Future Feature Discussion

34 minutes ago, NotMario said:

Sort-of. You guys have put a lot of effort into features that strike me as being awfully similar to grid tie. (Parallel mode, GTM) Were i able to ask for a pie in the sky, i'd love to see a grid-tie mode w/ anti-islanding and variable output limiting. I don't know for certain this is possible, but those features (particularly parallel slave) seem tantalizingly close in my mind. Just using a user-defined setting rather than reading a frequency shift from a master - if i'm not oversimplifying. The capability of using an inverter as primary, and for augmenting grid power would be amazingly useful and would likely broaden your customer base. (Even if you do have to rewire it) Of course, even if you were able to do it, i kind of doubt it would be a release that came out this week!

(...trying to hide astonishment at clear logical thinking that is based on realities 🤯)

What you describe is totally and completely possible...in other words, an LF inverter "grid tie" setup is technically completely possible (even if not fully practical!)  However, the number one issue is legalities with municipalities and the fact that we aren't U.L. listed yet. 

I'm not sure of the easiest/best way to handle anti-islanding, as it's a real doozy of an issue!  ('Miss' a half wave every second or so??  Put a forward error on the frequency sync, and if mains doesn't hold it back, it'll exceed 62Hz and shut itself down?)

However a "true grid tie" setup isn't of considerable value, as the inverter will have to shut down when the mains goes down.

What likely is going to be the first step would be a "zero export" methodology intrinsic to the inverter.  In short: loads are connected to the inverter output, mains connected to the AC Input.  Based on the battery voltage (i.e. available power), the inverter could "regulate" the power UP TO the current output loading (zero export meaning an electric meter will never "run backwards").  If the battery voltage falls below a user setpoint, the inverter will just run in pass-thru mode (i.e. not putting any power "into the grid").  And in the event the grid goes down, the inverter will just drop back into "inverter" mode, powering the loads until either AC mains comes back (UPS mode), or the batteries run out (UVP setpoint).

Yah, that won't be in 1.1r6, that's for sure 😉.  Maybe a bit later, after I completely redesign the WiFi support and pages...

34 minutes ago, NotMario said:

Also, your "Output OFF" setting doesn't work on my firmware version. The only way i can turn output off is in the System mode setting.

Fixed in the 1.1r6 I hope to release this week 😉.  Assuming you have a Rev. C board on 1.1r5??

34 minutes ago, NotMario said:

FWIW, the GS's efficiency didn't seem radically different. What efficiency would you expect out of a GS?

I'm hoping for mid-80s out of the current iteration (i.e. typical run of the mill).  Don't have any actual measurements, though...

A 10% difference in efficiency is hugely significant in the heat generated in the realms these inverters run (80% efficient will generate DOUBLE the heat of a 90% efficient unit).

4 minutes ago, Sid Genetry Solar said:

(...trying to hide astonishment at clear logical thinking that is based on realities 🤯)

What you describe is totally and completely possible.  However, the number one issue is: legalities with municipalities and the fact that we aren't U.L. listed yet.  An LF inverter "grid tie" setup is technically completely possible.

I'm not sure of the easiest/best way to handle anti-islanding, as it's a real doozy of an issue!  ('Miss' a half wave every second or so??  Put a forward error on the frequency sync, and if mains doesn't hold it back, it'll exceed 62Hz and shut itself down?)

I almost mentioned legality concerns! Of course that would be a problem, lol. I know absolutely nothing about that. 🙂
(The remainder of my post here is ignoring this legitimate problem, since this is hypothetical anyway)

Yeah, not sure what the standard is for anti-islanding delay. You are definitely the expert here, but the second option sounds preferable to me. This would slightly distort the sine wave, wouldn't it?

4 minutes ago, Sid Genetry Solar said:

However a "true grid tie" setup isn't of considerable value, as the inverter will have to shut down when the mains goes down.

Interesting how people see things differently. So for one thing, it would be useful for so-called net-meters - if you have abundant renewable power to sell to the grid.

In my case, during the summer i can generate solar power literally 24 hours a day - and summer is when i use the least power - i simply cannot use it all!

It's also useful in the opposite case. During winter, when i get 2 hours of daylight, i generate enough to "offset" my power bill a little bit. (though you're right you can do a passthru configuration instead - see next quote) However, i do like the idea of minimal wiring - only one circuit required for grid-die. And in a pinch, you could wire it in such a way you could flip a few breakers to use the AC-out instead of AC-in (*gasp* nasty, i know)

4 minutes ago, Sid Genetry Solar said:

What likely is going to be the first step would be a "zero export" methodology intrinsic to the inverter.  In short: loads are connected to the inverter output, mains connected to the AC Input.  Based on the battery voltage (i.e. available power), the inverter could "regulate" the power UP TO the current output loading (zero export meaning an electric meter will never "run backwards").  If the battery voltage falls below a user setpoint, the inverter will just run in pass-thru mode (i.e. not putting any power "into the grid").  And in the event the grid goes down, the inverter will just drop back into "UPS" mode, powering the loads until either AC mains comes back, or the batteries run out (UVP setpoint).

The only thing i dislike about this is how much extra wiring this can take. (in my case probably 1000$ worth of wire alone - and a whole lot of gray hair caused by cussing at 6-AWG wire for a week) It also obviously limits how much power you can use because it has to go through the inverter. You could limit this to a single subpanel of course, but then you can't augment the other panel.
I assume this method could also run a positive export (legalities notwithstanding). Essentially what you're doing here is a "grid tie inverter with limiter sensor" all in one box. You could use additional setpoints for backfeeding as well.

This could also be accomplished by having a wireless sensor (hall sensor behind your meter) that would use the inverters web API to dynamically modify the export current limit, then you can do a "zero export" (or whatever target you want) without the additional wiring. Granted wifi isn't the most reliable method and you'd have to have some pad. This also gets around the passthru amp limit, since you're only using the AC-in.

Anyway, the grid-tie idea, like i said, was mainly a pie-in-the-sky on my wish list. I shop for grid-tie inverters occasionally and inevitably think to myself "if only there were a decent grid-tie inverter." So i never end up buying one. There's always something it doesn't do - can't do 240, made for solar rather than batteries, no way to limit output power, very expensive, etc. What i really appreciate about the GS is the ridiculous number of features. I mean - 480v 3-phase - seriously? Awesome... I mean i'll never use it, but try finding a commodity inverter that can do that. And since i saw these modes that clearly required grid-tie-like behavior, i thought why not add grid-tie as a feature in it's own right. By no means is it a deal breaker for me.

Basically, i think both methods have merit.

Addressing the legalities again, maybe having it as a warranty-gated feature would sneak around that.

1 hour ago, Sid Genetry Solar said:

I'm hoping for mid-80s out of the current iteration (i.e. typical run of the mill).  Don't have any actual measurements, though...

A 10% difference in efficiency is hugely significant in the heat generated in the realms these inverters run (80% efficient will generate DOUBLE the heat of a 90% efficient unit).

Yeah, 80% and 90% is a world of difference. Heat also has it's own secondary effects like longevity and power expenditure on cooling. It snowballs real quickly into real world costs.

The next time i can test a predictable load, i'll get you some numbers for high and low load situations.

Anyway, tons of thoughts, so might have been a stream of consciousness there.

9 minutes ago, NotMario said:

And in a pinch, you could wire it in such a way you could flip a few breakers to use the AC-out instead of AC-in (*gasp* nasty, i know)

Even if the inverter ran as a true "grid-tie" system, it would always backfeed out of the AC Input.  This is firstly because the AC Input relays are available for use as "disconnection devices" in case the inverter errors out and shuts down.  And secondly because the current regulation is done on the Input side current sensor.

If AC Mains is hardwired to the output of the inverter, it has no way of disconnecting the AC in event of a fault.  This also means that there's basically unlimited current on the DC terminals of the inverter as well...even if the unit is off.

The only non-load that should be connected to the OUTPUT of the inverter would be a grid-tie system--because it shuts down when the AC does.

15 minutes ago, NotMario said:

Yeah, not sure what the standard is for anti-islanding delay.

It's covered in the UL 1741 documents.  I know what the requirements are, that's why it's so difficult 😉.

15 minutes ago, NotMario said:

Addressing the legalities again, maybe having it as a warranty-gated feature would sneak around that.

Maybe possibly.  To be legal, the inverter would have to be UL certified.  This costs tens of thousands of dollars--and any minor adjustment/tweak to either the inverter design (including the chassis) or even the firmware in the CPU (including updates)...would have to be recertified to be legal.  There'd be UL-cert updates, then non-cert updates...which probably falls into a serious legal gray area.

UL cert is a huge expense (5 figures for starters)--we've got to start turning a profit before something of that would be considered.

18 minutes ago, NotMario said:

I assume this method could also run a positive export (legalities notwithstanding). Essentially what you're doing here is a "grid tie inverter with limiter sensor" all in one box. You could use additional setpoints for backfeeding as well.

Sure thing.  It's all about what the regulation and control software runs the "throttle" at...

18 minutes ago, NotMario said:

So for one thing, it would be useful for so-called net-meters - if you have abundant renewable power to sell to the grid

Net metering seems to be more of a political thing than anything else.  It comes and goes from state to state depending on which way the wind blows......if it's "in", it's great for the homeowner.  If it's "out", that expensive grid-tied system's ROI just went through the roof.

...I can see both sides of the issue, but I'm not connected to the grid at all...

20 minutes ago, NotMario said:

This could also be accomplished by having a wireless sensor (hall sensor behind your meter) that would use the inverters web API to dynamically modify the export current limit, then you can do a "zero export" (or whatever target you want) without the additional wiring. Granted wifi isn't the most reliable method and you'd have to have some pad. This also gets around the passthru amp limit, since you're only using the AC-in.

True that.  Another option would be to utilize the hardwired comm port coming on the next round of GS inverters, and avoiding wireless stuff altogether.  Or if it became a significant interest, we could add a port and direct input for external current sense on the GS inverter itself (as there's spare CPU inputs on purpose for cases like this!)

29 minutes ago, NotMario said:

What i really appreciate about the GS is the ridiculous number of features.

I've always been frustrated by the LACK of features on consumer appliances.  Got a projection atomic clock right now that alternates between time and outside temp on the projection.  I could care less about the outside temp while I'm in bed--and would much rather it was constant, rather than alternating (=changing brilliance).  You guessed it: there's no setting.

So when I design things...you get settings on EVERYTHING reasonable 😉.

My understanding of most grid tie inverter's anti-islanding is the push/pull approach.  They try to push (or pull) the phase / frequency of the waveform and see how much current it takes to do it.  > 'so much' (or exceeds the amount of current the test is set to provide) and it determines the grid is present.  < 'so much' and it turns off the output after a number of seconds.

19 hours ago, Sid Genetry Solar said:

Even if the inverter ran as a true "grid-tie" system, it would always backfeed out of the AC Input.  This is firstly because the AC Input relays are available for use as "disconnection devices" in case the inverter errors out and shuts down.  And secondly because the current regulation is done on the Input side current sensor.

If AC Mains is hardwired to the output of the inverter, it has no way of disconnecting the AC in event of a fault.  This also means that there's basically unlimited current on the DC terminals of the inverter as well...even if the unit is off.

I may have not been very clear here. I was talking about as a reaction to a grid power outage when in grid-tie anti-islanding mode on AC-in only. In my normal [mode] case, i disconnect from grid, set to normal mode, then close the AC-out circuit on inverter. If i wanted to double the inverter as a grid-tie, i would open the AC-Out circuit, Set grid-tie mode, turn on Grid, close the AC-In circuit.

I would never suggest wiring [a closed circuit with] the grid ("mains") to the AC output - i know that's going to try to charge the batteries at unlimited current. Very bad idea. (The possibility of doing this by accident is what makes the approach kinda nasty - by wrong breakers or wrong order)

19 hours ago, Sid Genetry Solar said:

Net metering seems to be more of a political thing than anything else.  It comes and goes from state to state depending on which way the wind blows......if it's "in", it's great for the homeowner.  If it's "out", that expensive grid-tied system's ROI just went through the roof.

Absolutely. Though, all is not lost if you had a hypothetical GS inverter in Grid-Tie. Configure it for Zero Export with or without passthru backup. Or switch it to normal mode/passthru grid-tie and use as backup - or full off grid. (i guess a lot of grid-tie setups don't have batteries though)

19 hours ago, Sid Genetry Solar said:

I've always been frustrated by the LACK of features on consumer appliances.  Got a projection atomic clock right now that alternates between time and outside temp on the projection.  I could care less about the outside temp while I'm in bed--and would much rather it was constant, rather than alternating (=changing brilliance).  You guessed it: there's no setting.

So when I design things...you get settings on EVERYTHING reasonable 😉.

Hey, the very fact that i'm able to have this conversation [about a feature] with you is refreshing. Guess that's a small business for ya.

One thing that wouldn't take much time would be to add a nice stylesheet to the inverter status webpage. Oh, and put each "section" in a div (with a 'section' class defined) so the stylesheet can group them in a pretty way. If you want, i can make a stylesheet for you - likely colored based on the LCD interface. It'd probably be like 10 lines long. In fact...

<style type="text/css">
        body{
            font-family:monospace;
        }
        .section {
            background-color:black;
            color:#c2e276;
            display:inline-block;
            margin:0.25em;
        }
        .section h2{
            background-color:#c2e276;
            color:black;
            border:black 1px dashed;
            margin:0;
        }
    </style>

builder += `<div class="section"><h2>${topicName}</h2>`;

builder += "</table></div>";

Boy, did i take this topic on a tangent. 😄 Sorry about that, OP. I wasn't expecting this much interaction with the designer. That's a good thing, though.

36 minutes ago, NotMario said:

One thing that wouldn't take much time would be to add a nice stylesheet to the inverter status webpage. Oh, and put each "section" in a div (with a 'section' class defined) so the stylesheet can group them in a pretty way. If you want, i can make a stylesheet for you - likely colored based on the LCD interface. It'd probably be like 10 lines long. In fact...

So I have a bit of a surprise up my sleeve for a future update (not 1.1r6 unfortunately!)...a completely redesigned tabbed interface that unifies things considerably (instead of having separate HTML paths and system access modes for the functions):

<a class="ipsAttachLink ipsAttachLink_image" href="/monthly_2022_02/image.png.e54bde74aec5b8e064225c1b2d17f2d6.png" data-fileid="1082" data-fileext="png" rel=""><img class="ipsImage ipsImage_thumbnailed" data-fileid="1082" data-ratio="109.49" width="685" alt="image.thumb.png.9ee756b24a90517d21d8c30c6ac1d222.png" data-src="/monthly_2022_02/image.thumb.png.9ee756b24a90517d21d8c30c6ac1d222.png" src="/applications/core/interface/js/spacer.png" />

(actually designed this out in September last year, but ran into too many stability issues with the redesigned WiFi control on the WiFi board...so shelved it temporarily).

Worth noting: the ONLY picture in the above screenshot...is the Genetry logo (which is a compressed 2-color transparent PNG consuming an exorbitant 9.3kb).  Everything else is done with CSS shaders and gradients--for that matter, the tab action is completely handled in CSS, with NO JS support code.  (There's JS handling code, but my point is that the tabs click--and switch--without anything except CSS.)  And no, I'm not an experienced web developer by ANY metric--I spend 90% of my "web design" time digging through Google searches for solutions and examples that I can cobble together.

There is one thing I'd kinda like help with...I know I could figure it out myself at some point, but hey, if someone else wants to do the work...!  I've got some fancy JS gauges (no dependencies) used on the Genetry remote server page, and it'd be nice to dump them onto the "Status" tab (instead of just showing JSON decoding), so it looks a lot more refined.  The problem is that the Minified Gauge.js is ~44KB.  This is peanuts for most people--but the ENTIRE webpage and associated files must fit into a 190KB partition.  And I don't want to "just barely fit"--no, I'm trying to make things absolutely as small as possible...not only for an efficiency standpoint, but also so that there's comfortably room to add features.

As the "Gauge.js" is extremely customizable, I'm certain that an easy 3/4 of it could be trimmed out by removing all the customizations/features that I don't need for the GS project.  Download it here: https://canvas-gauges.com/

53 minutes ago, NotMario said:

I would never suggest wiring [a closed circuit with] the grid ("mains") to the AC output - i know that's going to try to charge the batteries at unlimited current. Very bad idea. (The possibility of doing this by accident is what makes the approach kinda nasty - by wrong breakers or wrong order)

Well...not quite.  Both the input and output AC lines on a Rev. C GS inverter have separate current and voltage sensors between them.  Transformer is directly connected to the output, and the relays connect the input to the output.  What I meant by "unlimited power" has to do with the fact that the inverter can't disconnect a load from the output.  On a GS6 with 120 (or 240v) input to the transformer, the transformer will output 32vAC * 1.414 (square root of 2) = 45.25vDC.  Too low for charging most batteries--but if you short those battery terminals out by accident, it's gonna be bad!

My point was more that if mains is connected to the output...even if the inverter's turned off, there'll be "unlimited" DC voltage at the battery terminals.  More of a safety thing...

CSS-only tabbing was all the rage about 10 years ago. I remember making a [windows 95 style] menu that used no JS. It was very hard to do back then because the CSS features varied a lot between browsers.

About your gauges idea. Making that framework smaller is probably less practical than just building a small custom JS framework. Of course that depends on what kind of gauges you want to have. Some basic "loading bar" style gauges would be 20 lines of code or so. It might actually take more code to decide/declare which gauge (if any) is appropriate for each item.

You can "gimp" me a picture, depending on how complex it is, i can probably whip up something space conscious.

This is what the gauges look like on the remote server screen.  Notice the "range colored bar" (i.e. red for invalid, yellow for "alarm", green for "normal"), the pointer, the "ticks", and the actual readout number.

I know that making something like this look nice isn't easy...but if appearance is the goal...well...!

Doesn't have to be this fancy--these were just the simplest (standalone) gauges I could find.

2 hours ago, NotMario said:

CSS-only tabbing was all the rage about 10 years ago.

Perhaps you're thinking about CSS drop-down menu tabs--which are quite commonplace.

I only managed to find a single example online that did this particular CSS trick, though.  In the earlier screenshot, if you click a tab, it automatically Z-orders the corresponding "box" to the top.  I can click between all 3 tabs and view the corresponding page...with no JS.

My biggest bugaboo is that examples these days tend to live on "extensions", "frameworks" (Node.js, jQuery, Ajax, etc.) and other nonsense that end up considerably bloating the project for no real gain--at least in my scenarios.  (I end up tossing most of the examples I find for a particular function due to this alone.)  For example, the jQuery framework alone requires almost HALF of the available partition space in the WiFi board--and I can do just fine without it by using basic HTML, JS and CSS.

So you like those speedometer style gauges. They're quite a bit more complex.

It so happens i have a gauge class that's roughly similar to that. It works by stacking elements and CSS rotating them.
It doesn't do those colors on the edge though. I'd have to chew on that a little to figure out a way to add that feature.

The class is a few kB fully expanded. I've never tried it, but it should minify and compress pretty small.

That's partly why I'm wondering if the "canvas gauge" JS library could be stripped down--as it has support for bar-style gauges, an HTML tag control interface, and LOADS of features I don't need. 

But like you said, that's probably a lot of work.  I do like the very polished end result though...

I jumped off the deep end...
Those rings were a real challenge. lol

The js stands at 5.5kB uncompressed. With compress you could drop it to <2kB total.
It does have a pretty extensive stylesheet which takes about 2kB uncompressed. Stylesheet lets you customize pretty much anything.

Of course it doesn't require any frameworks. (though my demo uses a perlin noise library)
No it doesn't have bar-style guages. A class for that could be made in around 20 lines.
 

<a href="/monthly_2022_02/gauge.png.227c49e2745aae926b2b8b2b5c46367f.png" class="ipsAttachLink ipsAttachLink_image"><img data-fileid="1086" src="//forums.genetrysolar.com/applications/core/interface/js/spacer.png" data-src="/monthly_2022_02/gauge.png.227c49e2745aae926b2b8b2b5c46367f.png" data-ratio="51.86" width="617" class="ipsImage ipsImage_thumbnailed" alt="gauge.png">

11 hours ago, NotMario said:

I jumped off the deep end...
Those rings were a real challenge. lol

The js stands at 5.5kB uncompressed. With compress you could drop it to <2kB total.
It does have a pretty extensive stylesheet which takes about 2kB uncompressed. Stylesheet lets you customize pretty much anything.

Of course it doesn't require any frameworks. (though my demo uses a perlin noise library)
No it doesn't have bar-style guages. A class for that could be made in around 20 lines.
 

//content.invisioncic.com/g308908/monthly_2022_02/gauge.png.227c49e2745aae926b2b8b2b5c46367f.png

Whoa, that's really getting somewhere 😉.  I am not at all surprised that those colored range gauges were challenging...but am very happy with the size.  I am assuming by "compressed" you mean "minified"?

Yes, I'm picky picky (haha!)...maybe a bit of tweaking to the pointer for that nice "pointer shape", and a bezel on the outside.  (Latter shouldn't be too hard...but the former might not be the easiest 😉).

This definitely looks very promising.  And is very tiny to boot.  Love it.

Don't need a bar-style gauge at all.  My comment to that effect was that the other "canvas-gauge" file had another heap of unused code for the bar-style gauges.

I used a "graph" JS thing for all the system temperatures--but technically, those could all be small round gauges.  Bar-style gauges for fan speeds--but again, that could be round gauges.  All about keeping it simple while making it look good.

Thanks for your effort on this--it looks extremely promising.  And you definitely have the idea 😉.

So, compressed i actually meant gzip (non minimized, additional minimization would probably get you sub 1kB). You might be aware, pretty much all standard web clients support gzip. What you would do is store the JS in gzip and you could either:  have the web server set content-encoding:gzip, and just send it as-stored (the browser decompresses for you). Or you can implement libz and decompress when serving the file. The latter option can of course use any decompression technique you want - but of course the decompression functionality, itself, uses space, though a lot of embedded environments will already have it built in. Incidentally, you can do this with anything served by the web server - so all assets can take advantage of it - big savings to be had.

Minimized reduces it like 40%? There are a bunch of different ones, so quoting a "minimized" size is like quoting a price in a nonsensical currency.

Anyway, i'm going to polish the invocation a little bit. Right now you set a classed div with a json inside describing the gauges parameters. A one-liner bind command will build the gauge. You can then keep a pointer to the div to update(value) it. I really want a better way to supply the parameters - the json looks very weird to me directly in an html tag.

I'll make you a bar gauge. These are so easy to do i never kept one to reuse, lol.

Try it out:  https://jsfiddle.net/ozmu5ag6/3/