Photo-voltaic Power Generation System

Since troubleshooting the Schneider/Xantrex Automatic Generator Start (AGS) is turning out to be an extended and significant effort, I'm collecting information here for ready access.

Background

The Xantrex AGS is used to autostart a 5KW diesel generator. Brief grid outages of a few seconds are common here, and the 16KWh battery bank is capable of carrying the critical loads for several hours. The desired behavior of the AGS is that the generator will be started only when the Xantrax hybrid inverter is inverting, and the battery voltage has dropped to a level consistent with a significant (20-25%) state of discharge, or the critical loads are at a higher than normal level. The AGS documentation¹ indicates such a configuration is possible but in practice we have been unable to achieve anything even close. In particular the AGS is prone to starting the generator for no discernable reason, often shutting it down immediately but also sometimes leaving it running for hours when grid power is available and the batteries are fully charged.

In the hope of aiding the troubleshooting process a Conext Combox device was purchased and installed, and configured to log relevant status parameters at five second intervals (the finest granularity possible). The generator itself was disabled by disconnecting power from the internal DC bus used by internal remote start controller and the Xantrex AGS was configured to support automatic generator starts triggered by either battery voltage or high inverter loads. The complete configuration of the AGS and the XW 6048 hybrid inverter are recorded in text file form here.

After a week the logs were pulled (manually by removing the SD card; the web based log viewer in the Combox is worthless) and analyzed. For the period 2014-08-22 through 2014-08-31 inclusive, there were 16 false starts; 12 on the 22nd and one each on the 23rd, 25th, 27th, and 31st. Inverter mode was logged only one time during that entire week, on the 31st corresponding to that false start (and only for 15 seconds). These logs can be seen here; both the raw CVS logs and selected content converted to spreadsheet format.

Gory Details

The AGS and XW 6048 were upgraded to the latest available firmware before this latest set of logs was obtained.

The installation of the AGS doesn't appear to be suspect, but the entire wiring scheme from the generator internals to the AGS unit is documented in excruciating detail. For ease of troubleshooting multiple 24V LED indicator lights were added to show all relevant control signals. The AGS interconnections and indicator lights can be seen in this photo. Note that once it became clear that debugging the AGS was going to be a non-trivial exercise I purchased a second Xantrex SCP (System Control Panel) accessible from the location where the AGS and wiring were installed².

Per the documentation the AGS is not supposed to start the generator via the "Strt Load" trigger unless the threshold current (10.0A here) is exceeded for five minutes, and only when the XW is in inverter mode. That clearly isn't the case as almost all of the false starts occur when the XW is not inverting.

One theory that has been advanced by Xantrex tech support is that the XW is inverting for an interval too brief to be noticed or to be logged by the Combox (which cannot log more frequently than 5 second intervals). I think that unlikely because of the Bogard Engineering "Trimetric" State-of-Charge controller that is monitoring the battery bank. It is configured to recognize the batteries as fully charged when the battery voltage exceeds 54.3 volts and the current is less than 7.0 amps. However, it continues to accumulate watt-hours so the percentage SoC indication increments beyond 100%. If any current is drawm from the battery bank, even for a millisecond, the SoC indication immediately resets to 100%³. I initially thought that behavior annoying, and complained to the manufacturer who confirmed it as a "feature not a bug", but now consider it a useful feature as I can tell at a glance roughly how long it's been since the last grid glitch.

Initially I thought the AGS was triggering the generator start based on AC output load even when the XW was clearly not in inverter mode, but after much experimentation and observation I don't see any pattern at all. Attempts to cause a false start by sustaining a load above the "Strt Load" threshold for more than five minutes have failed, and false starts have occurred when the load was minimal or even negative (with a grid tied inverter backfeeding the critical loads panel connected to the XW 6048 AC Out).

Note that the Combox does not properly recognize a power export from the critical loads (when the backfeed from the interactive inverter into the critical loads panel exceeds the power consumption from that panel, which usually occurs during daylight hours). Both the logs and the web portal graphical display show zero for negative power flows. However, the SCPs do correctly report negative power consumption from AC Out and the false starts have been observed with power consumption positive (and above the Strt Load threashold, when not inverting), positive (and well below Strt Load), zero, and negative. Whatever is triggering those starts isn't obvious.

