Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

How to make a 30 amp J1772 extension cord for public charging

This site may earn commission on affiliate links.
What don't we put some positive energy into developing a safety circuit?

I need a circuit designed that will limit the maximum EVSE pilot signal "duty cycle / pulse wave modulation" to the following:

1) EVSE pilot signal > than XX duty cycle, output XX pulse wave modulation (PWM) only
2) EVSE pilot signal <= to XX duty cycle, let the EVSE's PWM pass through unchanged


I've been thinking about this:

1) Luckily, it's theoretically possible for this to be self-powered (leached from the pilot signal itself): the car only draws current from the positive half-cycle of the square wave, and since our circuit needs to reduce the duty cycle of the square wave (in the cases where it is doing anything at all), then the power delivered to the car is less than the power drawn from the EVSE [talking here of the very small amount of power in the pilot signal itself, not the AC!]. Hence in the case where we are reducing the duty cycle, there is power available to run our circuit; in the 'do nothing' case, we either need to arrange that our circuit consumes a negligibly small amount of power, or else we always need to reduce the duty cycle very slightly (which would be a shame, not a show-stopper). If the pilot signal had been the other way up, then this would have been a non-starter.

This is also a good secondary safety effect: the circuit can't get it wrong and significantly increase the duty cycle over what the EVSE specified without running out of power.


2) Analogue approaches: it's easy enough to turn the duty cycle into a voltage and threshold it (though maybe issues with accuracy), but building something that modifies the square wave having decided to do so is tricky - simply interrupting the current to shorten the high pulse isn't guaranteed to work, as it needs to be actively driven negative (even though the car isn't drawing a load on the negative half-cycle, the input it uses to monitor the duty cycle is entitled to expect it to go negative. Not impossible, but it's all getting a bit complicated and hard to demonstrate that it works in all cases.

3) Digital approaches: the required logic is a trivial piece of software if you can get a small micro in there - the challenge is to power it while keeping all the other conditions correct:
  • Mustn't draw more than about 1mA (and preferably less) when not pluigged in to the vehicle, to prevent the EVSE thinking there's a vehcile plugged in.
  • Must draw enough current during positive half-cycles to keep the voltage in range - so that the EVSE neither thinks the car has gone away nor thinks it is demanding ventilation [ideally, we pass through the ventilation request, but would probably be OK just to detect it and shut down, with this adapter specified not to work for vent-required charging).
  • Must track duty cycle changes during operation, in case of a smart EVSE.
  • Must detect the car having gone away (or requested end of charging) and reflect this back to the EVSE.
  • Avoid exceeding the load capacitance permitted on the pilot (2.4nF). This will be a bit marginal, since we will be presenting a complex load, but unlike the other values the spec limit here is rather arbitrary.

By way of scale, in the charging state the car draws approx 6mA @6V [50% duty cycle], and 3mA @ 9V before starting to charge. So a micro such as MSP430F5172 which uses 100uA while running and about 10uA in sleep is just about feasible - if running all the time, would have to reduce the duty cycle by 1.7% (charge at 29A rather than 30A) to make up for the micro's consumption; if it can be kept mostly in sleep mode then the effect becomes negligible (and can probably be lost in system tolerances).

Simplest approach would be to separate the two sides: just rectify the pilot to supply +/- power and then talk to the EVSE on one side and generate pilot from scratch (using that supply) on the other side. However, this would look very bad for the effective capacitance. Better approach is to run the output synchronous with the input: still rectify the EVSE's pilot to power the micro and the negative rail, but take the positive output directly from the input with a FET, drive the output to the negative rail for the other half cycle, and have an additional means to load the input to burn off excess power. Probably it can all be done just by accurate timing, but a micro with an ADC (as above) would allow for monitoring the voltages and tweaking the duty cycle if necessary to keep them in tolerance. Also with that particular micro gives you a temperature sense 'for free'.


