Academic Analog Circuit Puzzle

Beau Schwabe

I was looking into MLT-3 signaling used in Ethernet communication, and stumbled across something that sparked an idea that I have had for simultaneous bi-directional communication for a long time. I fiddled around here and there and came up with an analog circuit using a few discrete components performing the correct decoding function needed to do this.

Anyway, the idea is, that data is transmitted on the "A" or "B" I/O and feedback on the other I/O determines if the 'opposite' side is in agreement or not. If they agree the Feedback signal is "high" otherwise the signal is "low".

I have a solution using 9 discrete components that I will post later, but I wanted to see what others would come up with.

LoopyByteloose

I am not sure I understand. Does the schematic you present work or not work? And since there are two inputs to the gates for the feedback, I get the feeling that something is missing in the way of wiring unless the EX-OR gate inputs are tied together and merely serving as a logic inverter.

Heater.

I don't get that diagram either, If A is high and B is low then the level on the line between the R's is halfway up or down and indeterminate.

I suspect one input of the XORs should be the actual value of the transmitted data (A or and the other XOR input should be what is seen on the line.

Beau Schwabe

Yes, the schematic works.

The X-NOR has one Tri-state input that works on three different voltage levels (Low, V/2, and High)
If you look at the output of the X-NOR with respect to "A" and "B" through the resistor divider the output function is an X-NOR.

The "Feedback" signal is just so the system sending knows the state of the recipient. Assuming the system knows the signal that it is sending (High or Low), by looking at the Feedback, it can determine the state of the recipient.

Heater.

Ah, a subtle interpretation of "tri-state", should have looked at your truth table a bit harder.

So the challenge is to build that tri-level gate thing then.

Beau Schwabe

Heater,

"So the challenge is to build that tri-level gate thing then" - Yes, a window comparator essentially that is suppressed at Vdd/2, the resistors give it the X-NOR characteristics.

BTW) No Op-Amps allowed here.

kwinn

If high speed is not required and current loop (10-20mA) is used an AC input optoisolator could be used at each end. For higher speeds 2 high speed optoisolators could be used at each end.

Heater.

Yep I was thinking discrete. Two LEDs with two series resistors across the input in opposite polarity. Select suitable LEDS such that full voltage lights up one or other, half voltage lights up neither. Have a photo transistor or such watching the LEDs and pulling on a resistor to provide the output.

6 components so far. No matter if the output polarity is up side down as we can invert it in software:)

wjsteele

Beau Schwabe (Parallax) wrote: »

BTW) No Op-Amps allowed here.

Well... I have to start over, now! :-)

Bill

Heater.

Thinks...

Only one LED can be on at a time, so we only need one current limiting resistor feedingh, down to to five components:)

Phil Pilgrim (PhiPi)

I think this will work:



If A<B, the top NPN will conduct, pulling /B low. If A>B, the PNP will conduct, pulling /B high. If A=B, neither conducts, so the bottom NPN biases /B to /A.

-Phil

robot mogul

I do not comprehend the general idea of the analog circuit puzzle because of the visual planning. I do grasp the concept of the certain transponder effect, also known as the closed circuit transciever, reciever prospect. I also think that PhiPi's plan is incorrectly drawn and explained. In my opinion there should be two contact points, one at the very beginning, creating a three point meeting closed circuit and two at the end of the analog puzzle, creating first the same as the top, a three point meeting closed circuit main, and another three point meeting closed circuit end.

Beau Schwabe

Here are some other references that uses a similar visual topology.

http://mountains.ece.umn.edu/~sobelman/papers/mthsieh_iscas04.pdf

http://cva.stanford.edu/books/dig_sys_engr/lectures/l9.pdf

http://www.ee.ucla.edu/~ingrid/Publications/2002epep.pdf


Phil,

I'll have to simulate your design, but I think the B-C junction of the NPN and the PNP will form a back to back parallel diode regardless if the transistors are on or not.

Phil Pilgrim (PhiPi)

Beau,

I think it could also be done with MOSFETs if the signaling voltage was high enough.

-Phil

Heater.

OK how about 4 components:

Two LEDs in parallel, opposite polarity.
Put this arrangement in series with R.
Use a photo transistor with resistor to positive supply to detect light from the LEDs
Junction between photo transistor and resistor is the feedback signal. A LED ON gives you a low on the feed back pin.

If you are transmitting a 1, one LED or other will be ON if the other end is transmitting a 0. Else it's OFF.

If you are transmitting a 0 one LED or other will be ON if the other end is transmitting a 1. Else it's off.

We have enough information here to decode what the other end is sending whatever state our end is in.

Put the same circuit at both ends.

Always assuming we can light two LEDs in series with the given drive voltage.

Phil Pilgrim (PhiPi)

Heater,

I think that will work. I know the objective was to do it with discrete components, but there are optoisolators that have front-to-back parallel LEDs built in: three components (2 resistors, 1 optoisolator).

-Phil

Heater.

Yep, what speed do you think we can get out of it?

Does it have any possible advantage over using two wires?

Edit: Isn't that 2 components? The series resistor for the LEDs is already given as R.

Phil Pilgrim (PhiPi)

Maybe so. I was counting both the series resistor and the pullup for the opto.

-Phil

Beau Schwabe

Heater,

"Does it have any possible advantage over using two wires?" - One possibility, is that if you convert an existing infrastructure you suddenly have twice the throughput.

kwinn

This was the best I could come up with other than the AC optoisolator.

Beau Schwabe

kwinn,

That beats mine by one component ...

Optimized for 3.3V operation:
3-NPN's
4-Resistors
2-Diodes

Beau Schwabe

kwinn,

Actually you can reduce your version down to 6 components if you eliminate the NPN transistor and convert the PNP's to NPN's



From looking at the solutions, there seems to be two general ways way to do this:

1) Is to detect the polarity on the line

2) Is to detect Vdd/2 (a window comparator)

I'm not sure which is better over the other or if it should really matter, or is one is faster over the other.


Here is the version I came up with using the #2 comparator style.

The Window is targeted from 1.4V to 1.8V when supplied with 3.3V

kwinn

Beau, mine has 7 components since it uses the existing resistor between the base/emitter of the pnp's. I would be very interested to see the 6 component design you mention. That would make the use of a dual transistor in 1 package possible.

kwinn

Heater. wrote: »

Yep, what speed do you think we can get out of it?

Does it have any possible advantage over using two wires?

Edit: Isn't that 2 components? The series resistor for the LEDs is already given as R.

All the AC optoisolators I am aware of are fairly slow so max speed would be around 9600 baud, maybe as high as 19.2K.
As to advantage over 2 wires, none, probably a speed disadvantage in most cases. The advantage comes if you are doing a retrofit or upgrade, need full duplex, and only have 1 free wire in an existing cable. A situation I have encountered several times in the past.

Phil Pilgrim (PhiPi)

Here's a design that uses a single transistor:



The resistor divider between Vdd and Vss sets a bias point for the feedback slightly above Vdd/2. When A is high, the MOSFET pulls the bias point slighlty below Vdd/2. That way, the voltage at the feedback point will be > Vdd/2 when B is high and < Vdd/2 when B is low, regardless of the state of A. (The divider resistors have to be large compared to R.)

-Phil

Beau Schwabe

kwinn,

The schematic below is a derivative of what I was talking about with reducing your design down to 6 components.

Here is a 4.5 component design using the voltage detection option #1 I mentioned earlier (.5 - the resistor could be shared as one)