-
Hi all just so you know I am busy rather than asleep here are some updates to my ongoing GTI designs, basically I have one incarnation in service whilst designing its improved replacement, an ongoing process.
Those familiar with the typical LCL filter used for grid interconnect will understand that whilst it has many advantages it has severe penalties caused by its resonance and phase shift making the feedback around the current loop very difficult to stabilize. This does not really have a parallel amongst simple feedback discussions such as type I, II & III. 99% of related treatise assume your a mathematician and have a copy of mathcad neither of which is true in my case. However reading material endlessly can procure results and I found some references mentioning the use of LCL capacitor current as a method to assist loop stability and the lights switched on for me!!
Capacitor current has a leading phase wrt to line current and solves the basic problem of attaining a feed-forward effect that I have long sought after, as in predicting line current before it happens
I hope the enclosed plot illustrates the effect of mixing a proportion of capacitor current into the feedback loop ( the really nasty waveform is just line current, the better is with a proportion of capacitor current).
The next issue is that of sine reference, I must admit to being guilty of simply using the grid voltage!! Fool I am sure many say well OK and now I understand if your grid has a lot of harmonic distortion or your grid coupling impedance is highish your GTI acts as a harmonic amplifier!! So I shall finally resort to synthesized sine reference by micro phase locked to zeroX but why the hell does reference material scare people off with crazy wording such as "rotating reference frame" etc, for gods sake its a simple pll right!! maybe I grew up to long ago
Finally I would like a word about the venerable ACS712 hall current sensor, I used this in innocence in the presence of high powered ferrite inductors and transformers OMG I know I should have known but as warning guys it doesn't like external magnetic fields! Anybody know of a good analogue isolation barrier as I am thinking of dumping them in favor of resistive shunts, I have loads of high voltage auxiliary power available ?
Edit so there is the HCPL-7510 from Broadcom (Eww) but it's optical and there must be similar from real designers somewhere
-
The HCPL-7510 was an Avago part that I think originated with HP/Agilent, it is probably fine (Those guys were good at isolation amps).
TI have some of the old Burr Brown iso amps still in the catalogue, may be worth a look?
Or you could maybe go old school with an IL300 or suchlike?
A 'Rotating reference frame" is just saying 'phasor diagram' to me, nothing too scary there.
Regards, Dan. -
Use a closed loop magnetic current sensor like the ones from LEM - these are quite insensitive to external fields.
The HCPL-7510 is also a good and reliable device, though there are newer parts with better linearity and bandwidth available from HP/Avage/Broadcom. Better not use the magnetic or capacitive coupled ones from TI / ADI (to name the most known brands), they might be nasty sources of high frequency EMC. Stay away from coreless or open loop sensors in your environment. With some exeptions, e.g. when properly used, the Sensitec ones may perform well in your environment though beeing coreless.
Anyway, I'd prefer a LEM current transducer of appropriate size. Easy to use, +/- 15V or single supply 5V versions available, and isolation is for free.
Edit: If it is just the capacitor current to measure, there's no need for a DC path, you could use a simple small CT here.
Another one: If you have paralleled capacitors in your LCL, you can use these as a current divider (smaller cap in parallel to a larger has proportionally less current) together with a larger value (easier to get a larger signal amplitude) sense resisor.
-
Many thanks for your ideas and help I must admit to being slightly biased against certain companies that have acquired technology through aggressive acquisition followed by downsizing of employees but sometimes there is no choice
As for the name changing game, just Grrrrr
I like the idea of a small CT for capacitor current and I have never had a problem with small CT's in SMPS's so that should work As for the output current it is good to have the LEM recommendation although I recall these can be expensive OTH in a noisy environment a shunt can loose considerable power to provide enough signal.
I must confess to currently (pun) working in other area's of the project, namely the MPU that requires considerable overhaul and for me a completely new assembler to digest
-
Things are once more moving to the prototype construction phase having created yet another new topology and control scheme. I have been looking at PCB costs and they are ever rising, the pain in part is caused by the size of the components mounted on the PCB so I have decided to revert to my valve ancestry and chassis mount the major components, in this case being the transformer, inductors, large capacitors and also avoid using a PCB for mounting heatsinks. The only reason for a PCB at all is the voltages are to high (~500V) for stripboard construction and a compact layout is required around the switching devices/drivers.
