I finished the Velassi version 2 in January 2022 (the first Velassi -without motor- was finished in April 2019 and rode about 5600 km).
This new model corrects some errors of the first version (let’s call it V1) and is intended exclusively for electric drive (Tongsheng TSDZ2).
The main changes:
- Steering as vertical as possible for greater stability at low speeds
- 20” rear wheel (instead of 26”) and suspension with pneumatic shock absorber (instead of rubber blocks)
- Slightly longer seat ramp to accommodate taller riders
- Slightly higher bottom bracket and reclining seat
- Rider closer to the front wheel (short cranks only), but the wheelbase has been retained to provide a sufficiently advanced center of gravity when loaded or with a taller rider.
- Reinforced rear structure to accept a luggage rack.
- Structure modified for greater torsional rigidity.
- Pine and fir structure instead of oak for better bonding and lower weight (but larger cross-sections). However the greater frame length and reinforcement ‘swallowed’ the weight gain.
- Bottom bracket layout modified to accept a mid-drive motor.
- Handlebar pivot bearings set further apart, as the stress on the handlebars is very high (you tend to pull on them).
- Better quality brake cables and sheaths, and adjustable pull levers (on the V1, water was getting into the sheaths and causing malfunction, so much so that I ended up installing a hydraulic rim brake at the rear).
- Single-roller chain tensioner.
- A kickstand
- Chain guard and guide tubes
I’m keeping the Nexus 8 gearbox, which should be much better suited to a mid-drive motor than it was to a muscle bike (poor low-power efficiency and insufficient range). On the other hand, I’m switching to a narrow chain, as the 1/8” chain doesn’t pass through guide tubes.
The handlebar and rear suspension bearings are still IGUS plastic bearings with a 16mm-diameter axle, but at the rear they are now spherical bearings for better load distribution and to take account of construction inaccuracy.
There is no noticeable wear to the shaft or bearings on the current Velassi suspension, but the stresses will be greater on version 2 (larger lever + luggage). The first version of the handlebar pivot shaft on the current Velassi, on the other hand, has been severely worn, but the steering was poorly designed and had to be modified.
The softwood construction was much easier, as sanding is easy. My laths were also straighter, which facilitated assembly and good geometry.
I had thought of making a mounting marble and didn’t do it for the main frame. It was a mistake and for the rear chassis assembly I’m going to make a marble to compensate for the rear suspension arm misalignment.
On the other hand, the characteristics of commercial laths are extremely variable, with weight variations of over 25%. By sorting the laths by weight, I saved over 350g. A less dense lath will, however, absorb more resin (which is favorable for bonding).
The 3D model is made in OpenSCAD and is an evolution of the previous model, but I still spent a lot of time on this new model.
I’ve been using OpenSCAD for seven years now, and I’ve become quite comfortable with it. Above all, I use my own library, which I created at the start, and which saves me a lot of time.
The model is parametric, but there are so many parameters that modifying the bike to make it a different size takes quite a lot of time.
I had originally intended to publish this model, but making it out of wood is relatively tricky, and I’m moving towards publishing a riveted aluminum construction that seems more within the reach of the average DIYer, safer and less restrictive. I’m going to try out this kind of construction on another type of bike (a compact recumbent with top steering for urban use), but the geometry of the Velassi V2 seems right for a touring bike, and it seems interesting to publish a version that’s easier to build.
The shapes of the wooden parts are projected flat in DXF format, then dimensioned in 2D software (at the time with NanoCAD -Russian software- but I switched to LibreCAD by running Linux).
Manufacturing process
The aluminum parts are cut with a disc cutter and a hand-held hacksaw, drilled and then sanded with my tank/disc combination. There’s quite a bit of sanding involved, and I weigh the parts as I make them. It’s better to drill at the beginning as long as you have flat surfaces. A drill press is essential and I recommend a fairly large model (like mine, 80mm diameter column, below that it’s too flexible). An improvement would be a drill with a variable speed drive. I had made the aluminum parts long before starting construction.