All sounds quite doable, though I don't personally have a need for it since I'm in Europe where the type2->type2 cables are already self-identifying.
 
Actually there are several models of EVSE being sold that come with a NEMA 14-50 plug to supply power. The NEC only requires that the cord be no more than a foot long in addition to a max of 40A.

That in itself is not a a problem. An EVSE that plugs into a NEMA 14-50 should generate a 40A or less pilot signal, like the UMC. I was saying that I hope no one is selling an EVSE that plugs into a 14-50 and generates a pilot signal greater than 40A. I doubt anyone is doing that.
 
But Flasher every part of this extension is open to view while in use. From the Tesla-issued J1772 adapter sticking out a few inches from the ChargePort all the way to the J1772 EVSE plug laying on the ground. So not the best idea maybe to disappear around the corner into a burrito shop as soon as you use it for the first time, but keep your eyes on it and feel the connectors for undue warmth. J1772 plugs get abused/vandalised so inspect/clean before use.

Thanx Tony for this!!

Now what we need next along these lines is a similar EXTENSION CORD for the SCs. Yes, 400 volts DC at 125kw. After seeing all 6 of the SC slots getting ICEd yesterday brings me to this conclusion. An extra item would be an orange holder to keep the SC end off the ground.
--
 
Tony has asked this question (shared his idea) on several EV boards. On another board it was pointed out that you can't use the 120/240 line voltage to run the 'idiot-proofing' current pilot. In short, until the car asks for power, the 240v isn't there. I suppose in theory you could just pass the generic pilot and then bring on a circuit that pulls the pilot down to 30A if necessary, but I'm starting to think this idea isn't a 'general purpose' product. It'll be fine if I make one for myself for my LEAF (i'll never pull too much power with that!) and is likely safer than a 240V extension cord on the unprotected side of the EVSE.

on further read of the NEC - I don't see that this J1772 extension violates anything. It isn't part of the EVSE. I know of no rule that says you must have a fixed receptacle on the EV (I'm happy to be shown evidence either way). One could even argue there is a cable management system - the user! - but I'm not a lawyer, so I just have to apply common sense and good engineering.

I probably wouldn't leave ANY extended system unattended unless in a very controlled environment anyway.
 
Last edited:
This is true, but you should be able to power it off the pilot signal itself.

Have you calculated how much power is available? If you want to reduce the duty cycle from 60A to 50A, the theoretical power available is less than you need to run the MSP430 while using the peripherals and an op-amp. Might be able to eliminate the op-amp by putting the positive pilot through with a FET like you suggested. You still have to drive the square wave to -12v for the period of time that it's shortened. A lot of components are needed and they all cut your efficiency.

You could probably run the logic entirely off the -12v pilot regardless of how much you need to shorten the duty cycle. You would just need a very efficient DC to DC converter.
 
Last edited:
Have you calculated how much power is available? If you want to reduce the duty cycle from 60A to 50A, the theoretical power available is less than you need to run the MSP430 while using the peripherals and an op-amp. Might be able to eliminate the op-amp by putting the positive pilot through with a FET like you suggested. You still have to drive the square wave to -12v for the period of time that it's shortened. A lot of components are needed and they all cut your efficiency.

Any time you are actively shortening you have power to spare and will need to be burning it to prevent the voltage drifting up and the EVSE turning off. My calculation above gave a maximum 2% duty cycle reduction needed to 'pay for' the micro. The tricky case is a very low output EVSE (Blink?) where the duty cycle is very low in the first place and you don't want to shorten it. However, in this case the micro can be almost entirely in sleep - just needs to wake up once in a while to make sure it's not a smart EVSE that's suddenly started offering more power. Your 60A to 50A example is the worst case for shortening as this is just at the corner of the steep part of the curve (where 2% reduction adds up to 5A reduced charging current), but your example itself is still covered and this end of the duty cycle range is better since the micro's consumption is spread over more of the time.