Another lesson learnt is to divide the power circuits into two PCB's as this will make re-spin half the cost. I made the mistake before of trying to fit a prototype into a fairly tightly dimensioned enclosure and although this worked the PCB costs were much higher than had I given myself more space for a more distrubuted mechanical design.
The control function has not been an issue as it is constructed on stripboard and plugged into the power PCB, very simple to re-work at almost zero cost.
Hope this gives food for thought for others trying to reduce PCB costs in high voltage power circuits
-
I have completed the preliminary layout of the DC (input side) board and it shows a substantial size reduction is possible. Size is now 125*105mm and an estimated cost of $27 Both the main transformer and first stage inductor are now chassis mounted to save pcb real estate.
Outdoor metal cases for high voltage circuits are also very expensive in larger sizes however I discovered a neat steel lockable letterbox on an auction site for just £8 This is more than big enough being A4 sized and 85mm deep. I have no sheet metal working facilities nor anywhere to buy stock so cannot fold up my own.
Onwards and upwards the AC board (output side) is waiting to be placement tuned and routed!
-
Taking the plunge just to show what sort of issues arise here is a very early part routed AC board, the fight is on between track widths to accomodate current and spacing to accomodate voltages the TO220 mosfet packages are the worst area and this about as much power as you can squeeze out of them at nominal 240Vrms, any higher and the spacing vs track size just doesnt work even with conformal coating.
A problem with KiCad is you cannot specify spacing between particular nets, just a global spacing for a given net, causes headaches in multi-level high-voltage designs like this so a lot of manual gap checking
-
Work suspended for Chinese new year! That was bad timing on my part but hey everybody needs a holiday
-
Chinese new year over! DC boards (first sent out) arrived today, turns out held up longer in UK by snow than air-freight from China
Whilst waiting built and tested as much as possible of the control board, built on the ubiquitious stripboard as all low voltage. Socket at top is for processor.
DC-board just loaded with 5V regulator (just a linear from ~12V as consumption so low). 12V is derived from the panel voltage by a low-power switcher in the immersion (water heater) boost converter, better to only have one low power auxilery high voltage supply in the system, reduces complexity and costs whilst improving efficiency
-
Looks nice.
-
DC board fully loaded (except high energy storage caps) and tested along with control board. Lots of thick heatsink cutting today with an angle-grinder to make heatsinks for both boards, frozen fingers out in the garage! AC board went out for fab yesterday.
Run out of work so will have to start the case
-
AC boards arrived a while ago, stepwise loaded as each section tested, holdups due to unforseen MPU instability but now reached the stage where the main transformer and inductor are ready to be added (external to the pcb's) so the serious testing of the complete GTI can begin!
-
And here is the trial fitting in the case (letterbox) layout of everything designed to ensure the essential PicKit3 was going to fit in situ Run out of fiberglass sleeving so won't be wiring the magnetics till late next week! On the hunt now for something to keep the spiders out of the fan to avoid spidercide
-
Looking good. Is this going inside?
-
Looking good. Is this going inside?
Hello and thankyou It's going outside high up on a brick wall under an open sided shelter, the box has a steel door not shown here, the top vent (letterbox) will be filled with wire mesh to keep insects out and propped open 30deg to allow air out but keep water drops out (there shouldnt be any). Boards will be conformal coated after debug.
-
Conformal coating is a good idea. Condensation.
-
Fiberglass tube arrived so magnetics wired, here we are on the workbench starting debugging at low voltages/energies so far functionally ok but can be difficult to create realistic operation in a test harness. Something else I hate about windows is mouse freezing with heavy USB activity (both PicKit3 & Owon scope) if anybody else gets this try unplugging the (USB) mouse for a few seconds, usually works and saves a reboot!
-
Busy testing, enclosed pic shows anti-islanding in action red being grid and yellow GTI DC input current. As an interesting aside the grid has a non-symetrical voltage variation that the GTI is trying to correct, this variation varies in frequency and comes and goes at different times of day acompanied by a 2-3dB drop in broadband SNR. Looks to me a bit like a cycloconverter!
