the wikipedia article, unfortunately, is of not much help in understanding how this thing works.
Ok, so if a single chip can return its bit string, can two chips connected together return a concatenated bit string based on components’ strings?
Is there a way to apply a bitmask, let’s say zeroing out the 2nd and 4th bit? Or, looking at the issue more generically, another transformation function that maps to a bit string of the same length, such as OR’ing with a bit string of equivalent length?
ETA: my question is not specifically about some existing RFID chip on the market (obviously the answer would be no, nobody wants to do things like that), but rather about the potential of making chips with such properties for the same sort of trivially low price that they make the existing ones.
Two chips cannot be concatenated together to form one big super chip with a really long bit string.
You could make a chip that takes an incoming bit string and performs some sort of transformation on it then returns a bit string containing that transformation. This is commonly done as a security feature in some RFID systems. The algorithm is usually a lot more complicated than simply applying a bitmask.
On second thought, maybe “concatenation” is just a special case of a transformation that does not preserve the number of bits. Let’s say we use chip B to transform string of chip A, and it just so happens that the transformation is appending a bit string unique to chip B.
Ok, now how do this sort of transforming and capable-of-being-transformed chips fare in terms of price? If I wanted to make a million chips A and a million chips B, would they end up being just as nice and cheap as the vanilla $0.05 passive RFID chips?
Or would let’s say the need to have an interface allowing the two chips to be put together and taken apart “manually” massively increase the cost of the system even in the best economy-of-scale case?
Or could we implement this setup for a lot cheaper by making the interaction between chips A and B be some analog electromagnetic interaction happening at close proximity through the plastic cover and hence not involving precise alignment of metal wires?
I’m not really understanding what you think you are accomplishing by having a second chip in the system.
RFID is fairly simple. There are several types. The simplest, which is hardly ever used any more, uses RF diodes which are tuned to particular frequencies. You “program” the tag by clipping out the diodes that you don’t want to respond. This gives you a very limited number range because each bit requires an additional diode and frequency.
Most RFID tags these days use an integrated circuit of some kind. There are two types, beam powered and battery powered. The beam powered ones divert the incoming RF signal into a capacitor. Once the capacitor is charged, it then functions like a battery and powers the integrated circuit. In either case, the integrated circuit can be the simple type that just spits out a bit stream, or it can be more complicated and do some number crunching on the incoming bit stream and use that to generate the response bit stream.
It’s basically a command/response protocol, so I don’t really see where having a second RFID chip in there accomplishes anything. If the second chip is doing some kind of number crunching, you’ve still only got one command going out and one response coming back. Any number crunching that could be done and passed from one chip to another would be much more easily accomplished by simply making the single RFID chip a bit smarter (i.e. more logic gates inside of it) and having it do both calculations.
You want to keep your number crunching to a minimum, since the more processing you do means more battery power or more power required from your incoming RF beam. More number crunching also means a more expensive integrated circuit, though your costs per chip will be pretty small if you make a bizillion of the things.
By the way, this is a good example of what is inside the plastic.
The spiral copper bit is the antenna. The chip is the little black thing in the center, attached to the antenna. The little squiggly spot in the center that isn’t attached to anything is just the Texas Instruments logo (which makes it pretty easy to identify who made this one )
If you want to store and pass data from one chip to another you’d be a lot better off using a distributed network protocol like what is commonly done with RF PICs.
You need a custom device to do this. You also need to think through your requirements - how many connections do you need to support? Is order important? How about X,Y,Z connectivity? Distance?
yeah, but that sounds expensive. I think a distributed network protocol will require microprocessors in its nodes, right?
Whereas I want to be able to use the off-the-shelf passive RFID scanners to detect whether or not two dirt cheap passive chips are “touching”. So I guess the ideal of it would be:
scanner sends a signal to chip A
chip A transmits its bit string to chip B either through a wire or through some short distance capacitance / other E&M mumbo jumbo effect
chip B does the transformation and computes its output
scanner picks up the output of chip B
So you think that’s impossible to do while keeping to the passive chips approach? Or possible but probably a lot more expensive than the cheap chips we are used to nowadays?
I don’t see that being possible using passive RFID chips. The signal that the chip sends back is significantly weaker than the signal the scanner sends to it. The signal from chip A wouldn’t be strong enough to get chip B to start up. Chip A also needs to be smart enough to detect whether it is connected to chip B or not, and now you are starting to get into the realm of decision making microcontrollers anyway.
How does chip A know that it is chip A and how does chip B know that it is chip B? How does chip B know that its signal is coming from the scanner vs. chip A? In your scenario, both chips would respond simultaneously to the scanner, which would garble the signals up pretty badly.
ETA: RF PICs are a couple of bucks each. I don’t think a custom solution is going to be any cheaper than that unless you literally produce millions of the things.
to clarify my thought there, no chip “knows” anything. Chip of type A always transmits its bit string, whatever it is, to chip of type B, if any such chip happens to be sufficiently “nearby” to pick it up. Chip of type B always does its built-in transform on what it gets from A and transmits to the scanner. If no chip A happens to be nearby presumably chip B does not transmit anything or else transmits “I am not connected” string.
So you think that no electromagnetic effect would allow transmission of this data from A to B given the power levels in these passive chips?
ETA: could chip B be “started” by the power transmitted by the scanner, but nevertheless work with the bit string it “senses” in some way from chip A?
The output from a passive chip is nowhere near strong enough to power another passive chip. Active chips can work at a distance of ten to twenty feet, and are often used in such a way on railway cars. Passive chips have a range of a few inches at best because the signal coming back from them is so weak. Chip A’s output is a couple orders of magnitude too low to trigger chip B.
The things you are trying to do are easy enough for an RF PIC, but are far beyond the processing and decision making capabilities of current RFID chips and protocols.
You could theoretically engineer a custom integrated circuit to do the things you want to do, but it would take very careful design and very large production numbers (millions) to get the price down to anything close to what you want.
just to be clear here, when you say “power” do you mean “be the only source of power” or do you mean what I am describing in post 11, i.e. no passive chip can generate a signal strong enough to be detected by another passive chip? Presumably, or so my thinking goes, once the signal is detected, it can be amplified using the energy supplied by the scanner.
So what sort of a price do you think it will be with million per year production?
Perhaps this question should be asked more generally - suppose we are planning to produce some small chip that contains some analog and some digital processing features. Is there some rule of thumb for establishing the lower bound on its price of production in the best possible mass production case? E.g. if passive RFID chips sell for $0.05, is that a common price for small mass-produced analog+digital chips?
I suspect that it would have both problems (not enough power from the incoming RF to power the device and not enough power to be a detectable signal).
An amplifier adds complexity. Current RFID chips use fairly simple receivers because they expect the signal level from the scanner to be pretty high.
You may be able to do this sort of thing with a custom chip but it’s definitely going to be a difficult design challenge. I also don’t see this scaling up very easily (i.e. what happens when you add chips C and D?).
My experience is all in low volume production. When I have been involved in custom chip design we only had maybe a thousand made, so they were pretty pricey.
Generally speaking, a fairly simple chip can be produced for maybe $0.25 to $0.50 each in volume. Getting something down to $0.05 each not only requires the device to be really simple, but it requires a huge engineering effort as well. Keep in mind that RFID has been around for decades and the designs have gone through dozens of refinements.
I guess a stumbling block for my understanding is that I don’t understand the difference between a bit string originating from inside a single chip and the bit string originating from some similar other chip.
Let’s consider an unrealistic setup as follows. We will remove the transmitter part from chip A and simply feed its bits into chip B via copper wires. Then chip B, for simplicity, will do no processing of any kind but will simply retransmit it. I guess this illustrates the possibly faulty mental model that I have - this setup sounds to me perfectly isomorphic to the “normal chip” setup and hence it sounds doable, at least if all the copper wires are nicely connected on the workbench by an expensive technician. Then, from this setup, it is a natural leap to the “chip B transforms the bit string before transmission” situation.
Ok, so apparently my mental model is wrong. So what am I misunderstanding about the situation? Are those bits in chip A actually never sent to the transmitter in any way because they are “built into” the transmitter? Do these bits ever actually show up in some on-chip wires/cables or do they manifest purely as the analog response to the scanner signal?
An RFID chip these days is apparently a lot simpler than you seem to think it is. The el-cheapo ones are basically just a transmitter. That’s it. The bit pattern is essentially built into the transmitter and never goes out across any wires at all external to the chip. The tag is literally just a chip and an antenna (and a capacitor). That’s it. Disconnecting the transmitter would involve redesigning the semiconductor chip.
The more expensive RFID chips that do security processing and such have a much more elaborate integrated circuit that does have a receiver built into it, but it’s expecting a fairly high level signal and wouldn’t be able to hear the signal transmitted from another chip. These are going to cost you a bit more than a nickel a pop though.
Also, if A and B aren’t identical (A only receives, B only transmits) then you don’t get your economy of scale from producing a few bizillion of them. Are you going to make half a million of A and half a million of B? Is there ever going to be a C or a D?
In terms of manufacturing, yes, it is implied that chips of type A and B are produced in equal numbers. There is no C or D because we only want to deal with physical couplings between two items, like in Lego. So let’s say the male slot gets type A, the female slot gets type B and we use our scanner to find the couplings.
Ok, some more questions.
You mentioned the power limitation for passive chips. So, what power is available in the chip when it is being scanned? Could we conceivably increase that power by either increasing the power of the scanner’s transmission or else by increasing the size of the receiving antenna (for applications where we don’t mind a bigger chip)? In terms of scanner power, is it now hitting the limit of the danger to humans, or is the problem more along the lines of the chip getting fried by a big current?
In terms of combining the bits from two chips, what if we put into chip A a transmitter without antenna and have it connect to the antenna of chip B when they are “attached”. Then we will design the two transmitters to fire away with a small delay, first A and then B, in such a way that the scanner would recognize the two bursts as happening via the same antenna. I guess the feasibility of this would depend on how the scanner does the multiplexing between signals of different chips, which I never saw explained - so any input on that is also appreciated.
I do realize that even if the above happens to be “feasible” on the workbench it probably wouldn’t be realistic in real life Lego. Nevertheless, I think that it is a good stepping stone for me for understanding the mechanics of the process.
code_grey, I honestly do not understand what you are trying to achieve.
Also why should the RFID chips have any number crunching capabilities at all? It is a lot more efficient if each chip feeds their bit string to a central computer and let it do all the number crunching.
As an engineering challenge, I think this would be easier to implement by getting the scanner to work out the relative positions of the two responding devices all by itself, maybe using a combination of triangulation and interferometry, etc. I believe these techniques can be employed on a fairly small scale.
Maybe I’m misunderstanding something - but is the basic problem getting the RFID reader to detect when chip A and chip B is in close proximity? If so, is it close enough if both chips are within reading distance of the RFID reader? (I’m assuming we’re talking close range 14443 proximity chips here - that is, 10-15cm reading range at most).
If so – why not just read both chips, and let the RFID reader determine the result. Giving the series A and series B chips unique serial numbers according to a sorting algorithm will allow the reader to determine if it, in fact, has two chips within its range, and if they are both type A and type B.
I helped designe a RFID reader a few years back, and we could read up to three RFID chips in the same field, without collision issues (we could read four or five, with a bit of a hit and miss collision detection algorithm).
)