next topic »

[Reply #60 on:]

Hi Marco,

about the tri-state buffer, I think that a more suitable equivalent scheme may be this one, where basically both the drivers of the CMOS output are disabled.

Rather, from the graphs looks that the voltage transferred from the source to the bus is too low, and that may create problem due to hysteresis in the comparator.

The FSK wave equivalent circuit for our scheme is an RLC where the bus voltage is equal to the coil voltage, so we can use a frequency analysis for the ratio between the coil voltage (VL) and the input voltage (V).

Is too hard write the complete formula in this forum, but considering the first harmonic of the FSK wave at frequency omega (rad/s):
omega->0, VL/V->0 ;
omega=omega0=sqr(1/(LC)), VL/V=omega*L/R and I=V/R (max current) ;
omega=omega1=R/L, VL/V ->1 ;
omega>>omega1, VL/V = 1.

To size properly the components we may start with the max current I=V/R, using R=270 Ohm we have no more than I=18mA at omega=omega0. The omega0 condition is the resonance of the RLC and from this frequency the ratio start to increase.

Assuming to use two waves at 4 KHz and 7 KHz, we may define omega1=3 KHz (about 20 K rad/s), so the FSK waves will have a high gain on the bus. The L = R/omega1 is about 13 mH.
So, we have to use a C value that give us omega0<omega1. Using C = 1uF we have omega0 = 1.5 KHz.
Using R=270 Ohm, L=13mH, C= 1uF the current will be always lower than 18 mA and the FSK waves will have a VL/V ratio near to 1. Than the bandpass filter will cut the noise that may be on the bus.

I don't know which are the commercial values available for the coils, but something in that range shall be used. Does you have the commercial values table?

I haven't tried to simulate the circuit with that numbers, so I may be wrong on that. I will try as soon these numbers.

Please let me know your opinion.

Thanks for sharing.

Regards,
Dario.

[Reply #61 on:]

Hi Dario.
Effectively the inductors I've placed on the first revision of the simulation file are not correct. Yesterday I've got the same conclusions you wrote in your last post. I've simplified the overall schematics in a piece of paper and I could now better understand how currents flows on the circuit. I've then manually calculated the reactances for inductors and for the equivalent circuit.

The problem is that, adding more boards, the overall impedance "seen" by the transmitting board falls down: each board is a "resistor" that adds in parallel to the bus. The overall impedance seen by the transmitter is a set of parallel impedances made by the power supply and the boards.

Increasing the inductors size will reduce the problem because it increases the impedance introduced by each board in the bus.

I've found on the market some inductors with physical size compatible with PCB restricted size. They're in the range of 6.8mH, with an overall current of 0.15Amps. For the power supply I can find a 33mH rated at 1.5Amps.
For higher values, the physical size of the inductor is too big and the inductor is too costly.

If you run the simulation you can see that so big inductors introduces a lot of "artifacts" at the beginning and at the end of the modulated signal. A 50Hz superimposed on the voltage bus is catastrophic :-(

With a commercial three state buffer we can manage up to 30mAmps... so the budget is very limited.

I'm start to thinking that the overall proposed solution is not affordable due to costs and/or technical problems.
I have to think about.

Thanks!
Marco Signorini.

[Reply #62 on:]

Hi Marco,

The problem is that, adding more boards, the overall impedance "seen" by the transmitting board falls down: each board is a "resistor" that adds in parallel to the bus. The overall impedance seen by the transmitter is a set of parallel impedances made by the power supply and the boards.

Increasing the inductors size will reduce the problem because it increases the impedance introduced by each board in the bus.

Mmmmh... I didn't got the point.

We have one big coil that is the load for the bus, that shall be rated in 10-30 mH and one more coil for each board that shall create us problems.

The filtering used for the power supply is an LC an so can be sized as bandpass filter that shall be out of the FSK frequencies. The DC/DC converter it self is equivalent to a resistance of 240 Ohm. So we have L-C//R (where - is series, // parallel).

Using a uH coil and 220uC capacitor we should be in the safe zone (I haven't calculated this, just smelling the numbers!).

If you run the simulation you can see that so big inductors introduces a lot of "artifacts" at the beginning and at the end of the modulated signal. A 50Hz superimposed on the voltage bus is catastrophic :-(

Could you please post some capture of these "artifacts", because in my simulation (with my scheme) and 35mH coils I've a quite nice signal. Also the 50Hz should be a low enough frequency to be out of our filtering band.

In your previous schematic, using 220uH coils and 1uF capacitor you were in the rage 10-86 KHz (omega0, omega1) so the 50 Hz should be out of these problem. Maybe you are finding this using new values?

With a commercial three state buffer we can manage up to 30mAmps... so the budget is very limited.

If you have time, please try the numbers that I've proposed. I will try to do the same within this week.

I'm start to thinking that the overall proposed solution is not affordable due to costs and/or technical problems.
I have to think about.

Never give up!!! We are not so far from the goal!

Regards,
Dario.

[Reply #63 on:]

Hi Dario.

