Wednesday, 19 August 2015

For Hardware entrepreneurs - How to go from Idea to world-class Product

Last year at Lumos, I have gone through full hardware development process and when I look back, I have learnt a lot about the product development. I realized that I should share my experience with fellow hardware entrepreneurs.   
Again, this is from my limited one year experience and few other leanings might yet to happen. Consider this blog  for only simple or medium complexity consumer products.

Design Process

Here, It is assumed that  you have the product idea and you have limited knowledge about product development.
My product development process is matching this to a large extent. But, the article looks at it from business perspective while I am putting it from engineering perspective.                        

Proof of Concept  (POC) :

Proof of concept is a dirty prototype of the core functionality. Don't try to accommodate extra features. Generally, proof of concept is making the product using Arduino, raspberry-pi, General purpose PCB (GP), breadboard etc.

Why should you do proof of concept before starting product development?
  • Help you establishing technical feasibility. Lot of time, you assume that two kind of products are available and your idea is just fusion of both. But, in reality, it might be not possible technically or it might be significantly complex to implement. 
  • Help you understand important technical components very well. i.e, you might assume that you just need to use a camera for your product but during the POC, you need to set 100 parameters for camera function. You need to understand camera well before configuring these parameters.
  • You have something in hand to understand customer needs. You can judge customer's interest from this. You can talk to a customer about importance of a particular feature. Also, you get sense of what other features are required to make it complete products etc.
  • With your initial customer feedback and some idea about complexity, you can start drafting the specifications.
  • You can show the POC to a potential investor (your uncle or an angle investor or your manager!!) and try to fetch some seed money. Without POC, no one is going to fund your product design activity (ignore if you have rich parents or you are rich parent !!).
  •  More than above all, you get confidence on technical feasibility and market existence before you start spending huge money.

End outcome of the POC are many. Most important one for product development point of view is specifications of the product. During POC phase (or end of it), you would have interacted with at-least 100 potential customers and know their requirement (Yes, 100 is a random number :P. The number depends on type of market and customer segment). You should able to create a list of customer needs and put them into these 3 buckets.
  • Must haves : Required by all potential customer segments.
  • Optional : Required by some of the customer segments
  •  Lateral
The lateral needs (features based on it) are very important. These are not asked by any customer but you think it would delight them. The product needs to be well thought and lateral features usually create extremely satisfied customers.

Now after this, you can study competitor products or solutions also to understand what customer needs they are addressing. Lot of time the customers you meet forget to tell one of 'must haves'. Looking at competitors might help here.

After you draft the needs, start drafting the specifications. Consider both customer needs and complexity involved in implementation. Try to define the "To manufacturing" time (deadline for product development or design closure) along with specification of the product. A good approach is to launch first product with minimum required features and keep adding features on next version of  the product. This might help in staying ahead of competition.

My master project guide, Dr. Andre Pittet, used to suggest to go for 'wish specification' first and then cut down the features based on time availability and complexity of implementation. Also, your customer development team should get feedback from the customer on each feature.
Your specifications needs to be complete. Everything from operating temperature, humidity level to amount of mechanical load it can take should be specified.  Do specify target cost also. This is Lumos charger specification for example.

FAQ on POC

My complete solution includes cloud connectivity, storage and visualization too. Should I get it up at POC level ?

No for most of products. If you get the data on serial terminal of the PC is good enough. ( Beware, I am biased here, I don't appreciate complexity of putting data from serial terminal to cloud, making it secure, giving nice visualization etc.)

How much time do I need to spend of POC ?

Typical time for simple or medium complexity product is 1-3 weeks. Most of this time goes into the component procurement. You might just spend 10 hours for it.
But, if you are developing a device which measure blood pressure just by holding it in palm(complex technology) then it requires good effort. Specifically because it is not proven technology and requires testing on multiple kind of people.

Do I need to talk to customer before writing specification ? Apple never does any user survey and develops amazing product, why can't I ?

I won't go into whether Apple does the user survey or not here. They are product genius and know what user need very well and deliver beyond user's expectation. But, I would recommend doing customer development (which is little different from market survey, read the startup bible 'the lean startup' to understand what is it) before start putting heavy money and time. They main reason here is that in most cases the founders don't represent the customer segment or founders are the only customers who need the solution. Talk to few hundred potential customers before quitting your MNC job.

Instead of customers, can I talk to some retailer of similar products who will be selling my products to customers ? They know the customer segment and their requirement very well.

Yes, do talk to them. But, never assume they know all customers requirement or represent customer segment. i.e. If you are developing home automation system, which you are going to launch with leading builder. You understood all the requirement from the builder but never talked to any home buyers (potential users). This will lead to low market acceptance because the end users are not going to be happy.

Design Phase :

Design phase start with detailed specifications and end with a product which you need to exactly replicate. It involves all design for manufacturability, certification consideration. Typical timeline for this is 6 months to 3 year. Best way to quicken the development is to stop changing specifications too often. 
For electronic product, there are mostly two design activities, electronics design and mechanical design. You might have few more. i.e. we have the fashion design activity too.
Here, I am assuming that you have the block diagram of your solution. I won't address much on how to find good solution for the problem you are addressing.

Electronic Design Activity :

The process chart is  self-explanatory. The first 3 steps, part selection, schematics design and PCB design go together so indicated in different color, same way the last two.  Very important consideration here is that,  you will have at-least two revisions of above process. Some time you need 4 or more revisions because small inaccuracies.  Electronic design needs focused mind !!



Important Parts selection

After specifications, you start with important and critical parts selection. This involves your camera or any other sensor, micro-controller, battery, power regulators, battery charge controller etc. This is most critical part of design phase. I recommend to start putting these parts into the schematics also.
While doing the part selection, with keeping eye on the functionality, Keep eye on following parameters also.

  • Cost
  • Error budgeting ( What would be worst case error in measurements)
  • Power calculation ( based on datasheet numbers )
  • Future upgrades(You can add complex software or add the 'wish features' later).
  • Selected part's market life ( the part you selected might get obsolete before you go to manufacturing)
  • Certifications ( I am going to address certification in a separate blog)
  • Design complexity
Here are few tips for part selection
  • Design support (reference design, application notes, forums, software support) is important consideration for each part
  • Involve the component distributor like Avnet or WT from start of your design. They will make sure you get all required support.
  • Get quote for 1000 numbers to estimate the BOM cost
  • You will at-least get half of the digikey price for 1000 quantity (from distributer like Avnet or from China)
  • If you are getting component like battery from China, make sure you ask for the certificates (MSDS, UN 38.3 etc) upfront. May be before placing the sample order.
  • Before selection microcontroller find out the development tool cost i.e. TI has excellent BLE chip cc2540/CC2541 but TI's pre-compiled libraries for BLE stack is available for IAR only. IAR license cost few lakhs !!
  • Prefer ARM based micro-controllers. They make your system scalable and they supports GNU tools.
  • If you have complex functionality, get the EVM boards and test your functionality on the EVM board.

Schematic Design

After you have selected critical part, schematic design includes passive component selection, doing power calculations, cost optimization, selecting ferrite beads and provide provision for future expansions. I am putting some basic tips and learning.

  •  Follow reference schematics or EVM schematics to extend possible. If you are not following any of IC manufacturer recommendation, think it well and test it well.
  • Provide provision for EMI/EMC related components like common mode inductors at all your power inputs.
  • Provide ESD protection at all the ports where humans can touch. Don't overdesign. Cost increase significantly with your specific requirement. i.e. think thrice before you use tantalum capacitor. Check if it is really required.
  • If you are selecting passives and planning manufacturing  in India. Make sure all you get from India.  You will get all resistor and capacitors at Micro-components SP road (phone : 98451 57490). Micro-components get all components from 'Yageo' so quality is ok.
  • There is one digikey equivalent for India : comkey.in. Check it out. Could get all connectors and inductors from here.
  • Wurth Electronik is one of the best supplier for magnetic components. They are not dam cheap but be sure about the quality. Advantage with Wurth is that they are the only big manufacturer without strict minimum order quantity restriction. Their India team is very energetic and supportive.
  • Capacitor cost 0.5 rupees per piece (for quantity of 100-1000) unless you have special requirement i.e. 47uF in 0805 package
  • Resistor (1% or 5%) cost 0.2 rupees per piece (for quantity of 100-1000) (specific low value or higher power rating cost more)
  • Get the schematics reviewed by many people. 

PCB Design

Steps for PCB design are following.
  1. Make or get footprint of basic components. If you make the footprint, get it verified with other team members. 
  2. Get mechanical layout from the industrial/mechanical designer. Which includes PCB size, position of all the connectors, LEDs, Switches etc. 
  3. Floorplan the components from the mechanical design and routing patterns. 
  4. Decide the stack-up of the PCB. This include how many layers of routing, copper thickness, PCB core material type and thickness.
  5. Get all schematic net's maximum current and maximum voltage and out it up in an excel. This will define your routing trace thickness(current) and spacing (voltage) .
  6. Route it
  7. Get it verified from third person
  8. Generate gerber and release to manufacturing
It is very difficult to give generalized tips on PCB design. Here are few tips.
  • If you can, go for only one sided component mounting. It saves the assembly cost.
  • Check what minimum trace width your supplier can support
  • Before you select the passive package size, check the cost factor
  • Follow the IC manufacturer's layout guideline thoroughly 
  • Try to get one not disturbed ground plane without any traces. You can merge analog and digital ground in this plane. Approach here is to minimize ground resistance so that digital current can't create ground noise.  Not disturbed ground plan help on EMI, EMC also
  • Decide PCB board manufacturer now itself to design up to his capability.
  • Make sure design has taken care of the manufacturing tolerances. Sometime the PCB cutout tolerances cause problem of the board fitment in an accurately designed enclosure.
  • Be sure of what kind of surface finish you want. Generally popular one are HASL, lead free HASL, ENIG, OSP and gold. Read about each of them and wisely choose one. Very few manufacturer support OSP in India. But, it is one of most cost effective and good quality solution for mass production. Make sure your PCB assembly manufacturer is ready for OSP.

Prototype Manufacturing:

Manufacturer your designed board !! Generally we prototype 10 boards.
Good prototype manufacturer :  PC process Penya, Bangalore or PCB power, Gandhinagar, Gujarat
Lead time : ~ 1 week
PCB assembly :
We don't make stencil at this stage. But, if you have complex PCB, you might have to do that. We do manual mounting and soldering. for complex PCBs, you might have to go to EMS.
Generally, we follow board bring-up method for PCB assembly. In board bring-up we mount components in stages and check the functionality in stages.  i.e. mount first the power components, check if  intended voltages are present at the not mounted IC pins, followed by micro-controller and then sensor.

Software development

Yes, now your board is up. Start developing the software. Lot of time you would like to start the software development on EVMs or reference boards. No generalized tips here other than be careful while porting software from one board to other board.

Testing 

One of the most important part of design cycle is testing or design verification(DV). If you are a consumer company, this becomes extra important for you. You should make sure that your product works without any bug.
There are many DV approaches. Take any approach but make sure you have detailed testplan and all test are performed. Some physical test like drop test, over temperature, short circuit at output also need to be included.

FAQ on Electronic Design

Why do you say that there would be at-least two revisions of the PCBs ?
Because product development is complex and you don't have good simulation tools. Many parts of design are going in parallel and mistakes are likely to happen. If you have done similar product with same parts previously then there is possibility that one revision is enough.
One thumb rule is that when you can't explain what is problem with your board to your business co-founder and he insist that not to go for any revision now, then you are going for the last revision.

Why is board bring-up method is preferred in prototype manufacturing ?
If you get fully assembled board from your EMS and start testing it, mostly you will see IC heating up or capacitor burning. If not, you will have an issue which will make you nuts debugging. Irrespective of where you do assembly, make sure your functionality is getting up one by one.

Mechanical Design Activity :

Mechanical design/ Industrial design is most important part of your product. Users interact with it more than your underlying electronics. It is very important aspect if you are making a consumer product. Also, manufacturability is much more complex here that an electronic part manufacturing.
If you don't have any one to design the mechanicals then get  a good industrial design firms. Typically, an industrial design firm has an industrial designer who handle studying user requirements and creativity and one mechanical engineer who understand manufacturing processes and materials. Industrial designer, mechanical designer and electronic designer should work in synergy.  Here is simple process for mechanical design from an electronic designer point of view.
  • Understanding User requirement 
  • Create few early sketches 
  • Understand engineering requirement for each sketches 
  • Make early cad 
  • Review early cad (by the CEO and entire team) 
  • Go for final CAD with DFM consideration
  • Go for 3D printing and find all issues (PCB fitment etc)
  • Iterate to point 6 
  • Finalize the CAD 
  • Create guidelines for mold manufacturing
  • Go for production

Some last steps before design closure

  • Certifications : I will cover it in my next blog
  • Design for manufacturing check : get it checked with experienced design for manufacturability person. Ask feedback from your manufacturer.
  • Setting up quality process (what points voltage/current needs to be measured to make sure all components works)

Manufacturing                                     

This is a vast topic and general tips won't make sense much. Take all data mentioned here with some doubt. Match your product with our product and see all tips given here make sense for you.

Mechanicals Manufacturing

This process is much complex than electronic manufacturing. If you are going for injection molding, the major complexity is to do mold manufacturing. Generally, molds are costly. Good manufacturer quote >4 lacs for any size mold. Here are few things are learned.

  • If your target quantity is <10000 unit. Don't do mold. One way is to avoid injection molding totally buy using metal enclosure or off the self enclosure.
  • There are some manufacturer who can give prototype exactly looking like injection molding prototype. 3DPD and cycloid do that as per my knowledge. P.S: I have not tried this yet.
  • One other option is protolabs, which supplies CNC machined components for low volumes.
  • I refer this mold manufacturer, Tamam components, contact them on tamam2010@rediffmail.com. They are cost effective and reliable on lead times. Part quality is also ok. 
  • Don't assume that once mold is done, you will get exact parts. There are variations like molding time, pressure and temperature. Make sure that you go to manufacturer place and do experiments to get best combination. If you can't do this then get help of some experienced person. This is very important step for quality assurance.

Electronics Manufacturing


Electronics manufacturing is reliable if your EMS or PCB assembler follow the standard quality practices. Wait for my blog where I explain all layers of PCB and PCB manufacturing and assembly process. Here are some simple tips.
  • Don't forget to lock your code after programming. Some-one can read and copy your product as it is.
  • Make sure you visit the PCB assembler's factory
  • First make one board and test is well before going for mass. (To make sure that pick & place machine is programmed correctly)
  • Quality test must happen at PCB assembly manufacturer's place
  • Define maximum failure rate with your PCB assembler before you place Purchase Order
  • Go for pilot production (100 numbers) before going to mass ( > 1000)
  • Managing working capital for manufacturing is not easy task
  • Make sure that final assembly process (putting PCB into the box) happens accurately.
  • Don't go for 100% manual assembly process. Manually soldering through holes are ok but SMD components assembly must happen through reflow. There are  does manual placements, which creates lots of failures.
  • PCB board typical lead time for mass production is ~2-3 weeks
  • IC components lead times are in the range of 6-8 weeks

No comments:

Post a Comment