Batteries
in [Mandrake]
|
From: Moe Trin on 27 Sep 2007 20:34 On Wed, 26 Sep 2007, in the Usenet newsgroup alt.os.linux.mandrake, in article <SOzKi.6204$J03.5590(a)trndny03>, Adam wrote: >Moe Trin wrote: >> Where does the air go from/to the hard drives? >I assume the air around them is pulled out by the fan on the left side >of the case, which is only a few inches away. Where might the air be coming in? Gaps around the drive[s]? >I suppose I could blow smoke into the system to get some idea of >airflow, but I don't think that's really a good idea for a computer, >unless maybe I use dry ice in water. Also, I'd have to remove the >side of the case to watch it, which would ruin the test. All smoke - which is a particulate - could result in residue. The dry-ice is water vapor, somewhat better - but neither is likely to be harmful in a short test. As for the case, it's normally replaced with a plexiglass version. >> If it's _possible_ I like to have airflow blowing in the the board >> at an up angle > >I suppose I could take the 80mm fan that I attached to the left side of >the case, and attach its top two corners to the bottom two fan mounting >holes in the case, and somehow point it up somewhat, so it would be >blowing air at the general direction of the RAM. I've also seen people install the fan at the bottom, blowing at the motherboard, and counting on the air being deflected up past the RAM. Some of this is a pay-your-money-take-your-pick situation. >I suppose it would be getting air from the vicinity of the HDs, though, >so maybe that's not such a great idea. That's why I've got the air going out past the drives, rather than in. >When I added that case fan a few years ago, I measured temperatures >(Motherboard Monitor software under Windows, sorry) with it going in >both directions, and having it blow out was a little better. It >actually didn't make very much difference in either direction, only a >few degrees less than no case fan at all. Don't forget that you are looking at the temperatures at an unknown spot somewhere on the motherboard (obviously, the CPU temp is inside the chip), and that's somewhat akin to a single thermostat in a house. It's a compromise, and only reflects what's going on "there". The orientation of the board is also a factor. The original IBM PC power switch location (right side at back) forced things so that if you decided to place the computer on it's side, it HAD to be left side down, which put the power supply (with the fan) on top. Putting it right side down was possible, but put the power switch (which you may not have been using anyway) in an awkward position, and would have created a less desirable heat removal situation. >I wonder if we haven't been looking in the wrong direction. Basically >everything works correctly except that 512MB DIMM. Is anything really >that urgent replacing it (or using "mem=1428M" until then)? If it's working, AND the system isn't being used for anything critical, then you can probably get away with it for now. I would put replacement high on the 'to-do' list. ]]>VBat: +2.03 V >> Getting back to the battery thread >Both the BIOS and lm_sensors consistently report 3.1V for the battery >voltage, and it's one lithium CR2032, so that sounds about right to me. So was that a typo? >>> System Temperature 40C/104F >>> CPU Temperature 33C/91F >> >> those look out of whack. > >Yes they do. When I start up a cold system, they're both below 30C, but >within seconds both are above 30C, according to the BIOS. Well, I >suppose the CPU temp is okay, just the MB temp that's too high. At a "cold start", they should show (very roughly) room temperatures. The CPU should ramp up quickly, because the POST, and in fact everything up through kernel load - is keeping the CPU very busy. Even when it's doing nothing, it's almost certainly using an "idle loop" (jump around in circles, waiting for something to happen) as opposed to using the x86 'HLT' instruction (which stops the CPU until the "next" interrupt occurs). The motherboard temps _PROBABLY_ should show something similar, but a lot depends on where the sensor is located relative to what-ever is heating things up. >Both the BIOS and lm-sensors show pretty much the same thing, so I >assume that's correct, to within a few degrees anyway. Gotta find >that temperature sensor on the motherboard. Correct, in that they are reading the same. Not necessarily correct in respect to absolute values. >According to the BIOS, they're all within 2%, so I don't really think >that's something I need to look into right now. Agreed >No. When the system has only been on for an hour or two, both the BIOS >POST and Memtest86+ don't find any problems. When it's been on for >several hours, they both find memory problems That _tends_ to indicate a temperature problem. I don't know how hard Memtest86+ is beating on the system, but the way a more common test is run is to beat the snot out of the system - in the GCC-SIG11-FAQ (also known as the Sig-11.FAQ) by 'bitwizard', the suggestion is to run a loop that repeatedly compiles a kernel. The problem with these tests is that "unused" memory could be bad, and not be detected no matter how hard you flog the system. Memtest86+ should be better for this, as it's a single function "operating system' that really doesn't have to worry about other things, but I have no experience with it. >But even when the system's cold, it hangs during boot unless I use >"mem=1428m", so something is flaky in that memory even when cold. Which might reflect something about the way the kernel is being loaded. >When the system's been on for a few hours, both the BIOS and Memtest86+ >find memory problems at all of those settings. BTW Memtest86+ (v1.70) >reports L1, L2, and RAM speed, and those settings result in 668, 689, >711, 718, 726 MB/sec for RAM speed respectively. If "Normal" is 100%, >then "Safe" is only 3% slower, and "Ultra" is only 5% faster. Not sure exactly what that's measuring - that doesn't sound as if it's doing very much. >I hooked it up to the motherboard's parallel port today, and