Where are the "systems integrators" now?
Phil Pilgrim (PhiPi)
Posts: 23,514
When I first started designing electronic systems for control applications in the 1980s, I was able to pick and choose among a host of available modules to cobble together into a working whole. I didn't have to design or fab any PCBs -- just bought board-level microcontroller modules, Opto22 stuff, etc., off the shelf, and combined them into a working system. I was a "systems integrator," and I could charge a lot for my services. It wasn't the cheapest way to go, certainly, given the option of designing everything from scratch, but it was the most efficient -- especially given the moderate volumes I was dealing with.
Fast forward to the late 1990s, early 2000s. I soon realized that I could design stuff for other systems integrators. Optical sensors were my specialty, and I was able to come up with some successful designs, which I private-labeled for Parallax and which they were kind enough to market for me. (They still do.) These modules were sold primarily to what I would consider the "middle class" of users: high-level hobbyists and entry-level OEMs, i.e. systems integrators.
But, lately, this middle ground of users seems to be in decline. Hobbyists and even entry-level OEMs have become more cost-conscious. Hobbyists want cheap stuff from China, regardless of any lack of features, quality, or support. OEMs want tailored designs that rely upon jelly-bean parts, rather than another company's IP. Then there are OEMs who don't care so much about costs, as long as they can wring a high profit from very exclusive products to a limited clientele. (Think research labs, government contractors, etc.) This situation was brought to bear recently in erco's post about Hamamatsu's latest spectrophotometer module:
https://forums.parallax.com/discussion/168232/almost-as-good-as-phipi-s-m-m-sorter#latest
That module is amazing (!) but expensive and not easy to obtain (think Brooks Brothers). At the low end of the color-sensor spectrum is Parallax's ColorPAL (think Men's Wearhouse), which seems to sell well to hobbyists and educational concerns. In the middle is the TCS3200 color module (think Nordstrom's). It once ignited a flurry of interest among the "middle class" users -- especially during the brief tenure of the M&M sorter. I know of at least one OEM that still uses them, but interest in it has faded over the years.
So I'm wondering if, like the American middle class, the middle class of electronic modules -- the ones that systems integrators once relied upon -- is disappearing. And, as a consequence, are systems integrators also a vanishing breed?
Thanks,
-Phil
Fast forward to the late 1990s, early 2000s. I soon realized that I could design stuff for other systems integrators. Optical sensors were my specialty, and I was able to come up with some successful designs, which I private-labeled for Parallax and which they were kind enough to market for me. (They still do.) These modules were sold primarily to what I would consider the "middle class" of users: high-level hobbyists and entry-level OEMs, i.e. systems integrators.
But, lately, this middle ground of users seems to be in decline. Hobbyists and even entry-level OEMs have become more cost-conscious. Hobbyists want cheap stuff from China, regardless of any lack of features, quality, or support. OEMs want tailored designs that rely upon jelly-bean parts, rather than another company's IP. Then there are OEMs who don't care so much about costs, as long as they can wring a high profit from very exclusive products to a limited clientele. (Think research labs, government contractors, etc.) This situation was brought to bear recently in erco's post about Hamamatsu's latest spectrophotometer module:
https://forums.parallax.com/discussion/168232/almost-as-good-as-phipi-s-m-m-sorter#latest
That module is amazing (!) but expensive and not easy to obtain (think Brooks Brothers). At the low end of the color-sensor spectrum is Parallax's ColorPAL (think Men's Wearhouse), which seems to sell well to hobbyists and educational concerns. In the middle is the TCS3200 color module (think Nordstrom's). It once ignited a flurry of interest among the "middle class" users -- especially during the brief tenure of the M&M sorter. I know of at least one OEM that still uses them, but interest in it has faded over the years.
So I'm wondering if, like the American middle class, the middle class of electronic modules -- the ones that systems integrators once relied upon -- is disappearing. And, as a consequence, are systems integrators also a vanishing breed?
Thanks,
-Phil
Comments
There is opportunities for new inventions to enter but the general supply chain is pretty cut and dry "give me something that just works with as little configuring as possible."