I certainly didn't intend there to be an opamp in the circuit. I ought to draw this up properly, but the circuit I had in mind comprises:

Three outputs from the micro
1) P-FET connecting input pilot to output pilot. When left turned on, this will pass through the EVSE's pilot with no other action required: the body diode will conduct the negative half-cycle.
2) Another FET to drive the output to the negative rail (with FET#1 turned off) for cycle shortening.
3) diode+ FET + resistor to source excess current from the pilot when shortening - equivalent to the circuit in the car, but either make the resistor slightly larger or drive the duty cycle of the two FETs to add up to less than 100% to compensate for current taken by the micro and so keep the voltage level seen by the EVSE correct.

At least one input to the micro: means to sample the output pilot, to detect the car requesting/not requesting charge (ADC input - probably needs another output to enable it so as to not waste power in voltage divider when not actually making a measurement). This input also lets you see if you are causing the pilot to droop due to too much power used in this circuit, and so adjust the duty cycles to compensate. Theoretically, you could get the EVSE-side pilot timing through this one signal, but probably better to have a separate input for that purpose (which can be high impedance since it's only a digital level).

The micro then needs power - series resistor, diode, cap, LDO. Series resistor chosen to avoid triggering the EVSE at startup. Capacitor needs to hold up the micro across the negative half-cycle, but there's reasonable headroom to do so (6V down to 3V for the micro). Note that this gets charged up close to 12V at startup (car not connected, EVSE putting out continuous 12V), thereafter tracks the EVSE-side pilot high voltage (with the beneficial side-effect that it stops the EVSE-side pilot springing up when we turn off the feed through to the car in the shortening case).

Finally, there's another diode/capacitor to hold the negative voltage. This part is slightly unsatisfactory, as there is no way to pre-charge it and so the first few cycles seen by the car will be less than full-height in the negative direction.

Add a few more parts for level shifting the FET gates and we have a total of about 8 transistors, 2 diodes, 20 resistors, 4 caps, LDO, micro, and probably a varistor or suchlike for transient protection. Most of the Rs&Cs can be 0402 size, so the whole thing probably fits in say 30x30mm PCB.

You could probably run the logic entirely off the -12v pilot regardless of how much you need to shorten the duty cycle. You would just need a very efficient DC to DC converter.

As you say, there might be benefits from being more inventive in the powering arrangements.

You could run the micro entirely off the negative side without a DC-DC just by having the micro's Vss be the -12V rail (since the control FETs need their gates level-shifting anyhow), but this would have the disadvantage that the micro doesn't start up until the car has started charging. Not actually much of a problem so far as the functionality is concerned, but it does mean that the circuit doesn't fail safe: if the micro dies, then the car continues to see the full pilot.
Hanging the micro between +6 and +3 might reduce the component count. Or maybe you can take advantage of the 1kHz to drive a crude buck converter to improve efficiency without many extra parts (the current-limiting resistor in the +ve supply becomes an inductor instead).

Or you could cop out and power the thing from a CR2032 button cell, giving about 5 years standby or 90 days continuous charging. But I think that the self-powered option ought to work.
 
Thanx Tony for this!!

Now what we need next along these lines is a similar EXTENSION CORD for the SCs. Yes, 400 volts DC at 125kw. After seeing all 6 of the SC slots getting ICEd yesterday brings me to this conclusion. An extra item would be an orange holder to keep the SC end off the ground.
--

You're welcome. I've been thinking about this idea for CHAdeMO, but hadn't really considered it for Tesla since all the components are proprietary. But, after hacking several of the UMC's open, I can see that there "ain't no magic" in there. Those pins can be built and crimped, normal everyday wires used, etc. Since Tesla uses the SAE "Frankenplug" protocols, all the documentation and specifications are public (well, for a small fee, anyway!).

Because Tesla uses a relatively simple locking mechanism, it will be FAR easier to build than a CHAdeMO one.

