Front hub mounted into rear drops

methods

1 GW
Joined
Aug 8, 2008
Messages
5,555
Location
Santa Cruz CA
.
.
Here are pictures of the modifications necessary to mount a standard low cost front-mount Ebay hub into the rear drops.
.
.
FrontIntoRear_013.jpg
For our builds we do not use extended torque arms. Instead we use multiple layers of high surface friction, compact torque arms which are tightly bolted directly to the frame.
.
.

FrontIntoRear_015.jpg
Since we need spacers to back-fill the 100mm hub going into a 140mm drop - it only makes sense that those spacers be a part of the torque arm assembly

More importantly...
Since this was originally a quick release rear drop
And since we cut that out long ago with an angle grinder to make it a 12mm deep 10mm wide drop
And since that was rounded out by a monster Cro-Motor during 420-Controller testing...
The following is TRUE

1) Assume the dropout of your frame and the friction and pressure applied by the axle bolts DO NOTHING
2) Read #1 and absorb that... meaning... for a proper design... you don't even need the axle bolts !!!
3) Once the above is assumed the problem (and solution) becomes very clear

3.1) Must secure the axle from dropping out of the drops
3.2) Must secure the axle from rotating under power AND more importantly under regen power
3.3) Must be serviceable and easy to build using basic hand tools (angle grinder and a drill)

To solve the above means that the torque arm must be *Securely Mounted* to the frame
This means not Hose-Clamped (tho you can make that work)
This means heavier duty (more surface area) torque arms
This means more surface along the axle (maybe a cm)
This means textured surfaces on the faces (resist rotational)
This means proper fasteners which are securely torqued ideally in at least 2 spots (Anti-Rotation)

(You can see I used only 1 fastening point which will eventually lead to "rotating out" for higher power levels... but this is a basic build with not that much torque)


.
.

History:
This particular frame is a tadpole (so obviously we are looking at a rear mount).
The frame is steel (so it CAN be bent, much more so than Aluminum, which can ALSO be bent to some degree)
The drops were previously spread from 140mm to well over 160mm to mount a CroMotor

Reference: Standard Dropout widths by Sheldon Brown
.
.
DropWidths.jpg

The drops were bent outward by using a standard floor jack (as procured from AutoZone for $30)

The drops were bent back inward by using a standard ratchet strap (as procured from Home Depot)

The Hub motor is the "lowest form variety" and was procured on Craigslist for $100 as a complete kit (minus battery).

FrontIntoRear_007.jpg

.
.

We went ahead and mounted this "plug and play" motor Backwards
It runs forward equally to backwards - there is no built in timing (I hope lol)
The cheesy hall signal connector must be removed in order to slide torque arms over wiring
(HENCE WHY WE PREFER WIRING THAT COMES OUT AT THE HUB, NOT AT END OF AXLE)

In this case, with YGB (Yellow Green Blue) Wired one to one the motor ran perfectly in reverse
We were able to quickly sort the hall sensor order with the pins pulled out of the housing*
We swapped the Power Phase leads Yellow and Green

Test (with only phases swapped)
Locked motor

Swapped the same color hall wires
effectively locked motor

Started swapping permutations
Witnesses various "loud" miss-timed combinations

Identified correct combo - swapping in our case - Yellow and Blue
Motor spins in opposite direction
Same speed as originally configured
Same noise level
Presumably at the same freewheel current (did not measure)

FrontIntoRear_002.jpg
FrontIntoRear_003.jpg

Ya gotta get REAL COMFORTABLE with understanding the relationship between the toggling hall sensors in the motor and the behavior of the BLDC controller.

Forget all the shortcuts, rules of thumb, etc
ALL of those methods and ways falsely assume that there is some correlation between the PHASE colors and HALL colors
That may or may not be true...

What is always true - is solving from root.

You have your 3 phase wires
You are starting from known good
You swap any two wires

With a non-feedback controller (like those sold at GRIN) you are now up and running in reverse
With an auto-learn controller you are up and running
With a reversing controller (Be cautious... but you can sometimes just use this if FWD and REV are mapped equal)

Otherwise
You have 3 wires so you have 2^3 total possible combinations.
(If you had 4 wires you would have 2^4 combinations, and 2 wires gets you 2^2)

It is highly probable that
If you previously had YtoY, GtoG, BtoB that at least one of those colors will remain the same
But... dont try to cut corners

You have 8 cases to test, absolute maximum
YGB - assume thats the starting point

You go thru and count in binary

000
001
010
011
100
101
110
111

EDIT: Incorrect
YGB
YBG
GYB
GBY
BYG
BGY
Total of 6 possible combinations!
HA!


