Feature locking. Product differentiation? Scientific and industrial instruments

[NeoPTLD]

As computer enthusiasts, we're all too familiar with feature locking used for the purpose of product differentiation. That is, they're manufactured the same, but performance or features are "soft locked" on lower end model to sustain premium value on higher end model.

Many industrial and scientific instrument have very primitive computer.

Here, I'm messing around in settings on a $1,200 MSRP laboratory scale. I'm messing with it, because it was acquired for next to nothing. The sensor itself is damaged from prior abuse. It works, but reading is not repeatable.

There are several models of in this series, all using the same casing and specifications say that they all weigh the same.

In service manual, it describes two sub-models. One is 0.01g reading, other is 0.001g reading. They use different sensors.

In service parts section, there are only two sensors. The 0.001g and 0.01g type.
The unit I'm working on is limited at 200g, or 200,000 count. and reads "overload" when this is exceeded.

The $3,000 model which looks just like it has a million count or 1000g x 0.001g.

The service manual describes using service software to set up the model type and serial number via RS-232C after replacing the board, then using set of weights and temperature to program calibration into the EEPROM.

In service menu, there's an option to display A/D converter output. So, I go in there and low and behold, it provides one million count resolution and weighs all the way up to 1,000g.

Back in the 70s and 80s when memory was absurdly expensive, there's certainly a reason why a million count A/D would cost more than a 200,000 count, but I see no reason for this now.

Given access to factory service software, I can easily see how sensor unit from the low-end model(from say fried board) traded-in can be used to fix up the top of the line model with damaged sensor.

I suppose you can say this is a form of digital rights management to limit features of software controlled products to that of price paid by the customer.

How common is this kind of lock-down in industrial process control, manufacturing and instrumentation equipment?

I can see this as very attractive for vendors seeing that even if it can be defeated easily, they're unmotivated to do so as it voids calibration and make instruments uncertifiable.

 

[CycloWizard]

There are two separate issues here. The first is the quality (i.e. dynamic range, sensitivity, linearity, repeatability, and so on) of the sensor. The other is the resolution of the data acquisition system. If I have a low-quality sensor, then having a high-resolution A/D is meaningless since all of the extra bits will just fluctuate due to noise. With a balance, the range is also limited by the mechanical design of the sensor: overloading can induce permanent deformations or even failure.

That said, I'm not sure how common feature lock-down might be. In my experience, it's usually on the software side where features must be enabled by entering an appropriate license. There are usually legitimate hardware differences between models built for lab use that differ because of real differences as mentioned above.

 

[NeoPTLD]

There are two separate issues here. The first is the quality (i.e. dynamic range, sensitivity, linearity, repeatability, and so on) of the sensor. The other is the resolution of the data acquisition system. If I have a low-quality sensor, then having a high-resolution A/D is meaningless since all of the extra bits will just fluctuate due to noise. With a balance, the range is also limited by the mechanical design of the sensor: overloading can induce permanent deformations or even failure.

That said, I'm not sure how common feature lock-down might be. In my experience, it's usually on the software side where features must be enabled by entering an appropriate license. There are usually legitimate hardware differences between models built for lab use that differ because of real differences as mentioned above.

They do have another product series that only goes up to 300g, which is just as capable, smaller and cheaper.

This one is based on a platform capable of going to 1,000g, which explains why this line is costlier.

In the NTEP certification document, it lists the same exact weighing cell model # for all models within this series.

 

[Mark R]

I've seen this quite a bit on medical equipment like MRI scanners.

You'll have a mid-range model and high-end model. Both will have pretty much similar physical specifications e.g. size, magnetic field strength, etc. but there will be some differences - e.g. the mid-range model might have 8 RF receiver/DAQ modules, whereas the high-end model might have 64 RF receiver chains.

However, there will frequently be functional differences - e.g. the low end model might not offer NMR spectroscopy, magnetic susceptibility imaging or water diffusion tensor estimation.

The additional RF front-ends make life easier for the techs, so are desirable (but are insanely expensive - I'm told they cost around $20-30k per channel) - e.g. for a whole spine scan, they can simply apply a 64-channel phased-array whole body antenna array to the patient on a 64 RF channel scanner and click GO. On an 8 channel scanner, they'll be scanning the neck with one antenna, disconnecting it and plugging in the thoracic spine antenna, scanning the thoracic spine, then disconnecting the antenna, plugging in the lumbar spine array....

However, options like spectroscopy, diffusion tensor estimation, etc. are purely software options and applicable regardless of the hardware load; something I have seen is that at launch, the options are only available on the top-end model. Then after a year or two following release, they'll be made available for the entry-level model, so that clinics looking to upgrade to the latest model are forced into buying the top-end model if they want to get these sophisticated features.

 

[NeoPTLD]

However, there will frequently be functional differences - e.g. the low end model might not offer NMR spectroscopy, magnetic susceptibility imaging or water diffusion tensor estimation.

Do these require additional extensive calibration that actually adds production cost substantially or are they fairly trivial?

For lab scales, I believe calibration & adjustment by technician is one of the most expensive step in production.

There are factory calibration parameters that must be programmed in to adjust for individual variation in components. In fact, its these steps that make newer units cheaper to make. The parts don't need to have as exacting values as long as they're fairly close and extremely stable. The variation just needs to be memorized into the EEPROM.

The 1kg model and 200g use the same steps. Just different weights for linearity and span calibration.

 

[Mark R]

Do these require additional extensive calibration that actually adds production cost substantially or are they fairly trivial?

The data acquisition for some stuff is trivial - just minor changes to the scripts that run the hardware; you may not even need to change the configuration GUI on the acquisition computer as the basic setup parameters are the same. No additional calibration, etc.

The difference is all in the software that:
1. converts raw sensor data from the scanner into a dataset of interest (e.g. reconstructing a spectrum is different reconstructing an image)
2. software to post-process images (e.g. magnetic susceptibility imaging works by filtering an image into different components, and selectively enhancing contrast of the components, then recombining)
3. or software to display and interact with the data (e.g. tensor data is highly multi-dimensional and can be explored in many ways)

When I worked with a group that had a research agreement with the manufacture, they had direct access to the hardware scripting engine, so could develop new or clone existing techniques, although the data processing usually ended up having to be done in matlab.

 

[NeoPTLD]

Is the higher price on the hardware, or is it for the additional licensing on post processing software?

I don't expect that post processing feature on the software to be free. This is different from software limiting the available function of hardware.

 

[Colt45]

I think it's pretty common. A lot of the time, at least part way through the life cycle, it's cheaper to just make the 'better' model, than to make both, in some cases.

I don't see what memory has to do with this..? You only need 3 bytes of ram to hold a million.

However, a higher bit ADC generally costs more. But 200k/1M is only 2 bits difference.

perhaps they both used the 20bit ADC, and units that had less accuracy / couldn't be calibrated / whatever / were 'binned', the lower 2 bits thrown out, and boom, 200k count.



I guess the other option is that the ADC and on out is the same in both models - only difference is the.. strain gauge, or whatever the element is... If it's lower tolerance, there is no point printing the extra bits, as they have no weight.. they're just error.