Unable to wash clothes with my GS.

Well I got two more batts comeing in today, and well I'm at the moment still running the main GS, and I just also noticed, its saying right now its getting 26.2v and useing about 3a.  when I check my battery voltage with a meter it showing 25.4v. . .

seems like the GS is thinking it smarter then me?

And seems like shawn took down the upgrade c board on the web site.  and he don't have a email listed on the web site either. . .

We're here to take care of actual customers, not try to push them off as quickly as possible...so there would be no charge for the boards.  Yes, moving wires to the new board...and a little wire fabrication (a new 12AWG power wire to the new control board).  Nothing that can't be done with some basic electrical knowledge--and I don't mind helping out via video call just to double-check it's done right.

Will the new board have the relay?  If you got a picture of the finished wired board, and you saying a 12awg wire mod?  I think shawn solder one of the wire on the master, I think its the ground.

Is there any chance on sending the wire in question that is already modded with what it needs?

I'm not afraid to changeing the board my self, I would have somebody helping me, just need the photo so I and them know where the wires need to move to.

Do you want me to send the old boards back to you?

hmmm, seems like the  private messaging on here has been removed?  I used to have a spot with all our private talks saved.

Do you need my mailing addY/

will switch to PM--it's still fully here.

Well, Here is the latest updat eon the Saga of my washer.  Sid Has got more data, and we are closer to solving the issue.  Also, I was incorrect on the China invter being a LF, its a HF but, at the same time a PJ LF work just as good.  So, there might be a firmware update comeing out soon to resolve my washer, not sure if he will push it out to everybody or not.  Also dang we got my fans to go down to 11% when they spin up instead of 25%, now they not screaming all the time lol.

So, May your Washes be in your Favor!!!

Yeah, the one unit was a Reliable Power HF inverter--these having the boost voltage regulation in hardware will have a solid speed advantage over a firmware-controlled voltage regulation.

My guess about the PJ LF "running the washer fine" has to do with the PJ inverter outputs being extremely dirty (especially the side without any filtration!)  And as the washer seems to be extremely sensitive to voltage drop, the very dirty PJ output power actually causes the washer to think the voltage ISN'T sagging with all of the dirty spikes.  This means that the clean GS output power is actually working AGAINST us with that washer!

It's a very sensitive washer, for who knows why.  Was interesting to note that if <a contenteditable="false" data-ipshover="" data-ipshover-target="/profile/3-the-blind-wolf/?do=hovercard" data-mentionid="3" href="/profile/3-the-blind-wolf/" rel="">@The Blind Wolfturned on his 240v central air, the washer seemed to behave fine.  But if the central air was turned off, the washer went back to hiccuping and tripping.  I noted that the battery voltage dropped from 26.6v to 25.4v with the central air on (~1800W per inverter)--which indicates that battery voltage sag under load is a significant player in the equation.

Have to figure out how to get the inverter to react faster WITHOUT going into oscillations for certain customers.  Oh the joys....!

Also the battery bank wasn't but about %50 charge.   I ran them down and didn't get them back up fully charge due to it being a stormmy day and today is no diff. I only generated 214ah yesturday.  I got the stuff for my bib batts, to let it put out 400a, just got to drill the holes on the one cell so I can use a stud or set screw to hold the copper bars that I'm going to cut down and drill holes in to join each cell.  That might help on the sagging issue.

20 hours ago, Sid Genetry Solar said:

Have to figure out how to get the inverter to react faster WITHOUT going into oscillations for certain customers.  Oh the joys....!

If i might suggest...

Have a running standard deviation (of output voltage) over the last second (aka 50/60 samples) or so.
Use the standard deviation to calculate a sample size for the hysteresis on the regulator. (with min and max of 1 and 3)

Thus if the regulator begins to oscillate, the standard deviation will increase, causing the hysteresis to increase to compensate. 

Of course, it may require a little experimentation to define the stdev to hysteresis size formula.

On 6/19/2023 at 6:11 PM, NotMario said:

If i might suggest...

Have a running standard deviation (of output voltage) over the last second (aka 50/60 samples) or so.
Use the standard deviation to calculate a sample size for the hysteresis on the regulator. (with min and max of 1 and 3)

Thus if the regulator begins to oscillate, the standard deviation will increase, causing the hysteresis to increase to compensate. 

Of course, it may require a little experimentation to define the stdev to hysteresis size formula.

Worth noting that the processor on the GS control boards isn't capable of doing serious math at any rate of speed--everything has to be very simple.  For example, "hard" math is a 32-bit multiply (as it's an 8-bit processor).  Any sort of floating-point or trig functions of any sort--or even division for that matter--is simply not feasible at any rate of speed.  (I handle division by doing "fractional multiplication" on binary powers.)  A 16x16 multiply that has to run 24,000 times per second had to be coded in pure assembler.

There's several things to consider as possibilities for adjustment.  For one, there's a hardcode "reaction curve", which can be adjusted via firmware update...which exists to try to prevent oscillations.  It's possible this curve isn't quite right.  But it's also possible that said curve just simply isn't sufficient to handle the wide range of different system setups out there.

One possibility would be to have an "error result" register of sorts, where we could have some sort of "error result" register that indicates the result of a correction (perhaps just a threshold trigger "error before correction and error after correction").  If the error overshoots, the curve multiplier would need automatically dialed back.  And if the error undershoots, the curve multiplier could be automatically pushed forwards.

Obviously, while analytical methods such as that would not be immediate, each individual "system" would likely have a fairly specific error range (resulting from the battery type, the system setup, the configuration, etc., etc.)
One HUGE caveat I see with such a method...would be "wave scalping" appliances that put a huge load on the AC line for a few cycles, then disconnect it for a few cycles.  This would easily drive a "error result" routine completely off the road--that is, without adding significantly further complexity to also analyze the actual load.  Oh, and analyzing the load also requires determining which way power is flowing in order to properly interpret the "error result."

It starts to get extremely complicated very quickly!  Which is why I'm trying to find some simple solution to the problem.

16 hours ago, Sid Genetry Solar said:

Worth noting that the processor on the GS control boards isn't capable of doing serious math at any rate of speed--everything has to be very simple.  For example, "hard" math is a 32-bit multiply (as it's an 8-bit processor).  Any sort of floating-point or trig functions of any sort--or even division for that matter--is simply not feasible at any rate of speed.  (I handle division by doing "fractional multiplication" on binary powers.)  A 16x16 multiply that has to run 24,000 times per second had to be coded in pure assembler.

There's several things to consider as possibilities for adjustment.  For one, there's a hardcode "reaction curve", which can be adjusted via firmware update...which exists to try to prevent oscillations.  It's possible this curve isn't quite right.  But it's also possible that said curve just simply isn't sufficient to handle the wide range of different system setups out there.

 

One possibility would be to have an "error result" register of sorts, where we could have some sort of "error result" register that indicates the result of a correction (perhaps just a threshold trigger "error before correction and error after correction").  If the error overshoots, the curve multiplier would need automatically dialed back.  And if the error undershoots, the curve multiplier could be automatically pushed forwards.

Obviously, while analytical methods such as that would not be immediate, each individual "system" would likely have a fairly specific error range (resulting from the battery type, the system setup, the configuration, etc., etc.)
One HUGE caveat I see with such a method...would be "wave scalping" appliances that put a huge load on the AC line for a few cycles, then disconnect it for a few cycles.  This would easily drive a "error result" routine completely off the road--that is, without adding significantly further complexity to also analyze the actual load.  Oh, and analyzing the load also requires determining which way power is flowing in order to properly interpret the "error result."

It starts to get extremely complicated very quickly!  Which is why I'm trying to find some simple solution to the problem.

Interesting. There are some O(1) shortcuts i've come up with over the years to accelerate an approximation of stdev and avg operations. They usually involve a single multiplication, addition, and division 32-bit op. The calculation(s) need done at most once per AC cycle.

Fortunately, the numbers involved are so small, floating point is not necessary. Using a multiplier of 100 gives you 2 decimal points for free for 16-bit words.

avg = function(n, i = 0){
	var s = i
	var n = n
	this.add = function(v){
		return s = ((n - 1) * s + v) / n
	}
}

stdev = function(n, i = 0){
	var a = new avg(n)
	var l = i
	this.add = function(v){
		d = v - l
		l = v
		return a.add(Math.abs(d))
	}
}

Where n, i, v are "sample count", "initial value", and "current value".
There are 4 dynamic math operations for each stdev calculation, and 2 storage ops. Maybe even doable on your 8-bit toy cpu. 😄

Obviously, these wont compute their statistically accurate equivalents - but we just need to approximate their smoothing properties.

Food for thought, as always.

Well, I'm ordering probley another 100ah 24v head way cells these will use a straight cable to a 500a relay controlled by a 30a bms.  Had three more GS error shut downs and I've been hearing sometimes one of the GS beep three times but everything keeps going, not sure what that about, cause they are the only things I got that even make beeping sounds.

Alos I sort of did a 3 minute dryer run, and well looks like I need to see about getting a dryer that don't suck!!!!!. dang over 300a to run that thing. If that was running and my ac was going, I would almost max out my system, and thats not with the water heater running, so looks like I'm going to need to get a third GS. . . Hey sid have you figured out how to do a three GS config with 240v 😛  I might be saving up for another GS in about a year form now. going to work on my battery bank and more solar. . .

going to work on my battery bank and more solar. 

I  do not think headway cells   make a good powerwall  system .   I do not think  a dryer should be run with an  inverter or use my welder with  an   inverter  or use a inverter as a battery charger .    More solar is always good  but the connections  is expensive  ,   

3 hours ago, The Blind Wolf said:

Had three more GS error shut downs

What were the errors?

1 hour ago, dickson said:

I do not think  a dryer should be run with an  inverter or use my welder with  an   inverter  or use a inverter as a battery charger . 

Why ever not?  If you have the power and ability...go for it.  For that matter, I know of two customers in CA who utilize 24v GS inverters for powering their mobile welding rigs.

The charge/AC input limitations are PJ limitations, not GS inverter limitations.  Charge efficiency in GS inverters should be 75% or higher; a PJ should be expected under 60%.  (I find it quite funny that none of the Chinese LF inverters specify charge efficiency--probably because it's close to 60%.)

Why ever not?

My only purpose was to run   a  4 ton heat pump .   After one day of test  it can run  the whole house  and   dryer  and light welding and a 5 hp air compressor .     I have enogh  solar and separate  dedicated battery chargers  to chage the battery for use all night .    I  will use my 30000W  Powerjack  when summer is over  and no need to  use the 4 ton heat pump .     The Powerjack  produve  so much  heat thet I use it for heating  the room .    The inverter is  cool and the fan throttle is 10 percent  most of the time  .  I have  a portable  AC  to keep the inverter  cool  in the room .    

5 hours ago, NotMario said:

Maybe even doable on your 8-bit toy cpu. 😄

Next major revision control board will not be completely backwards compatible...but it will utilize a 96MHz 32-bit ARM-core CPU.

We do not have any plan to EOL the Rev. C board design; the ARM "D" design is more of a "futureproofing" in case we can't procure the U.S. made CPUs at some point in the future.  Just to have another option available "just in case."