- - - Updated - - -

You could run the micro entirely off the negative side without a DC-DC just by having the micro's Vss be the -12V rail (since the control FETs need their gates level-shifting anyhow)

Unfortunately, not every EVSE complies with the standard for a 12v negative wave, therefore this wouldn't work. Sadly, the EVSE in the most popular EV that the world has ever had (Nissan LEAF) for model years 2011 and 2012 only sends positive voltages on the pilot pin. This EVSE won't work with Tesla (and Tesla powered, like my Rav4 EV) cars. Nissan fixed this for 2013, however the 2013 LEAF can't accept over 75 amp duty cycle on the pilot! Allegedly, they are fixing that. Heck, the Tesla Roadster can't accept a pilot signal over 70 amps.
 
Last edited:
Or you could cop out and power the thing from a CR2032 button cell, giving about 5 years standby or 90 days continuous charging. But I think that the self-powered option ought to work.

If a battery is required, I recommend cells that are commonly available around the world, and that the output be 3.3 volts so that we can also power LED lights in the plug and receptacle, in addition to the Tesla remote charge port opener transmitter.

- - - Updated - - -

DRAFT 2:

J1772extensionCordDraft2.jpg
 
Any time you are actively shortening you have power to spare and will need to be burning it to prevent the voltage drifting up and the EVSE turning off. My calculation above gave a maximum 2% duty cycle reduction needed to 'pay for' the micro. The tricky case is a very low output EVSE (Blink?) where the duty cycle is very low in the first place and you don't want to shorten it. However, in this case the micro can be almost entirely in sleep - just needs to wake up once in a while to make sure it's not a smart EVSE that's suddenly started offering more power. Your 60A to 50A example is the worst case for shortening as this is just at the corner of the steep part of the curve (where 2% reduction adds up to 5A reduced charging current), but your example itself is still covered and this end of the duty cycle range is better since the micro's consumption is spread over more of the time.

arg thanks for your insight. I'm curious about your calculation that only required about 2% shortening to power the micro. Whenever I start using peripherals with any micro, the power consumption usually increases by several times what the micro needs to stay on. After you listed the parts below 2 diodes, 20 resistors, 4 caps, level shifters, LDO, all of which consume power, I'm wondering if you can really keep the pilot from being pulled too low. Many EVSE are already on the low side of +-12v. Maybe you're not worried about this because we can just shorten 3 or 4% to be safe but once you go above 50A, like you said you end up shortening 5 - 10A (2 - 4%) just to run our circuit. That's why I keep thinking we should just take it all from the negative rail, but that's a little tricky.

What did you mean by "low output EVSE (Blink?) where the duty cycle is very low in the first place." You mean one that is putting out, say 15A pilot, and after powering our circuit there's only 12A left over? There would be a lot of power to harvest from the negative rail in that case but hard to design a circuit to harvest the excess power from the high rail when duty cycle is long, and neg rail when it's short. This may not be an issue most of the time but if people are counting on this to scale their amps down, they get used to it working so they stop paying attention, and then it doesn't work 2% of the time, then you would melt something 1 in 50 uses!

Also, are you sure you need level shifters for the FETs? Would 3.3v not control 12v fast enough? Or because we have to control the negative rail?


I certainly didn't intend there to be an opamp in the circuit. I ought to draw this up properly, but the circuit I had in mind comprises:

How will you measure the voltage on the EVSE-side pos rail? If you use large resistors in the voltage divider to reduce power, won't it be hard to do this accurately without an op-amp?

Thanks for your thoughts.
 
I'm curious about your calculation that only required about 2% shortening to power the micro. Whenever I start using peripherals with any micro, the power consumption usually increases by several times what the micro needs to stay on.

The MSP430 follows its datasheet quite closely - they give you separate consumption figures for the various functional blocks, although the big-ticket item is high frequency clock. Certainly if you turn everything on it uses (relatively) lots of power, so the game is to be inventive in getting the job done in small bursts and spend the rest of the time asleep.

