Author Topic: News from the GTI design workbench, Blog  (Read 9573 times)

0 Members and 1 Guest are viewing this topic.

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #25 on: April 13, 2018, 06:51:07 pm »
Having fixed all the blown up bit's including the control board testing has resumed at 120Vrms grid. I know how much you guys like pictures so todays shot is the error amplifier output when the DC-bus is to low a voltage to support the peak sinewave voltage causing the ea to saturate whenever the grid nears it's peak. This problem will simply be solved by stepping up the DCbus to the next test level (re-arrangeing transformer taps). What I like about the waveform is it's so nice and clean with no nasty resonances!
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #26 on: April 22, 2018, 06:51:58 pm »
Been a bit quiet lately, apart from being the go crazy seed planting season I have had some very hard to trace mosfet popping episodes in the grid H-bridge  |O. This was discovered to be caused by the PIC driving both halves of the H-bridge on (active low) simeltaniosly some 20mS after power up even though at this time and condition no code exists to do so  :--. This is extremely disheartening! Further work discovered it was around the time another channel of the HSPWM was having its IOCTL register changed (the unfolder/H-bridge is driven by an HSPWM channel). After some additional attempts at localising the problem by using scope triggers the fault mysteriosly disapeared  :-//. This coincided with a long programming period (usually they are very short). So essentially the mosfets were lost due to an aberation in the PIC HSPWM, there are indeed some failings of the IO-override bits I use in the Errata though not this specific case however the Errata dates from August last year and obviosly will only contain those test cases so far reported.  Given I have reported several and had no responce it is quite likely other people have encountered and reported this bug already  :--.

As for a fix/work-around I am still considering this before sacrificing any more mosfets  :-BROKE
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #27 on: May 09, 2018, 06:36:24 pm »
So what can be the worst of all evils  >:D Juggling with changing the MPU to avoid the bugs has gotten me back pretty well where I started but here are some of the alternatives and why I dismissed them

1/ STM32F303 I was so impressed by the data sheet I went ahead and bought one, it was only in the latter stages of evaluation I realised a showstopper, the PWM although it did everything else did not support half-bridge or push-pull SMPS's and neither did any of there chips! Turns out there PWM is a very common piece of ARM related IP used by many other ARM MPU vendors so wiped all of them out too including the often mentioned NXP.

2/ Cypress PSOC5 Again very good spec and no worries regarding PWM weirdness as incorperates programmable logic where a custom design can be implimented. First indicator of problems was no errata on the website and when tech support contacted said no such document existed!! THEN I downloaded the IDE, OMG like a childs toy plaything, you are supposed to draw schematics of the design then it goes off and adds whatever boilerplate code it thinks is appropriate, the final straw came when I finally found the RTL description in a seperate document to the functional description with a flat index 45 pages long not even alphabetical, upshot, unuseable IDE and no alternatives (I asked).

3/ PLD was possible except all analogue peripherals would have to be external, but the killer was development system cost.

4/ A simpler PIC that I have used in the past with well known PWM's. Sadly re-discovered why I had chosen the PIC24EPxxxMCxxx in the first place as the PWM is an advance on all those preceding it.

It is very difficult to fix power mosfet drive problems with external logic as it must be secure against mis-operation in all power up/down circumstances. In the end I have decided to persevere with the PIC24EP as the least of all evils  >:D and I am at least now able to quantify some of those even though Microchip show no interest at all. For the time being I think I have devised a discreet circuit with known power cycle parameters that will work-around the PIC problem that was causing MOSFET loss, so the parts are on order and I hope I can replace the lost power mosfet 2nd time around without damage to the pcb, this being a hard ask on a high voltage part. In the meantime there is software to re-write to accomodate the changes and steer clear of the bugs :)

I hope my brief assesment of some alternatives might help others :)
« Last Edit: May 09, 2018, 06:45:46 pm by fourtytwo42 »
 

Offline capt bullshot

  • Super Contributor
  • ***
  • Posts: 3033
  • Country: de
    • Mostly useless stuff, but nice to have: wunderkis.de
Re: News from the GTI design workbench, Blog
« Reply #28 on: May 09, 2018, 07:19:47 pm »
I remember some Renesas controller having a PWM unit supporting half bridge, push-pull and phase-shift PWM modes. Can't tell which one it was, AFAIR a rather small one. We (at work) considered this one for a resonant converter, but did it for whatever reason in another way (using a CPLD). At least you're guaranteed their PWM has quite nothing in common with the ones you mentioned.
Safety devices hinder evolution
 
