I never considdered name brand componants .... and yet I would make that choice carefully for just about anything else.

I will try the mods on the 317 thanks.
It definatly sounds like it will help without causing any additional issues.

Any thoughts with respect to sensing the AC voltage at U1b. I have been wondering if using the full Stator (240VAC when it right) rather than the 10% tap. Do you think that would make any diference? Is the 10% tap likly to be behaving as a shunt in any way and therfore showing some current waveforn superimposed on the native 10% of the overall output voltage.

Al

 

OK so staying with a linear aproach to power supply, using a BJT in place of the MOSFET, I need to reduce the overall current consumption so ....

Increasing the LM317's output divider will do just that. Since your MOSFET gate current supply voltage really isn't terribly critical, you should be OK.

How about a totem-pole from JFET's, or even smaller MOSFETS.
Any thoughts ... I will be ordering from somewhere anyway so a couple of extra componants makes no diference.

Just consider this as a band-aid to something better in the future. It would be nice to work with the lights on, right?

 

I never considered name brand componants .... and yet I would make that choice carefully for just about anything else.

Without lots of time for testing for suitability, failure rates, etc. - you're really better off to stick with name-brands from established manufacturers. They have a reputation to uphold.

Any thoughts with respect to sensing the AC voltage at U1b. I have been wondering if using the full Stator (240VAC when it right) rather than the 10% tap. Do you think that would make any difference? Is the 10% tap likely to be behaving as a shunt in any way and therefore showing some current waveform superimposed on the native 10% of the overall output voltage.

Just use the tap. That's how it was originally engineered, so why would you want to try something different?

The 10% tap will already have a relatively low voltage on it.

On that line of thought, one really easy thing you might consider is to use a DPDT relay with coil rated 220v energized by the output of the genset to switch the MOSFET supply input from the output of the aux stator to the 10% tap winding. It'll take maybe 50mS-200mS before the relay contacts to stop bouncing, but as long as you still have >10v on the output of the LM317, you'd be OK.

When the main output voltage drops below the relay's threshold, it'll automatically switch the MOSFET gate circuit power input back to the aux stator.

 

Here's what I'm talking about:

Lousy choice for a simulation name on my part.

Basically, it's just adding a DPDT relay to switch where the power's coming from. V2 represents the 10% tap from the genset. V1+V2 are the 220v out.

When the output starts coming up to the relay's rated pull-in voltage, the contacts switch over.

I haven't simulated it; don't have the correct syntax for the 2nd set of contacts for the relays, and am out of time.

 

Realy apriciate the ongoing help ... I have water again
(Despit things running a little hot.)

I'v rambled a bit below thinking as I went I supose.

I like the whole dual supply idea ... Any reason I cant have two regs feeding the same rails and set the one fed from the 10% tap to a higher voltage so that the first one shuts down when the rail exceeds its set-point?

For that matter how about just taking the reg off the MOSFET source all together and sizing the zener so I get a maximum of 9V at the source.
Then put the existing 317 on the rectified 10% tap delivering say 11V to the same rail and the MOSFET would turn off anyway wouldnt it?

I think I'd need an aditional diode to protect the 317, by the time you factor that in I cant help wondering if a 7812 wouldnt be a better choice with the diode drop that wpuld give me an 11.3 rail right?

Perhaps I just thought myself into a silly place ... back to facts:-

I have attached some aditiona linfo.
The circuit now with the changed devider on the reg. (Green cap not fitted yet)

With the welder as the load circa 50A into the batteries ...

Scope trace showing the ripple I need to get rid of on my sense input Ub1

Scope traces showing gate voltage and freequency.

Scope traces showing clean gate egdes.

Q1 was running hot 60 / 70C I guess but ran for 15 mins or so without any further issue.

With the pump as a load, another 15 mins or so, the PWM is much closer to 50% and Q1 runs significantly cooler.
(I estimate the pump is about 1KW)

Q2 runs cold all the time WoooWooo.

I am getting there folks and its down to your collective efforts.

As always looking forward to any and all comments

Al

 

My latest thoughts on PSU after your comments on dual supply.

I still intend to make Q1 a BJT but the theory is the same, I think

Am I just missing something or will this solve the problem, startup via Q1 subsiquent supply via the bridge on the 10% tap with Q1 forced off naturally.

I havnt drawn them in but I would be using the caps and choke suggested by SgtWookie.

Al

 

Hi again folks.
My genny for started up after being off for 18 Hrs. Due to the loss of any residual magnetsme this is probably as low as the startup voltage gets and it worked just fine.

I still need to solve the hot PSU issue but I think the posts above will address that one way or another.

SO ..... Last significant issue, and its a big problem:-

I am using the 10% tap on the main output stator to sense the output voltage.
I need a stable DC level without ripple to feed the PWM comparator with. Simply adding a huge cap with a smaller bypass isnt going to do it because the overall responce is then too slow and the system oscilates.

I have played with loads of values and what is there now is as good as I can get it but I still have way too much ripple and its worse with the inverter charger as the load.

If I need to do something active so be it but what?

I have a few ideas but no experiance or background knolidge that is any use.

I'm stuck and strugling folks, having to adjust the existing circuit every time the load changes significantly.

I realy need to solve this ASAP, as always any and all input will be much apriciated.

Al

 

I'm posting from work so I can't spend anytime researching this.

I remember something about an op amp set up to act as a capacitor, one that has quick reponse and doesn't load down a line drawing a large current at startup, but otherwise acting like a large capacitor to smooth out ripple in a DC voltage.

Might that be something that could help?

 

That's another new concetpt .... but I have to say it sounds like it's worth a look.