After you listed the parts below 2 diodes, 20 resistors, 4 caps, level shifters, LDO, all of which consume power, I'm wondering if you can really keep the pilot from being pulled too low. Many EVSE are already on the low side of +-12v. Maybe you're not worried about this because we can just shorten 3 or 4% to be safe but once you go above 50A, like you said you end up shortening 5 - 10A (2 - 4%) just to run our circuit. That's why I keep thinking we should just take it all from the negative rail, but that's a little tricky.

I'm largely ignoring the option of powering from the negative half-cycle, because of the issue of it not becoming available until the EVSE has been woken up, and also because of Tony's information that common EVSEs don't actually go negative anyhow. I'm also ignoring cunning schemes with inductors, DC-DC converters etc. because they are getting just too complicated and the overhead is likely to wipe out any gains.

I've been a bit casual with terminology in discussing this: my calculation was based on the current guessed to be consumed by the micro (100uA). All of this current has to come from the positive pilot, but the pilot is at 6V and the micro wants it at 3V, so half of the power is being dissipated in the other components. Admittedly the LDO adds 2uA of current that doesn't go through the micro, so I should really be using 102uA, except my 100uA wasn't anywhere near that level of accuracy in the first place.

My 2% calculation rounded this up a bit, and made the simplifying assumption that when the FET is turned on, all the current is going into the car, while when it is turned off all of the current is going into charging our reservoir capacitor. Depending on component values, this assumption could be actually accurate: with the car's load disconnected, all the current will certainly be going into our reservoir capacitor (as there's nowhere else for it to go) and the pilot voltage will rise slightly, then when the car is turned on again the pilot voltage will be below the reservoir capacitor voltage (+diode drop) so no current flows through the diode. In practice, the series resistor limiting the inrush current spoils this a bit and we probably draw a bit of current while the car is 'on' (and proportionately less when it is 'off'), so the pilot voltage we achieve brackets the target voltage rather than just drifting up.

The one significant thing not properly estimated here is the dynamic current needed to charge/discharge the FET gates. I need to actually produce a complete design and pick some real components so as to simulate this.

What did you mean by "low output EVSE (Blink?) where the duty cycle is very low in the first place." You mean one that is putting out, say 15A pilot, and after powering our circuit there's only 12A left over? There would be a lot of power to harvest from the negative rail in that case but hard to design a circuit to harvest the excess power from the high rail when duty cycle is long, and neg rail when it's short. This may not be an issue most of the time but if people are counting on this to scale their amps down, they get used to it working so they stop paying attention, and then it doesn't work 2% of the time, then you would melt something 1 in 50 uses!

This is thinking about the case where we don't want to do any shortening at all, but equally don't need to do much work either (also, in the limiting case we can't shorten the pulse if it's already at the shortest possible). So we would hope to draw our standby current in parallel with the car and not have the pilot droop to much; in this configuration, the shorter the pulse, the more current we have to draw out of it to sustain our average consumption. So with the minimum 10% duty cycle, we are drawing 10x our standby current during the high cycle. If we we only cared about relatively high current EVSEs then we could simply make the thing refuse to work at all on the very short pilots, but the example of the de-rated Blink stations (and also 120V EVSEs) means that would be a major shortcoming.

I absolutely agree that this circuit has to behave safely in 100% of possible EVSE scenarios - otherwise it's much worse than no circuit at all: as you say, a cable that claims to be foolproof but isn't quite is much more dangerous in practice than one that's known to have limitations and so the user is expecting to use due diligence. However, that doesn't mean it has to have optimal behaviour in all combinations, just safe behaviour: so long as the failure mode in any of these corner cases is that it refuses to start charging at all, or charges at less current than would ideally be possible, that's OK. For example, I think we could reasonably declare that we don't support ventilation-required vehicles - so long as the behaviour if you do manage to find such a vehicle is no charging rather than illicitly charging without ventilation.