The following users thanked this post: fourtytwo42

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #29 on: May 10, 2018, 07:57:06 am »
Thankyou for the tip :) I think thats the old NEC, I seem to remember 78K from them many years ago!
I have had a look and there a bit bewildering at first with 20+ product lines and a selector thats too slow to work but early days, I just have to get a feel for them and learn there quirks, certainly plenty of choice :)
It seems there development boards are quite high-cost that may explain why I missed them in the first place.

EDIT Parts arrived so control board modified (getting busier) and ready for test BUT have to replace mosfets first  :-/O
« Last Edit: May 10, 2018, 02:57:22 pm by fourtytwo42 »
 

Offline ahbushnell

  • Frequent Contributor
  • **
  • Posts: 734
  • Country: us
Re: News from the GTI design workbench, Blog
« Reply #30 on: May 10, 2018, 11:27:25 pm »
TI has a Piccolo DSP which is designed for power supply and motor control.  And they have the big brother the Delfino.  Even one with 2 cores and it will run at 200 MHz. 
 
The following users thanked this post: fourtytwo42

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #31 on: May 11, 2018, 04:35:16 pm »
Thanks for the Piccolo tip :) Slight difficulty with protoboards here, seems to incorperate programmer/debugger on the same board but NOT as a breakoff module...... Anyway adds another string to my bow thanks :)
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #32 on: May 15, 2018, 10:03:08 am »
And there we have it, handling and fitting the mosfet with an anti-static foam tutu works  :-+
Fairly easy to remove afterwards with tweezers.
Of course having an EARTHED soldering iron helps too  :-[

EDIT :_ Just realised posting this here doesnt entirely make sense as I originally posted the weird friday mosfet problem in the tech forum, sorry!
« Last Edit: May 15, 2018, 01:45:12 pm by fourtytwo42 »
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #33 on: May 16, 2018, 07:07:13 pm »
Progress today, blew another complete set of unfolder mosfets  :palm: but that was after running a while and a VERY nasty test, removing and reconnecting the grid!

This time I cannot blame the PIC as my workaround circuit discretely avoids it's pwm problems so I just have to admit the mosfets are not man enough for real world spikes and surges. Of course it's pure chance what the phase of the grid is on reconnection and there's loads of spiky stuff as the contacts slide together so who knows what might be happening and the chances of capturing anything usefull on the scope are about nill  :-\

So I decided one instant improvement in beefiness is to replace the mosfets with igbt's, just a bit more surge capacity in the same package (TO220). I am beginning to wish I had used thyristors here in the first place ghaaaa nothing like belt n braces  :-+ anyway the igbt's are compatible with the same pcb assuming I can get the mosfets out without irreperable damage (one of them has been changed twice aready). Also changeing up from 800V mosfets to 1200V igbt's and there are 470V tvs's in there so whoaaaa must be current surge!!
Time will perhaps tell  :-\

Sorry no pics, mosfets look just the same as working ones and only a mild pop from a sand filled ceramic fuse announced the failure....
« Last Edit: May 16, 2018, 07:13:53 pm by fourtytwo42 »
 

Offline ahbushnell

  • Frequent Contributor
  • **
  • Posts: 734
  • Country: us
Re: News from the GTI design workbench, Blog
« Reply #34 on: May 17, 2018, 02:33:08 am »
Schematic? 

Andy
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #35 on: May 17, 2018, 07:23:36 am »
Here you are :) V1 is the grid and V+ is the DC source. There are TVS arresters across both the grid and DC output of the rectifier. However there is not one across any quadrent of the bridge and I am thinking that might be a mistake. In reality the grid connection is via an LCL filter not shown here and the TVS is on the H-bridge side of the LCL.
 

Offline f4eru

  • Super Contributor
  • ***
  • Posts: 1086
  • Country: 00
    • Chargehanger
Re: News from the GTI design workbench, Blog
« Reply #36 on: May 17, 2018, 10:57:46 am »
Hey, fourtytwo42, your project is quite interesting ! I'm doing something similar, but island mode in the first version, with grid connection possibility for later dev.