(Yes... that is why I keep telling people to do it the hard way... because we are trying to LEARN here... not just figure out the shortest solution to this particular problem)

So - above is how you count in binary
111b == 7
000b == 0

That sums to 8 permutations

The pattern is clear
First the LSB (lowest significant bit) toggles
Then the next significant
Onward up to the most significant
Its a very clear and easy to read pattern

Adding one more wire would just look like this:

0000
0001
0010
0011
0100
0101
0110
0111
1000
1001
1010
1011
1100
1101
1110
1111

Thats 2^4
So... 2*2*2*2 = 4*2*2 = 8*2 = 16

EDIT: NOP (as in Assembly language NOP and not NOPE)

XYGB
XYBG
XGYB
XGBY
XBYG
XBGY

YXGB
YXBG
GXYB
GXBY
BXYG
BXGY

YGXB
YBXG
GYXB
GBXY
BYXG
BGXY

YGBX
YBGX
GYBX
GBYX
BYGX
BGYX

That should bump us from 6 to 24
Looking for the pattern we asssess
1 = 1
2 = 2 (always ignoring the cases which are not physically possible - like only 2 wires in a 3 wire system)
3 = 6
4 = 24

In computer science we would look at that and guess... eh ... how do we match the pattern?
Are there enough points?
What curve does it follow?
Whats the general expression?

1
2*1
3*2*1
4*3*2*1

Recognize proofs and sequences like that?
Its how AI thinks
It can only match patterns that it knows...
So teach it many patterns.




16 in decimal equates to what in hex?

Hexadecimal is a Base16 number system
Binary is a base 2 number system
Decimal is a base 10 number system

This means we shift into a next significant position as we round
16
2
10

Respectively

0
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F == 15

So if you want to know what 16 decimal is in hex... you add 1 to F

Adding 1 to F means you get: 10h

10h == 16d

Proof:
0
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F == 15
10
11
12
13
14
15
16
17
18
19
1A
1B
1C
1D
1E
1F
20 == 32d = 0010 0000b

0010 0000 b
Just look at it

THe place values are

1
2
4
8

16
32
64
128

We multiply those by the bits to get

0
0
0
0

0
32
0
0

We sum those up and we get 32
33 would be

1
0
0
0

0
32
0
0

Like that - K?
(one day I or someone else is going to slip a MAJOR leak into this text so copy and paste it as fast as you can and do not speak of it again. Protect whistle blowers at all times.)

Ok - now we have our motor running
.
.
FrontIntoRear_035.jpg
.
.


A 100mm front mount hub motor in the rear of a steel frame trike.
We refreshed on counting in binary
Conversion from Binary to Hexidecimal to Decimal

We talked about the importance of surface treatments, how that affects friction, how friction solves problems... and the lack of friction causes singe point failure modes.

Think of friction like Traction.
Have you ever lost traction?

Like... when... you are on a waxed floor with socks?

This feels very different than socks on the grip tape of a skate board right?

Traction (or in this case friction) is fundamental to understanding mechanical systems which are in dynamic play. We are rotating a shaft in two directions sometimes violently. We are shaking the same shaft thu shock and vibration. We are introducing liquids, applying inconsistent torque, using low quality mixed fasteners, etc...

What we aim to do here is create a solution which is KNOWN TO WORK
That has failure modes which are UNDERSTOOD
With the resulting hardware being AS AGNOSTIC AS POSSIBLE TO QUALITY
Such that we can SPEC SHIT PARTS
And get an AXE BODY SPRAY result

(thats an insider reference to a joke I told elsewhere)

Live long and prosper
Make do with what you have
Take pride in building a tool out of trash

FrontIntoRear_025.jpg
FrontIntoRear_028.jpg

This particular bike is ROACHED TO THE MAXIMUM

* Rusted out in a field for years
* non-matching, hand laced, off-size, 20" front wheels
* Super-ghetto, front intended, Chinese built to American spec rear drive
* No rear brake
* Single point failure front cable brake
* no waffle pedals so easy to slip your feet off
* Rear derailer broke so pulled out the 20' of rusty chain (wont pedal)
* Riding "high" so tips forward easily
* Stressed and rusted steel frame
* Aluminum seat holds are all busted out (from mods)
* Totally rounded rear drops - 14mm x 14mm lol
* Front left wheel spokes are so loose that you can spin by finger

