Exercise Machine Brake System - US Patent 12017107
To deliver this piece of modern art, containing immersive guidance on a highly efficient form of whole body exercise, into homes, we had to design a new drivetrain which would fit in to a very small circle at the front of the machine. The original rower’s fully closed envelope leaves space for a gear stage. As you can see in the image below, the original rower contained a gear stage, which would never fit in to the small circle left in the Wave rower form. All electronics, our eddy current brake, any excess webbing, and an exhaust fan, would have to fit in this new small circular area.
Maintaining Flywheel Speed The Hydrow Wave would continue to use an eddy current braking system for resistance. Such a resistance mechanism contains a cast iron flywheel with a steel core, which spins in a magnetic field generated by two electromagnets. In our case, each electromagnet consists of electrical current flowing through two copper wire coils wrapped around a three-pronged core. The stength of the magnetic field generated by each electromagnet is determined by the wire thickness, number of wire wraps around the ferrous core, and the amperage flowing through the copper wires.
Given a contant magnetic field strength, the resistance torque produced at the eddy current brake is primarily determined by the thickness and radius of the flywheel core, the distance between the flywheel core and the electromagnet cores, and the rotational speed of the flywheel.
Simply eliminating the Hydrow’s gear stage would mean an almost three times reduction in the rotational speed of the flywheel given the same linear handle speed.
Hydrow’s original brake had been designed by a CM, and so initially there was some interest in experimenting with copper wire thickness, flywheel core geometry or material, or perhaps adding additional electromagnets. I was instructed to design a testing rig so that we could change these variables and learn from the process, given that a purely matematically derived solution was unlikely to map well to the real complex world of electromagnets and their effects on a spinning flywheel. When I produced a quote for the cost and time associated with building this rig, management decided to call off the process, and assume the same flywheel - electromagnet configuration as our original rower.
Thus, the only means of delivering the same resistance experience on the Hydrow Wave as on the original hydrow rower, given no gear stage, was to find a different means of reducing the drive webbing’s mechanical advantage on the flywheel system.
Spiral Scroll One way to reduce the mechanical advantage the drive webbing has over the flywheel is to reduce the radius of the scroll which the drive webbing wraps around. The original scroll has a radius of 41-42mm - to eliminate the drivetrain’s gear stage, we would have to reduce the scroll’s radius to 16-18mm.
I began with a quick-turn small radius scroll which I designed in CAD and printed on an Ender FDM printer in my bedroom. I installed this scroll on to an original hydrow rower that I took apart and retrofitted. The setup was awkward, but I was able to perceive a feeling of a ripple effect, where the resistance I felt at the handle felt ‘choppy’ and wrong.
After some pondering, I realized that every time the drive webbing stacks on top of its termination point the radius of the scroll makes a step change by the thickness of the webbing (1.2mm). On the original 42mm scroll, a 1.2mm change in radius would create a 2.9% in the linear drive force on the drive webbing (given a consistant torque between radii). On the new 15mm radius scroll, a 1.2mm change in radius creates an 8% change in linear force transmitted to the drive webbing.
As I rowed this smaller scroll, every time the scroll turned over and the webbing stacked (approximately 10 times per stroke), I experienced a 6-8% step change in the mechanical advantage I have over the drivetrain. Hydrow’s firmware already accounts for the stacking of webbing and its affect on mechanical advantage, and hydrow modulates the strength of the magnetic field in the rower’s eddy current brake in a speed-based model, so this step change in mechanical advantage does not translate directly to torque change. However the effect is jarring and undesirable.
An intern menitoned use of a spiral shaped scroll to smooth the transition between webbing wraps (shown in the below image at left), so I modelled a spiral scroll in CAD. To enable me to test rowing different scroll iterations more easily and quickly, I pulled an old testing rig out of storage, and retrofitted it to become a “works-like” rig for the braking system. I used this platform to test a rope-drive scroll whichI designed as well. The rope showed signs of wear after I rowed on this testing rig platform for only 45 minutes, so we discontinued development of the rope concept. The below image at right shows an integration of the spiral scroll in to the testing rig architecture. The scroll was integrated into a complex cnc-machined part with bearings and a clutch bearing. This complex scroll part was mounted to a long shaft that I designed for testing of a variety of assembly cases.
Using this testing rig pointed to three major areas of refinement which would become my primary focus for the next stage in the development process - 1. refining the scroll geometry to eliminate perceivable bumps, 2. choosing a layout for the bearings in the new brake assembly, and 3. designing a return spring which could retract the webbing back in to the rower for this new application.
Scroll Refinement My first step in refining the scroll was to choose a means of terminating the webbing. On the “works-like” rig, I had simply burned two holes in the drive webbing, and run screws through these holes and in to the scroll and center axle. Machining multiple tapped holes in to the scroll and drive shaft seemed like an inefficient drive solution. I drew up a few concepts, but ultimately chose to adapt the original hydrow’s thru-bolt solution (below, at left) to this new design (below, at right). In each below image, the yellow arrow-terminated line shows the wrapping patch of the drive webbing, and the green line indicates a stitched region between the drive webbing a doubled-back section which forms a loop around the scroll’s thru-bolt.