Do you have the complete schematic with drivers, output filters ?
Your trouble probably comes from from the output filter vs protection etc etc etc...
What's your strategy to regulate the output current ?
how are your PWMs cycled in the positive and negative cycles ?
Do you have enough Tlow when the voltage is high to recharge the drv charge pump caps ? (what's your max pwm ?)
how do you sync to the grid ? PLL ?
do you have a sync waiting delay after reconnection ?
etc etc etc...

What's your DC filtering ?

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #37 on: May 17, 2018, 02:21:55 pm »
There are several reasons I would not publish actual schematics here nor discuss anything in more than general terms.  Those that come to mind are safety, commercial confidentiality and the intolerance some members exhibit to others when discussing high voltage circuits but thank you for your interest and good luck with your own project, perhaps you would like to start a blog on it too :) I should add there are silicon vendor examples of GTI's published on the web.
« Last Edit: May 17, 2018, 03:57:15 pm by fourtytwo42 »
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #38 on: May 17, 2018, 04:01:22 pm »
IGBT's fitted but ohh what a mess! Unfortunatly I could not get hold of the TO220FP packages I wanted so these are standard TO220's with bushes, mica washers and copious amounts of heatsink gunge! Good enough for test but the insulation is definetly not so good  ::)
For comparison thats a TO220FP diode to the left.

EDIT I am releived the igbt's seem to have solved the problem though I did also add a TVS from the AC to DC sides of the unfolder and tweek the software to try and avoid noise at grid restart at the same time! But I am not hurt by the extra TVS, the igbt's were cheaper than the original mosfets and software is free as they say hahaha
NOW I need some igbt's in TO220FP because the creepage on the present assembly gives me the creeps  :-DD
« Last Edit: May 18, 2018, 08:30:49 am by fourtytwo42 »
 

Offline f4eru

  • Super Contributor
  • ***
  • Posts: 1086
  • Country: 00
    • Chargehanger
Re: News from the GTI design workbench, Blog
« Reply #39 on: May 19, 2018, 12:02:43 pm »
Quote
k you for your interest and good luck with your own project, perhaps you would like to start a blog on it too :) I should add there are silicon vendor examples of GTI's published on the web.
Yep, I understand.
I'm also not allowed to publish it as it'sa commercial thing, but I can share a few insights, I'll not write a blog.

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #40 on: May 20, 2018, 08:44:21 am »
Today's problem is classic transformer flux walking, the yellow trace in the picture is the primary current and you can clearly see the knee of magnetic saturation, this has been caused by several cycles of unequal conduction that in turn are caused by instability in the control system.

I am pretty sure this is what led to the tracks being blown off the board last year but now the primary overcurrent protection has been completely redesigned the problem can be examined without loosing the pcb!
 

Offline f4eru

  • Super Contributor
  • ***
  • Posts: 1086
  • Country: 00
    • Chargehanger
Re: News from the GTI design workbench, Blog
« Reply #41 on: May 20, 2018, 11:33:40 am »
Interesting, I didn't get a problem yet with flux walking, perhaps my primary is balanced better.
For the overload on tests, to not blow up everything, and test with no proper cooling, I usually pulse the system 10ms on, 1s off until everything is stable

Offline ahbushnell

  • Frequent Contributor
  • **
  • Posts: 734
  • Country: us
Re: News from the GTI design workbench, Blog
« Reply #42 on: May 20, 2018, 03:24:10 pm »
Today's problem is classic transformer flux walking, the yellow trace in the picture is the primary current and you can clearly see the knee of magnetic saturation, this has been caused by several cycles of unequal conduction that in turn are caused by instability in the control system.

I am pretty sure this is what led to the tracks being blown off the board last year but now the primary overcurrent protection has been completely redesigned the problem can be examined without loosing the pcb!
If the current is not too high you can put a capacitor in series with the transformer and if there is a imbalance a DC voltage will correct for the walking.
Or you can use a current probe and detect the flux walking by the current on one side.  The probe has to be able to measure a DC component.  We buy probes from LEM. 

