Power Over Fiber for Small Sensors

[sodoyle]

TL;DR Summary

I am interested in using power over fiber (PoF) to power small sensors.

I have a "high votage" converter with components at different voltages distributed throughout. I have some sensors that need to be powered in one location that's some distance from its power supply. I want to minimize (or eliminate) the high voltage wires routed throughout the converter.

My initial approach was to "tap" off of one of the capacitors balancing resistors to power the sensors from here instead. I've been having problems with this because when I connect the DC/DC converter across the resistor, the voltage drops and I'm not able to regulate 5 V out. I tried using a buck-boost to get around this, but it still hasn't been successful as of yet. I'm still working to solve this with a DC/DC converter for now.

One thing that came to mine is power over fiber (PoF). I've read about this some in the past but I'm still pretty unfamiliar with it. It seems like an ideal candidate for my application in theory. I know that there will be a whole range of new design challenges introduced with this method, but I just wanted to poll the audience and see if anyone has experience with PoF.

I will continue to do my own research on this. Any resources, advice, or even just general discussion is greatly appreciated.

 

[Baluncore]

Power over fibre is very limited.
If you have a voltage balancing string of resistors, you have much more power available than is possible using PoF. I cannot help thinking that you are looking for problems in a field in which you have little experience. How many threads will it take until you actually specify the actual problem?

Why do you really need a floating power supply?
What instrumentation does the floating supply need to support?

 

[sodoyle]

My apologies for the multiple threads if it would be better for me to have only one. I felt like the topic was far enough off that it would be best to discuss under a new one so it wasn't jumping all over the place.

I need a floating supply because I want to minimize the high voltage wires routed throughout my converter. I'm trying to eliminate the need for that, even if it takes a more creative approach as I'm trying to do here.

Initially I thought about tapping the balancing resistors. That's created other challenges as I've mentioned below. I was thinking of ways to get around tapping the balancing resistors all together which is why I was questioning PoF. The sensors are only roughly 100 mW using 20 mA @ 5 V. This will actually be to power a voltage sensor for the capacitors. Although I can get more power using the capacitors, it doesn't seem like it's necessary for this application.

You're right about me having little experience in this. I've never tried to tap balancing resistors to supply another converter. It seems to be more challenging than I initially thought so I was trying to think of other options. Other suggestions so far are good for giving me ideas for future designs, but are not suitable here. For instance, I do not have the ability to use 8 series capacitors at lower voltage instead of just these two at the higher voltage. One suggestion was to create my own buck converter using a 1.7 kV SiC MOSFET. That is an option that I'll investigate, but I'm still looking at other options. I've also looked at a charge pump and buck-boost. The problem I seem to be having is the voltage drops to only a few volts across this resistor instead of the 20-30ish V without the converter across it. I'm guessing that can be explained as the impedance of the converter is much lower than that specific balancing resistor. Placing the converter in parallel with it drops the equivalent impedance to roughly that of the converter so the voltage drops.

 

[DaveE]

The problem I seem to be having is the voltage drops to only a few volts across this resistor instead of the 20-30ish V without the converter across it. I'm guessing that can be explained as the impedance of the converter is much lower than that specific balancing resistor. Placing the converter in parallel with it drops the equivalent impedance to roughly that of the converter so the voltage drops.

Is this a surprise? You can analyze/model this without worrying first about what kind of PS you will put there. Really, some up-front systems engineering is in order here. I fear you will never get to a good design by trial and error. I get the impression that you are in over your head, reaching for technological solutions without analyzing the problem.

We'll keep answering specific questions, as best we can, but I'm not sure that will result in a good design.

 

[DaveE]

Plus, one more question regarding this. I need to know before I respond anymore.

Please describe the human safety issues involved with this. What is the isolation between people and your 2KV source. Are they fail safe? Are safety agency approvals part of your design requirements? Should they be?

 

[sodoyle]

Is this a surprise? You can analyze/model this without worrying first about what kind of PS you will put there. Really, some up-front systems engineering is in order here. I fear you will never get to a good design by trial and error. I get the impression that you are in over your head, reaching for technological solutions without analyzing the problem.

We'll keep answering specific questions, as best we can, but I'm not sure that will result in a good design.

Looking at it now, it's not really a surprise. It's just something that I now know to consider. I was hoping to keep this thread more so about PoF so I could get some useful insights there while I continued investigating powering from the balancing resistors. I assumed my questions were getting repetitive there so I wanted to investigate further on my own and come back with questions that were more thought out if necessary. The more I've look at the balancing resistors approach, the less feasible that solution looks for a few reasons. 1) The equivalent resistance of the resistors would have to be pretty low for the current needed. At 2 kV, the power would be pretty high which would lead to thermal issues. 2) it would be hard to maintain voltage balancing between the capacitors because of the large current draw from them. Do these seem reasonable?

Plus, one more question regarding this. I need to know before I respond anymore.