A number of the parameters the Combox logs appear to be dubious, like "XW AC Load Current" and "System XW Load Power" for which improbable values are sometimes logged. The Combox also occassionally writes spurious garbage in the CVS log files. It's possible the Combox is also incorrectly logging the generator status parameters as well, though the generator starts are real enough.

Detailed Log Analysis - short-cycle misfires

Schneider/Xantrax tech support has asked for an analysis of the logs (which were also provided to them in full). So here's my analysis of a representative false start on 08-27. I picked that day because there was only the one false start that appears to be most representative of such events (08-25 is comparable). The raw log file is 0827.CVS and the relevant records extracted to Excel spreadsheet format is aa href=" AGS and XW.

The false start begins at 06:53:00. The Generator On reason is "AC Current High" (3), the Generator Off Reason when the run signal is dropped 30 seconds later at 06:53:30 is "AC Good" (21). Note the XW Inverter Status remains "AC Pass Through" (1025) throughout. At that time the SYSTEM Load Output Power is 926W, XW Load AC Current is 9.5A, XW Load AC L1 Current is 3.4A and XW Load AC L2 Current is 6.2A. The total of L1 and L2 is 9.6A, close to the XW Load AC Current (in some of the logs XW Load AC Current is recorded as zero but L1 and L2 usually appear to have sensible values).

Note in particular that:

1. the XW 6048 never enters "Invert" mode (or if it does it's for less than five seconds⁴).
2. the AC Load is never greater than 10.0A (the Strt Load setting), not for five minutes per the Xantrex AGS documentation and not for even five seconds
3. the run signal is dropped in only 30 seconds, whereas the Xantrex AGS documentaiton states that the Stop Load trigger requires a full minute below the Stop Load current threshold (7.0A here).
4. AC2 (generator input) is not logged because the generator was completely disabled throughout this logging period (I didn't see the point in continually abusing the generator with multiple short-cycle false starts).

If I can get some sort of resolution of this type of incident I'll move on to the more complicated events that were logged on 08-22 and 08-23 where the run signal was sustained for hours with grid power available and without the XW 6048 ever entering "Invert" mode.

Hypotheses

I'm getting input from a number of interested spectators. One suggestion is that EMI may be an issue. That is plausible given that the utility power here is fairly "dirty", with brief outages and voltage sags being rather common. I don't have the equipment to quantify the problem, but we see the lights flicker and when I'm in my office I hear the squealing and beeping from multiple UPSes.

That said we haven't noticed any problems⁵ during the year and a half before the AGS was installed. That is powered only from the Xanbus and so presumably would be isolated from line power noise.

The XW 6048 is about 20' from a main lug service entrance panel which is fed from a main disconnect about 50' away where the neutral-ground bonding occurs. The grounding conductor goes to two ground rods by the utility meter (another 10' or so). Those two ground rods are connected to the ground plane system of an adjacent radio tower via a 4" wide by 25 gage (0.022") copper strap. The ground plane system consists of a 30' diameter ring of copper strap and six ground rods. The ground lug bar in the SE panel is also connected to a new ground rod installed with the PV system that is about 20' away. The 240V feed to the XW 6048 (grid input or "AC1") is properly grounded to the SE, and that to the main disconnect.

Schneider/Xantrex tech support asks periodically about the Xanbus (Xantrex proprietary version of Modbus) network cabling. The Xantrex documentation specifies a maximum length of 131' (40m) for the entire network. Recently they also began documenting a requirement that no single segment (between any two XW devices) be less than 6.5' (2m) in length. The network topology is as follows (also at the end of the detailed configuration document).