it >installed and printed the test page with no problem at all. It's just >slower than I expected. Right now it's printing its own manual (PDF >file) and taking about two minutes per page. That seems a lot slower than I'm used to - I'm thinking it should be closer to 2 to 4 PPM after the initial wakeup. >Also I don't see an "OFF" switch on it. I don't have access to a 5 or 5M, but do recall that they had one, vaguely on the top on the display. >BTW it was mfd in June '97, has 20M RAM installed, and draws a whopping >10 amps! Not totally un-real, it's meant as a production style printer in a high usage situation. The last ones I saw were in the local community college in the computer lab. >I'm a little confused. No LAN, just the one system, and its built-in >Ethernet port is already connected to the DSL modem/router. Well, you _do_ have an existing network, but I certainly would not even hint at putting the printer on that. Look at the output of /sbin/ifconfig and /sbin/route -n and this will show what you have now. >The RJ-45 jack on the printer (which is an add-in card) is labelled >"10BaseT". Expected - 100Base* wasn't all that common in 1997. We were just starting to use it, and had two subnets (out of 11) running it. But then, we still had one 3Base5 (the original Ethernet) accessible, though by that time nothing was actually using it. >I gather that if I want to hook it up that way, I should install the >"new" NIC card (as eth1?) and run a cable from that to the printer. Yup - cross-over cable. >I just checked the web sites of the local retailers, and they do >indeed have things labelled "crossover cables" at prices from $4 to >$28. So is that what I should buy? Yes, though the top price seems high unless that's a long cable. >I gather I just have to buy one of those, and install the NIC card. >What are the advantages of that over using the parallel port? Would >it increase printing speed over its current 0.5 PPM? In your case (single system), it's not going to do that much for you. I was under the impression (for unknown reason) that you had more than one system. >> Does it have the option of an external battery? The internal >> batteries are user-replaceable, and _MAY_ be hot-swappable. >> http://www.apcupsd.org/ > >Found downloadable manual for that, too. I don't think it can use an >external battery, and it looks like a replacement internal battery will >set me back at least $80. Time to check eBay. or at least your favorite search engines. The 1500 is a large beast meant to handle the heaviest loads running unattended. For a single system in general use at home, there are UPS available that cost about the same as that battery. >I see it has serial and USB ports on the back. Is that to notify the >system that power's going down? (Sorry, I know almost nothing about >UPSs.) -rw-rw-r-- 1 gferg ldp 41481 May 23 13:03 UPS-HOWTO >Apparently the UPS originally came with software, which of course I >don't have. Will I need to search for that too, or does Mandriva >have something included? They may have, but see section 3.3 of that HOWTO. And while you're reading that, check section 5.5 about batteries. >My motherboard's one serial port is used by the faxmodem, but I have >some USB ports free. Will a standard USB A-B cable be sufficient? I can't say about USB, as I still use RS-232. I share the UPS, so have the serial cable going to one box, and a network interconnect to warn the other systems when power goes out. >Thanks! One of the benefits of being an active LUG member, I guess. >And also of showing up to get the stuff a day before everybody else. Yeah, there is that ;-) Old guy
From: Robert M. Riches Jr. on 27 Sep 2007 20:35 On 2007-09-27, Unruh <unruh-spam(a)physics.ubc.ca> wrote: > > Actually I doubt that you could push Gigabit signals down a coax. I also > find it astonishing that you can push Gigabit down a twisted pair, but it > works. The old-style (large) satellite dishes routinely pushed RF signals of 1GHz to 1.5GHz down a piece of coax. I installed the 102 feet of RG-6 between the dish and the receiver at one location, and it worked just fine for transponder 23. Properly done, you could push a lot more than 1Gb/s through a piece of coax. Now, handling bidirectional traffic and detecting collisions might be a bit tricky, but probably solvable. -- Robert Riches spamtrap42(a)verizon.net (Yes, that is one of my email addresses.)
From: Moe Trin on 27 Sep 2007 20:37 On Wed, 26 Sep 2007, in the Usenet newsgroup alt.os.linux.mandrake, in article <JeCKi.106905$bO6.104684(a)edtnps89>, Unruh wrote: >>Moe Trin wrote: >>> Sounds good. > >AARGH. Moe, you should know better. Bill - he's not running life support on it. It's not a critical system, and there are a few other things going on in his life. >They have lead acid batteries. They only last for a finite length of time, >just like in your car. Depending on the amp hour rating it is probably just >an motorcycle battery in there. -rw-rw-r-- 1 gferg ldp 41481 May 23 13:03 UPS-HOWTO Old guy
From: Moe Trin on 27 Sep 2007 20:39 On Thu, 27 Sep 2007, in the Usenet newsgroup alt.os.linux.mandrake, in article <RAFKi.91518$Pd4.80096(a)edtnps82>, Unruh wrote: >"Robert M. Riches Jr." <spamtrap42(a)verizon.net> writes: >>On 2007-09-26, Unruh <unruh-spam(a)physics.ubc.ca> wrote: >>> ( 4,5 7 8 are not used No idea why cat 6 is not 4 wire rather than >>> 8 wire, but...) Because Cat6 is aimed at GigE >>100mb ethernet uses all eight wires, doesn't it? There are several types of 100 MB Ethernet - several do indeed use four pairs, though they are quite rare. The "common" version does not. >AFAIK no, neither does Gigabit. Try searching for 802.3ab - GigE does require 4 balanced pairs Old guy
From: Christopher Hunter on 28 Sep 2007 01:49
Unruh wrote: > Actually I doubt that you could push Gigabit signals down a coax. I also > find it astonishing that you can push Gigabit down a twisted pair, but it > works. The input to my 22 GHz spectrum analyser front end is a BNC! I have 5, 10 and 18 GHz signals around here going down cheap coax with little loss. Chris |