Fortunately, the particular case we are talking about here fails safe if we get it wrong (and I think we can get it right). If we take too much current while charging, the voltage droops into the 'ventilation required' range, so the EVSE either aborts charging or (if any actually exist) turns on ventilation. If the car is disconnected unexpectedly, the small current we are taking is much too small to hold the pilot voltage in the active region by itself, so the EVSE turns off as normal.

In fact, for this case I am confident (based on previous experience) that I can get the standby current below 10uA average while still monitoring the pilot regularly to check that it hasn't changed. For the minimal 10% pilot that's 100uA average during the high pulse, hence 0.1V change in the pilot level, which should be comfortably within tolerances.

I am less confident of the accuracy of my 2% penalty when actively shortening - my original thinking was around Tony's challenge of the extension cord built from 30A-rated components, so with the practical population of EVSEs it wouldn't matter that much even if you had to shorten by 10% (so long as the pass-through mode works without shortening).

However, I've just realized that this same circuit could be used to solve the "Roadster can't charge from >70A EVSE" problem (given your interests, you probably had this in mind from the start: I'm just catching up:smile:). For that application, being able to trim only a very small slice is critical, given the much higher ratio of charging amps to duty cycle at that end of the range. On the other hand, you are now only ever shortening when the duty cycle is over 90%, so actually you can afford to draw most of the power as you need it rather than storing it up from the shortening period (per the above logic that it was OK to take 100uA during the 10% pulse to meet a 10uA average requirement, you could take 100uA from a 90% pulse to meet a 90uA average requirement). So with slightly different resistor values and/or slightly different software strategy it probably still works.


Also, are you sure you need level shifters for the FETs? Would 3.3v not control 12v fast enough? Or because we have to control the negative rail?

3.3V swing is enough, but it's a matter of the reference voltage. If the micro's Vdd/Vss are +3.3V and GND, then that's not a useful range for controlling either of the FETs. If you hang it upside down and make the micro's Vdd/Vss Vpilot and Vpilot-3.3, then the top FET (linking pilot-in to pilot-out) can be driven directly, but the bottom one (driving pilot-out to -12V) still needs shifting somehow. But maybe that can just be with a capacitor (since it's only ever driven for short pulses). Complicates the power arrangements doing it this way up (need to find a low-leakage negative LDO), but maybe worth it.

It's getting to the point where I need to do some real work on this rather than just discussing generalities...

How will you measure the voltage on the EVSE-side pos rail? If you use large resistors in the voltage divider to reduce power, won't it be hard to do this accurately without an op-amp?

I normally use resistors in the 50K-100K sort of range, and then have a FET to turn it off when not actually measuring (which only takes a few us).

The inputs stage of these ADCs is effectively a small capacitor, so for a given accuracy there's a trade-off between bigger resistors in the divider using less power but needing a longer sampling time so end up being turned on for longer.

Possibly we might save some parts by sharing this with the FET/resistor already needed for burning off excess power when the shortening time is longer, though that limits what we can measure.
 
Last edited:
on further read of the NEC - I don't see that this J1772 extension violates anything. It isn't part of the EVSE. I know of no rule that says you must have a fixed receptacle on the EV (I'm happy to be shown evidence either way). One could even argue there is a cable management system - the user! - but I'm not a lawyer, so I just have to apply common sense and good engineering.

It calls for a "cable management system that is listed as suitable for the purpose". The term "listed" is defined in article 100 as:
Listed. Equipment, materials, or services included in a list
published by an organization that is acceptable to the au-
thority having jurisdiction and concerned with evaluation
of products or services, that maintains periodic inspection
of production of listed equipment or materials or periodic
evaluation of services, and whose listing states that either
the equipment, material, or service meets appropriate des-
ignated standards or has been tested and found suitable for
a specified purpose.

