The human eye’s color perception is notoriously variable (see, for example, the famous dress), which makes it difficult to standardize colours. This is where spectrophotometers come in: they measure colours reliably and repeatably, and can match them against a library of standard colors. Unfortunately, they tend to be expensive, so when Hackaday’s own [Adam Zeloof] ran across two astonishingly cheap X-Rite/Pantone RM200 spectrophotometers on eBay, he took the chance that they might still be working.
They did work, but [Adam] found that his model was intended for testing cosmetics and only had a colour library of skin tones, whereas the base model had a full colour library. This was rather limiting, but he noticed that the only apparent difference between his model and the base model was a logo (that is, a cosmetic difference). This led him to suspect that only the firmware was holding his spectrophotometer back, so he began looking for ways to install the base unit’s firmware on his device.
He started by running X-Rite’s firmware updater. Its log files revealed that it was sending the device’s serial number to an update server, which responded with the firmware information for that device. To get around this, [Adam] tried altering the updater’s network requests to send a base unit’s serial number. This seemed promising, but he also needed a device-specific security key to actually download firmware. After much searching, he managed to find a picture of a base unit showing both the serial number and security key. After substituting these values into the requests, the updater had no problem installing the base model’s firmware.
[Adam] isn’t completely sure how accurate the altered system’s measurements are, but they seem to mostly agree with his own colour calibration swatches. It’s not absolutely certain that there are no hardware differences between the models, so there might be some unknown factor producing the few aberrant results [Adam] saw. Nevertheless, this is probably accurate enough to prove that one of his roommates was wrong about the color of a gaming console.
We’ve seen a few projects before that measure and replicate existing colors. The principle’s even used to detect counterfeit bills.
Until it’s tested to be accurate 100%, I’d be careful about using the altered model. It is possible the cheaper cosmetic model has a defect that prevented it from accurately reading the color and was rejected and relegated to cosmetic use only.
It’s probably lacking proper Calibration, which takes time and is therefore expensive to do
I very much doubt so. It´s much much likely just the firmware, along with some per-device calibration settings: same sensor, same hardware.
Since it´s a Pantone device, i cannot imagine each device is not factory-calibrated. Pantone´s whole business is all about accurate colors, after all…
Even if the hardware’s the same, and it’s calibrated, It could easily be binned sensors – the <1% ones get the regular firmware, with the 1% – 5% ones getting the cosmetic firmware.
If he could find someone with a base model he could easily test the accuracy of his.
Why does potent hardware deliberately limited by software + some licensing bs reliably make my blood boil? I guess I will never in my life consider it wrong to break into one’s own hardware so its true potential can be unleashed. Good for this person that they done it! Less waste.
This is quite common, most unfortunate, but also understandable.
One of the justifications is that the “cheap” version cost more then BOM + Assembly costs, so the company can make a (modest) profit from that, while the more expensive “fully featured” version allows the company to generate enough profit to pay the salaries of the engineers who design the next product.
But the limits are vague, with managers and CEO’s skimming of millions in “bonuses” while the overall company is struggling.
Worst case I heard was from a car, which could have a simple on/off for the windscreen wiper, or an interval switch with a bunch of settings. The switch was the same, but in the “simple” version, an extra piece of plastic was added to block the extra positions.
Another example are the Rigol and Siglent oscilloscopes. Extra bandwidth (officially) costs money, while in practice, up to around 200MHz, hardware costs are negligible. The software block is quite easy to circumvent, and I think those companies also do that on purpose. “Unlocking” the software is attractive to hobbyists, while companies just pay extra for the bandwidth.
I’m guessing that ST does something similar with their uC’s. The sell over 3000 models (on paper), but apparently many have double the flash size, can have an USB peripheral that should not be in it according to the datasheet and other changes. For them, this reduces cost, because they need less masks, and silicon variations. And when a model becomes popular enough, they do create a set of masks without the extra peripherals so they can make more chips on a wafer, and it still has (just) the peripherals as stated in the datasheets. The “extra” peripherals on the silicon are probably also skipped during production tests, and faster production is lower costs.
Rigol also goes quite far in this. They’ve got a power supply sold in two versions, with different resolution in current and voltage resolution / accuracy. The expensive version has a color display, while in the cheaper version, the display is limited to monochrome. I don’t buy rigol equipment anymore. It’s too “toylike”, with numeric buttons in a circle, a DMM with an atrocious font, and in their newer scopes they put android on a uSD card.
Kind of the, BT with your WiFi because overall it’s cheaper to include than having two separate chips.
There was a time that chips with varying RAM or flash sizes (or other features) were selected at test time: Those that failed a test got packaged and sold as the less-featured version.
That’s probably less true now that process yields are better, but I’ll bet some memory is still made/sold that way.
CPUs certainly are. I’d guess anything near the edge of technology is.
“High end stuff” Probably yes, but microcontrollers are generally made on 20 year old machines and with 20 year old technology. And the process has been well optimized.
Also, when machines are made very precise, and well maintained, they do not wear. Really wear is 0, because there is no mechanical contact. And that is why even 30 year old semiconductor machines are still running.
This is both a blessing and a curse. It’s cheap to manufacture IC’s on those machines, because their development has already been paid for many times over.
It’s a curse because it puts a hard limit on the production capacity of “old technology” semiconductors. The knowledge of how to make those old machines is mostly lost, and re-creating that technology and making new machines is prohibitively expensive, because production has to compete with those old machines.
In the early days of semiconductor manufacturing yield was sometimes reduced to 0. And that for months on an end. Can you imagine that? A whole semiconductor factory running, and no product to sell at all? One of the reasons for this was nearby farmers were spraying their crops with sulfur containing pesticides, and trace amounts of sulfur leaked into the clean rooms.
There are some very interesting documentaries about early semiconductor manufacturing on Youtube.
There is a market segmentation here.
One full featured model for the printing industry, which needed the features, and at the times still had fairly big money.
Another less capable model with a limited feature set, for an audience with a tighter budget. It may also have a lower production cost due to simpler calibration, or simpler (out-of-spec for the full feature set) hardware.
Both groups benefit from this setup, because economy of scale is at work here. Development cost is shared between both groups.
Many things today are no longer differentiated by the hardware, but by the software. Software development is expensive, although many people keep forgetting this. If something is “free of charge”, most of the time you pay in some other way.
yes, and just having to only get all the required approvals so legal sell something for one device is going to save money
I agree, though I’m not sure that’s the case here.
The company was providing a product specifically to ID skin tones for makeup and coded it accordingly. Adding in the features and libraries for any color complicates the usability (it may also cost more if the color libraries are licensed, which they probably are). It is like using a #2 Phillips head screwdriver for work that require it, rather than a multi-head screwdriver. A simple firmware update process may also be an advantage in keeping the tool working as needed. These limitations may be benefits the customers want for this application.
when you buy PC you are getting “potent hardware deliberately limited by software + some licensing” …
Writing custom firmware which leverages the hardwares full potential would be a cool hack, but this is not.
This is just using illegal downloads.
Man, are they putting dog whistles in kid’s cereal boxes now?
Go defend the big corporations, maybe they’ll love you if you do.
For the dress:
https://en.wikipedia.org/wiki/The_dress
I don’t do asocial media and had not seen it before. The phenomena is quite curious. It would be nice to have some more pictures of that dress in different light conditions.
Another lighting trick is in supermarkets, where the grocery section has very weird lighting (Just look up, the lights are different from the rest of the store). This is done to make the groceries look “better” (From their point of view). From the consumer point of view: it hides brownish colors so it’s more difficult to see whether the groceries have started to deteriorate because they’ve been lying in the shop for a few days.
It wouldn’t surprise me if there is an automated calibration procedure during production. That may be more cost effective than precision in parts, assembly, and optics for manufacturing. The calibration for the skin tone model may be more limited, not covering the full color gamut, which could hurt accuracy even after updating the firmware for broader operation.