Hacker News Reader

Comment by zipy124 | Hacker News Reader

The largest problem I see is shipping times. If I need to download a new software "part" (library or other), the shipping time is the download time, nearly instant.

If I need new hardware pieces, its either next day shipping, a few days by air-freight or three weeks on a boat from china.

This limits prototype turnover time, and means iterating quick is much harder.

Finally you have the problem that hardware is expensive and an additional cost. A hardware startup has all the same costs as a software company but with the addition of hardware.

View on HN

Parts are pretty much instant, pcb turn around times are 3 to 15 days depending on how complex they are.

Even in the bad old days of punch cards and priesthoods the turn around for software was faster.

I have a little pcb mill in the garage that I use for prototypes.

To this day I've not met another EE that knows what a voronoi mapping is, or why you'd want one. In a previous startup where I was the software engineer I got through more prototypes for the analogue signal paths in an afternoon than the two other EEs had in the previous week.

Only for simple pcb. If you are making multi-layer pcb with complex stacks, pcb manufacturing and soldering (with associated tooling setup, validation and so) are easily 2 months of turnaround

Well to be fair if you add "validation" the turnaround for any noncomplex piece of software can reach months pretty quickly as well.

But yeah I'm not gonna argue that sw isn't faster than hw in 99% of the cases.

Up to 4 layers is pretty standard these days. It's more than 8 where you start running into issues.

That’s a big part of it. But mostly it’s that your dev+deploy+evaluate cycle is so much slower. With web software you can write a feature or bug-fix and push to prod in minutes - and repeat that many times a day. With hardware each equivalent cycle is weeks or months (especially in my vertical - farms).

That's what I meant by prototype turnover time, your iteration cycle.

Some of us write software for embedded devices where a mistake in programning means a critical machine in a remote area needs someone to physically go there and thus spend months in test before we dare deploy. Some of us care about quality and won't allow a customer to see bugs thus even though we can deploy in seconds our reputation won't allow it. There is also software that bugs can kill people so again we won't deploy without evtensive testing - much of this is regulated and we would go to prison for deploying at will even if it works.

those who work in any of the above react in horror to stories of how the web deploys to production so fast. we know we are not perfect and don't understand how you would risk it.

In some shops this is solved by staggering scheduling so people work on several projects. Some aspects of HW can be simulated (think analog with SPICE-likes, or logic level like the chip designers and FPGA users do) so this reduce the need to iterate every time in hardware.

A lot of the iteration work can also be done on the board you received. You don't have to wait for your new board to see if those additional decoupling capacitors will do something, you add them on your current board by hand... You would be amazed at how far rewiring can go, sometimes entire BGAs are removed, installed dead-bug style and each pin wired by hand.

Where I've worked in hardware dev, the biggest bottleneck is software.

It's because hardware developers don't take the software side seriously and either assume they can just do it themselves or don't think it's worth a decent salary

Couldn't be more untrue

It's always somebody's fault, and I haven't analyzed it that closely. But empirically, the priority of software is to finish the previous project, while hardware moves ahead. We don't touch production code. Some of us (myself) write our own test scripts. Very little hardware runs without some kind of computation happening. When one of my designs goes on the shelf, I want it to be provably reproducible and tested.

We definitely value software, and the devs are our friends. It's not such a huge place that we're totally isolated from one another.

The pace is not really parts supply constrained in my experience. It just takes much longer to build and validate even relatively modest design changes.

True but that reality applies to all your competitors. Your investors should care about your performance relative to the market.

Not to your competitors in Shenzhen, they can get basically any part within one day or can get any prototype made easily.

They also get it stolen even faster.

Investors are generally wanting to see little existing competition so that’s not really the issue.

They’re more concerned with factors that will cause the company to self-destruct. Running out of money before hitting PMF and growth is the most common failure mode for any startup, and is much more likely with any hardware startup, due to the dramatically slower iteration times.

Do you see potential for better software design and simulation tools including VR to reduce the need for physical prototypes?

There's very good simulation tools available but scrappy startups can rarely afford them. You also need to people who know how to use the tools appropriately.

Reply on news.ycombinator.com