Adaptation of this layout to the new smaller-radius design required careful selection of the thru-bolt’s radius from the center rotational axis, the length/depth of the webbing stitching cutout, and the amount of clearance surrounding the thru bolt. All of these parameters were far more critical on the new lower radius design, as a small lump represented a large percentage of the overall radius and thus a perceptible change in the mechanical advatage a user would have on the drivetrain.
I designed a scroll, using the webbing thickness and stitched length to drive these critical parameters. I ordered a few parts to be SLA printed, and installed them in a prototype brake I had assembled at the time. Of course, this first-pass scroll still exhibited perceivable ripples in resistance while rowing.
So, I designed a test where myself and a coworker took measurements from a dial indicator at 16 points in the rotation of the scroll. From measuring an initial and final radius, we could use the change in dial indicator measurement to determine the absolute radius of the webbing stack at each point. Measuring the radius of the drive stack at 16 points over 3-4 rotations showed trends in the high and low points in the drive webbing’s wrapping. I ran this test multiple times to ensure good data.
In the below image, the dial indicator, which was held in place by an arm fixture, is represented in black. A yellow, arrow terminated, line shows the trajectory of the wrapped webbing
This test quantified the effects of a few realities of the design that I could not estimate without experimentally gathered data:
- In this test, the drive webbing was under the tension of a return spring, which retracts the drive webbing in to the rowing machine. This tension likely slightly reduces the thickness of the polyester drive webbing, and this affect was hard to estimate in magnitude without such a test.
- In this test, the drive webbing experience the sinching and compressive force of the multiple stacked layers of webbing present. This would also change the behaviour of the webbing’s stacking.
- I could not accurately estimate the extent to which the webbing sinks in to small gaps underneath. In theory, the webbing can bridge gaps if adequately small. For example, I should expect that a paper thin incision in the scroll would not affect the wrapping of the webbing. However, what about a 2mm gap between the end of the webbing stitching and the end of the clearance cutout designed in to the scroll ? what about a 3mm gap between the scroll’s thru bolt and a wall of the scroll ?
- I couldn’t be sure how the webbing would behave around the thru bolt.
This test gave me direct feedback on my design, and helped me to devise a few new 3D printed iterations of the drive scroll. Ultimatley, my management approved one of the designs I devised and we produced a tool for injection molding the scroll in mass-production.
Brake Layout, esp BearingsHydrow’s original braking mechanism consisted of a main shaft supported by two bearings mounted in to frame-mounted steel plates, and a static connection between this main shaft and the flywheel:
In a rowing machine, the user accellerates a flywheel during the rowing stroke, and then the user has to retract the handle into the machine to begin the next stroke. During this retraction phase, the flywheel must continue spinning, decoupled from the user’s retraction motion. This means the rowing machine’s drivetrain must contain a one-way clutch somewhere in the system.
The original rower nested this clutch in one of the components in its gear stage. Since the Wave drivetrain would contain no such gear stage, I designed a means of nesting a needle roller clutch bearing inside the brake itself.
To test this configuration, I built a few prototype resistance units. I disassembled I think close to ten original hydrow resistance units, and brought the flywheels from them to a local machine shop. I sent the machinist ‘before’ and ‘after’ 3D files of the flywheels, and a 3D print with critical dimensions and tolerances. I ordered the large steel assembly plates from a local waterjet supplier, and made up a package of 3D files with 2D prints for a local machine shop to manufacture bearing cups and shafts. I ordered bearings and fastners from the web, ordered scrolls to be 3D printed.
Critical Tolerances Commentary
Critical tolerances were exptremely important in this assembly. For example, when these flywheels were post machined, the machinist made three holes - one for the small non-drive side bearing (seen to the left of the needle roller clutch in the above image), one for the needle roller itself, and one for the large drive-side bearing. These ball bearings to the left and right of the needle roller clutch bearing ensure that the shaft spins in the center of the needle roller clutch. These bearings bear the ~35 lbs of vertical load of the flywheel, and locate the shaft to be concentric with the needle roller clutch, so the needle roller clutch can resist relative rotation between the shaft and flywheel in one direction. What this all means is the two bearing bores use an H7 tolerance on their nominal diameters, and the center bore for the needle roller is specified with an N6 fit. There must also be a position (per ASME 2018) specification for this machining process - I do not remember what I used exactly in this design, but based on the tolerances of the system if I was speccing this part today I would call the machinist in question how they made such measurements for position (do they have a CMM in-house, would they have to fabricate a tool / jig for the application) and get a cost estimate for a position spec relating each bearing bore’s axis (measured surface) to the central needle roller bore’s axis (datum) each to within 5 microns, as a starting point.
The two supporting bearings nested in cups on either side of the assembly also required some careful consideration. As you can see, the drive-side bearing is larger than the non-drive bearing because the drive-side bearing sees increased radial load. What you can’t see in this 3D file is the drive side bearing is held in its cup with a K6 trasition fit, while the non-drive side bearing is held in its cup with a F7 precision roll fit. These two bearings have a relatively fixed location on the shaft, so they cannot also have a fixed location in the bearing cup and mounting plate assembly. The sliding fit in the non-drive side bearing cup creates a ‘pin and roller’ location scheme, and prevents axial overconstraint in the system.