Xanbus wiring	T == RS485 120ohm terminator
		Cat5e, minimum cable length 6.5' 2m), max total network length 131' (40m)
		Documentation says use T568A termination but Xantrex tech support confirmed
		that consistent use of T568B is fine (that is the in-house standard)

	[SCP bmst]	[AGS     ]	[XW  6048]	[Combox  ]	[SCP up  ]
            T\______7'______/\______13'_____/\______7'______/\______49'_____/T

Total length of the network is approximately 76', give or take a couple of feet. The two 7' segments are factory made and the two longer segments were site terminated.

All cables are Cat5e and either factory terminated with molded connectors or site made with an RJ45 crimp tool and molded boots, and checked with a TENMA 72-6709 Network Cable Tester Pro Cat5 cable tester for wiremap, twist, and shorts. The Xantrex documentation calls for T568A termination but the local standard here for the several thousand feet of installed Cat5e is T568B. Most factory made Cat5 Ethernet cables come that way too. I know from working with Ethernet that "A" and "B" terminations are equivalent as long as both ends of the cable are done the same⁶, and Xantrex tech support agreed with that at the time the Xantrex XW 6048 was first installed.

Current Status

While still hoping for a resultion via Xantrex tech support I'm facing the realization that it may be necessary to kludge some other solution. Some options are considered in the following section. I don't want to invest yet more time and money in workarounds until, hopefully, I have a better understanding of why the AGS is misbehaving as it does. At the moment the false starts appear to be completely unpredictable which is going to make any workaround more difficult.

I will update this section with any new information.

UPDATE 2014-09-04: Xantrex tech support has concluded that the AGS device is defective and will replace it even though I can't find the proof of purchase paperwork. They wanted the paper sticker bar code on the back.

UPDATE 2014-09-05: I received a call from a new person at Schneider/Xantrex who says the logs indicate a problem with the AGS or XW 6048 hardware. I sure hope it's the former as replacing the XW 6048 would be a pain. He said Schneider would replace the AGS, even though I've lost the proof-of-purchase paperwork.

UPDATE 2014-09-15: The new AGS was received on Saturday. It came pre-loaded with a very new firmware version, 3.00.00 BN 2. When installed the Conext Combox didn't recognize it but after digging around on the Schneider web site I found a very new version Conbox firmware upgrade, Ver02.00BN0656. On installing that the Combox did recognize the new AGS. Unfortunately the new AGS is exhibiting exactly the same false start behavior as before. It didn't take long to capture one; event 0915.xls occurred only a few minutes after I enabled Automatic mode on Monday morning. The generator was enabled and connected to the AGS up so there was an actual start. As before the XW did not enter invert mode. The load as shown by "XW Load AC Power" did not exceed 1100W from the point where Automatic mode was enabled until the false start. However, I noticed something interesting: the false start occurred exactly five minutes after automatic mode was enabled and "XW Load AC Current" exceeded 10.0A (the AC High start trigger threshold). It's as if the AGS/XW were ignoring the fact that the XW wasn't inverting.

I'm a bit puzzled at the discrepancy between "XW Load AC Power" at <1100W, "XW Load AC Power - Apparent" at <1400VA, and "XW Load AC Current" at ~11.5A which is roughly 2700W. That's quite a discrepancy. Autostart is disabled again pending feedback from Schneider/Xantrax tech support.

I still hope to eventually resolve the issues with the Xantrex AGS, but I meantime I called Ralph Hiesey of Bogart Engineering to confirm the suitability of the Pentametric state-of-charge monitor for generator starting. That ended up being a very long and informative call touching on almost every aspect of my complex PV grid-tied off-grid system and I learned a number of things of interest. I was pleased to hear that the Pentametric could share a shunt with the Trimetric meter that is already in use, saving me the hassle of redoing that installation in the upstairs hallway wall. After that call I ordered the Pentametric PM-5000-U and PM-101-CE products and downloaded the documentation with the intention of configuring the Pentametric to autostart the generator based on state-of-charge levels, either as an adjunct to the Xantrex AGS or as a complete replacement. I'd still really like to be able to start the generator based on high inverter loads but that may not be possible.

