-
Start thinking about COGS...
05/09/2019 at 02:01 • 0 commentsThere are two primary numbers to be concerned with: Cost of Goods Sold (COGS) and the retail price to consumers. COGS is what it costs you to build and sell the product. The retail price is what a consumer pays for it. The difference is profit -- hopefully!. You need profit to make the product viable to produce and to -- again hopefully! -- fund the development of your next cool product.
COGS isn't always an easy number to calculate. Some parts of it are well defined, like component costs. Other aspects can be harder to nail down. COGS also includes:
- Cost of prototyping your design -- there might be more than one prototype build.
- Cost of the bare, unpopulated printed circuit boards (PCB).
- Cost to assemble the circuit boards -- stuff the boards.
- Cost to ship bare and/or assembled boards to you -- you may incur more than one shipping cost.
- Cost to test the assembled circuit boards -- you don't want to ship non-working products.
- Cost of the packaging your product is shipped in -- boxes, bags, tape, labels, and packing materials are NOT free.
- Cost of the shipping to your customer, to possibly include cost of you taking boxes to the shipper.
- Any and all advertising costs -- if nobody knows about your product you won't sell any.
- Spare product in case you have to repair/replace defective units -- what percentage extra should you build? What is your warranty, if any?
- Your time -- Is your time worth zero? What do you think your time is worth?
- Wages and benefits for any employees you may have.
- Overhead for your place of business. Electricity etc. isn't free.
- Whatever I've forgotten = intangibles… There’s always something.
Your retail price is determined by your COGS and the profit margin you add. Determining that margin number can be tricky. Too high a margin and your retail price might result in fewer sales. Too low a margin and the net profit can be depressingly small for all the work you've invested. If all you do is break even on your costs – or worse lose money – there is little incentive to go through all the trouble to create a product and bring it to market. Your time and effort are worth something! Profit can be used to fund development of your next cool idea too. Prototypes cost money.
It is really easy to research these topics and collect all kinds of advice regarding "rules of thumb" and how much margin you should add to make a product worth producing and selling. I encourage you to do that research when determining the margin for your products. You should also research what similar products cost -- if any exist. This is a great indication of what the customers are willing to pay.
Another thing to consider is whether or not your product can command a premium price compared to others in the market. Is your offering differentiated in some compelling way that would entice customers to pay more for it? The value you can add to your product is an important factor when determining retail price.
Yet another thing to consider is whether or not you will be selling the product directly to customers or using resellers. Companies like Amazon can potentially get your product in front of millions of eyeballs. They don't do that for free though. Their fees go directly into your COGS and you should adjust retail pricing accordingly.
-
Which smart LEDs to use?
05/09/2019 at 01:57 • 0 commentsThe two primary smart LED types favored by makers are the WS2812 and the APA102. The WS2812 requires only a single 5V logic level signal to drive a daisy-chain of LEDs. Hundreds of LEDs can be controlled with a single GPIO pin. The downside is that this serial data signal is relatively slow which impacts how fast you can update individual LEDs and how many LEDs you can string together for a given update rate. The timing of this signal is also somewhat precise and not easy for simple microcontroller boards to handle.
The APA102 LED requires two 5V logic level signals to drive a chain of LEDs -- one serial data signal plus a corresponding clock signal. As with the WS2812 style, hundreds of LEDs can be controlled with two GPIO pins. One downside is doubling the number of GPIO pins required to drive the LEDs. A major upside is that adding the clock signal allows you to run at a substantially higher update rate.
So, tradeoffs.
A quick look at major suppliers in China showed that the WS2812 style is about half the price of the APA102 style. The WS2812 and its' variants such as the SK6812 family are readily available from numerous sources in the original 5050 package (5.0mm x 5.0mm) and the smaller 3535 (3.5mm) package sizes. Very tiny. Having secondary suppliers is important – you need alternate sources on the chance that components from your primary supplier become unavailable.
The APA102 style is also readily available in the larger 5050 size package, but can also be found in the smaller 2020 (2mm) size. Very, very tiny indeed! It was interesting to notice that some of the vendors claiming to have the APA102 2020 parts would advertise an incredibly low price then show a picture of a 4-pin smart LED.
APA102 style LEDs all have six pins!
Caveat emptor applies at all times when dealing with component suppliers. Pay attention! If most of the vendors are charging $0.15 to $0.18 per LED in bulk quantity and one vendor is offering them for $0.01, it is likely a scam. If it sounds too good to be true it probably is.
The color of the LED package may also be important. Most come in white packages but sometimes black can be found. This can play into the visual aesthetic of your product.
Digging further, it appears that many of the controller boards that makers favor, such as Arduino, Teensy, Raspberry Pi etc., have software libraries available to drive either type of LED. Therefore, I could choose either type of LED. Unfortunately these two types of LEDs are not interchangeable.
I had to choose one or the other to move forward.
It would be easy to argue back-and-forth about the merits for one type of LED over the other. Since my primary motivator for this kit is reduced cost, I chose to go with the less expensive WS2812 mini 3535 style LED. In large quantity the WS2812 style LEDs are about half the cost of the APA102 devices. Back of the napkin math suggested that a x8 cube with 512 LEDs, the raw LED costs for WS2812 would be more than $40 less expensive compared to the APA102 style. That is a significant price difference.
-
Set some initial goals.
05/09/2019 at 01:53 • 0 commentsPeople tend to view goals as admirable things to strive towards. An engineer tends to view these as constraints that begin to define the design. After looking at YouTube videos of various cube sizes I decided to design an 8x8x8 cube. If designed properly I figured it would lend itself to bigger and smaller versions using the same techniques.
Choosing a x8 size (shorthand for 8x8x8) yields the first constraint: 8x8x8 = 512 smart LEDs.
