Phantom Power or sharing power line with differential pair communication.

Thread Starter

sc0tch

Joined Nov 6, 2018
64
Hey everyone,

So a project i've been working on has been working great in the lab, but once i moved into the real world where cables were longer and such there was alot of interference, so i changed a bunch of stuff to communicate using differential pairs rather than sending a single signal over a cable. Unfortunately now I either need to add cables or find a way to power the board using the same cables as the UART signals. My thought was simply to drive the signal from the master device at higher power (8v and GND) using a mosfet or optocouple off the microprocessor, and at the remote daughter board treat it like AC current and rectify it and store some energy into a cap while also using a voltage divider to bring the signal down to logic levels.
Master Board.png View attachment 165548
Daughter Board.png

*EDIT: I did forget to put resistors on the optocouplers in this schematic, and if the general premise of the schematic will work i will likely change the transmits to use a D-Latch to ensure sync with clock.

My other idea to tackle this was to send 5v0 across a pair while treating 5v0 as logic low and use a charge pump to push the voltage higher to say 12v0 for logic high, but not sure how I would handle the differential pair for this while also having a ground connection.

I have quickly diagrammed the first idea.

Thanks very much guys, y'all have been a great help in my projects so far! Also just got my first FPGA developer board to mess around with(based on a Xilinx spartan 6)
 
Last edited:

AnalogKid

Joined Aug 1, 2013
10,990
It is not clear which of your schematics is the "local" and which is the "remote". Also, the upper one has two outgoing signals but no incoming, while the lower one has two incoming *and* one outgoing. Where does that outgoing signal go?

1. What is the highest frequency or the narrowest pulse width of the signals?

2. What is the cable length?

3. What are the voltage and current requirements of the remote board being powered?

4. Since the two circuits must share a common ground for the phantom power system to work, the optocouplers do not get you any isolation. If you are using them as high-side power switches, consider p-channel MOSFETs for much better efficiency. This is particularly important if you stay with 78xx regulators on the remote board (see below).

What do D2 and D7 do? They are in series with the bridge, so the bridge is not "bridging". Also, the two 3-terminal regulators need bulk filter capacitors on their inputs.

BTW, there is a whole universe of differential signalling parts, generally named by the name of the governing standards: RS-422 and RS-485. These have a very large common-mode range (much better than an LM339) specifically designed to work with long wire runs.

Also, the reason phantom power works in audio systems is that the signal voltage is very small compared to the phantom voltage. That is not the case for you. Your problem can be solved, but it will take more than two resistors and a cap.

78xx regulators have about 10 mA of static current. Newer devices have much a lower value.

Note: positive voltage source symbols should always point upward, and GND symbols *always* point down.

ak
 
Last edited:

Thread Starter

sc0tch

Joined Nov 6, 2018
64
It is not clear which of your schematics is the "local" and which is the "remote". Also, the upper one has two outgoing signals but no incoming, while the lower one has two incoming *and* one outgoing. Where does that outgoing signal go?

1. What is the highest frequency or the narrowest pulse width of the signals?

2. What is the cable length?

3. What are the voltage and current requirements of the remote board being powered?

4. Since the two circuits must share a common ground for the phantom power system to work, the optocouplers do not get you any isolation. If you are using them as high-side power switches, consider p-channel MOSFETs for much better efficiency. This is particularly important if you stay with 78xx regulators on the remote board (see below).

What do D2 and D7 do? They are in series with the bridge, so the bridge is not "bridging". Also, the two 3-terminal regulators need bulk filter capacitors on their inputs.

BTW, there is a whole universe of differential signalling parts, generally named by the name of the governing standards: RS-422 and RS-485. These have a very large common-mode range (much better than an LM339) specifically designed to work with long wire runs.

Also, the reason phantom power works in audio systems is that the signal voltage is very small compared to the phantom voltage. That is not the case for you. Your problem can be solved, but it will take more than two resistors and a cap.

78xx regulators have about 10 mA of static current. Newer devices have much a lower value.

Note: positive voltage source symbols should always point upward, and GND symbols *always* point down.

ak
Thanks for the response. Maximum frequency would be around 10-15khz over a maximum of a 10m cable.

My daughter board's power requirements are 5v- at .3 mA(just hall effect sensors feeding the mCU) and 3.3v at 1.4 mA to power a pic that i plan to use to controll the UART interface.

Yes the two boards must share a common ground so I will change the optocouples to mosfets as there won't be complete isolation either way.