UPDATE 2014-09-16: Schneider/Xantrex tech support says that since the new AGS didn't fix the problem it must reside in the XW 6048 inverter, and they will replace that under warranty. Since the legacy XW 6048 has been discontinued I will receive the newer XW+NA 6848. After a quick glance at the specs that appears to be at least as good and probably a step up in performance⁷. I just hope the replacement won't require extensive rework of the EMT conduit.

UPDATE 2014-09-20: I was told by Schneider/Xantrex tech support on 2014-09-17 that I would need to upgrade the SCPs to firmware version 865-1050-01.02_00_00.BN0013.xf0. That was a bit puzzling as the SCPs are currently at version "1.06.00 BN 19" which appears to be newer, but on 2014-09-18 I gave it a try. The Combox will not recognize 865-1050-01.02_00_00.BN0013.xf0 as a valid upgrade option, whereas it does recognize the 865-1050.01_06_00.BN0019.xf0 I used for the earlier upgrade to "1.06.00 BN 19". I reported that result to Schneider/Xantrex tech support but have not received a response yet (it typically takes a couple of days to get a response, if any, to such queries).

In the meantime I have received and partially installed the Pentametric battery monitor (PM-5000-U input unit and PM-101-CE Ethernet interface). A fuse holder is on order for the battery bank connection but in the meantime I've powered it up with a standalone DC power supply and have configured the Ethernet connectivity. Unlike most such gadgets the Pentametric does not provide a web server interface, instead Bogart supplies the source code to a dedicated utility that uses a custom homegrown protocol on TCP port 1701. That did compile successfully (with one tweak) on an Ubuntu 12.04 system. The Pentametric will drive an external relay (also on order) by programmable battery voltage and SoC thresholds. If we can get the AGS to work it will be used for load based generator starts and the Pentametric for SoC based starts. If not then the Pentametric will be used exclusively for generator starting.

UPDATE 2014-09-24: The frustration continues. The new XW+6848 inverter is scheduled to arrive tomorrow (I received a call from the shipping company), but I'm still unable to upgrade either SCP to the "1.02 BN 13" firmware version I'm told is necessary for that new inverter. This remains the case even after upgrading the Combox yet again, to Ver02.01BN0673 which was only released on Sept. 22. The latest Schneider tech support contact doesn't appear to believe that I'm doing the upgrade properly, so I've prepared some screen shots as proof:

A screenshot showing the current Combox software version (upgraded earlier today).

A screenshot showing the unzipping of the SCP firmware file.

A screenshot showing the start of the upload of that file to the Combox.

A screenshot showing the completion of the upload.

A screenshot showing the pick list that does not include the just uploaded firmware version. Note the older "1.06.00 BN 19" still appears as I don't know how to remove it without completely wiping the Combox configuration (and anyway the "1.02 BN 13" upgrade was tried first).

UPDATE 2014-10-06: The resolution of the firmware upgrade issue was simple once finally known: I was instructed to rename the file 865-1050-01.02_00_00.BN0013.xf0 to 865-1050.02_00_00.BN0013.xf0 (i.e., remove the "-01"). That did the trick.

The new XW+6848 arrived as promised and was installed on 2014-09-27. I did the installation strictly by the book, including use of a torque wrench for tightening the terminal connections⁸. No false starts at first, even with "AC Current Level To Start" at 3.0A and "AC Current Trigger Start Delay" at 0.5 minutes to encourage any misbehavior. Hurray!! I was starting to be cautiously optimistic about an end to this long ordeal. But then on 2014-10-05 I was demonstrating the multiple means of stopping or disabling the generator to my wife, in preparation for an upcoming trip that will have me overseas for several days (at this point my wife doesn't trust the autostart controls at all). I engaged the manual start inhibit switch that I've wired into the transfer switch panel. This is wired across leads 5 and 10 of the AGS, and generates a F203 fault code and reverts the generator state to "Manual Off" when closed⁹.

Immediately after closing that switch, clearing the fault code, and re-enabling Automatic mode the AGS enabled the run signal. This with the XW still in AC Pass Through mode. Arruggh ... the same false start issue that has plagued us all along.