...which typically means UL in the US, or CSA in Canada.

The maximum cord length past the EVSE is 25' (625.17) unless there is a listed management system, which an extension cord + UMC would violate.

Again, you can probably still make the argument that the EVSE is an appliance is not covered under the NEC, but you'd only find out once you've been denied a claim by the insurance company and take it through the court.
 
Last edited:
... but you'd only find out once you've been denied a claim by the insurance company and take it through the court.

These unqualified "jail house lawyer" admonishments pop up many times on the internet, usually just getting louder and more repetitive when their position doesn't appear to be accepted. I'm loath to continue this debate on a forum about electric cars in a sub-forum labeled "TECHNICAL". But, since the issue seems to be the only issue that the above poster has with this project (and so far has no bonafide technical comments to offer), I feel compelled to comment in such a way that ends the debate.

First, the above admonishment has been repeated ad nauseum; I'm confident that the readers now fully understand these claims as stated. At this point, each reader can choose to follow this advice, or not, but to keep posting it over and over doesn't make it true; it detracts from the actual purpose of the thread.

Secondly, I actually called my home owners insurance company, USAA, and posed this specific scenario. I was absolutely assured that my company would pay the claim provided there wasn't intentional malice (intentionally trying to burn the house down). Just having a 26 foot cord or a 126 foot cord where a 25 foot one is specified by NEC was not a factor for them. Again, they made it clear that an electric vehicle charge cord longer than 25 feet would not be a factor, even if it was the cause of the fire and in "violation" of NEC, and any other industry standard. They did mention that following industry standards for wiring was encouraged, but something that didn't meet those standards wasn't the "smoking gun" to prevent them from paying the claim... again, without malice.

They did mention that if an older house had "knob and tube" wiring, an inspection would be required of the house, and the homeowner would be given time to repair anything UNSAFE, but they would still be fully covered the entire time the policy was in force. You might want to check your ACTUAL insurance company if you consider this an issue, rather than be swayed by unqualified internet blather.

Thirdly, it's a big world out there, and rules, regulations, insurance, etc, are different everywhere. Obviously, it's wise to understand the potential implications that are specific to your situation in your political jurisdiction.

If you have an insurance company that would use such an issue as a reason to disqualify your claim, my company suggested that "you get what you pay for", and I couldn't agree more. Perhaps the poster has one of those insurance companies. Now, can we move on? If there is continued retorts about insurance and legal issues, I ask the moderators to move those posts to an appropriate non-technical part of the forum.

Thanks.
 
Last edited:
There's a considerable simplification that can be made to this circuit's requirements:

I have so far been treating the requirement as "shorten the high pulse" - and so need to drive the output low before the end of the EVSE's high pulse, hence needing to store enough charge in a negative voltage rail etc to drive a negative transition.

However, this is logically equivalent to "stretch the low pulse", which is much easier: let the EVSE drive the pilot negative for as long as it wants, and if that's not long enough for our requirements, just leave the output floating for the 'stretch' period (while the EVSE is driving the input pilot high). At the end of the stretch period, just connect the input and output together, and the EVSE drives the output high. The car is not supposed to load the negative side of the pulse (the EVSE is supposed to check that the car presents an asymmetric load, so as to detect the difference between an actual car and a high-resistance short**), so the output should float at the negative value until we drive it back the other way - the extra capacitance of the longer cable helps here.

So the circuit now looks like:

  • a P-FET to control whether a high level on the input pilot passes through to the output or not.
  • a diode to ensure the output is always driven low when the input is driven low (the FET's body diode might do, or you could add a better diode in parallel).
  • another diode/resistor/FET (equivalent to what is in the car) to keep drawing current from the high pilot while the output is disconnected for lengthening purposes.
  • a one-shot triggered by the falling edge of the pilot, giving a pulse of length equal to the minimum output pilot low time we require: for the duration of the pulse, the 'pass-through' FET is turned off and the 'dummy load' is turned on.