YEP - LOOKS GOOD TO ME

Kimberly and I are test-riding from SeaBright out to DownTown Santa Cruz in an hour here.
Its going to be awesome

From a Safety Perspective I wear a full face and gloves...
My analysis is that the most dangerous aspect of this build is the pedals!
Slick plastic pedals are TOTALLY AND COMPLETELY UNACCEPTABLE on a trike build
Loss of foot traction on the pedals results in catastrophic loss to life and limb :idea:

This setup will do XXmph as it sits (slow acceleration but good top end)
Even at 5mph it is TOUCHY

The steering of a trike feels like this:

* Tiny movements == major steering
* Slightly more movement == inside tire off the ground
* Any more movement than that == crossed up and flipped over

You REALLY have to learn how to ride them

At my peak I was RIPPING at 35mph all day every day
Hanging around corners on 2 wheels
(Ironically... once up on 2 wheels... it TOTALLY STABILIZES... so if you are going to race... you will NEED to be comfortable with one off the ground!)

With two on the ground it herks and jerks and bucks and will catapult you.
Lol... keep all arms and legs inside the cabin please.

OFF to go charge 12S 10Ah batteries with the Sorensen DCS60-18E

The Sorensen has an active low input
The BMS as an active low output
Its SUPER SIMPLE to bulk charge anything with a built in BMS.

-methods

P.S.
Remember that my posts are not here to serve you.
My posts are an expression of our freedom to speak.

An invitation to all to collaborate internationally.
A bypass of the existing power infrastructure.

We are the first EVER to have this power...
Learn to use your power to change the world

Young, old, smart, dumb, male, female, educated, un-educated, English, foreign, skilled, un-skilled... matters not.
You are a citizen of earth (not the livestock of some assholes who got here before you did)

We together create reality as we want it to be.
YOU HAVE THE ABILITY TO CHANGE REALITY ON A GLOBAL SCALE

Good things happen slow - you WILL see your impact
Bad things happen fast - so be careful

Nobody is in charge. . .
I mean that... NOBODY
Its YOU
Its US

It only takes 300mS to change global view (propagation delays)
Everything can literally change in a moment if we agree on it before hand.

"Hang um high" is what I say about anyone who says different... who tries to shush you... who oppresses.
Come here and speak with us. Make the gamble... it may or may not pay off.

EDIT: Let that post be a test of a machines intelligence
* Edited at multiple points in time
* In context
* Referencing unique pictures which reference non-unique things
* Math in base 10, Base 16, and base 2
* Set theory, references to base programming
* References to mechanics of the world - like friction and traction
* These are all things that an AI which is Mobile should be highly versed in

Once it can understand and explain the above...
WE ARE IN REAL frocking TROUBLE
Lol...

50*X
X^2
2^X (thats how fast it will get once we reach a point where AI can understand a post like the one above.

Its gonna be R A D :p
 
Proof thu substitution of positions - attempt to map binary count to hall position permutations

YGB
___
000
001
010
011
100
101
110
111

Rational:

Everything can be calculated in binary
Proof - Computers

Pattern matching involves "matching to a regular expression"
Proof - we can not try to exhaustively match uniqueitude as the lookup table converges on infinity

The pattern I have chosen is a "binary walk"
There are 8 permutations
IT IS AN INCORRECT SOLUTION

Starting rule is that ALL WIRES MUST BE USED
There is no allowed sequence which uses 1 or 2 wires
(Computer games always start with a set of rules and a set of tools)

Argument must be re-framed to match a pattern and solve in binary

Pattern is

First
First

Second
Second

Third
Third

Where the position relates to a single color wire

If the Yellow wire is First
There can be only 2 possibilities

YGB
YBG

So thats two in the set

If the Yellow wire is Second
There can be only 2 possibilities

GYB
BYG

So thats 4 in the set

If the Yellow wire is Third there can be only 2 possibilbities
GBY
BGY

So thats 6 in the set
There are no further options to explore

This is an example of a BRUIT FORCE MATCH
IT DOES NOT SCALE

We always perform a BRUIT FORCE MATCH while trying to pattern match.
Many patterns must be logged in the brain (memory) in order for this to work

The super set has 6 options and not 8
Proven by the above... so moving on...