In the prototyping phase, I certainly made mistakes with tolerance stackups - the first prototype produced a strange grinding in some hard-to characterize conditions. I assumed bearing constraints to be the culprit, and after a few configuration iterations I had a brake that spun freely and sustained high loads under rowing. We took this layout (in the image above) to production, and to my knowlege, after selling somewhere around 40 thousand units, the production Hydrow Wave brake had very few if any defects with bearing life or wear.
Return Spring Design During the design of the Wave brake system, I sourced 25-30 iterations of the return spring. In the original Hydrow rower, the ~80mm diameter drive webbing scroll turned 4-5 times per stroke depending on user rowing stroke length. The gear stage translated these 4-5 rotations in to 11-12 roations at the brake flywheel. Without any gearing stage, a return spring mechanism would have to turn with the main shaft 11-12 times. I considered designing a bungee, or sprung rope to wrap around the shaft and pull the handle back into the machine at the end of each stroke, but this would require additional shaft length, and area to store excess bungee. So, optimizing a stainless steel spiral power spring for more turns became the primary path.
A spiral power spring has a few key parameters:
- Spring thickness: this is the thickness of the stainless steel sheet stock used to form the spring
- Spring length: this is the total length of stainless steel material used to form the spring
- Spring width: this is the width of the stainless steel material used to form the spring (shown as the width “in to the page” in the image above)
- Spring OD: the spring is installed in to a housing, and the inner diameter of that housing defines the spring’s OD. The springs in the above image are held to form using thin metal wires, but for example’s sake I am speaking about their properties as if these springs were installed in a housing
- Body coils: when the spring is at rest, in general there will be many wraps of spring material resting against the outer housing. These are called body coils. When the spring is wound during a rowing stroke, these coils will peel one by one away from the outer housing.
- Free coils: when the spring is at rest, there will be a few wraps suspended between the center shaft and the body coils. These are called free turns or free coils.
The application of each power spring also contributes a few key parameters:
- Preload turns: The application requires a minimum torque, at the location of minimum spring rotation. For a rowing machine, this is the minimum required spring torque to hold the hydrow rowin machine handle in place in the rowing machine without any user interaction. The number of spring turns required to create this torque output is the number of preload turns. this number of preload turns varies based on the spring stiffness.
- Working turns: The application requires the drivetrain’s main drive shaft to make a certain number of ‘working turns.’ In our application, the spring had to make 12 turns at a maximum per stroke, thus definining the ‘working turns’ at 12.
How to affect spiral power spring parameters:
Each spring is made from a piece of rectangular stainless steel. The cross section of steel used defines spring thickness and width, while the overall length of steel stock used defines the length. The material used to form each spring is cold worked by one of two methods. Some springs are formed by tightly wrapping the spring material around a shaft, and fixing them in place for a short period of time (seconds, or at most minutes) In this process, the size of shaft used to wrap the spring will affect how much of the spring presents in the final design as “free coils” vs “body coils.” A smaller shaft will produce more free coils and fewer body coils, while a larger shaft will have the reverse affect.
Another method involves feeding the spring through three rollers. In this configuration shown below, raising the center roller will bend the spring material more, again creating more free coils and fewer body coils in the spring’s resting position.
I further lengthened the spring, which did effectively delay the phenomenon where the final spring body coil peeled away from the spring housing and the spring formed this violent ball around the brake shaft. However, at a certain point there was so many coils of material in the sprint housing that the layer-layer sticking and slipping would sychonize by a sort of harmonic oscillation in to popping movements throughout the spring, which had a similarly violent effect compared to the original problem. I experimented with a few different greases, varying grease additives and overall viscosity. A Dow Corning MOLYKOTE NGLI #2 grease yielded better results than other viscosities and additives. However, the spring noise never reached an acceptable level.
I also experimented with using both thicker, narrower stock and thinner, wider stainless stock from which to form the springs.
Ultimately, we elected to design in a small approximately 2:1 gear stage connected to the main brake shaft using a small timing belt tensioned with an eccentric tensioner. This design change meant the spring makes only 5-6 rotations per stroke rather than the original 10-12. We increased the width of the spring significantly, enabling the spring to backdrive the inertia of the shaft and bearings. This design passed our 3 million cycle testing requirement, so we took it to production.
Finalizing the return spring design concluded the mechanical development of the Hydrow Wave Brake system. We applied for and acheived US patent 12017107 for this design. I worked alongside a validation engineer to run tests of multiple units of this design to 3 million rowing cycles. During the validation of the design we changed scroll materials a few times, and tested a few different webbing materials and thicknesses.
We released the design to two manufacturers, seperately, for supply chain diversity. The design passed smoothly through DFM review, and was subsequently released for production. I worked with our quality engineer to ensure the SOP instructions at our CM called out important assembly instructions, as well as correct torque setting where critical, and proper part numbers of specified retaining compounds. Approximately 40,000 units have sold, with almost no mechanical defects in the field.