For the wooden parts, I print on paper the shape of my parts from the DXF plan. I glue the paper to the wood (gluing the wood, not the sheet, as this greatly increases the size of the paper). Because the humidity of the glue changes the size of the paper, I have to check the dimensions carefully. Straight cuts are made with a plunge saw (Festool TS55, a fantastic machine) and small cuts with a radial saw, sometimes using a hand saw for safety reasons.
The parts are assembled with 3mm wood screws, which are used to position the parts and hold them in place during gluing.
I had originally intended to remove the screws, but I gave up because it’s tricky: the glue has to be hard enough, but not too hard, because then you can’t remove the screws. Basically, you have to remove them after four to six hours of polymerization with the resin I’m using. On the Velassi V1, I had paraffined the screws to make them easier to remove, but sometimes it’s impossible. As a result, I’ve abandoned the idea of removing the screws, but I regret it a little because in some places it’s rather unsightly. I first make a ‘blank’ assembly and then start sanding (or planing the corners) on this blank assembly. The parts must be completely sanded before gluing (because once you’ve started the resin, you can’t sand any more - it’s toxic for the duration of the work). You need to prepare the assembly very carefully before you start gluing: for adhesion reasons, you need to apply one coat of resin per day, as the resin doesn’t stick well to itself once it’s fully cured (after a week). Doing one coat a day ensures chemical adhesion between coats. As I can’t sand between coats because of the resin’s toxicity, I make the necessary adjustments with a cutter.
To apply a new layer over a fully cured one, you’ll need to sand/scrape (with a cutter or hacksaw blade) the cured layer to ensure mechanical adhesion between the layers. If it’s for structural bonding (as I had to do to add the luggage rack support), you need to remove all the resin and return to the wood (with a rasp, for example). It’s important to bear in mind that some elements are always forgotten (e.g. inserts for attaching a steering spring, a bottle holder, cable guides, a reinforcement piece…) and that this extra work has to be done immediately, while the resin is in the final polymerization stage.
Moisture resistance
To be moisture-resistant, an epoxy wood structure needs three coats of epoxy resin and two coats of varnish or polyurethane paint. The principle is to block any transfer of moisture with the wood and be totally waterproof. The Velassi V1 didn’t receive these three coats of resin and was never varnished, because I thought (rightly) that I’d have to make modifications, and the moisture penetrated the resin layer and grayed out the wood in certain places. It’s also worth remembering that unsealed wood changes dimensions, and I’ve had problems with my steering bearings because of this (also because I had to remove the resin in certain places to adjust the fork, so I was no longer watertight at the steering ‘column’). So I took care of this on the Velassi V2. The bike was completely disassembled after 2500 km, modified a little to fit panniers on the front and given two coats of two-component matt polyurethane varnish. I used two full cans (once the two components have been mixed, the can lasts 48 hours, but as you can apply a coat every fifteen minutes or so, you don’t have to wait that long).
Provisional assessment
This new version is much improved, more stable and easier to drive. The impulse given by the motor at start-up also makes it easier to use, and you’re quickly up to a speed where the bike is more stable.
Despite the smaller rear wheel, the pneumatic shock absorber ensures incomparably superior comfort. The smaller wheel also allows much better passage of the cable sheaths, which no longer suffer when the rear wheel frame is folded.
On the other hand, small amounts of damping are not taken up by the shock absorber, as there is too much internal friction, so you need (which is the case) big tires (width 55 front and rear). The rear tire (Schwalbe GT365) has excellent grip on mud compared to the Schwalbe Marathon on the Velassi V1.
At the moment, the shock absorber doesn’t seem to be deflating. Its set pressure is around 8 bar, so I can eventually reinflate it with an ordinary mini-pump.
The choice to reduce the rear wheel diameter (from 26” to 20”) was also linked to the installation of a mid-drive motor, and I wanted to reduce the efforts on the transmission (proportional to the wheel diameter). I also left the motor’s power limitation (as delivered) with a maximum current set at 10A, which corresponds roughly to a maximum power output of 400W. This motor can supply much more (around 700W like most mid-drive motors) but my desire to limit effort was welcome, as the first chain only lasted 600km and the gearbox seems to suffer a lot in off-road use. I find this power of just over 400W more than sufficient for my purposes. Further chains (strongair model) laster 3600 km.