You said:

In your previous schematic, using 220uH coils and 1uF capacitor you were in the rage 10-86 KHz (omega0, omega1) so the 50 Hz should be out of these problem. Maybe you are finding this using new values?

Yes. If you have time, please download the latest github simulation version and change the 24V bus supply with a 24V+5sin(50Hz) generator. Look at the voltage before and after the passband filter. It was late when I did it yesterday, so probably I did errors, but I remember something I wouldn't see... :-(

If you have time, please try the numbers that I've proposed.

I did it and I can confirm that the current flowing from the driver is compatible with your results. Unfortunately is not so easy to find a 13mH cheap and small inductor... so I used a 6.8mH inductor in the simulation. The current handled by the driver is reasonably small enough for a commercial unit.

... but...

I don't think that increasing the inductor is the correct way to have something cheap and reliable. If I well understood, we can have the same effect we have increasing the inductor size if we increase the injected signal frequency. If we can rise the modulated signal frequency to something about, for example, a MHz, the inductor reactance will be multiplied by a factor of  100. This let us able to reduce the inductor size.

I know it's impossible to demodulate by software any FSK signal within a MHz range but we can use a MHz carrier to replace the high level of generated FSK. I mean... it's like placing an OOK modulation on top of the FSK.

The receiver side would be a pass band centered on the MHz carrier followed by an integrator and the usual Shmitt thrigger. The recovered FSK signal would be demodulated as before.
The circuit is a little bit complex but I think it could perform better than what's actually doing.

What do you think about?

Thanks!
Marco Signorini.

[Reply #64 on:]

did anyone get domestic home to work over the internet?

[Reply #65 on:]

Hi Marco,

in the while, I've worked a bit on the drivers for the FSK and I have some code with collision avoidance that is working. Is not complete, because still miss the checksum, and for sure require more test. I've used it in a small setup with three nodes and a few centimeter bus, without additional hardware controllers.

I've decided to include this code as preliminary in the release A3.1 of Souliss, that is now available for download.

In the mean time, I've spent some time on the hardware side, has we discussed there are some ideas, but still there is some work to understand how fit your requirements on the size of inductors.

Regards,
Dario.

[Reply #66 on:]

Hi Dario, hi All.
Sorry for the late answer to this thread I started a long time ago... but holidays and other jobs have taken precedence.

I would like to summarize the point where we are and what we understood by discussing on this topic.

We started the idea to develop an Arduino based system targeted to antitheft systems. The first approach was to develop a multi purposes Arduino with some relays, digital and analog inputs, digital outputs, a power supply and an RS485 based communication interface onboard. The main goal was to produce a board able to fit into a standard European 503 box so people could place it anywhere in the house. The RS485 was chosen because it's a well known standard, used on many commercial antitheft systems as main field bus.
We developed two version of that systems. The first version was build around a linear power supply that was replaced by a switching power supply on the second version. I've personally build two units fully working (some pictures are still at the EtherMania Blog) and I've developed a library useful to easy control the onboard SPI expander.
Suggestions came by this thread allow me to change a little bit the original target: the final unit I've on my desk mounts a set of Arduino's compatible headers that allow people to use the board not only for antitheft systems, but as a main block to develop/prototype domotics applications.
And that was the incipit...

but... we wanted more....

We started speaking to replace the RS485 transceiver with a custom communication system that should provide the power required by each board and allow multimaster communications between boards. This should be done through a simple twisted pair, thus reducing the connection complexity. The idea is to have a distributed system where all boards are connected together by a "simple two wire cable" altogether connected in a multiple star topology and, luckily, not terminated.
We don't need super high transfer rates: we need to be able to transfer only actuation commands and other possible parameters like the status of a remote switch, temperature and so on.

Summarizing the full thread contents I can say that:
- we opted for a modulation system based on FSK. Two different symbols for the 0 and 1 bit statuses. The modulation is simple. The demodulation could be done through an hardware comparator embedded on the Atmel micro and measuring the timing between incoming frontends. We can do some hardware assisted time measures so they are enough accurate for the purpose.
- we defined a physical layer based on a set of capacitors, inductors, drivers and comparators/filters. The first were used to decouple either the Tx either the Rx stages from the DC on the bus; the inductors were used to provide high impedance to the FSK signal when feed into the main and distributed power supplies. Push pull drivers were used to "inject" the FSK over the DC line and comparators and active filters were used to filter out and square the incoming Rx signal.

This solution was very interesting and strong enough to produce good results in the simulated environment. We were able to transmit and receive the simulated FSK signal between nodes connected together by simulated cable (with some added resistors and inductors to simulate a real working situation).
Unfortunately this is not something that could be easily replicated in the real world. The proposed circuits work well when large inductors are used. Inductors are big and not cheap. Large inductors are not "friends" of electronic systems because they introduce spikes, high voltages and currents.

So we ended up with a NO valid solution but with a lot of information and experience that I would like to use to start a new thread because I think that here we're now out of topic.

The new topic is here: http://arduino.cc/forum/index.php/topic,121868.0.html

[Reply #67 on:]

start a new thread

Start away, I lost interest when the interface got into the unfamiliar (for me) territory of inductors and transistors.

_____
Rob