I had wondered about using opamps to isolate the ripple and then sum its inverse back in, somhow ....
That's a bit beyond me right now though and as I dont even know if it would work I suspect tinkering without some basic idea of how to do it isnt goint to help much.

Off to Google simulated capacitor right now thanks.
Al

 

Forget what I just said. I think you could do this with an active low pass filter. Design it for a LOW freq, and it should give you more of a slew voltage than an actual ripple, anyway it should make the change slow enough to stop the jumping around and keep the reference alot more stable.

If you go further with this idea there is a ripple cancellation filter design that I found a pdf on, but GoodGrief the author was Math happy. I think it might have been a thesis or something.

so anyway here is a pic without values to give an idea of what i'm on about.

 

Interesting, thanks.

I think I will chuck this at a breadboard and play with the concept.
I wish I knew how to calculate the impedance of an RC network, more research I think !!!

Any pointers folks ....
Would I calculate the caps first and then treat them as resistors at a given freequency?
How about diferent waveforms, that must have an effect right?

Perhaps I should learn to use LTSpice ... where to start?

Al

 

Google "LTSpice download".
You should arrive at Linear Technology's download area. You don't have to sign up for newsletters, but many of them are quite interesting.

Join the LTSpice User's Group on Yahoo! Groups. Lots and lots of models for you to download and install.

Like anything else, there is a learning curve involved. The software isn't exactly intuitive, and has a few quirks to it, but you get used to them.

I have had no time in the last few days to comment; I've been very busy, and today is another very busy day.

 

I have downloaded LTspice and havnt worked out how to get started yet.
The Group on Yahoo sounds like a good plan, thanks.

You have been a huge help Sgt, thanks so much, you definatly shouldnt feel the need to explain that you have other things to do apart from educating me.

I am about to start a thread asking very basic LTspice questions in the general chat section. I have looked at several tutorials but they have all started from the premmis that I have some idea what I am doing .....
I am sure there are loads of folke like me who want to learn the basics but are also flowndering.

Al

 

Hi Al, Only new here but you are on a simmilar mission that I am. My prob is simmilar & ive managed to disasemble the AVR module & hopefully drawn out the circuit corectly, but need help with un known value of some resistors in the circuit as they were broken in the dismantling proces. Also need a more knowlegable person to cast there eye over the circuit to see if its viable or there is any mistakes. Daryl

 

This AVR looks identical to the one on my genny, despite being from a smaller unit.

I have a couple of observations and comments but please remember I am not an expert, in fact I’m not that good at electronics in general particularly BJT’s, so these are just opinions and could be way off base.

C1 can’t be where its currently drawn, it would explode. Given that its value is probably correct I suspect it is somewhere the other side of R1.

The maximum power dissipation of D2 will shed some light on the value of R6, which I suspect is probably a relatively high wattage resistor anyway.
My first attempt at building an AVR had the MOSFET gate driven like this.

The transistor turns the MOSFET off when it is on and since the MOSFET would fail if it were in any state other than fully on or fully off the base of the transistor must have some form of hysteric drive.

The divider R2-R3 probably provides this changing bias via C2 during the MOSFETS transient phase.

The network R1,C1,R5,VR1,D3,R4 is clearly providing a reference fed back form the main stator, but I am unsure exactly how.
It could be a DC level, in which case the system would oscillate at some unknown free running frequency but I think it is more likely that the 100Hz ripple is being used as a drive for the system.

As always folks … Any and all comments welcome.
Al

 

Thanks for the reply Al, You could be right about C1, D2 looked about .5 Watt Zenner in physical size. I suspect R2 & C2 are a feed back to make the circuit oscilate? as the Mosfet needs to be switched on/off. The 30V dc is for the voltage regulation as this voltage varies with load. Im of to bed now as its nearly midnight here. Daryl

 

I am going to try and simulate this in Spice, just the sensing bit. I think that D3-R4 is effectivly shifting the level of the signal at R1,C1,R5 without reducing the magnitude of the ripple.

If I get Spice to give me anything, I'm good with it, I wiil try and have a post ready for when you get up. Twelve hours of difference might make this conversation a litle disjointed.

I am hoping that the sesnsing portion of this circuit can be utilised to drive the power board I already have without the need for my PWM implimentation.
If I cant get Spice to work I will have a go physically on breadboard!
Al

 

Hi Al, Its another day Have re drawn circuit as you were right the capacitor was in the wrong place. Also posted pics of both sides of the PCB, as I have decided to clear it of all parts & fit new parts, mutch easier & cheeper to do. Daryl

 

Hi Al, Great sucess to day, decided to remove all components & start again with mostly new parts ( If you replaced all parts it would only cost $14 Aust ) The new circuit is now working well, changed the C1 capacitor which was faulty with a 35V cap as the voltage there is 17V. The Zenner diodes are 1W, R1 is 1W & R6 is 2W. The voltage only varies from 245V no load to 240V with 700Watt load & nothing gets hot on the circuit board. Very happy with the results. Might be worth copying for your Genny. Daryl

 

I'm impressed, nice job.

How did you get the board so clean? the potting on mine is like rock.
I also have several burned tracks but they should be relativly easy to repair.

I have been playing with this for weeks, on and off.
The rectification and field switching circuit I have now seems to be working well, even with my inverter based welder attached.

My origional AVR ran hot from new and the generator output was never that smooth. It didnt look too bad on a DVM but small spikes and dips were visible with a light bulb, never mind a scope.

I think I am going to take the sensing portion of your circuit and see if I can build something simmilar to drive my power board in place of the PWM I have now, it will be interesting to see if the spikes come back.

I am hoping to improve things with the new AVR and then add under and over volt protection, but thats another project.

Glad your sorted ...
Al