Sine wave power inverter

[psdayama]

Hi Saugata,
I think that for controlling the amplitude a hardware approach is better
as so many delays and errors can creep in. There is a Buck Boost circuit with
two mosfets and an inductor and few diodes. The DC input to H bridge can be
controlled with B/B circuit with reqd. time delays generated by Micro output(which doesn't have to be so fast as changing per cycle of sinewave)
The shape and frequency then can be generated by normal 2 PWM outputs(Actually one is enough just route the output to another branch with a gate and this also have dead time built in).
I still wonder why anybody is not responding about my querry of transformers. The output of Inverter is serries of pulses at high frequencies with polarity reversal at 50-60 Hz. So is the transformer used designed for
50 Hz operation or high frequency like say 20KHz? There is huge difference
in size of transformers.
regards,
purushottam dayama

Hi Suded,

If I have understood right then you are suggesting that I build a hf oscillator around the SG3524 then use the PIC18F452 to supply the closed loop control voltage to the SG3524 (the PIC HWPWM acts as a DAC).

Now the issues are,

1. I already have two HWPWMs built into the PIC

2. The SG3524 allows dead-time programming through a simple resistor but while switching phases I again need a few logic chips to get it done.

3. I have designed the software around interrupts. I am doing an AD-Oversampling which bruce, tim and melanie suggested in the MELAB forum.For a pulse by pulse current limiting (so that the MOSFETs don't blow up) I am using a comparator to sense the current and generating an INT to turn off the MOSFET drives.(Heh! Heh! a digital copy of the SG3524 current limit scheme).
So why use the SG3524 as the ad-sample and dac conversion will add up to the loop error.

The only thing that concerns me is the THD introduced while driving inductive loads as my routines are not so fast to correct that in realtime. I tried but the results are erratic (thats because of lack of knowledge from my side). The PIC is running at 40MHz (10MIPS,10MHz@HSPLL). Now I am concentrating on reducing noise on the AD line, using interrupt based AD (PBP tends to waste time while doing AD).

BTW can anybody shed some light on the time consumed by LCDroutines. Because while processing interrupts with PBP I found glitches while the LCD was being updated.

Thanks to psdayama and doctor for the links. As far as PSPICE goes I am not sure whether my knowledge is sufficient to handle them (I don't have an engineering background)

Regards
Sougata

[sougata]

Hi Dayama,

You are right that only a single PWM can be used and done exactly the same. It is modulated with a sine table (32 samples). The transformer needed is a 50Hz type. The transformer, output choke and capacitors form a low pass filter (actually the sinewave is integrated) and the output waveform is a sinewave.

Your approach of using a B/B converter would rather complicate things up and also be costly. That is basically a dual conversion design if I understood right. With a MCU you can tune , upgrade your design without actually increasing the production cost. If you are in India then you may be aware of the latest buzz of "DSP" sinewave inverter. Here the approach is same.However a DSP gives you much better ADs and fast computation with dedicated hardwares (for e.g a 32 x 32 Multiply) so you can keep your sinewave undistorted with inductive or mixed load.

I could not post the whole code due to professional restriction but here is what it does :

1. On Interrupt it fetches the sinetable vaue (indexed) for PWM

2. Offset it with ERROR from previous and current sample (a PID style error loop)

3. Turn on the MOSFETs with the calculated value.

What I don't do is dynamically correct a voltage error in the same index pointer. My ADC conversion time does not allow me to do that. To mess up things more I actually take 4 samples and do and average to deal with noise. Then I have a nifty LCD routine, a battery monitor, RS232 interface (to shutdown my computer and inverter when not needed)

BTW where is Pramod who kicked this thread into existence again.

Regards

Sougata

[sougata]

Hi Dayama,

It is possible to reduce the size of the transformer. Looks like a class-d amp solution to me. A simple filter would reconstruct the sinewave and remove the HF component.Did you find anything or anyone having done it on the net. I will also search and probably would like to develop one. But some practical start point is needed.

Regards

Sougata

[psdayama]

Here is the link:They make readymade PWM amplifiers but don't ask cost!!!!!
http://eportal.apexmicrotech.com/mai...m_overview.asp

I asked their reps in Mumbai and they said each IC costs Rs.25,000/-
But one can make PWM output like that and then have just High freq. txr or
direct output with filter.
regards,
Dayama

Hi Dayama,

It is possible to reduce the size of the transformer. Looks like a class-d amp solution to me. A simple filter would reconstruct the sinewave and remove the HF component.Did you find anything or anyone having done it on the net. I will also search and probably would like to develop one. But some practical start point is needed.

Regards

Sougata

[psdayama]

Hello everybody (Saugata,Mel xxxx)
I wonder why there is no activity here since last month!
I have just now made a inverter of Delhi type with sinewave output.
The transformer for 500VA output is whopping 10kg in weight.
You all may be discussing sine wave with PIC but practically
there is no use if such big transformers have to be used and adding
to cost about 60% of the project.
I have hit an idea to circumvate this and will discuss if anybody
is interested.
HAPPY DIWALI!!
dayama

[sougata]

Hi Dayama,

I actually wrote the code for a customer's specific requirement and know any further development is entirely personal and my current schedule does not permit actually working on it. Currently I have a running code with fair sinewave output. However THD measurements and transient response characteristics were not done. Also to use it like a UPS I need to design a fast transfer circuit. The sugar cube relays are fast enough but fails with increasing and mixed type load. I have incorporated an interrupt based AC Fail detection circuit that checks the following:

1. Improper Line Frequency
2. Unhealthy AC-Line Voltage
3. Sudden brown/black out

When I compl



As far as the transformer issue goes I beleive it will take much time and effort which is beyond my scope right now. Also in Kolkata it is very difficult to source good quality ferrite cores.

As I told you previously we need to have a high frequency carrier and modulate it with a 50Hz sine table. (Which I am doing anyway). Use of the high frequency should theoritically be used to drive a ferrite cored transformer (read low cost light weight) so a low pass filter in the secondary should reconstruct the modulating 50Hz.

BTW I am attending the Microchip Masters Seminar at Lonavala on 29th nov. are you?

Thanks for keeping this thread alive and giving me a kick to ponder further.

Regards

Sougata

[psdayama]

Hi Saugata,
Its OK to be late than never. I think U should stop the work on high frequency
as there are more complcations. Instead I suggest that whats Ur code for
current work is alright.
I am planning to make inverter with Half Bridge
output which is direct to Load and not thru transformer. Problem is everybody is inverting at low frequency of 50Hz from battery. And use same
transformer for charging battery when on mains.
If we make high frequency >40 Khz inverter with
square wave and output is rectified and made high voltage DC then with dual supplies (+/-) we can use IGBT/MOSFET half bridge to drive sinwave output
directly without transformer. This will save the output transformer cost and ferrite transformer would cost hardly 100/-. Also for charging battery same type SMPS circuit can be used and its wattage can be max 120W which would cost hardly 200/-.(Old PC SMPS 250W are available for 150/- in market)
I am still designing the circuit. I will post the block diagram here as soon as
it is ready.
When you are working with this type circuit U can have slightly expensive
circuit which doesn't require a relay -sugar cube or any type as U are just switching DC sources with help of diodes and they are very fast!
Basic plan is
MAINS POWER---|----->SMPS WITH CHARGE CONTROL--> BATTERY->-|
|_____>RECTIFIER --->|<RECTIFIER--<SQR.INVERTOR<|
|
50HZ PWM INVERTOR WITH HALF BRIDGE OUTPUT
|
LOAD WITH SMALL FILTER

This ckt will work like true online UPS still expensive just as much as ordinary
invertors.But U can have Ur DSP alright and also PICs other a/d lines can be used for battery charge management as well!
For cheaper type the Mains rectifier for dual supply can be ommited and
instead sugar cube relay used for switching outputs. Half wave +/- rectifier
requires costly HV elcos.
For good ferrites ask for dealer of Cosmo Ferrites. U check their list on web page. There is another manufacture Hinodaya.
I won't be attending M'Chip Masters course as I am just beginer in this field but not to programming. May be I will attend next time when I hope to get some nuances and I am working on Transformer Ratio test set(comparator
type) and want to use PIC for advanced features.So I think that I will also be
familiar with PICBASIC then.
Wish all the best for conference.My place is abt 2hrs from Lonavala
regards,
psdayama

[sougata]

Hi Dayama,

What you are proposing is a dual conversion design. When you have a battery bank of greater voltage it makes sense other wise for a 12/24 volt system consider this.

1. The Primary current will in the order of 50 AMPs so it is necessary to use copper bands for the primary windings.

2. Frequency has to be kept high (ferrites) consider the switching losses of the primary MOSFETs

3. You have two different switching stage so a single PIC may complicate your programming issues.Of course you can use a ready made SMPS controller like the SG3525 but then this forum may not be the right place to discuss.

4. Microchip has a very good appnote on such type of design.

If you are using PBP then this forum will help you out in solving problems.

One of my laboratory power supplies runs on a PIC for a Buck regulator.

We get only a limited supply (not all models) of cosmo ferrites so it is beyond my scope of experiment. However I would definately like to meet you while I am in Mumbai.Send me a private email if you are interested.

Regards

Sougata

[barkerben]

I made a 24v, 250W Quasi sine wave inverter using a basic PIC (can't remember the exact details) a couple of years back.

It used quasi sine rather than full sine - in other words, a square wave with dead time. Having said that, regulation was less than 10% at full load, and efficiency was over 95%. Quite good fun really...

[robinjacob]

Hi Sougata,
Have you finally designed the sine wave inverter? I am really interested to see the design. If you can please send me a schematic.

Thanks and Regards
Robin Jacob.

[mincing]

Dear Souhgata,

Thank you for giving so much of info.What was the ratings of UPS you have made.and what were the efficiencies.
Could you please help with Transforner design information.

Also I have seen a round core trnsformer some ............

also if you are using full bridge topology how do you drive the bridge..........

If you could help with LLC filter inductor design and transformer ,that would be really great.

Thanks in advance

[Charles Linquis]

Many years ago I made a 400W . 120V *PURE* sine wave inverter using a PIC. The input voltage was 48. I used a normal inverter (with a ferrite transformer) to give me a regulated 185VDC to feed the PIC-based H-bridge sine wave section.

I used high-side drivers and FETS on the "top" two legs of the bridge and ordinary (low side) drivers and FETs on the lower two.

The high-side drivers can't be "PWM-ed". So I alternately turned them on (one for each half-cycle) while using 78Khz PWM on the lower two. Simple CMOS gates prevented "punch-through" by not allowing the upper and lower FETs on the same side to conduct simultaneously. A small (47uH) inductor was in series with the load and a .1uF was across the load. A scope showed the output to be a perfect sine wave with no glitches whatsoever.

The PWM value was from a lookup table (the program was written in assembly).

The switching FETs needed only a small heat sink.

[babalola]

hi guys,

sine lookup table is a vey important issue as far as digital control of inverter is concerned,can any of you expert treat us (we the beginer )in details the basics of creating this important table?.there is a formular used in microchip reference design,

vref= vp* sin(2II*K/N).
my concerns are:

1. What inform the choice of N( no of samples) other than the niquist sampling theorem?
2. how far can we go in the choice of this value?
3. what is the relationship between this no of samples and the total harmonic distortion of the sine wave?
4. is the above formular valid in all cases?

i dont know if this topic has been mentioned in the past, but i believe detail understanding of this important topic will assist greatly who will become a power inverter designer with microcontroller.

regards,
babalola

[EMTRON]

Dear sougata, am very grateful and i appreciate you for your help to every one on this forum. am also happy to join this reportable organization to improve my knowledge in electronics and PIC programing field.thanks for acceptance.

[Panzer]

Hi Saugata,
....
I still wonder why anybody is not responding about my querry of transformers. The output of Inverter is serries of pulses at high frequencies with polarity reversal at 50-60 Hz. So is the transformer used designed for
50 Hz operation or high frequency like say 20KHz? There is huge difference
in size of transformers.
regards,
purushottam dayama

Hi psdayama,
As the inverter is working as D class Amp - the end of the MOSFER matrix must be smooth with low-pass filter. Then you will have a nice sine wave only 50 Hz - and the transformer will work nice.
If you are trying to combine the transformer to be puss one over the 20KHz then will become some more complex and for sure the PIC as a 8-bit machine will be not enought.
The usual case is using a schematic of push-pull converter (as 3525) for rising the voltage level to 30% over the output one and then to converted to AC with MOST matrix only.
This way will save you a lot of pain in the ass.
WBR
Panzer