My Next Idea


As a preface, I will say right now that I don’t think this is a venture fundable company and at best, would be a lifestyle business for a few years.  I think, however, that there is room in the market for my invention and without any further hesitation, here it is:

Personal atomic clocks.

Seriously.  Hear me out:

There are people who spend tens of thousands of dollars on home audio and call themselves audiophiles.  They buy low oxygen, ancient Western Electric cables to use to connect their speakers.  They buy dedicated amps that offer zero features by modern standards.  And, this market is big enough to support SEVERAL magazines on the topic.

Likewise, there is a market for watches that cost several thousand dollars to tens of thousands of dollars.  Again, everyone has heard of Rolex, Longines and Breitling and these brands regularly advertise in broad interest magazines.

Knowing this, why do people look at me when I say that there is a demand for a personal atomic clock?  The market may not be huge but it is one that is willing to spend money. Even better, they can be made with off the shelf parts for little money. The blunt truth is that someone who is willing to buy used parts and is willing to tinker can make their own for less than $100 from parts for sale on ebay.

Let’s step back for a second.

What is a clock?  Simply, a clock is an oscillator and a counter.  The oscillator moves or ticks at a regular interval and the counter keeps track of the ticks.

An atomic clock is one that derives its frequency reference from some sort of an atomic event.  So, instead of gravity driving a pendulum, you measure the energy emitted from a rubidium source.  Dave Jones of the EEVBlog has a great video on how these work so I won’t go in to much detail here.  Rubidium oscillators are a common atomic reference while cesium standards are the industry standard for precision.

Chipscale Atomic ClockThese devices can be small.  The standard rubidium oscillator is a few cubic inches.  Symmetricom has managed to shrink this in to what they call a Chipscale Atomic Clock – a device so small it can be mounted right to the PCB.

And how much does a rubidium oscillator cost?  About $50 used on eBay.  And how much does a counter cost?  Maybe a dollar if you use a common microcontroller.  The total price may be closer to $30 if you use something prebuilt like an Arduino.

rubidium oscillator

The one last part of the equation that is missing is the display.  If you know me, you already know that I am going to say Nixie tubes.  Nixies would actually be perfect for this application because a modest investment could mean that one manufacturer would own the remaining stock of IN-18s.  Talk about a competitive advantage – no one could produce a similar clock.  There may be one or two popping up on eBay but my rough estimate of the market is that for around $100,000 you could buy the remaining known stock.

Of course, there are dozens of types of nixie tubes and great care must be taken to ensure that there is a sufficient supply for mass production.  Sourcing tubes would likely require several trips to former Soviet countries to meet with dealers and source NOS.  However, initial prototypes could be built with tubes for sale on eBay.

So we now have an oscillator, a counter, and a display.  From an electronics perspective, we now have a minimum viable product.  But we want to make the BEST clock on the market, not just a good one so there are three things that we will do different.   First, we will include a redundant frequency reference.  This will increase both the short-term and long-term stability of the device and will as provide a compelling selling point.

Second, we will include a series of external frequency references – probably 1 PPS and 10 MHz.  Now, if someone has another oscillator or reference that they want to add, they can easily integrate it.  If you’re a lab and have a house reference, this will work perfectly.

Lastly, we will include a bi-directional time reference –ideally both NTP and GPS.  This not only saves the user from having to manually set the time but it also allows them to have an additional frequency reference.  This clock should also act as a frequency reference and output the same 1 PPS and 10 MHz outputs so it can be used as a reference for other devices.

To this point, I’ve only been talking about the electronics.  The last part is the quality of the enclosure.  I am not a fan of the Sony-Black design (I don’t hate it as much as Steve Jobs apparently did) but I am a fan of laboratory or industrial looking test equipment with sharp edges.  I want it to be heavy.  I want people to know that this device is not a toy.

It's possible to build the BEST clock for the consumer market.  So what?

Well, okay, the point of making a product is to sell it, right?  We would need to perform some market tests but based on anecdotal conversations with serious nixie clock collectors and want-to-be time nuts.  Based on my connection to the nixie clock market as well as to the manufacturers, I believe that a target price is $10k is realistically achievable.  I have been shown proprietary designs of unreleased products that sell for similar price points and personally know many of the buyers who have expressed purchase intent in the unreleased products.

How would one go about building this?  Well, there are two parallel design tasks as far as I can tell.  We will assume that we are using a common development platform like a PIC or an Arduino and that some components, such as the power supply can be built from reference designs or purchase as assembled components.  That leaves the tasks of designing the enclosure and the electronics.

The enclosure can be developed without the complete electronics.  The outline of the major components are known and, ideally, this will be a large enough device to ensure that space will not be an issue.

The electrical engineering task will take some research.  None of the components have particularly obscure requirements.  The nixie tubes require a high voltage current limited power source but that should be relatively easy to engineer.

The biggest challenge will be writing the software.  How does one account for more than one frequency reference?  Do you simply average the outputs?  Do you use one to correct another?  Does one simply act as a backup?  These are not questions with obvious answers but these problems have been solved.  Generally speaking, each frequency reference is used to “steer” another oscillator to produce a more accurate signal.  This synthesized signal is what is used to measure time as well as provide the output reference.

Lastly, and this is a major consideration, the insides of the device must look nice.  After I finished Steve Jobs’ biography I had a new appreciate for the level of design that went in to every product.  I want to carry that same level of detail in to my product.  I want users to be able to open it up and I want them to be able to see the inside workings of their ultimate clock.

So what’s next?

Nixie clocks are inherently limited by the number of tubes on the market.  It may be possible to manufacturer new ones but eventually other display technologies will need to be used.

Aside from the display, it may be possible to create variations of the product.  For example, what about a $1000 alarm clock that is based off of the Symmetricom Chipscale device?  There are many other options and those will be the subject of another blog post.

How rubidium oscillators work

[youtube width="483" height="360"]I55uLRRvLCU[/youtube]