Andy
 

Offline ahbushnell

  • Frequent Contributor
  • **
  • Posts: 734
  • Country: us
Re: News from the GTI design workbench, Blog
« Reply #43 on: May 20, 2018, 03:25:40 pm »
Interesting, I didn't get a problem yet with flux walking, perhaps my primary is balanced better.
For the overload on tests, to not blow up everything, and test with no proper cooling, I usually pulse the system 10ms on, 1s off until everything is stable
In the early days of power electronics people didn't know about flux walking.  Products were shipped and most worked but some would blow up.  It took awhile before it was solved. 
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #44 on: May 20, 2018, 06:13:50 pm »
Thanks for the replies :) I dont worry to much about the walking as it's an effect rather than the cause, just pleased the current limiter catches it :) Yes the trafo is capacitor coupled, it's a half bridge but the capacitance has to be large enough to deal with the power concerned. 10mS is certainly long enough to destroy any semiconductor I know of!

The problem here is actually the control loop does not attinute Fo suficiently allowing the modulator to occasionally lock into this unfortunate situation. Anybody familier with GTI's will know the control loop is a delicate balancing act and of course having changed its responce some new issues arise.........

I should add I do use a LEM in one location but they take up a lot of space compared to a simple current transformer, I originally used the Allegro things but they are badly affected by external fields, the LEM is sooo much better :)

I remember the old days of doing 48V input push-pull with bipolars, now that had some serious unexpected saturation issues way back in the early 80's :) Worked in the end just made the trafo bigger so it could handle more abuse  :-DD
« Last Edit: May 21, 2018, 07:48:19 am by fourtytwo42 »
 

Offline f4eru

  • Super Contributor
  • ***
  • Posts: 1086
  • Country: 00
    • Chargehanger
Re: News from the GTI design workbench, Blog
« Reply #45 on: May 21, 2018, 03:46:13 pm »
why not use a simple shunt resistor instead ?
I usually put a shunt on the lower leg of the H bridge, that can be acessed and measured very very easily.
Which driver do you use ?
I use the NCP5104

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #46 on: May 21, 2018, 05:28:48 pm »
why not use a simple shunt resistor instead ?
I usually put a shunt on the lower leg of the H bridge, that can be acessed and measured very very easily.
Which driver do you use ?
I use the NCP5104
I am sorry I think we are talking different era's or technoligies, shunt resistor ? H-bridge, what H-bridge where ?
How many watts is this inverter of yours, from what input voltage and as I dont know what country your from what grid voltage ? and does it use a ferrite or iron transformer ?
« Last Edit: May 21, 2018, 05:30:32 pm by fourtytwo42 »
 

Offline f4eru

  • Super Contributor
  • ***
  • Posts: 1086
  • Country: 00
    • Chargehanger
Re: News from the GTI design workbench, Blog
« Reply #47 on: May 21, 2018, 08:34:23 pm »
Hello,

The inverter I design currently has a 48-> 310V     40 kHz DC/DC followed by a 310V H-bridge, in order to create a 50Hz 220V sinus.
output is approx 250 VA.

The current sensing is through a single lower leg resistor.

For the h-bridge, you have also shown a h-bridge in your sim schematic....
« Last Edit: May 21, 2018, 08:37:09 pm by f4eru »
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #48 on: May 22, 2018, 04:31:31 am »
Ahh now I understand :) But if you sense current in the output using a resistor don't you then have to transfer that signal across the isolation bridge to your controller on the input side ? Thats why I use current transformers or the LEM they have built in isolation and with a resistor you have the signal level vs losses connundrum.

My grid voltage is a little higher than yours around 240Vrms here in the UK, guess your in mainland Europe then ? But I have seen it rise to 250V here and fall as low as 225V thats what you get for living in a small village sharing one large 11Kv distribution  transformer on a pole, I think it's either 50 or 75Kva.

My GTI runs from ~180Vdc input and is designed for 1600W that being 50% more than my presently installed PV.

I didnt get to look at it yesterday, to much sun outside and gardening to do, trying to clear the house of seedlings and get them outdoors hahaha
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: News from the GTI design workbench, Blog
« Reply #49 on: May 22, 2018, 06:21:59 pm »
This evening I went trafo hunting, this is similar to truffle hunting and for engineers can be just as rewarding  :-DD
In this case I am upgrading the test rig to increase voltages and powers towards rated and had exhausted the present capabilities. So I was trafo hunting to increase the available input voltage and found this beauty, once part of a linear computer power supply. I used to have a lot of these and bigger but you know how it is moving houses and wives, things get downsized even the junk pile  :-DD
 
The following users thanked this post: ahbushnell


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf