Home
JAQForum Ver 20.06
Log In or Join  
Active Topics
Local Time 07:01 22 Jan 2022 Privacy Policy
Jump to

Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.

Forum Index : Electronics : Nano Power Inverter - Roll Your Own Style

     Page 13 of 13    
Author Message
Solar Mike
Guru

Joined: 08/02/2015
Location: New Zealand
Posts: 920
Posted: 11:09am 31 Oct 2021
Copy link to clipboard 
Print this post

Fod3182 is an excellent generic device for H-Bridge drivers
3A, 50-210ns switching, variance any 2 devices = 90ns. With your scope you could easily match them up on a plugin breadboard, prior to running them live; that variation could stuff up dead timing.

Si8261BBC-C-IP is PDip plug in replacement, Si8261BBC-C-IS is the SOIC version, both with 8v under voltage lock out.
4A, 20-60ns, variance between devices = 25ns.

The Si device has superior spec's for matching multiple drivers. Last I looked RS still have some in stock.

Cheers
Mike
Edited 2021-10-31 21:17 by Solar Mike
 
wiseguy

Guru

Joined: 21/06/2018
Location: Australia
Posts: 681
Posted: 10:07pm 31 Oct 2021
Copy link to clipboard 
Print this post

Thanks for the info Mike.  If the theory of the differences in the optos being the culprit, causing the grunt is correct, I have been playing with the figures to see worst cases.

After a short study of the data sheet & playing devils advocate on all the extremes of the Si8261BBC specs working against you I calculate a potential worst case difference between 2 units of 218nSecs.

This is small fry compared to the FOD3182 which comes in at ~530nSecs and they don't publish max and min values for rise and fall times only typical times were used for the calc. By the way the 90nsec diff part to part is actually twice as bad, its +/- 90nS.

The Si8261BBC is a nice part but its simulated reverse diode voltage is -0.3Vmax. I am using Warps inverse optocoupler drive so the board would need a bit of redesign to work to its best performance.

I reckon my best option without a redesign is to match/select the optos.  The small capacitance that is used for the dead-time which is in parallel with the optos might also need to be selected/matched, they can be +/- 20% depending on the type & specs.

I have reversed the high side optos on my inverter and will finish reassembling it today, will be interesting to see the result.

hopefully I am now on the right track to solving this grunt, time will tell...
If at first you dont succeed, I suggest you avoid sky diving....
Cheers Mike
 
Warpspeed
Guru

Joined: 09/08/2007
Location: Australia
Posts: 4384
Posted: 11:26pm 31 Oct 2021
Copy link to clipboard 
Print this post

I remain unconvinced.

This is all well trodden home brew inverter ground.
The thundering heard have been building grunt free inverters of very similar design using similar parts for a very long time.
Something rather unique is at work here, its a problem nobody else seems to have had.

It will end up being something very simple and totally unexpected.
Cheers,  Tony.
 
wiseguy

Guru

Joined: 21/06/2018
Location: Australia
Posts: 681
Posted: 07:38am 01 Nov 2021
Copy link to clipboard 
Print this post

  Warpspeed said  I remain unconvinced.
This is all well trodden home brew inverter ground.
The thundering heard have been building grunt free inverters of very similar design using similar parts for a very long time.
Something rather unique is at work here, its a problem nobody else seems to have had.
It will end up being something very simple and totally unexpected.


I'm sure not many of these well trodden thundering herds are using an inverse coupled optocoupler drive using the FOD3182 driven by the nano code (apart from maybe Murph) and with my deadtime, layout and my (ultra?) low impedance 3xsendust stacked choke/2xtoroid setup and 13+ AMP FET drivers with 2R2 gate resistors?  Obviously something in the above lot must be causing this issue.

Also when you suggested that 2 x RC filters should give 2 x sinewaves, not in the configuration we are using, this would result in 2 x half sinewaves 180 degrees apart. LHS is held at GND whilst we PWM RHS. Next hold RHS at GND & PWM LHS.

When reassembling my inverter I accidentally reversed an 8 pin connector and promptly took out 2 resistors and killed 2 optos, one that I had just swapped! - which negated a bit what I was hoping to see, I actually got out of it quite lightly.  Any way Tony as you predicted after replacing the optos & swapping the optos around did not seem to influence anything by any measurable amount, note I did not try swapping the dead-time capacitors across the optos, but I cant see them influencing much.

All CRO scans below are 1 sec/div You can see the grunt starting well before the end of ramp it typically seems to kick in at about 70-75% of soft start ramp

Typical startup energy envelope with grunt.


Reverse transformer after above capture soft stopped.


Second soft start after transformer reverse - back to normal shape.


Run same test with 80 milliOhms of series R


Run same test with 200 milliOhms of series R


With more than 100 milliohms of series resistance in the choke transformer primary circuit the envelope looks pretty normal and there is no sign of grunt.  This outcome is obviously quite unacceptable.

I have exhausted my current ideas & have no idea what to try next, so for now I will shelve it for a few days, time for some alcohol maybe lots of it !
Edited 2021-11-01 17:40 by wiseguy
If at first you dont succeed, I suggest you avoid sky diving....
Cheers Mike
 
Warpspeed
Guru

Joined: 09/08/2007
Location: Australia
Posts: 4384
Posted: 07:53am 01 Nov 2021
Copy link to clipboard 
Print this post

Best just to walk away from it for a while and relax a bit.

Some fresh thoughts and ideas may come to you.
Cheers,  Tony.
 
Murphy's friend

Senior Member

Joined: 04/10/2019
Location: Australia
Posts: 230
Posted: 09:32am 01 Nov 2021
Copy link to clipboard 
Print this post

Mike, could it be that you made your inverter a wee bit too complex?  

Below are two schematics of my 6KW opto isolated inverter which I built to your original schematic but with quite a few parts omitted.
It is completely silent on start up or shut down (I never start it with any load).







I do not have a CRO capable of seeing tracces like the ones you posted BTW.
 
wiseguy

Guru

Joined: 21/06/2018
Location: Australia
Posts: 681
Posted: 11:06am 01 Nov 2021
Copy link to clipboard 
Print this post

Hi Murph, The main difference between my build and your (and others) build as I see it is the choke.  Your power stage certainly looks very simple when you omit the other dozen or so FETs  

Ok I admit to some "complexity" but they are not related to the issue of the grunt, they give me some extra features that I wanted (no grunt does not count as a feature)
You also have an active filter in yours but as I've tried two totally different feedbacks with identical results I'm sure its not that.

Did I take too many turns off my main toroids, from memory I wanted 235V out from low battery of around 44V, and that was the winding ratio I used ~ 1:7.5. The gain is wound back a fair bit running from 48V I guess open loop it would be around 255V maybe my feedback is still too slow and it over drives briefly.

I love Sendust chokes and how efficient and cool they run with 50 - 100kHz+ frequencies it just laughs at you. That is when its not grunting apparently.......

If I had used Iron and wound lots of turns I suspect I would not have a grunt. But my challenge now is how to make them work for me.

I'm not beaten yet but I do need some space from it for a couple of days so I can regroup and tackle it again.

Starting an inverter unloaded giving it a short spell before connecting the load is very wise.  I am thinking of more complexity and adding a zero cross type relay so it connects the load at the zero crossing. (it then shouts "suprise"). Then a relay of course to bypass the zero crossing losses - dont you just love complexity haha.
Edited 2021-11-01 21:10 by wiseguy
If at first you dont succeed, I suggest you avoid sky diving....
Cheers Mike
 
Murphy's friend

Senior Member

Joined: 04/10/2019
Location: Australia
Posts: 230
Posted: 01:48pm 01 Nov 2021
Copy link to clipboard 
Print this post

  wiseguy said  Hi Murph, The main difference between my build and your (and others) build as I see it is the choke.  Your power stage certainly looks very simple when you omit the other dozen or so FETs  



You are right, drawing in those extra mosfets does mess up a simple schematic so I left them out.
FYI, the 6KW inverter has 4 parallel mosfet strings (16 units) and the 3KW one only 3 (12 units) in parallel.
I use removable little PCB mosfet carriers nowadays, have blown up way too many to solder them directly to the big PCB any more. The little circles in the schematic shows where they connect.
Funnily, ever since I used yours and warpspeeds inverter ideas the blowups have vanished, its definitely a winning design as far as I'm concerned.

I  use a re cycled Aerosharp choke core.
 
wiseguy

Guru

Joined: 21/06/2018
Location: Australia
Posts: 681
Posted: 11:44pm 01 Nov 2021
Copy link to clipboard 
Print this post

Murph, for the value of others thinking of copying your circuit, I may have said this to you before, but the correct value of the 4 resistors from the gates to the LEDs should be no higher than 150R for VCC 5V.  The data sheet specifies a recommended current of 10mA despite their typical 3 - Maximum 7.8mA threshold on current at a forward voltage of maximum 1.8V.