I found that to be repeatable; each time I set to Manual Off mode and then back to Automatic the AGS will now enable the generator run signal (the generator itself has long been disconnected from AGS control). This after a week during which no false starts were observed. I have no idea why this is happening now. I did manually run the generator (disconnected from the AGS, but with AC2 power visible to the XW) a couple of days ago. Yesterday I power cycled the Combox (by disconnecting the Xanbus cable) as it had stopped responding. Then the run inhibit switch test. Perhaps one of those events was the provocation?

As an experiment I disconnected all the Xanbus devices except the AGS, the XW, and the Combox (AGS-XW-Combox). I used brand new Cat5e cables. Note that with this configuration everything has been replaced since the false starts were first experienced: the AGS has been replaced, the Combox is brand new with firmware upgraded twice, the XW+6848 is brand new as of a week ago. The cables are new (two factory made 7' Cat5e patch cables), and I even dug out two new RS485 terminators. So every single component of the configuration has been replaced.

The false start is still occurring! With the "AC Current High" trigger, with the XW not inverting.

As a second experiment I replaced the Combox with an SCP (SCP-AGS-XW, again with two new 7' Cat5e cables). Same result. So, we can rule out the cables as the problem occurs with multiple cables. We can rule out the Combox and the SCP as the problem occurs with those components completely absent. This problem has occurred with two different AGS devices and with two different XW devices.

At this point I'm left with the conclusion that the Schneider/Xantrex/Conext hardware/firmware is defective by design. Remember this false start issue has persisted across multiple hardware and firmware upgrades, to the point where every single component, both hardware and software including all cables and terminators, has been replaced. yet, the problem persists. I don't think we can still credibly assume any type of single-sample hardware defect.

I'm not terribly optimistic about a solution at this point, though I will cooperate fully with Schneider tech support if they have any more ideas (any solution will involve a firmware upgrade at a minimum, I think).

Note to Schneider tech support: I reiterate my earlier offer to configure direct access to the Combox. I'll just need your source IP address(es) to get you through the three firewalls it sits behind. You could then review the device configurations and cause and observe the false starts yourself.

UPDATE 2014-10-09: After a phone discussion with Schneider/Xantrex tech support my contact concurred that the firmware was buggy and said that resolution would need to be escalated to the firmware maintainers. I disabled the load based triggers, in addition to disconnecting AGS harness leads 5, 6, and 12 as recommended.

I left the "Exercise" trigger enabled, and this morning I experienced the first exercise event with the generator actually connected. Unfortunately the AGS immediately killed the run signal right after asserting it, and then continued to repeat those short starts every 115 seconds until I disabled the AGS. This log extract shows the first few of those abortive starts. I don't know how long it would have continued those short starts, but if continued for very long that would be pretty rough on the generator.

So, I can't use the AGS for load based starts and I can't use it for periodic exercising. I've not yet seen problems with voltage based false starts, but since the Pentametric also does voltage based triggers (in addition to SoC) that isn't of much interest even if it worked reliably. The AGS is thus useless until and if the false start problem is addressed, and will be disconnected until and if Schneider has new firmware that purports to do so.

UPDATE 2014-10-10: I received notice from Schneider tech support yesterday that the firmware developers in Canada have confirmed a firmware bug. No ETA for updated firmware was given, but it's a relief to have that confirmation as usually the hardest part of solving this type of problem is just recognizing that there is a problem. The "Exercise" mode problem is assumed to be a consequence of the same bug.

I was was also asked for a photo of the installation. A shot of the initial AGS installation is already on this web page, but I took one of the current configuration. Notice it's gotten a lot more complex as I struggled to debug the autostart problems; I added more indicator lights and a time delay relay in the hope of filtering out some of the shorter false starts¹⁰. I've also added the Pentametric SoC meter which is now the means of controlling generator start. Note some of the leads and an indicator light are dangling loose as I wait to see what the final configuration will look like. I will probably end up with seven LED indicators which will require another cast aluminum box as seen to the left of the SCP. The AGS will be used only for manual generator stop/start via the SCPs or Combox, with all start triggers disabled, and will be removed entirely the first time it misbehaves in that minimal configuration.

I've also posted an ASCII art¹¹ wiring diagram corresponding to that current configuration. That is intended for my use and so describes a lot more than just the Schneider AGS configuration; it includes the Inova Tech autostart board and the Pentametric SoC meter and relay.

UPDATE 2014-10-13: Schneider has confirmed that they have reproduced the false start issue which will take an indeterminate amount of time to resolve. In the interim they advise us to "...put the AGS into STANDBY when you don't want the gen to start automatically". Well, the whole point of the AGS is to start the generator automatically (that's the "A" in "AGS"). So the AGS is disconnected and we are using the Bogart Pentametric SoC meter to handle generator autostart. That has behaved reliably with no false starts.

For now the Schneider AGS is still installed, although electrically disconnected. If and when Schneider releases new AGS firmware I'll give it at least a brief try, but if it still shows any proclivity for false starts it will be removed entirely and discarded (I couldn't in good conscience sell it on Craigslist or Ebay).

Possible Workarounds

Unfortunately many days of consultation with Schneider/Xantrex technical support have not brought much hope of a resolution. With something like $10K invested in the Xantrex hardware and generator switching to a different product mix is unfortunately not a viable option, so I've been considering my options for external workarounds to what appear to be uncorrectable software bugs in the Xantrex hardware.

I have discussed the problem extensively with Peter Albrecht of Inova Hightech, Ltd. who has been most helpful. The short false starts of a few seconds duration, which seem to be most common, could be filtered out with a standard time delay relay. I will install one just to protect against short-cycle abuse of the generator, but that will not address the issue of the long term (many minutes to many hours) false starts that have been recorded. Peter has also offered to reprogram the Invoa autostart board to perform the delay function in software (e.g. ignoring a run signal until it persists for N seconds).

A NO relay with a 120V coil powered from the grid could be used to disable autostarting of the generator entirely when grid power was present. That relay would need a time delay so a grid hiccup during an outage wouldn't immediately kill the generator. However, that relay would prevent scheduled exercising of the generator (currently configured for 21 days) which I consider a valuable feature. The purist in me also dislikes the idea that a coil will be energized almost the time, even if the power consumption is only a few watts. Inova is developing a fancier autostart board for this generator that will include an exercise function, so that is a possible (albeit expensive) alternative if such a grid cutout relay is used. I'm assuming of course that the Xantrex AGS doesn't fail to start the generator when it should; so far I've only seen false starts but we haven't had any generator-worthy outages yet.

Yet another and even more expensive option would be to discard the Xantrex AGS entirely and replace it with an independent autostart controller. The Inova board provides true "two wire" start capability¹² so the external controller only needs to close a set of dry relay contacts to signal the generator to start. Some off-the-shelf hardware of that type is apparently available but at least some custom development would probably be necessary to handle the battery DC voltage trigger. Such a project would exceed my electronics skills and enthusiasm and so would require some outside assistance. I've made some preliminary inquiries and it appears that could be done for about what I've already spent on the Xantrex hardware². UPDATE: there appear to several off-the-shelf solutions that could replace the Xantrex AGS entirely, such as the Victron BMV-600S monitor and the Bogart Pentametric. Since I'm already using the Bogart Trimetric monitor and am happy with it I'll probably try the Pentametric first. The Pentametric has a relay that can be configured to close when the SoC drops to a preset level, and open when it rises to an "off" setpoint, while also triggering off high and low temperature compensated battery voltages. Triggering just off of battery voltage as the Xantrex AGS does isn't desirable (as the battery voltage varies with load as well as SoC of charge), and I'd ideally like a load based trigger rather than running the batteries down when the generator will need to be started soon enough anyway, but a SoC based trigger would be preferable to what I have now. The Pentametric monitor costs $360 (or $510 with the optional Ethernet interface).