Please describe the human safety issues involved with this. What is the isolation between people and your 2KV source. Are they fail safe? Are safety agency approvals part of your design requirements? Should they be?

The source and DUT are in a room that is separate from all people. Interlocks are installed so nobody can be present during testing and everything is operated remotely. Safety agency approvals are not part of the design requirements; however, safety will still be kept in mind. Still, lower voltage tests are always performed first and voltages are increased gradually.

 

[Baluncore]

You have a voltage balancing chain with more than sufficient current to drive an LED or converter. There are many systems that could generate a voltage dependent frequency and communicate that through an opto-isolator or a fibre. The choice will be determined by the answers to these questions.
1. Is monitoring the individual capacitor voltages all you need to do?
2. Do you want Lo threshold and/or Hi threshold warnings, or do you want a 1% accurate value?
3. What is the min/max voltage range across a single capacitor?

 

[Keith_McClary]

Have you considered batteries?

 

[sodoyle]

You have a voltage balancing chain with more than sufficient current to drive an LED or converter. There are many systems that could generate a voltage dependent frequency and communicate that through an opto-isolator or a fibre. The choice will be determined by the answers to these questions.
1. Is monitoring the individual capacitor voltages all you need to do?
2. Do you want Lo threshold and/or Hi threshold warnings, or do you want a 1% accurate value?
3. What is the min/max voltage range across a single capacitor?

My initial approach was to use fiber for communications from the voltage sensor to a central controller. I was looking at using the buck converter so I could reduce the current through the balancing resistors to reduce their losses. The idea of PoF was to get around draining the capacitors using the balancing resistors.

To your questions:
1) All I intend to do for now is voltage measurements. The voltage sensor is already designed so all I need to do is power the ICs. That is why I need 5 V, 20 mA.

2) I do need fairly accurate values.

3) The capacitors are rated for 1.5 kV each (3 kV equivalent) so I would like to measure at least 110% of the rated voltage. That puts the maximum voltage at 3.3 kV. The minimum operational voltage is 1.5 kV (750 V each) not considering startup or simply operating at lower voltages for testing.

 

[sodoyle]

Have you considered batteries?

I didn't consider batteries because a battery would still run into the same problem wouldn't it? I would have to constantly charge the battery since it would have a constant power draw so I would need a charge controller wouldn't I?

 

[Keith_McClary]

I would have to constantly charge the battery

I ran across this:
https://www.mouser.ca/storing_harvested_energy/

 

[berkeman]

I didn't consider batteries because a battery would still run into the same problem wouldn't it? I would have to constantly charge the battery since it would have a constant power draw so I would need a charge controller wouldn't I?

I had the same thought about using batteries at the sensor locations. Yes, you would need to charge them occasionally, but that depends on how often the system is run, how big of a battery you use at each node, and how efficient you make your sensor circuit.

Why does your sensor require 20mA at 5V? There are circuit design techniques to lower average power consumption of isolated battery-powered devices (think mesh RF networks). Can you say more about what these sensors are doing, and how often they need to send their data to the central data gathering node? How many hours per day does this system need to run? How many battery-powered nodes would there be?

 

[sodoyle]

I ran across this:
https://www.mouser.ca/storing_harvested_energy/

I can see where these may be useful. I need reliable power or these would be perfect-ish. PoF uses a concept similar to the PV energy harvesting mentioned here except you're sending light with a higher intensity to get higher power.

I've heard about solid state batteries (SSB) as mentioned in this article but I haven't actually seen them commercially available. I know they're working on them for EVs but I haven't even heard of anything for lower voltage applications like this. It seems like this article must've been written before 2013 since they say the cost of to change out batteries will approach $1 billion in 2013. I looked up the company (Cymbet Coorperation EnerChip SSB) and they appear to still be around. There's no technical information listed on their page though. Out of curiosity I'll dig a little deeper into SSB.

 

[jrmichler]

Have you considered using 4 to 20 ma two wire transmitters? One pair of wires to each sensor, good resistance to electrical noise, easy to use, and widely used in industrial applications. If you are not familiar with them, search 4 to 20 ma transmitter for info.

Do you need continuous readings for a short time, continuous readings for a long time, intermittent readings? For example, four AA batteries can power an Arduino with a Wheatstone bridge sensor for about 5 months if you only need one reading every 30 minutes.

 

[sodoyle]

Why does your sensor require 20mA at 5V? There are circuit design techniques to lower average power consumption of isolated battery-powered devices (think mesh RF networks). Can you say more about what these sensors are doing, and how often they need to send their data to the central data gathering node? How many hours per day does this system need to run? How many battery-powered nodes would there be?

I connected it to a benchtop supply and that's what it shows. It may be slightly lower if I took a more accurate measurement that the benchtop supply and used shorter wires, but it's in that range. It's already outfitted with all the necessary signal conditioning circuits and everything needed for communication by fiber.

 

[sodoyle]

Have you considered using 4 to 20 ma two wire transmitters? One pair of wires to each sensor, good resistance to electrical noise, easy to use, and widely used in industrial applications. If you are not familiar with them, search 4 to 20 ma transmitter for info.

Do you need continuous readings for a short time, continuous readings for a long time, intermittent readings? For example, four AA batteries can power an Arduino with a Wheatstone bridge sensor for about 5 months if you only need one reading every 30 minutes.

I'm not familiar with those but there's a lot online so I'll look into them.

I'm using this to measure the bus voltage so I need a continuous measurement all of the time. I'm sure that I would be able to swap out batteries in the lab environment, I'm aiming to design this so it can be operated without maintenance for long timer periods.

 

[DaveE]

You say it needs to be accurate, but that doesn't mean anything to me without numbers; 10%, 0.1%? Stable/repeatable with individual calibration, or truly accurate? Temperature and voltage range, etc.?

OK, here's a concept to consider. You use a HV resistor to charge a capacitor (I'd probably use 2 or 3 in case one fails, the safety guys want that). The capacitor is discharged when it reaches a high threshold down to a lower threshold (or for a fixed time). The discharge current is passes through a sensitive opto-isolator. On the LV side, you measure the frequency of pulses out of the photo-transistor. This could be a low power circuit, depending on the effort you put into the design. Everything on the HV side is operated at reasonable voltages from the current through the HV resistor(s). Look at LMC555 and 4N33 for starters, although I'm not sure the 4N33 is sensitive enough. You may want to use a pulse transformer instead. Of course there are A LOT of details left for you to figure out.

It is easier to deal with digital signals across an isolation barrier than accurate analog signals.

 

[sodoyle]

You say it needs to be accurate, but that doesn't mean anything to me without numbers; 10%, 0.1%? Stable/repeatable with individual calibration, or truly accurate? Temperature and voltage range, etc.?

OK, here's a concept to consider. You use a HV resistor to charge a capacitor (I'd probably use 2 or 3 in case one fails, the safety guys want that). The capacitor is discharged when it reaches a high threshold down to a lower threshold (or for a fixed time). The discharge current is passes through a sensitive opto-isolator. On the LV side, you measure the frequency of pulses out of the photo-transistor. This could be a low power circuit, depending on the effort you put into the design. Everything on the HV side is operated at reasonable voltages from the current through the HV resistor(s). Look at LMC555 and 4N33 for starters, although I'm not sure the 4N33 is sensitive enough. You may want to use a pulse transformer instead. Of course there are A LOT of details left for you to figure out.

It is easier to deal with digital signals across an isolation barrier than accurate analog signals.

I don't have an exact accuracy as of now (<5% difference). That's an interesting idea to measure the voltage but I already have a voltage sensor. I may be able to modify it by removing some of the unused signal conditioning circuits but I'm mainly trying to power what I already have. It would be nice to do another revision of this and have a better (I know, I know...what is better) voltage sensor so I'll be doing a literature review to see what all's out there and how I can build on it.

I do plan on using digital signals. The voltage sensor has an onboard ADC which communicates to a central controller through fiber.

 

[rbelli1]

5%

This would be easy using most common micro-controllers. Do you need to use "Your" voltage sensor?

It would not be too difficult to set up a programmable Bluetooth micro to operate off of two AA batteries and get several tens of thousand (hundreds of thousands maybe?) readings per set of batteries. Nordic and Cypress are two manufacturers I know off of the top of my head have chips that will do the job.

BoB

 

[Baluncore]

There is no need for batteries. The divider chain provides more than enough current to run a regulator and current to frequency converter. By charging a capacitor with the voltage balancing current, it can be discharged through an LED to drive an optic fiber or an optoisolator. That way there is very little supply current required.
Here is an example circuit that produces close to 1 Hz per HV capacitor volt, ≈ 2 kHz.

A bipolar NE555 draws too much current, so it uses the CMOS TLC555 or LMC555.
The Q1 Vbe in series with two 6.2V zener diodes is a 13V shunt regulator = Vcc. The 6k8 resistor provides a steady 100 uA current to bias the zener diodes.
Most voltage balancing current flows through the collector of Q1, to charge C1. When C1 voltage reaches 8.67 volts the 555 discharges C1 to 4.33 volt through the LED.
That 20 mA discharge pulse through the LED does not come from the power supply. Current through the LED is limited by the R4 = 220R.
The V to F conversion is offset by the regulator bias and the 555 supply current.

 

[minerva]

"Power-over-fibre" systems which provide small power supplies with very strong galvanic isolation in floating HV systems are discussed a little bit in Horowitz and Hill's The Art of Electronics 'X-Chapters', chapter 9x.23.

Broadcom make photovoltaic fibre-coupled devices designed for exactly this application.
https://www.broadcom.com/products/f...ponents/optical-power-converters/afbr-poc406l

You could make your own DIY system with basic solar cells and high intensity LED sources.