While I had many problems with brakes with the Vélassi V1, with the use of better components (especially sheath and cables) on the V2, the V-brake are still operating perfectly with high power – due to the adjustable pull levers.
Crankset width with a mid-drive motor for an ordinary gearbox
The width of a crankset is defined by measuring the distance between the right and left cranks (on the outside) where the pedals are screwed on.
On an ordinary bike, this width (called the ‘Q Factor’) is 168mm. Smaller riders may prefer a narrower width, and some bikes may have a reduced Q-factor of 156mm.
Cranks have an “offset” which is the distance between the face where the pedal is screwed on and the face at the connection to the bottom bracket axle. Typically, this offset is 15mm. It’s easy to reduce this offset on the left-hand side, but on the right-hand side of a muscular bike, the size of the chainrings means that space is at a premium. Reducing the offset on just one side creates an asymmetry that can cause problems.
A mid-drive motor that fits into an ordinary bottom bracket is much wider than an ordinary crankset (by more than 25mm), as the main gear has to be housed somewhere. What’s more, the bottom bracket is often offset to the right, and this in itself can cause problems.
The TSDZ2 comes (unlike its competitors) with 170mm cranks with ‘normal’ offset (15mm), which leads to a very high Q factor of 210mm.
Initially, I fitted shorter cranks (152mm) but with the same offset as the original ones, and thus obtained the same spacing.
Although the muscular effort is reduced with a mid-drive motor, this large width caused me major problems (persistent knee pain). I then replaced these cranks with 140mm long zero offset cranks, which no longer cause me problems, but which I find a little short. The Velassi V1 is fitted with 145mm cranks (replacing 152mm cranks) which fit me perfectly, but it’s not possible to find zero offset cranks of this length. To get short cranks on crankset motors, you mount unicycle cranks (which always have zero offset) or kids’ bike cranks without offset (by ripping off the original chainring, which is what I did for my cranks).
As far as I’m concerned, this motor (the TSDZ2) is simply unusable with its original cranks for people of average or small stature. Some dealers are aware of this and offer competing cranks (Bafang) with zero offset.
Fine-tuning
There were several stages of fine-tuning.
Parts were added to the frame:
A diagonal to reinforce the frame’s torsional rigidity.
Support the rear saddlebags and keep them spread out.
Front saddlebags (which are handlebar saddlebags).
Fit a steering spring
Put the bottle cage further back (the heel tended to touch)
Other elements have been modified:
152mm normal offset cranks have been replaced by 140mm zero offset cranks (due to knee pain).
A carbon tube has been inserted into the aluminum crutch articulation tube, which had become deformed.
The bottom bracket was sanded to bring the motor back to the left by about 7mm, partially correcting the problem of bottom bracket eccentricity (which significantly improved comfort). This also facilitates passage of the steering rod.
The front light has been repositioned behind the steering column axle to make it easier to transport on a rack and set up inside my tent (sisi!).
A mud flap has been fitted to the rear mudguard (for the same reasons).
The steering rod has been modified (conflict with the front brake) and now has a double bend to improve the angle of rotation.
The aluminum chain guard has been replaced by a plastic one to reduce noise when the bike jerks.
The chain tensioner has been replaced.
The chain has been replaced by a much stronger one.
The motor control board has been replaced (I’ll do a detailed explanation on the motor separately) with excellent results.
The bottom bracket has been lubricated with kerosene, as the bike used to squeak a lot before being dismantled for varnishing.
A new two-output USB converter has been added (under the battery isolator).
Dashcams (front and rear)
My self-built bike is of course not homologated as per European regulation -so I can’t resell it-. To be able to justify that my assistance doesn’t help beyond 25km/h and my actual speed in the event of an accident (to be sure I will be insured), I ride permanently with dashcams (front/rear) switched on, equipped with a GPS which records a speed watermark in the films. You can see that assistance ceases above 25 km/h, as acceleration drops off sharply, and in any case on the flat I drive at around 22/23 km/h at the assistance limit. The dashcam’s power consumption is quite high (around 7W), but that’s not a problem on an electric bike. They are heavy, with the whole unit weighing in at over 700g. The dashcams start up when I power up the bike, and I get a (barely audible) voice message indicating that they are operational.
