad-hoc and infrastructure modes

Network Protocols TCP, UDP, IPv4, ARP, ICMP, DHCP Clientv Network Standard Support IEEE 802.11b/g single stream n Operating Temperature Industrial -40 to + 85

RF Data Rates: 802.11n: 6.5, 13, 19.5, 26, 39, 52, 58.5, 65 Mbps802.11g: 6, 9, 12, 18, 24, 36, 48, 54 Mbps 802.11b: 1, 2, 5.5, 11 Mbps

Supply Voltage 4 - 10 Vdc (typical 5V) RF Output Power 15-17 dBm

Dimensions 42 x 23 x 5 mm

8.3. WIFI ALTERNATIVES

8.3.4 TiWi-SL

This self contained 802.11 b/g 54Mbps wifi module from SL Research is interfaced by SPI and is specificity constructed for simple connectivity with MCUs. It is also fully functional and compatible with TI’s Stellaris Cortex-M4 with TI SimpleLink technology and several other MCU types. Thereby it is a far better alternative then the CC3000 since it is a more general wifi module with higher speed capabilities but with the same tools as the CC3000. The main downside is that the support forum offered by SLR is lacking in comparison with TI. Below are a few interesting specifications from the manufacturers homepage [62].

• Self-contained solution with Wi-Fi driver, security supplicant and

TCP/IP stack all on-module.

• Small host footprint – as low as 6KB Flash, 3KB RAM.

• Low memory footprint enables Wi-Fi connectivity with MCUs. • Low MIPS/CPU overhead.

• Extremely simple API, primarily BSD socket; about 35 total APIs. • Universal IP connectivity enables Wi-Fi anywhere.

• Quick Wi-Fi implementation without previous Wi-Fi or RF

experi-ence.

• Quickly portable to any low-cost, low-power MCU platform. Wireless specification:

Standards Support IEEE 802.11 b/g Host Interface SPI

Vcc Min/Max 2.9 Volts / 3.6 Volts Temp Range -40 to 85

Transmit Power 20.0 dBm, 11 Mbps, CCK (b)_{16.9 dBm, 54 Mbps, OFDM (g)} Rx Sensitivity -89 dBm, 8% PER, 11 Mbps_{-76 dBm, 10% PER, 54 Mbps} Transmit Current 269 mA, CCK (b)_{187 mA, OFDM (g)}

Receive Current 92 mA b,g Power Down Current <1 uA

8.3.5 WizFi630

The WizFi630 from the company WIZnet is a very interesting wifi module that will allow data transfer speeds up to 150Mbps (90Mbps effective rate) that by far surpasses the other alternatives. This module can only be used with the AM3715 alternative from TI since it is interfaced by an Ethernet interface. The Ethernet connection is a big benefit for using the WizFi630 with the AM3715, since that will mean all four SPI ports may be used for the AFE.

The WizFi630 has several other connection possibilities but not something that will be usable to fulfil the stated requirements. All other connections will result in restraints on the transfer speeds, another concern is the lack of documentation that is basically non existent. Below are a few specifications of interest from the manufacturers homepage [63].

• Complies with IEEE802.11b/g/n

• Gateway/AP(Bridge)/AP-Client/Client(Station)/ • Ad-hoc Mode, WDS/Repeater supports

• Physical link rate up to 150Mpbs • Built-in 3 Ethernet ports

• 802.1x (only in AP mode)

• 802.11e and WMM (Wi-Fi multimedia) Wireless specification: Standards Support IEEE802.11b/g/n

Host Interface UART and Ethernet Output power 802.11b: 17dBm@11Mbps802.11g: 14dBm@54Mbps 802.11n: 14dBm@150Mbps/72Mbps Receive sensitivity 802.11b: -89dBm@11Mbps 802.11g: -74dBm@54Mbps 802.11n (40MHz): -66dBm@150Mbps 802.11n (20MHz): -70dBm@72Mbps Data rates 802.11b: 1,2,5.5,11Mbps 802.11g: 6,9,12,18,24,36,48,54Mbps 802.11n (20MHz): 7,14.5,21.5,28.5,43.5,57.5,65,72Mbps 802.11n(40MHz): 29.5,86.5,115,130,144,150Mbps

8.3. WIFI ALTERNATIVES

8.3.6 TiWi5

Another alternative from SL Research is the TiWi5 65Mbps wifi module, which con-nected by SDIO instead of SPI and will thereby allow the full usage of the available SPI ports for connecting the AFEs. The AM3715 can use this wifi module, and thereby it can also be used as backup during the prototyping by having one or the other as an alternative in the case of unforeseen problems. Below and on the next page are a few specifications from the manufacturers homepage [64].

• IEEE 802.11 a/b/g/n.

• Bluetooth 2.1+EDR, Power Class 1.5. • Full support for BT4.0 BLE and ANT. • Terminal for PCB/Chip antenna feeds. • Integrated band-pass filters.

• Cost saving module level certification accepted worldwide: FCC

(USA), IC (Canada), and CE (Europe).

• Compact design based on Texas Instruments WL1273L Transceiver. • Seamless integration with TI OMAP and Sitara application

proces-sors.

• SDIO host data path interfaces. • Low power operation modes. • RoHS compliant.