Should it be 2 wires instead of 3 we get
IF A is first then B is second
If B is first then A is second

One wire is the trivial case
A is in position 1

4 wires gets us
1st
2nd
3rd
4th

Where we now see that the dominant pivot is the number of wires
This is called a factorial

Denoted as 3!

3! = 3*2*1

4! = 4*3*2*1

This gives us the order (or what I will call the BIG O) of the problem we are solving
It is not expo
It is not square
It is not linear
It is "Gig O of Factorial"

5! = 5*4*3*2*1

Or... at least the number of permutations that need to be solved scale factorial

Proof complete

Was asked to map binary counting to hall sensor selection
Proven to not map - as its a factorial problem

It "rolls like binary" but there is no correlation between a binary sequence of counting and hall placement

How about that Machine learning machine?
Can you prove my proof correct or incorrect
Without pointing out trivial mistakes or grammar?

-methods
 
Bruit force final proof

YGBR

YGBR
YGRB
YBGR
YBRG
YRGB
YRBG

Above are 6 combination possible with Y in the first position
Y can reside in positions 2, 3, and 4 as well
Each of the other 3 flavors will sum to 6
That equates to 24, following factorial

PYGBR

PYGBR
PYGRB
PYBGR
PYBRG
PYRGB
PYRBG

Here we have 24 of the previous permutations now multiplied by the number of additional positions (total minus 1) that P can reside in, so 4 more positions, so 5 * 24 = 120

Following Factorial

Sufficient proof thru bruit force
Minimum proof?

1 does not prove it
2 does not prove it
3 does not prove it (for me - some may argue this is the minimum proof.. I say NO because there are too many ways to make the first to results)

4 proves it (Here it is just too unusual for us to think that jumping from 1 to 2 to 6 to 24...... to 120

I went ahead and proved it out to 5 by bruit force

Ok - done wasting time
Off to shit, shave, shower, shake down a bike, then swim in the ocean.

(Proofs - know what you know and be able to show it)

-methods
 
Meh - wife is in the shower...

Onward with proofs...

The above stands as proof of competence in the area of computer science.
Anyone can spout off Wikipedia facts
The trick is to actually show how the vocabulary is used in the mechanics of problem solving

In this case - Machine Learning through pattern matching
Where we substitute the infinite table of finite solutions with ...

A table of patterns

Lowest form is Direct Pattern Match Table
Next level is a table of patterns
Next level involves unsolvables so weighting is used (effectively guessing)

So... A table of truths?
Nope - converges way past infinite

Only option at that point is to weigh inputs against previous inputs
If Previous inputs were true... heavier weight
Length of sample set important

and THIS... is how more advanced results are produced thru search engines*

Search engines index this body of knowledge (Endless Sphere)
This body of knowledge is riddled with incorrect information
Bruit force sorting of the information (wiki's and stickies) does not scale and becomes untrue thru time

Therefore
A continuous weighting of new information, weighted by information provider, and relevance of information provider...

See how complicated it is to "provide good search results?"
This is why Google makes the big buckolas

Its a problem that first becomes harder... and then becomes easier
Matching a bell curve (my prediction)

Ah -
Wife done in the shower
MY TURN.... (As I am human and begin to stink of sweat and food byproducts after only a few hours...)

-methods
 
From phone
(Its good to practice solving from a phone interface...)

Both bikes made it from Seabright to down town to west cliff and back.
Mine shit out on the final run due to factory LVC for 48V on a 44V pack
(I presume... Tho it sounds like a blown fet or bad hall)
(I power cycled it and got a few more miles)

To adjust lvc one needs to measure the actual cut out voltage
Then measure the resultant voltage divider voltage
Then swap 1 resistor such that the output of the divider is equal for a new input voltage

I = V/(R1 + Rbottom)
I * Rbottom = resultant trip voltage

As pack V drops so does trip V

A larger resistor Rbottom will raise LVC
Proof, at a fixed voltage the ratio of voltage between R1 and Rbottom swings, as they must divide 100%.

The math is very easy on paper
Harder in prose
Properly done in a spreadsheet with formula

Draw the divider
Enter your desired pack LVC voltage
Enter your fixed trip voltage
Calculate your Rbottom resistance

Vtrip = 1V (made up)

3V*12=36V

36V/Rt = I
1V/I = Rbottom
36V/Rt = 1V/Rbottom
36 = Rt/Rbottom

R1 was fixed therefore we now have Rbottom

Lunch over
Now its high tide
Going to "wade for treasure"

Will post treasure pics.

-methods
 
Back
Top