This could be implemented in the software-based design previously discussed: just throw away the negative rail parts and change the software. But it's now simple enough that an analogue circuit is also feasible (though accuracy may be an issue).

Snags with this approach:

  • Output might float up, or be subject to noise, during the 'stretch' period. This is more of an issue with very long stretches (Tony's 30A cable) than with a short stretch (80A->70A fix for roadsters), but I think probably OK.
  • There will be a glitch on the input pilot at the point where the circuit switches from the dummy load to passing-through to the car (ie. the end of the stretch period). If this happens to be the point at which the EVSE samples the pilot state, this could cause an invalid reading and so EVSE to terminate charging. Again not really a problem with a very short stretch (EVSE is supposed to allow settling time after the rising edge and sample nearer the middle of the pulse), but could be for the 30A case. Possibly could be fixed with filtering.
  • The simple version presented above will go on drawing current into its dummy load for part of the cycle even if the car has disconnected, potentially causing the EVSE to fail to turn off. For the short-stretch case, there's a trivial fix: simply dispense with the dummy load altogether - the control circuit's power supply will be recharging during this 40us interval, and won't allow it to rise very far; if the car is not in fact present, it will carry on rising through the rest of the cycle (920us) and trip the EVSE as usual; in effect, the power supply is latching the level set by the car. For the longer stretch, this won't work and the dummy load is needed, along with some means to inhibit it when required.

So, for the full-blown solution supporting limiting at any point, the version using a micro is probably still the way to go (albeit simpler than before), but for the special case of reducing the pilot from 96% to 92% for the benefit of Roadsters, a very simple circuit (eg. built from a 4000-series CMOS monostable, that wouldn't need a regulated supply) might be possible.

----
** BTW this is another reason why we shouldn't be trying to power our circuit off the negative side.
 
[*]Output might float up, or be subject to noise, during the 'stretch' period. This is more of an issue with very long stretches (Tony's 30A cable) than with a short stretch (80A->70A fix for roadsters), but I think probably OK.

I think this will be ok, since we're trying to protect the 30 amp components, and if it actually has 28 or 32 amps, the protection is still valid.

But, for the 70 amp Roadster issue, that absolutely needs 70 amps or less or it just won't work.
 
There's a considerable simplification that can be made to this circuit's requirements:

I just spent a lot of time writing comments/questions to your previous post and then I saw this! It's so obvious yet didn't come to me right away. BTW I already found and built another solution for the Roadster problem, actually 2 solutions, one that only uses analog components. They depend on a couple of abnormalities with the way the Roadster charges and as such they aren't very elegant and wouldn't work with all cars. I'm more interested in this solution for other EV charging applications, not just extension cords.

I can't see doing this without the micro but maybe it's possible. The dummy load circuit has to match what the car is doing in order for it to work, so it needs at least 2 FETs for 2 resistor circuits, and we need a means to sample the load on the EV side and activate 1 or both dummy circuits. I know from testing a number of EVSEs that you can't count on them to measure the high level near the middle of the pulse. In fact most of them measure it shortly before the end of what would be the shortest pulse for the lowest duty cycle. That might easily place it in the dummy load time or during the glitch. Not sure if the glitch is something we need to worry about. All the EVSE circuits that I've designed take multiple samples (usually 64) and average them, but not sure how others work.

Why do you think the dummy load would stay on when the car disconnected? We're using a one-shot triggered by the falling edge of the pilot to turn on the dummy circuit. Wouldn't the dummy load get turned off at the end of our timing sequence the same as any other time?

I would guess that noise on the floating low signal would be a problem but I don't think that's hard to solve.

Could we solve the glitch problem by using silicon transistors instead of FETs to transition to and from the dummy load? Instead of switching on a new circuit of resistors, transistors could manage the current. As I think about this, it would be more complicated and introduce other problems. A filter approach might be better.