The value is determined by ((5-1.8)/.01)/2 = 160R so 150R is the nearest preferred resistor values and resulting LED current is 10.6mA.  However there is a further consideration, the recommended SN74LV86A data sheet suggests that guaranteed output voltage at VCC 5V and supplying 12mA will be ~4.3V.

Taking that into account now the sum becomes  ((4.3-1.8)/.01)/2 = 125R so 120R is the calculated worst case recommended resistor for the 4.  If you get a best case  FOD3182 opto/s with 1.1V min forward voltage the LED current for 120R would be ~ 13.3mA.  Well within their recommended operating range of 10 - 16mA.

In my much simpler nano Pico board, which is a work in progress - its essentially finished except I am holding off releasing it until I am happy it doesn't contribute to the grunt, there are little TO92 PNP & NPN buffer transistors at each gate output, which can be fitted so more feeble gates like the 74HC86 series can easily drive the opto LEDs. If the SN74LV86A is used though they would not need to be fitted.
If at first you dont succeed, I suggest you avoid sky diving....
Cheers Mike
 
phil99

Guru

Joined: 11/02/2018
Location: Australia
Posts: 341
Posted: 12:12am 02 Nov 2021
Copy link to clipboard 
Print this post

Some more things to take on your Contemplation Retreat.

Zero crossing switching is good for resistive loads, ideal for capacitive loads but worst case for inductive loads. For them lowest inrush current occurs when switched on at the peak of the voltage in either half cycle. So it depends on what the biggest load is likely to be at startup. A reasonable compromise is around the 135 degree mark (2 to 3mS after Vmax).

The "trim" experiments showed that the grunt can be altered. So perhaps dynamic adjustment of the trim in response to feedback of the primary current might actively suppress the grunt. By comparing the peak current of the current half cycle with the peak of the preceding half cycle and applying a trim of 1 unit (in the appropriate direction) if the difference is more than say 10% or by 2 for 15% etc. it could prevent the current peaks growing excessively, without affecting normal operation.
 
wiseguy

Guru

Joined: 21/06/2018
Location: Australia
Posts: 681
Posted: 09:00am 02 Nov 2021
Copy link to clipboard 
Print this post

Thanks for your comments Phil & I agree totally with your thoughts.

In my house, apart from the dreaded compressor, all fridges and other loads are power factor corrected type loads and in all honesty a compressor is the least likely item to be used whilst on inverter power & even so it has proved in the past that as long as the inverter is started and running THEN turn the compressor on it has always worked perfectly.

It initially took a lot of hard thinking to get my head around maximum current at zero crossing but it eventually gelled despite being a bit abstract.

Whilst doing the mental Gymnastics if the inverter primary is supplying current which at the peak of the current sine is actually the zero crossing of the 240V at the secondary what does that mean then if we put an inductive load on the output of the inverter, does that mean that the peak current of the load now corresponds with minimum FET current - AC thinking hurts my head.

I also agree with your comments about predicting and compensating based on primary current, but I would need to be a software programmer extraordinaire to achieve that. I am wondering what I may be able to do in an analog/discrete fashion, that is how I have got by over many years of being software challenged.

Maybe I could make a discrete trigger delay based on a number between 1 & 256, that way I could set the pulse number I wanted to see in the train for each 180 degrees and compare it to the corresponding pulse of the next 180 degrees to see how well they are matched.
If at first you dont succeed, I suggest you avoid sky diving....
Cheers Mike
 
Murphy's friend

Senior Member

Joined: 04/10/2019
Location: Australia
Posts: 230
Posted: 10:37am 02 Nov 2021
Copy link to clipboard 
Print this post

  wiseguy said  Murph, for the value of others thinking of copying your circuit, I may have said this to you before, but the correct value of the 4 resistors from the gates to the LEDs should be no higher than 150R for VCC 5V.  The data sheet specifies a recommended current of 10mA despite their typical 3 - Maximum 7.8mA threshold on current at a forward voltage of maximum 1.8V.



You are quite right, I have corrected the values on my board but neglected to do so on the schematic. Call it a senior moment..
Or was it done to giver those copy cats a tough time?    
 
     Page 13 of 13    
Print this page


To reply to this topic, you need to log in.

© JAQ Software 2022