Comment by robomartin | Hacker News Reader

Comment by robomartin | Hacker News Reader

  "And in particular, don't be deterred from applying to Y
   Combinator with a hardware idea, because we're especially 
   interested in hardware startups."

How are you planning on funding hardware startups? From your site:

  "Usually $11,000 + $3000 per founder. So $17,000 for two
   founders, $20,000 for three or more. Occasionally we
   invest more. The goal is usually to give you enough 
   money to build an impressive prototype or version 1, 
   which you can then use to get further funding."

A real hardware startup would require at least an order of magnitude more money than this. Unless they walk in with all the required tools, you could burn $20K just in software licenses (Solidworks, Altium Designer, Xilinx Foundation, Keil, etc.) and not have much left for other stuff. Heck, my DSO alone cost me about $20K. The computer I am typing this on probably has $50K in hardware and licenses on it.

I could personally consider the idea of presenting a hardware startup to YC, but I would need to know that this is not about (with the utmost respect) finding a few starving 20-year-olds that will kill themselves for a $20K investment. From my vantage point, if you are not throwing $250K+ into a hardware startup it just isn't going to happen. Of course there are exceptions to every rule. Then again, this ain't my first rodeo.

Now, if the idea is to throw some money at a project to cobble-together a smoke-and-mirrors prototype and then go raise a few million, well, for the right project this could work.

Again, I say the above with respect for what you do. I have done a lot of hardware/software/multidisciplinary development. It's very different from pure software web/mobile startups. Very different.

This is an important point and needs to be addressed. $20k doesn't go very far at all in the hardware startup world.

That said, it is possible to be frugal--to a point. My team and I try to use inexpensive solutions where we can. We're using Rhino instead of SolidWorks (OK until we need the hardcore simulation functionality), KiCAD instead of Altium (same, until we need the simulation), MikroElektronika's C compiler for ARM rather than Keil, MeshCAM rather than Mastercam, a cobbled together kit CNC mill rather than "real" mill, a Chinese-made laser cutter rather than an Epilog, and a wide array of machine shop equipment sourced from Craigslist.

Our product also touches biotech, so we have outrageous reagent costs too. We still have shell out for antibodies, fluorescent probes, proper biosafety lab space, etc.

I'd love to see YCombinator comment on accepting biotech/healthcare companies.

As you said, to a point. In my experience some of the approaches to frugality can quickly become counterproductive and eat away critically needed development clock cycles.

Here's an concrete example: I made due for years with cobbled-together CNC equipment of various kinds. I bought an old Bridgeport CNC mill really cheap. Within a month of use the spindle transmission had to be rebuilt. It did not have an automatic tool changer or tool probing. After that I took a Bridgeport knee mill and converted it to CNC. No tool changer or tool probe. It did OK. Both of these represented HUGE time drains and detracted from the main mission in major ways.

Later on I decided to do it right and leased a Haas VF6-SS. If I remember correctly the lease ran about $2,000 per month. Best decision I ever made. All of a sudden CNC was not a research project. The thing just worked and it could work 24/7 if required. It had an automatic tool changer and automatic tool probing. When properly married to Solidworks and a CAM program the setup was smooth as silk. This improvement in productivity was well worth the $2,000 per month and other costs (tooling, setup, training) in that it quickly returned everyone to being focused on what we were actually trying to design rather. It also opened-up design possibilities that did not exist before this machine arrived due to what was lacking on the cobbled-up machines (fourth axis being one example).

That's not to say that I am not for being frugal and creative whenever possible. For example, we used the CNC machine to manufacture custom assembly tooling that would have cost tens of thousands of dollars to purchase. We also made our own sub-table for the machine and saved tons of money.

Thanks for the recollection. I'd love to have the space/budget to lease a Haas machine. :)

You'll have to raise more money, of course. Currently, all YC companies are offered $150k automatically and companies that look promising to investors are able to raise a lot more. I believe it's possible to get most good ideas to the stage of "promising" on $170k.

That sounds like a awesome setup and the way more Biotechs should be done.

As another poster said, there are starting to be alternatives to these expensive and hard to learn software tools for the early prototyping work. I'm the founder of Tinkercad and we offer a web based solid CAD tool that is significantly quicker to get started with than Solidworks. There are similar tools emerging for circuit design.

A lot of our users are working on hardware startups. Once they converge on a working prototype many move forward to more traditional packages to optimize the product for larger production runs. The cloud based offerings are also developing very fast, simulation is a good example of a problem that benefits greatly from being run in a large cluster.

I applaud you for creating such tools. They obviously have a place where they fit very well.

As for what you term "hard to learn" tools, well, I'll have to disagree with you on that one. At a professional level tools such as SolidWorks are an absolute necessity, not a luxury. The list of needs is endless, from parametric modeling to assembly management, mechanical, thermal and fluid simulation, interference evaluation, draft analysis, configurations, documentation, sheet metal tools, molding tools, CAM integration, DFM integration, programmability, etc.

Parametric design alone (and table-driven parametrics) is a huge advantage in these tools.

And, as you get into complex assemblies of assemblies with hundreds or thousands of parts and thousands of fasteners things get interesting too.

The same is true of circuit design tools. As an example, once you get into serious high-speed digital design you need tools that can facilitate the design, simulation and layout of complex multi-layer controlled impedance data paths as well as power distribution systems. At the higher end you have to get into field solvers and various SI (signal integrity) type tools. With FPGA's you need tools that, at the very least, make the task of working with devices that have thousands of pins as easy as possible. Bank swapping, for example, is a nightmare without the right tools. Heck, dealing with any IC's with more than, oh, 100 pins per package on a less-than-capable EDA tool is a recipe for disaster. I did a PCB with an FPGA that had over 1100 pins. The design started on a lesser EDA tool because, well, long story, but the client didn't know any better and didn't want to spend the money for a better tool (I use Altium Designer, but there are others). After 16 hours spent defining the pinout and PCB footprint for the FPGA (with verification, etc.) he got the message. The same job on Designer took about 30 minutes tops.

As for simulation run on clusters. Yes, of course. There are services that will take SolidWorks fluid simulations and spread them across multiple nodes for speed. I've fluid/heat-flow simulations that have taken upwards of 18 hours on 64 nodes. Fun stuff.

Like I said above, there's a place for what I will term "entry level tools". This is not intended as a derogatory term at all. That's just what they are in the context of what "pro-level" (another grab for a term) tools can do.

I will definitely keep Tinkercad in mind to recommend to friends with projects that might fit what the sofware can do. I wish you huge success. Get it into schools. Kids need to learn to create this way.

Nomiku was built with 20K during HAXLR8R (http://www.kickstarter.com/projects/nomiku/nomiku-bring-sous...)

Reply on news.ycombinator.com