Nasty smell from the bench when I last fired it up, have yet to have time to trace source but did have shower just in case
-
After some trials and tribulations I have to say I am tickled pink by the sine linearity performance compared to the previous design. Still early days in testing but the enclosed scope shot shows a nice clean grid voltage waveform, note at this stage transformers are used to reduce voltages for safety.
-
Ongoing debugging at moderate powers and voltages, continuing to find what poor software I write like the enclosed scope shot that shows the main inverter getting unloaded intermittently near zero-cross, turned out to be entirely due to lack of thought and the potential consequences of writing some code in a particular way! Meanwhile also discovered (and reported to Mchip) another silicon problem in the HSPWM laughebly just before the easter holiday but fortunatly was able to sucsesfully devise a workaround. The more you use powerful features of chips the more you expose yourself to nasties but if it was not for the high level of integration considerable external logic and linear discretes would be required.
-
My BIG BANG theory is that it's related to overconfidence and always happens when things seem to be going swimmingly well
In my case I was retapping the isolation transformer to go up from 60Vrms to 120Vrms and I thought in the process I would ground the available centre tap for enhanced safety! All sounds fine, isolated scope, isolated laptop, what could possible go wrong 
Well I like my music but most times when I am testing it's switched off and due to the audio jack of the laptop being right next to the heavely used USB's unplugged from the sound system, but today I was happy, I knew exactly what tests I had to do so managed to get the audio jack in with the USB's. Everything ready, cannot remember the tune but when I plugged in the isolation transformer BANG!!! 
loud buzzing noise from the sound system, took a few seconds before I unplugged, EVERYTHING...nasty burning smell OMG 
.......HOW 
Well it turned out the problem was twofold, grounding the isolation transformer centre tap and having the sound system plugged into the laptop, those were both new things I had not done during previous testing
The sound system provided a path from the laptop to mains earth via the cable screen. The scope was monitoring something on both the grid side and solar side of the inverter connecting one leg of the grid to solar negative, thats ok normally as the scope is isolated anyway but now that referance was 60Vrms above ground and the final connection in the short circuit ? The PIC debugger referanced to solar negative and connected to the laptop USB
The moral of the story, check and check again everytime a test harness is changed or something like a sound system added to it!! The sound system survived, the laptop had a track blown off by the headphone socket, the pickit3 is a bit of a mess with at least 3 burnt chips and some burnt tracks, most of the GTI survived but lots of control board chips were destroyed and sadly as I live in the sticks it will take over a week just to get the parts
I dont want to post any pics today, too embaressed!
-
Bummer.
-
Such is the nature of power engineering, if you don't smoke something you are not trying.
It is **ALWAYS** something stupid that you wind up doing.....
I have lost count of the number of really expensive power devices I have killed over the years in screwups not completely unrelated.
It is when you do it to a 250A, 1,000V IGBT module that it really hurts, even when the resulting cloud of copper vapour stays safely inside the box.
Regards, Dan. -
Ahh so true, I wonder if amplifiers also count as power engineering I had some expensive blowups there too over the years.
I think if I lost an IGBT module that big it wouldnt be just the noxious fumes making me cry......
rambling alert! When I think I could sit back and write software to log temperatures or something that has no risk of destruction but then there wouldn't be the same challange and sense of acheivment for me at least, guess I like tiptoing around the edge of failure like someone walking around the top of a skyscraper or perhaps more meaningfully pushing the limits of innovation. I have always enjoyed electrically or mechanically oriantated engineering, something that moves or makes noise! Even this GTI crackles due to my less than perfectly wound magnetics as it reacts to itself or the grid changeing, quite a usefull aural feedback tool as long as you dont get overconfident and start listening to music during debugging sessions 
-
Slowly crawling out of the pit I dug myself, repaired the PicKit3 by replacing not just the tranceivers but a lot of discreetes as well including MMBT3904/3906's several diodes and even an mlcc capacitor SMT not being my strongpoint this took a while as I had to order all the parts in sequentially as it's not possible to find all the dead bits in one go! Again the control board dead bits have been replaced but only a test with the processor running will determine if it's fully functional again. What a fag, I will definetly try NOT to repeat such a stupid mistake again And so the trail of destruction continues more dead parts found on the control board that I do not have and have had to order, hopefully maybe these will be the last!