This is yet another interesting experiment that took place in the PCB Heaven techlabs these days. This experiment has to do with the digital caliper protocols. These calipers have a communication port for interfacing with a PC or other supported devices. As far as i know, there are basically 3 types of protocols for interfacing them. The first one is the Digimatic from Mitutoyo. The other two protocols are both for the Chinese calipers. To distinguish them, we will name them Chinese-BCD and Chinese-Binary protocols. The BCD and Binary suffixes comes from the way that the data are being transmitted.
Because the Chinese calipers are ridiculously cheap, i experimented with two calipers covering both the Chinese-BCD and the Chinese-Serial protocol. You can see the experiments here:
As far as the Digimatic protocol is concerned, i have not yet done any experiment. Instead, you can see the full documetnation of this protocol from the following file. This is a complete 52-bit protocol that is being transmitted only upon request (handshake), and the data comes in BCD nibbles:
Yes, I bought about $70 worth of stuff from them 2 weeks ago.
Some RTC (the model I suggested to you) and some high voltage supplies
for a Nixie project I am doing. For products that are basically just
to support hobby users, their documentation is very professional.
As both his company and I are in California, the shipping time was
only 2 days. Ordered Tuesday, order acknowledged Wednesday, shipped
on Thursday, received Saturday. Obviously will take longer to ship
to Greece. Just ask questions via the Contact page.
The RTC modules looks nice, and the battery lifetime is also brilliant. Have you ever bought from this site? They use paypal cart, i suppose it is secure.
"But for my application, i will use external power to either power the LCD or the micro that will convert to RS232. The same power will be used to power the caliper - that will no longer rely on batteries."
In that case, a normal NPN transistor, driven through maybe a 10K base resistor would be a good starting point for your experimentation.
"I will need your suggestions for a clock that i think of making, that will rely on batteries when the mains is out (this happens quite often here in Greece)"
Oh, now it makes sense. But for my application, i will use external power to either power the LCD or the micro that will convert to RS232. The same power will be used to power the caliper - that will no longer rely on batteries.
Actually, i will attach one caliper (liner) to the spindle of a carpenter's router, to measure the height of the cutter. I will not have any problem with the power supply. I will need your suggestions for a clock that i think of making, that will rely on batteries when the mains is out (this happens quite often here in Geece)
There are two issues. How much power is drawn from the caliper to
drive the level shifter, and how much power the level shifter draws from the system it is connected to.
The caliper battery has very limited power, so if you don't want to
affect the caliper battery life much, you need a CMOS type load.
For example, a CMOS chip, or the gate of a FET.
So a FET transistor might work, although you will need very low Vgs.
Maybe something like a 2N7000 or 2N7002.
If you use a NPN (or PNP) transistor to sense the caliper output, it
is easy to make it work with the voltage swing from the caliper, but
it will draw power that is many times the normal caliper power
consumption.
The suggested chips are designed specifically for these voltage
ranges. In my application, I also care about the current consumption
of the system my caliper is connecting to (it also has a very small
battery), and the suggested chips have an idle current of 0.5 uA.
Of course, if you can find a microprocessor for which 1.4V is an
acceptable level for logic high, then that would be a good solution.
Most micros use VSS*0.7 for logic high, so a processor that runs at
5V needs at least 3.5V. A processor that runs at 3.3V needs 2.3V .
This is why I mentioned that the caliper output at 1.5 for logic
high would be a problem.
I understand your comments about secrets. I work in Silicon Valley,
mostly for big chip companies. Lots of secrets, lots of re-inventing
of the wheel. In my hobby work though, I try and help others if I
can, and hope others will help me.
As for averaging of the readings, you must calculate the averages of
the full 24 bit numbers (as I did in my prior explanation), not just
the low 3 bits. This is because the rest of the bits can be affected
by the low 3 bits as it rolls back and forth over the major transition
from 111 to 000.
My current project (started 5 years ago, but been on hold for last
4 years) involves reading data from calipers (obviously). As the
voltage levels from the caliper are 0 to 1.5 volt, you may have problems interfacing them to a microcontroller, as the logic high
may not be high enough. May I suggest that you consider the following
parts to do level shifting (that's what I use): sn74AUP1T57 or 58.
These can shift the caliper signals to 0 to 3.3V signals.
Fist of all Philip, thank you for the links. Zero and Mode button! Of course!
It looks like you've done quite a nice job yourself! Where i live (Greece), people hide their so called "Secrets" from their discoveries. Not to accuse them as they make a living from this discovery. But that is the main reason i am now on the net. I believe that there is not actually a real secret. What someone has done, someone else can do it better. The difficult thing (that could be characterized as a secret) is to thing of a problem rather than finding the solution.
Anyway i will consider averaging the 3 LSBs. Now that i think of it again, i think you are right. Although with this cheap caliper such an accuracy would be pointless, it has to be done for the sake of the experiment. I am planning to make a PC interface or a remote LCD display.
Notation I will use: For a 24 bit number, the least significant bit
(LSB) is 2^0 , and the most significant bit (MSB) is 2^23
For point 1: I agree that the 3 LSBs are noisy, but they do have value.
For example, if the three LSBs (2^2, 2^1, 2^0) are toggling between
111 and 000, then the 2^3 bit is also changing as a by-product of the
noise in the 3 LSBs. Noisy data can be improved by the square root of the number of samples that are averaged. (from DSP theory, for random noise)
For example, if you average 4 measurements (add four 24 bit numbers, and then divide by 4 (down shift by 2) the new number will be twice
as accurate as before. In this case the 2^2 bit (of the average)
will be less noisy than the separate measurements. This is just
normal DSP type signal processing, not specific to your caliper data.
The reason the 3 LSBs are noisy, is that the measurement process is
noisy.
For point 2: By weight of bit, I am referring to how you interpret the binary data. For example, the binary number 1100 has bit weights of 8, 4, 2, 1, giving a decimal value of 12. For the data from the caliper, the LSB has a weight of 1/20480 inches, the next bit is 1/10240 inches, etc for each bit being a weight of double the bit to the right. Just normal binary interpretation, with a conversion factor to a physical quantity (distance).
I just collected some data from one of my Chinese calipers. I have seen several different data formats beyond the two you have seen. My current caliper outputs 2 x 24 bits, absolute then relative, at 76 KHz. Crazy stuff is the relative data (second set of 24 bits) are all inverted.
Here is the data, with the caliper reading 6.0000 inches
Data sampled on clock falling edge.
Data is received LSB first: 000111111111100001111111
Re-arrange as MSB on left: 111111100001111111111000
Invert all bits: 000000011110000000000111
Show as HEX: 0 1 E 0 0 7
Convert to decimal: 122887
Divide by 20480: 6.00034 inches
Caliper only shows 0 or 5 in 4th decimal place, so this dislays
on caliper as 6.0000
Hello Philip,
For point 1: I am not quite sure if these bits really mean something. The resulting number is accurate (except this 1.006 thing) w/o reading these bits.
For point 2: What do you mean by "weight of bits"? And also, for example i measured 100mm (3.937 inch) and got the result from the port 100.6mm (3.961 inch). How did you came to this 1/20480?
Nice article. With regard to the accuracy issue in the last section,
you may want to try the following:
1) Don't throw away the LSBs, instead average the readings to reduce
the noisyness.
2) Your article didn't indicate how you are interpreting the weight
of the bits. I believe the units are 1/20480 of an inch.
i.e. when the caliper is measuring 2 inches, the transmitted value
is 40960. Maybe this accounts for the error you are seeing.
Home
Contact
Forum
Projects
Experiments
Circuits
Theory
BLOG
PIC Tutorials
Tech-News
RSS
The Digimatic protocol from Mitutoyo
Rate this article!
Reddit this
HOT in heaven!
NEW in heaven!
