Title: "High Voltage Power Monitoring with AVR"

I'm a volunteer for a local streetcar (tram for our international viewers) museum. I would like to implement some power monitoring and logging with an AVR connected to an Ethernet chip to assist in quality control and performance monitoring. Our streetcars run on 600VDC and with two on the line we can pull up to 225A or so. The power supply system is an interesting beast itself, designed by a former member, which may merit future discussion as I'd be interested in tracking loss from the incoming 3 phase 208VAC though the a step up transformer to 3 phase 480VAC and then finally the DC supply.

For now though the goal is simply to measure and log the voltage and current on the 600VDC side. We already have some inconveniently located analog meters for voltage (60VDC meter) and current (300A/50mV external shunt meter) so I do have access to the 50mV shunt. The trick is that the entire supply is turned off and on so it cannot be used to power the monitoring system at all, that will have to come from the AC building power which is a separate system.

I'm somewhat concerned about the possibility of ground loops, etc. with the AVR being powered from a wall wart on the building supply and metering the DC supply system. So I'm looking for thoughts on how best to do this, if it really is a potential problem (haha) how best to isolate. I'm thinking we need something a little more complex than a simple voltage divider to measure the voltage and perhaps more than a in-amp on the shunt to measure current. One possibility I have thought of is to use voltage -> frequency converter ICs and then isolate those with opto-couplers. Other thoughts or suggestions?

 

The V-to-F with optocoupler sounds like a promising approach.

I think that if I were doing it and I was worried about ground loops, I would use an isolated wall wart (which think you already intend to use) and direct couple the measurement subsystem to the voltage and current being sensed. You can use an isolation coupling on the ethernet connection to prevent ground loops the the ethernet shield. Ethernet is not in my forté, but I think that other than the shield, ethernet is inherently isolated, at least for 10 and 100 MHz versions.

 

hi Ben,
This analog HCNR opto-coupler could be one option or the HCP OPA.

With the HCNR note the 4-20mA loop output option. Fig #15

E

 

This doesn't seem like a very difficult problem at all, you just have to pick the correct point to make the measurements.

Any decent wall wort is isolated from the AC line so you do not have isolation issues to deal with when powering the AVR.

That point for ease of connections is where you can put the shunt in the negative lead. Believe it or not a 500A shunt (which I would use), brand new, is on EBay for under $10. These things only put out .075 volts for 500A so for a micro running at 5V you need an amp with a gain of 5V/.075V to get good resolution.

Now where you connect that shunt is the picky place. I'm assuming you put it in your power plant on the negative power lead. (Note you may have code issues with this lead needing to be bonded to ground: if so let me know there is another way around for that.) You call the wire between the shunt and the power source ZERO. That is the zero voltage point for the AVR power from the wort.
It's a simple matter to make up a resistive divider from the hot line to ZERO to measure the system voltage.

With just two A2D inputs from the AVR and your system is under full surveillance.

There is always a drop along the system, you may consider a remote unit to measure the furthest point: that should give you the actual voltage at the car furthest from the power plant.

BTW, I call these things trolleys or more generally traction. I hope to get to Branford CT soon to ride some.

 

This doesn't seem like a very difficult problem at all, you just have to pick the correct point to make the measurements.

The good news is that most of that is already in place given that I am supplementing/replacing existing analog meters.

Any decent wall wort is isolated from the AC line so you do not have isolation issues to deal with when powering the AVR.

I thought that would be the case, so as long as I tie the AVR ground to the 600VDC ground we should be in good shape. The question I have though is once I start connecting the AVR to other stuff through an Ethernet jack will that introduce the possibility of a ground loop issue?

That point for ease of connections is where you can put the shunt in the negative lead. Believe it or not a 500A shunt (which I would use), brand new, is on EBay for under $10. These things only put out .075 volts for 500A so for a micro running at 5V you need an amp with a gain of 5V/.075V to get good resolution.

Yeah, the good news is there's already a 300A/50mV shunt in place for the analog meter. It would be very surprising indeed if we drew more than 300A, typical draws for a single car max out at about 125A and that's only for starting current, after we get going it's more like 50A. We are prohibited by organization rule (partially to protect the power supply but partially because of our industrial "demand metering" from the local utility which would spike our rate for the entire month if we overdrew) from operating more than two cars on the line at a time thus effectively limiting worst case current draw to about 250A if both cars started simoutaneously. This logging and monitoring system is, in part, because I'm curious just what our peak current draw is and how frequently that occurs.

Now where you connect that shunt is the picky place. I'm assuming you put it in your power plant on the negative power lead. (Note you may have code issues with this lead needing to be bonded to ground: if so let me know there is another way around for that.) You call the wire between the shunt and the power source ZERO. That is the zero voltage point for the AVR power from the wort.

That is where the 300A/50mV shunt is now, it's between the negative terminal on the 600VDC supply and the bonding wire which runs to the rail (which is obviously embedded in a concrete slab within our shops and also bonded to the rail the entire length of the line. One question though, why would you call between the shunt and the supply 0 and not between the shunt and the physical ground 0? Are you talking about for the purpose of voltage monitoring so that you don't count the current lost in the voltage monitoring process as current drawn?

It's a simple matter to make up a resistive divider from the hot line to ZERO to measure the system voltage.

No question that is easiest! As long as we don't have a potential problem (haha) with ground loops through the Ethernet this would certainly simplify things.

With just two A2D inputs from the AVR and your system is under full surveillance.

Well, full surveillance will be when I add a few CTs to the AC side of the supply so that I can monitor the 208VAC 3-phase incoming power and the post step-up transformer 480VAC 3-phase power which feeds the 600VDC supply. Then we'll be able to see where we're loosing the most power...

There is always a drop along the system, you may consider a remote unit to measure the furthest point: that should give you the actual voltage at the car furthest from the power plant.

Oh, distance drop would certainly be an interesting thing to measure. In real traction systems power is usually sent as AC to periodic substations where it is converted to DC (back once upon a time using rotary converters, we still have standing substation buildings in the city here from the original 100+ year old system). We run about a mile of track, the trick is that there is no power or communications at the far end other than our 600VDC line so any remote monitor would have to be fed from 600VDC and also power some kind of RF return link. It would perhaps be easier to build a monitor unit which traveled up and down the line on the car itself, that's actually a project somewhere on my todo list anyway as there are all other sorts of interesting things to measure on the car including air pressure, door positions, etc.

BTW, I call these things trolleys or more generally traction. I hope to get to Branford CT soon to ride some.

A lot of people do call them trolleys. Technically the trolley is the collector wheel at the top of the car which runs along the overhead wire to pick up power and the vehicle as a whole is a streetcar. Much more generally you can call streetcars, light rail vehicles, etc. traction.

 

I think that if I were doing it and I was worried about ground loops, I would use an isolated wall wart (which think you already intend to use) and direct couple the measurement subsystem to the voltage and current being sensed. You can use an isolation coupling on the ethernet connection to prevent ground loops the the ethernet shield. Ethernet is not in my forté, but I think that other than the shield, ethernet is inherently isolated, at least for 10 and 100 MHz versions.

This is actually something I have considered. It means having two wall warts one for the AVR circuit and one for the Ethernet circuit and putting opto-isolators on the serial data lines between them which is all a bit annoying, but is feasible. I am a network engineer by profession and I have anecdotally heard about ground loops on Ethernet anecdotally but I have also never seen them with my own eyes so I don't know if it's really a concern or not.

There are also such things as isolaters you can put inline on an Ethernet wire itself (see here) which appear to be unpowered and also appear to be grossly overpriced, I would love to see what magic is inside one of those.

 

I encourage you to research the electrical (PHY) specifications for ethernet. As I recall, it uses transformers to couple the data onto and off of the buss. You can get by with grounding the shield at one one end or the other so that ground loops are eliminated.

 

That is where the 300A/50mV shunt is now, it's between the negative terminal on the 600VDC supply and the bonding wire which runs to the rail (which is obviously embedded in a concrete slab within our shops and also bonded to the rail the entire length of the line. One question though, why would you call between the shunt and the supply 0 and not between the shunt and the physical ground 0? Are you talking about for the purpose of voltage monitoring so that you don't count the current lost in the voltage monitoring process as current drawn?

Here is how I envision your system:

Obviously with the rail itself embedded in concrete the ground bond exists in the rail itself as an Ufer bond. So for bonding purposes you already have the earth ground bond established. I'm slightly surprised you can break this connection to the 600V power with the shunt, that is based on my research for low voltage solar systems where it is a no-no to break the ground bond for any reason, even a small shunt.

But with an established system (that I assume is inspected to code) "it is what it is."

The point I indicate a MON RTN is what I would designate as zero for the monitoring system. This is so the voltage across the shunt is positive and thus slightly easier to amplify and send to the A2D inputs. The slight voltage drop from the shunt (50mV out of 600V) is negligible for the voltage sensing.

Since the monitoring system is floating (ungrounded) it may connect anywhere in the 600V system, this point is about as safe as the other side of the shunt.

You seem to have a good handle on the overall system, do ask for any specifics you may need to bounce off someone else.

I know about rotary converters. My dad worked in the power division of the Long Island Rail Road and we used to visit his friend who worked a sub station. It was a huge brick building (from the time brickwork was a work of art) with two of three stands (one spare) with huge rotary converters running. Very loud and that smell of ozone always present. The converter stands went thru the floor into the basement, and there was a permanent crane to jockey replacements if and when required.

The building is now long gone, replaces with couple of trailer sized units for the solid state converters. Damn big diodes!

What system are you working on? Doers it have a website to visit?