Troubleshooting Notes

In fairness to Schneider/Xantrex the XW 6048 inverter and AGS module are complex devices and they have been generally responsive in talking to me on the phone and via E-mail. Some of our troubleshooting sessions have taken the better part of an hour and they're still talking to me. In the hope that we may yet succeed in figuring out and correcting the false start problem I offer the following thoughts:

• I will cheerfully configure any logging the Combox is capable of.
• I have a logging multimeter (Fluke 289) which can do detailed logging of any one parameter of interest (e.g. voltage, current) for periods of up to several days. Since the false starts don't happen every day it may take awhile to capture the data of interest, but then again since the AGS was installed about a month ago we've not experienced more than three days without a false start event.
• I can configure direct access to the Combox by Schneider/Xantrex personnel. I'll need the source IP address to forward inbound connections through the firewalls.

What I don't want to do is keep "resetting to factory defaults". I've been asked to do that several times now. The very detailed configuration must be painstakingly typed in manually, a process that takes about twenty minutes during which time the critical loads are without power. I can accept an occasional mysterious glitch¹³ in heavily computerized hardware, but such a "reboot" is a last resort intervention that is not acceptable on a routine basis.

_

¹ "Xantrex XW Automatic Generator Start Owner's Guide" in file XW-Hybrid-Inverter-Charger_Installation-Guide(975-0239-01-01_rev-E).pdf, no direct link available.

² In hindsight I consider an SCP next to the AGS to be essential to getting the Xantrex autostart working, and the Conext Combox as well. Note that starts to add up though; roughly $222 for the second SCP, $365 for the Combox, plus the $170 for the AGS makes nearly $800 in Xantrex autostart-related hardware alone. Ouch.

³ I'm assuming that the battery voltage must drop below the 54.3V float voltage level for the inverter to operate.

⁴ Less than five seconds per the Combox log, but almost certainly zero per the earlier discussion of the Trimetric State-of-Charge monitor. I was not present on 8-27 to confirm that no very brief grid interruption occurred but I have witnessed multiple similar false starts where there was no perceptible drop in battery voltage (below float charging level).

⁵ With the possible exception of an instance where a nuisance alarm (low battery warning) sounding when it shouldn't.

⁶ Terminate one end of a Cat5 cable per T568A and the other per T568B and you have a crossover cable.

⁷ I've found several discussions online that claim that the rebranding of the Xantrex XW6048 to the Conext XW+6848 involved only software and cosmetic changes, and not any substantive hardware upgrades. According to those discussions the new rating of 6800W versus 6000W just reflects the use of new criteria for that rating.

⁸ Professional electricians laugh at me when I do that, but I'm comfortable with the use of torque wrenches where proper torque does matter (I own a dozen torque wrenches), and I want to "CYA" to deflect any possible warranty issues.

⁹ The transfer switch selects between two generators, a manual start gasoline genset and the autostart diesel generator. Normally the switch is set to the diesel generator, but will be switched to the gasoline one when exercising that generator with a load. The run-inhibit switch prevents autostart of the diesel generator when its output isn't visible to the XW inverter. It also constitutes another visible and easily understood means to disable autostart; just grab the transfer switch handle and pull it outwards slightly. I've wired a bright red indicator LED to show when this run-inhibit is in effect.

¹⁰ The time delay relay is currently superfluous with the Pentametric meter controlling the generator, as the Pentametric has shown no tendency for false starts. The TD relay was ordered before the Pentametric. It still comes in handy for testing because the built-in LEDs will show coil activation before the contacts close, allowing testing without an actual generator start without disconnecting any wiring.

¹¹ I date myself with the use of ASCII art, which I use for this sort of thing out of old habit. In the event of a catastrophic failure of the local infrastructure one can read or edit ASCII files on any type of computer, so I tend to keep key elements of mission critical info in plain text format.

¹² In principle the Xantrex AGS could have autostarted the generator directly, without the Inova board. However, that would have required running a harness of multiple wires out to the generator, and non-trivial custom wiring within the generator. The Xantrax AGS is also significantly less sophisticated than the Invoa board, which directly recognizes fault conditions such as low oil pressure and overtemp or overload conditions, and does a temperature compensated preheat interval.

¹³ In the year and a half of operation before the AGS was added I only did such a reset once, and that was at the request of Xantrex technical support for a nuisance alarm. That alarm has not occurred since but I don't know if the reset had anything to do with it.