My reason for the 339 was honestly that it was the only differential comparator in the default library for tiny cad, however i was originally intending to uses the 393(dual version rather than quad of the differential comparator) mostly because it's a chip i have a tape and reel of(and some experience in using), but more than willing to consider a better fitting component.

The diodes you referenced were a complete mistake of a last second whim of "should i throw something here to prevent contaminating my signal with noise from the voltage regulators" but now that i look at it they would never work whatsoever.

And sorry, forgot to add the capacitor after the bridge rectifier... had one there until i realize i had the polarity reversed in the schematic...

Also to note the signal is being feed over CAT-5E cable which was chosen for it's extreme cheapness($400/10,000ft) along with the low cost of connectors, but I am considering changing cable interfaces as this project becomes more and more complex.
 

dendad

Joined Feb 20, 2016
4,452
Is the Hall sensor analog?
If so, have it drive a 4-20Ma current loop. That is a standard in industrial control and can operate over quite long distances.
Do the processing at the controller end. Make your sensor dumb.
15Khz may be attainable. Give it a try and see.
But if it is not analog, as you are generating a pulse train, still go for a current loop and switch between the 4mA at rest and 20mA signal.
 
Last edited:

Thread Starter

sc0tch

Joined Nov 6, 2018
64
Is the Hall sensor analog?
If so, have it drive a 4-20Ma current loop. That is a standard in industrial control and can operate over quite long distances.
Do the processing at the controller end. Make your sensor dumb.
15Khz may be attainable. Give it a try and see.
But if it is not analog, as you are generating a pulse train, still go for a current loop and switch between the 4mA at rest and 20mA signal.
hey dendad. Each sensor in my application peaks around 60-100hz square wave, as the hall efect sensors are used with a magnetic impeller wheel to measure fluid flow. The issue is that we are trying to get as many sensors to a daughter board as possible...current proof of concept uses 4 on a dumb board but the cross-talk is triggering interrupts on other lines along the cable. The interferance is mostly if I try to use a twisted pair for 2 signals, but still some issues when using a twisted pair to carry a signal + GND or signal + 5v0.

The current board simply runs the 5V through the hall effect sensor and back to the main board where a LM393 compares it to a 2.5v ref to determine state.
 

AnalogKid

Joined Aug 1, 2013
10,990
For 15 kHz over 30 feet, I would thing just about anything would work for the data transfer. I've run single-ended TTL that far inside a football stadium. If you think differential signalling is required, then ok - let's go down that path.

Look into LDO (Low DropOut) regulators instead of 78xx - way better total efficiency. Also, consider a simple zener diode for such low supply currents.

ak
 

Thread Starter

sc0tch

Joined Nov 6, 2018
64
For 15 kHz over 30 feet, I would thing just about anything would work for the data transfer. I've run single-ended TTL that far inside a football stadium. If you think differential signalling is required, then ok - let's go down that path.

Look into LDO (Low DropOut) regulators instead of 78xx - way better total efficiency. Also, consider a simple zener diode for such low supply currents.

ak
Thanks for the insight, I'll try and troubleshoot some more with what I've been doing... Have a new oscilioscope coming so maybe i can get better data of why I'm getting crosstalk along the cable. Do you think interferance should be an issue at these low frequencies if using a twisted pair to carry two seperate signals?
 

ebp

Joined Feb 8, 2018
2,332
If you use LDO regulators and efficiency is important, chose modern types that use a FET as the pass element. Older designs use PNP transistors, so the base current for the pass transistor flows to ground instead of into the load as it would with an NPN transistor in a non-LDO design. When the input voltage for some of the PNP types is below that required for regulation, the quiescent current could rise to surprising high value - again, most of going to waste instead of to the load.

Watch out for capacitor requirements with newer LDOs. Some of the sort of "intermediate era" were very fussy about the ESR of the cap on the output - either too high or too low would cause problems. If you used a ceramic output cap, for instance, you might need to use a resistor of a few ohms in series. Most of the most recent generation of regulators will tolerate low ESR without a problem.

If you contemplate a shunt regulator (zener type) and conclude that a zener isn't really a great device at low voltage and low current, there are some 2-terminal shunt voltage references that are quite inexpensive and that might be useful. The low power version of the TL431 (TLC431 ??) could be used, but needs external parts. The standard TL431 might be OK, too. The standard TL431 is another device that is fussy about shunt capacitance - it is unstable with capacitance in a certain range. This is well detailed in the datasheet. I don't remember if the low power version has the same issue.

At the expense of needing to square it up at the receiving end (maybe), crosstalk can sometimes be satisfactorily mitigated by reducing the slew rate of the driving signal with an RC network or even just a series resistor.

Lots of phantom powered sensors use a current source at the the receiving end, Current limiting diodes (basically JFETs designed/chosen for particular IDSS) are convenient, through rather expensive. Current sources with BJTs or op amp circuits are quite simple, especially when high precision or very high output impedance isn't critical. Essentially it is similar to a 4-20 mA current loop, but for digital signals it can be quite crude and still perform very well. You can do things like simply using inductors through which to feed and recover power and use differentiated (not "differential") digital signals, but this requires a bit more elaborate circuitry to recover the signal (e.g. flip flop set on rising pulse, cleared on falling pulse). Of course if all you care about is the period between edges, it gets really simple.

[EDIT]: I should have mentioned - Because a current source, by definition and in reality, has very high output impedance, it isn't great when it comes to crosstalk susceptibility.

Also, since you are using a comparator as a line receiver, lots of hysteresis is your friend when it comes to rejecting crosstalk.
 
Last edited:

dendad

Joined Feb 20, 2016
4,452
You can try using RS485 drivers for differential signals to help cross talk. Also, low impedance current loops for coms can help too
 

Thread Starter

sc0tch

Joined Nov 6, 2018
64
If you use LDO regulators and efficiency is important, chose modern types that use a FET as the pass element. Older designs use PNP transistors, so the base current for the pass transistor flows to ground instead of into the load as it would with an NPN transistor in a non-LDO design. When the input voltage for some of the PNP types is below that required for regulation, the quiescent current could rise to surprising high value - again, most of going to waste instead of to the load.

Watch out for capacitor requirements with newer LDOs. Some of the sort of "intermediate era" were very fussy about the ESR of the cap on the output - either too high or too low would cause problems. If you used a ceramic output cap, for instance, you might need to use a resistor of a few ohms in series. Most of the most recent generation of regulators will tolerate low ESR without a problem.

If you contemplate a shunt regulator (zener type) and conclude that a zener isn't really a great device at low voltage and low current, there are some 2-terminal shunt voltage references that are quite inexpensive and that might be useful. The low power version of the TL431 (TLC431 ??) could be used, but needs external parts. The standard TL431 might be OK, too. The standard TL431 is another device that is fussy about shunt capacitance - it is unstable with capacitance in a certain range. This is well detailed in the datasheet. I don't remember if the low power version has the same issue.

At the expense of needing to square it up at the receiving end (maybe), crosstalk can sometimes be satisfactorily mitigated by reducing the slew rate of the driving signal with an RC network or even just a series resistor.

Lots of phantom powered sensors use a current source at the the receiving end, Current limiting diodes (basically JFETs designed/chosen for particular IDSS) are convenient, through rather expensive. Current sources with BJTs or op amp circuits are quite simple, especially when high precision or very high output impedance isn't critical. Essentially it is similar to a 4-20 mA current loop, but for digital signals it can be quite crude and still perform very well. You can do things like simply using inductors through which to feed and recover power and use differentiated (not "differential") digital signals, but this requires a bit more elaborate circuitry to recover the signal (e.g. flip flop set on rising pulse, cleared on falling pulse). Of course if all you care about is the period between edges, it gets really simple.

[EDIT]: I should have mentioned - Because a current source, by definition and in reality, has very high output impedance, it isn't great when it comes to crosstalk susceptibility.

Also, since you are using a comparator as a line receiver, lots of hysteresis is your friend when it comes to rejecting crosstalk.
Thanks for all the information EBP. Right now I am using 1MOhm resistor for hysteresis on a single line TTL communication. I guess right now I should hold off till i get a suitable O-Scope to find out exactly whats happening, and exactly where the interference is coming from.
 

Thread Starter

sc0tch

Joined Nov 6, 2018
64
Thanks for all the information! Finally got a new oscope... Just a usb one for the computer, but far better than the one I got on eBay last year... Found that most of my problem came from noise across the 5V0 rail whenever on sensor went from low-high. I went ahead and had added a ceramic cap onto power powering both boards and on each sensor to clear it up as best as possible.

I think I'm still going to make the circuit I discussed in this, mostly as a learning experience and to be ready to add more sensors to the project adding more cables later...

Obviously I don't have much experience in designing these digital communication circuits and jumped straight to an unreasonable conclusion of "crosstalk must be the issue". So for future reference what frequency would really be the line where single ended TTL communication shouldn't be used?
 

AnalogKid

Joined Aug 1, 2013
10,990
There are no power supply decoupling capacitors shown for any of the ICs on your schematic. These are critical for proper operation.

ak
 
Top