All of that tends to encourage brand lock-ins.
The way I see it today or technology stack is very high and very wide. From the guys turning sand into silicon all the way up to Amazon Cloud Services.
Every layer in the stack builds on the layers below.
Everyone is an "integrator" of the functionality the layer in the stack below them provides.
Me, I'm a "systems integrator" still. My thing recently is mashing IR cameras, radars, wireless communications, various bits of software, and my own magic sauce, into a system that kinda, sorta does what is asked.
Last of a dying breed I guess...
Back when I was doing "systems integration," if you needed a custom PCB, you'd hire someone who would lay it up by hand on gridded vellum using hand-cut rubylith and press-apply pads at 4x actual size. This was a pretty high bar, and it was easier and cheaper to cobble together ready-built modules to do what you needed.
Now, almost anyone can do a PCB layout using free, or almost-free, software. Along with the ubiquity of cheap, low-volume PCB fab services, this makes it possible for many to bypass ready-built modules for custom designs. So I'm sure this is a factor in the decline of "traditional" systems integrators.
-Phil
So, if I get your meaning of "systems integrator" correctly, he was the guy cobbling together custom solutions out of boards and modules and such to meet a customer requirement. Probably in short time and in low quantity.
Being closer to the customer he made good money doing that. Compared to the schmucks designing the boards and such. Being closer to the customer, and hopefully understanding their problem domain perhaps made him worth the money.
OK, so what if doing PCB layout is now very cheap and easy and getting boards made is very cheap and quick? Surely that just provides another tool for the system integrator to quickly and cheaply do what he does, as described above.
P.S. I did once work for a group doing Schematic Capture and PCB layout design software, CADSTAR at Racal. That was $10,000 a pop for the most basic setup. Today, KiCad is available for free and is much better!
You can still buy the modern day version of CADSTAR if you like: https://www.zuken.com/en/products/pcb-design/cadstar
1. Are more expensive than now-cheap custom-designed boards,
2. Bring more to the table than the design requires, again for more cost, or
3. Require expensive connectors, cabling, and/or larger housings to tie everything together.
BTW, I just completed a design for an OEM customer that required a microcontroller, a display, and a sensor. The sensor had to be a ready-made module, due to the complexity. The display was accomplished using a custom PCB I laid out using jelly-bean parts. The microcontroller? A Parallax FLiP module. The FLiP is an amazing value for the money that I couldn't hope to duplicate on a custom PCB for the same price.
-Phil
Components have been getting cheaper and cheaper for decades.
Industry in the west has been decimated.
Everything happens on a large, global, scale now. All is integrated.
I'm sure there is scope for the one man problem solver somewhere. The problems are different. The somewhere I know not where!
A lot of 'systems integration' can now be achieved via software. A lot of industry sectors that used to create their own equipment merely integrate COTS/MOTS with a few very specialist items, e.g., the defence industry. ASICs are less required than they used to be - FPGAs and flexible hardware have seen to that.
Perhaps it's just a generational thing, going back to browsing through thick catalogues and specs, mentally thinking how things might work together?
I see integration as all the rage, but it depends on what level you're aiming for. Sparkfun, Adafruit, Pololu, Parallax -- they're all making module-level integration boards that encapsulate a sensor or function. Many of the products are aimed at hobby and education, though I'm sure system designers are buying things too.
Once thoroughly in the mainstream these products are naturally overrun by generic replacements. In 2000 we were making our own ultrasonic sensors, or cribbing them from old Polaroid cameras. Now there are numerous sources for these, at various levels of competence and price. You can get them from China for less than a package of M&Ms.
On the consumer front the interest in any product diminishes over time. Initially there's interest just to experiment, but eventually that evaporates as new things come along. I see things like TOF sensors as capturing the imagination today. That won't last forever.
As to one such as Phil, unless someone comes to you with hat in hand with said hat containing a load of cash, are you going to develop something expensive hoping it may close a deal? The project that got me into the prop would have used a interferometric pressure sensor which could have been able to be mounted inside a cardiac catheter. Is cost about $50.00USD. The interface module for this was around $800.00USD. End of development without significant grant money.