Secondary constraints immediately followed. For example, one secondary constraint is the amount of current to drive that many LEDs -- recall that 512 smart LEDs = (512 Red) + (512 Blue) + (512 Green) = 1536 total LEDs that can pull up to 30.71 A of current.
This amount of current will add further constraints with regards to connectors, the sizes of circuit board, size of power supply, and circuit board trace sizes.
So, that one goal of a x8 cube created several significant derivative constraints.
Another goal would by physical size. There doesn't seem to be any consensus regarding the spacing between the LEDs. The spacing impacts how nice the display will look at a given distance. It also impacts the physical size of the cube. Physical size impacts the cost of the circuit boards to some extent as board cost is usually directly related to board area. Also, shipping costs of the final product relate to physical size and weight.
After looking at other cube designs I decided to settle on 0.75" (19.05mm) spacing to start with. At this point we're looking at a cube that will be a minimum of 6"x6"x6" plus whatever extra dimensions might be necessary. This is small enough to fit comfortably on a table top.
-
Design a product?
05/09/2019 at 01:46 • 0 commentsA primary motivator that initiates many product designs is to "find a need and fill it." If what you want to buy doesn't exist you have the opportunity to create it. Or, if what you want is too expensive, you have the opportunity to refine or improve existing products to make them more affordable. Maybe you can even come up with a novel method of solving the problem that will set your product apart from others. Chances are if you are looking for this product it is likely others are looking for it too. I certainly couldn’t find what I wanted so my only choice was to make one of my own. Can a well thought out and well designed kit at a compelling price point gain acceptance in this market segment?
Designing a product is an exercise in compromises. For example, a cube that requires zero soldering by the purchaser has higher manufacturing costs than one that requires some purchaser soldering, as I will show. I will also explore the trade-offs of various design architectures. As the designer it is up to me to make the trade-offs and compromises that result in the finished product. I have to balance the customer experience against cost of goods sold (COGS) and retail pricing.
Product development isn't easy. I went through many thought exercises, sketches and prototype builds to refine the design to its current form. Every decision has consequences that are related to cost, construction, aesthetics etc.
-
Quick Market Survey: What kinds of kits are available?
05/09/2019 at 01:43 • 0 commentsA market survey revealed that there are many kits available for discrete, thru-hole LED cubes in x4, x6, and x8 sizes. There aren't many kits for cubes larger than that. Most of these kits are inexpensive and include a pig-pile of thru-hole LEDs and instructions for you to use to solder it all together. Some kits include a base board to build the cube onto, and the base may or may not include a controller circuit to drive the LEDs. Thru-hole LEDs are cheap and all the expensive touch labor is performed by the purchaser. Some of the kits may include a separate controller board as well.
My quick market survey showed vastly fewer choices for addressable LED cube kits. These tended to be substantially more expensive and required little to no soldering skills from the purchaser. The ones I found also came with controller functionality built-in, and what I saw wasn’t how I envisioned controlling the LEDs.
-
The Problem
05/09/2019 at 01:40 • 0 commentsBuilding 3D LED cubes out of discrete, thru-hole LEDs is a popular right of passage project for many makers. Many of these cubes are small at 4x4x4 with some increasing to 8x8x8. Most feature single color LEDs. Less popular are cubes larger than 8x8x8 or with RGB full color LEDs. One aspect of these discrete LED cubes that poses a challenge is how to control the LEDs especially if the number of LEDs is large. An 8x8x8 cube has 512 discrete LEDs. If the cube is RGB then you have 512 red + 512 green + 512 blue = 1536 LEDs to control.
Regardless of how you choose to control them (individual wires, charlieplexing or some other method) you may have a lot of individual control signals to contend with. Assuming you have a controller board with the requisite GPIO pin count, you also have to contend with high current flow. While individual LEDs sip power, for large groups of LEDs the power requirements add up quickly.
As we showed above, an 8x8x8 cube has 1536 discrete color LEDs to control. If each of these LEDs pulls 20 mA of current, then at full power with all LEDs turned on the cube will pull 1536 x .02 = 30.72 Amps! Discrete LEDs also require current limiting resistors to ensure that they don't pull more than the desired amount of current. You can use current driver chips or transistors but either way the number of discrete components grows quickly.
One way to overcome the LED control issue is to use so-called "smart" or "addressable" LEDs. These are usually surface mount devices that include individual R/G/B LEDs and all the circuitry required to drive them, including current limiting, all inside a single, small package. Current is pulled directly from the power supply the LED connects to and not from a GPIO pin of a micro. Additionally, no extra transistors or driver chips are required to handle large current flow.
More importantly, the color value of the LED is controlled via one or two digital control signals. A discrete RGB LED requires four control signals -- R, G, B, and common anode (or cathode). Addressable LEDs, in contrast, require only one or two signals. Furthermore, the addressable LEDs can be daisy-chained together such that controlling a large number of LEDs still only requires one or two signals to control the entire chain. Some common LEDs of this type are the WS2812 which uses one digital control signal, and the APA102 which uses two digital control signals.
The downside of using smart LEDs is that they tend to be surface mount devices, not thru-hole mount, so building a cube with them presents different construction challenges. The discrete wires used to construct cubes built with thru-hole LEDs have to be replaced with printed circuit boards for the surface mount LEDs.
Another thing to consider is that the thru-hole LED cubes required coarser soldering skills whereas surface mount LEDs require finer, more detailed soldering skills if assembled by hand. However, most of the hand soldering can be alleviated by using reflow techniques for the surface mount LEDs. Regardless of how the soldering is accomplished, surface mount LED cube assembly is quite different from thru-hole LED cube assembly.