Wireless specification:

Standards Support IEEE 802.11 a/b/g/n (2.4 and 5.8 GHz)Bluetooth 2.1+EDR, Class 1.5 Bluetooth 4.0 (BLE)

Host Interface SDIO

Vcc Min 3.0 Volts

Vcc Max 4.8 Volts

Temp Range -40 to 85

Transmit Power (2.4 GHz) 18.3 dBm, 11 Mbps, CCK (b)14.4 dBm, 54 Mbps, OFDM (g) 12.5 dBm, 65 Mbps, OFDM (n) Transmit Power (5.8 GHz) 17.8 dBm, 9 Mbps, OFDM (a) 15.2 dBm, 54 Mbps, OFDM (a) 18.2 dBm, 6.5 Mbps, MCS0 OFDM (a) 13.5 dBm, 65 Mbps, MCS7 OFDM (a) Rx Sensitivity (2.4 GHz) -88 dBm, 8% PER, 11 Mbps, CCK (b)-74 dBm, 10% PER, 54 Mbps, OFDM (g)

-72 dBm, 10% PER, 65 Mbps, OFDM (n) Rx Sensitivity (5.8 GHz)

-87 dBm, 10% PER, 9 Mbps, OFDM (a) -72 dBm, 10% PER, 54 Mbps, OFDM (a)

-88 dBm, 10% PER, 6.5 Mbps, MCS0 OFDM (a) -70 dBm, 10% PER, 65 Mbps, MCS7 OFDM (a) Transmit Current (2.4 GHz) 247 mA, 11 Mbps, CCK (b)180 mA, 54 Mbps, OFDM (g)

166 mA, 65 Mbps, OFDM (n) Transmit Current (5.8 GHz)

296 mA, 9 Mbps, OFDM (a) 235 mA, 54 Mbps, OFDM (a)

298 mA, 6.5 Mbps, MCS0 OFDM (a) 219 mA, 65 Mbps, MCS7 OFDM (a) Receive Current (2.4 GHz) 93 mA

Receive Current (5.8 GHz) 100 mA

8.4. WIFI COMPARISON

8.4

Wifi comparison

Wireless comparison:

Module: Standards: Transfer (Mbps): Host Interface: Consumption (mA):

WL1271-TIWI 802.11 b/g/n 65 SDIO 165

CC3000 802.11 b/g 11 SPI 207

TiWi-SL 802.11 b/g 54 SPI 187

Compact-S 802.11 b/g 2.5 UART 130

WizFi630 802.11 b/g/n 150 UART, Ethernet ?

TiWi5 802.11 a/b/g/n 65 SDIO 219

Table 8.6. Wifi comparison.

The transfer speed is not the only aspect to consider when deciding on a Wifi module as is shown in the above table, but also the ease of implementation as well as which internal connectivity that will be utilized. If the AFE will be connected by the SPI it will be beneficial if the Wifi module also uses SPI, because of similarities in implementation and the possibility to have the data transfers on the same bus matrix.

Therefore the TiWi-SL or the CC3000 is the best option in my opinion for this thesis, there is also a technical specification that it will pass all requirements in both the data transfer as well as the usage in combination with TIs different MCUs, since Atmels alternative no longer is an option. The TiWi-SL is just a faster ver-sion than the CC3000, but the information and support from TI may be of higher value since both passes the speed requirements. Avaliable on TIs wiki page is all the needed documentation and example codes for implementing the TiWi-SL and the CC3000, there is even a recommendation for the usage of the access point(AP) TP-Link WR740N [65].

Chapter 9

Design proposals

The first research goal is to chose system hardware modules that will both fulfil all the requirements and make the implementation as simple and straight forward as possible. These first parts of the report has been created to try and be proactive and foresee any shortcomings of the hardware decision, below are the two designs of my choice presented.

9.1

Alpha design

The usage of the RHD2132 has already been recommended in the AFE section, RHD2132 in combination with the LM4F and the CC3000 will all work of the shelf according to the documentation. LM4F is recommended by TI to work with CC3000 and all software and tutorials are accessible through www.ti.com and this will mean that all the hardware will be supplied by the same manufacturer. The main benefit is that no blame will be able to be placed on the other manufacturer if the components where to be supplied by two different manufacturers. Integrum is already working with TI which may be beneficial. TI also has an extensive documentation, both in technical specification but also in accessible code and tutorials that should be able to help during the thesis. TI also has a wiki site just for the usage of the CC3000, because of all these benefits the choice falls on the CC3000 over the TiWi-SL.

9.2

Beta design

This final design proposal is composed of the AM3517 and the TIWI5 that are both specified to work of the shelf with each other. The big downside of this proposal is that the AM3517 is a overkill in the department of computational power and with several inbuilt options that are of no use for this thesis and maybe not even in future development designs. Note worthy is that AM3517 has several alternatives in connectivity that will leave the platform open for improvements in the future development. This design proposal also contain the RHD2132, the system will thereby consist of components that are verified to work together of the shelf.

9.3

Chosen design

In discussion with the leading engineers of Integrum it was decided to implement design option Alpha. In discussion when deciding between the RHD2216 and RHD2132 it was clear that the higher number of channels was of more importance then the right leg drive. Apparently the LM4F MCU was already being used in related projects and was accepted immediately. When it came to the Wifi module, the decision was made that because of the use of a TI MCU the implementation and integration between the MCU and Wifi should be simpler if TI CC3000 was chosen in comparison to the TiWi-SL.

This ends the first phase of the thesis involving information gathering and research, and choosing the final design based on that information and research.