Brent, I am not turning a blind eye to the inherent dangers of non-isolated equipment. Simiarly I also understand both the terrible power factor and inefficiencies of such a setup. I have my reasons and I'll take on your points one by one below. On Sat, Jun 03, 2017 at 11:43:36AM +1200, Brent Brown wrote: > On 2 Jun 2017 at 15:51, Byron Jeff wrote: > > > One example of what the developer describes as a "relatively safe" circ= uit is > > described here: > > > > https://www.youtube.com/watch?v=3DphK_nC4E_jA > > > > with a schematic here: > > > > http://i1063.photobucket.com/albums/t507/lookingfordbn/SNV32249_zps2171= 63f0.jpg > > I take his "relatively safe" or "fairly safe" comments tongue in cheek, a= nd suggest > thats what he means by the use of quote marks. That charger circuit I'd c= onsider > one of the least desirable topologies to choose in terms of safety. There= is no > isolation between input and output. There are no active measures to contr= ol voltage > or current. It is prone to one single point of failure casuing a signific= ant hazard (the > series cap failing short circuit - which I'd expect to be a common failur= e method). No isolation is a clear issue. In the given circuit, he addresses the safety of the issue by preventing the connector from being energized if it's above a certain voltage. That's the "relatively safe" part that the designer is referring to. In addition my plan is to double insulate the device just as double insulated tools do. So even though there is live wire circuitry, it'll be encased so that it's completely inaccessible. Active measures for voltage control are in my functional specification. Monitoring the terminal voltage and cutting off the charger at complete charge are my goals. For safety I'm trying to figure out the absolute failsafe to force de-energization of the charger when the terminal voltage exceed a certain value. As for the single point of failure, both the cap itself (a motor run capacitor) and the fuse both serve as failsafes. Run capacitors bulge out when they short and overheat. This separates the connection creating an open. In addition the fuse will overheat and blow de-energizing the charger. > > > Another example that uses a coil as a reactive element is the Bonn char= ger: > > > > http://www.evalbum.com/tech/bonn_charger.html > > Likewise, that one relies on a single passive component (this time an ind= uctor not a > capacitor) to somewhat reduce voltage and current when everything is work= ing as > expected. Dangerous. I wasn't planning on using this circuit. As to why, I'll discuss below. The point of the circuit is that is has both fusing and GFCI circuitry to lessen the possibility of overload or shocking hazard. > > > These two are the best examples that I can find with at least some mini= mal > > safety features added. But in both cases I still do no feel that it's > > enough for my application. The two requirements I'd like to design into= my > > circuit are now: > > I'd say they have less than minimal safety features... choosing a better = topology in > the first place gains you a large improvement in saftey. Safer examples w= ould likely > be more conventional designs: tranformer based, or switchmode isolated DC= /DC > converter. And here is the brick wall for me. Magnetics of all types to me are a toxic combination of incomprehensibility, extraordinary weight, and exhorbitant cost. Since the last two are fairly obvious in nature, I'll focus on the first. Magnetics of all types are vastly different than any other type of electronic component because the concept of common of the shelf doesn't really exist. Unlike other passive and active devices such as resistors, capacitors, diodes, and transistors, with magnetics the expectation is that you must roll your own in order to develop a part. How many folks would dip their own resistors or wrap their own capacitors? The consequence of this is that there is virtually no information on how to use obtainable magnetics in repurposed designs such as this one. It isn't the fact that I don't understand the concepts. Essentially I need to develop a 350ish watt supply in the mid 50V range to supply the 7A or so I would get from using 2x80uF run capacitors. I also understand the basic formulation of rectifying the AC to a high voltage DC bus for power factor correction, then chopping that DC through a full bridge DC/DC converter circuit at a higher frequency to shrink the transformer. All the magnetics would be the same as a PC power supply. We all have junk PC power supplies. But I've searched endlessly on how to pull the magnetics out of one and repurpose it for something like this. I never found anything of value because all such discussions do forward engineering from the specifications as opposed to backwards engineering from the available parts. In defining a buck converter, there are three parameters: switching frequency, current ripple, and inductance. Every design document out there takes the first two and computes the third. Nothing I've ever found takes a fixed inductance and required current and figure out what frequencies will possibly work. Then of course even if you do that, there's no guarantee that the magnetics will stay cool enough at the frequency to operate properly. In a lot of ways magnetics are black magic that requires juggling and tuning an exact mix of inductance, core material, wiring, and frequency to get anything to operate properly. So while the non isolated designs I put forth have their inherent safety ri= sks, their function is easily characterized: the run capacitor functions as a reactive element with an equivalent resistance of 1/(2*pi*f*c) when AC is put across it. So for my 80 uF run caps it acts as a 33 ohm series resistor at 60 HZ AC giving about 3.6 amps of available current with a 120V RMS AC input. So back to the magnetics tests: Incomprehensible? No. Heavy? No Expensive? My run caps were $5 each The only downside is the fact that the circuit is not isolated so live AC potential could exist at the battery terminals. This is the risk that I'm attempting to mitigate. > > > Same with the EVSE, which is designed to only engage power when a prope= r > > connection has been established. The J1772 pilot signal is a study in > > simplicity and elegance. With a single pilot line and little more than = a > > handful of diodes, resistors, and switches, it is possible to virtually > > guarantee that the EVSE is properly connected to the target before powe= r is > > applied. And with the live mains circuit I'm planning to use, this > > connection is essential for safety critical operation. > > Yes, agreed, simple and elegant. I have one in my garage ;-) Mitsubishi O= utlander > PHEV. You know all this, but, the fancy charger box with J1772 talks the = car and > tells it "I am an EVSE box connected to an AC power outlet and, if you ac= cept my > terms and conditions, you can draw up to xxA through me". It has a GFCI i= n there > too, but essentially the whole box is just a relay that switches the AC m= ains on or off > to the car. The battery charging circuitry is within the car. I'm well aware. My project for this battery is an electric riding mower. There's no need to have the charger on the vehicle. So I'm going to box the EVSE circuitry and the charger circuitry together with just a charging cable from there to the battery for the mower. > > > So I'm just looking for some thoughts on reasonably reliable design > > techniques to back up a microcontroller based system so that even if th= e > > primary circuit goes south, the safety requirements are still met. > > Sorry, I've only given thoughts so far on two designs I think are not saf= e. I guess I'm > saying that if the primary circuit in itself remains sufficiently "safe" = under typical > failure conditions, then it will be of lesser importance what happens if/= when your > micro fails. The micro will be controlling some of the issues of the original circuits. So if it fails, then issues such as overcharging will be back on the table. I'm still looking for suggestions for absolutely failsafes in the instance that the primary control systems fail to do their jobs. BAJ > > Brent > > > -- > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive > View/change your membership options at > http://mailman.mit.edu/mailman/listinfo/piclist -- Byron A. Jeff Associate Professor: Department of Computer Science and Information Technol= ogy College of Information and Mathematical Sciences Clayton State University http://faculty.clayton.edu/bjeff -- http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .