I'm going to replace the existing mini-PCI card in my laptop. The question is, which new card to get?
The senao, ubiquiti sr2 and wistron cm9 cards all look like contenders.
Which ones perform the best all-round under XP? With netstumber? Under linux? With Kismet?
I'd be using this just as a workstation, not as a router, access point or anything else. I'd be starting with just the built-in antennae that have U.fl cables but will eventually hack the case to allow for attaching an external antenna. I see that at least one card has a single U.fl connector so that'd hasten my addition of the external connector.
So whose card should I consider? Which ones are a pain to use and should be avoided?
And no, I do not want a PCMCIA card or a USB dongle. They're fine ideas but I'm interested in having the card remain built-in.
Under XP all of them more or less work. Under Linux, all of them more or less don't.
I've been trying to get fully-functioning wireless under around a dozen different distros with three different current chips for over a year now with no significant success. The Linux forums seem clogged with howling over poor wireless support. It's entirely justified.
If you plan to kludge an external antenna, I'd also avoid the RT2500-based cards which usually have tuned dual-diversity antennas, which I suspect might be a problem to hack.
Which maker and model laptop? (Your original message didn't appear on Newsguy). Be advised that current IBM, HP, and possibly Compaq laptops all have a "feature" which prevents MiniPCI cards that have not been type certified with the specific laptop from working. The BIOS detects the card and won't let the machine boot. There are workarounds but most are no fun.
Current favorite card is Intel 2200BG with Proset 10.x.
Incidentally, MiniPCI cards are much like WinModem PCI cards. The bulk of the "intelligence" is in the drivers and not in hardware on the card. The card determines the RF performance, but the drivers determine how well it deals with errors, PAD (packet assembly disassembly), retransmissions, user control, and such. Find the best driver, and then buy the card that goes with it.
That won't be a problem. It's a Toshiba laptop and it's already had several different cards in it over time. All have worked just fine. The only thing the 'genuine' toshiba card can provide is support for their on-board wifi on/off switch. But it's a software switch that only their driver supports. I've lived without that for quite a while. It's had a Dell 1300 Truemobile (b/g) card in it for a while an Intel 802.11b before that. I just discovered they hobbled the driver upgrade process by fiddling with the setup.ini file. Damned driver install wouldn't work until I whacked some values out of the ini file. Idiots, tough to say whether it's Dell or Broadcom to blame. I'm sure there's enough to go 'round. I can now control which antenna it favors, so now I can add that external pigtail.
The Proset stuff is what my wife's ultra-tiny Sony Vaio uses, can't recall which model card. It does seem to work well. That and her machine gets amazing battery life. Mine's a beast that doesn't usually last more than a hour on batteries. So I'm less worried about power saving issues. I want better WiFi use.
Hmmm, so are there any seriously negative issues with the drivers for the Senao and Ubiquiti cards? I looked at those for their power capabilities. I've never been particularly thrilled with how the inboard antennae on the laptop have worked. Getting a card with the possibility of more power holds some appeal.
I find it somewhat odd that the miniPCI cards would be described similarly to a winmodem. Given it's a PCI bus you'd think the cards would be as capable as a regular PCI card, certainly more than a PCMCIA slot device. Is that not the case? Or, much like modems, does it vary from card to card?
I'm always amazed at their power output. MMIC power amp efficiency at
2.4Ghz is about 10%. So, to get 400mw of output, the card requires
4000mw of power. Where does the other 3.6 watts go? In heat, or course. Fortunately, it's only in transmit so the card won't overheat or melt. |
1.3A at 3.3V in xmit.
The antenna problem is usually the type and location of the laptop antennas. Some are no better than a joke. For example, the antennas on the Compaq 2120US are under the display hinges, a truly hideous location: |
's a 1/4 wave meandering monopole, on high loss G10 circuit board. If you look closely, the black coax was seriously crunched by the hinge, which is what inspired the dissection and repair job. Hopefully, the Toshiba antennas are somewhat better.
The comparison with a Winmodem is mine and of course, since it's mine, must obviously be correct. Prior to Winmodem type technology, almost all of the data extraction and encoding for modems was done in hardware. The card presented itself to the ISA or PCI bus as a UART, with parallel ASCII data ready for consumption by the processor. The problem was that every time a character arrived, the UART would toggle a hardware IRQ line, the interrupt would cause the CPU to flush everything into the stack, and execute the serial interrupt routine. That was fine for 1200 baud modems, but as the data started coming faster and faster, the constant CPU interruptions for every character became a big load on the system.
So, the solution was to move much of the processing into the CPU. Since the data was already in the CPU, it didn't need to cross back over the ISA or PCI bus. There was no need to execute CPU cycle gobbling IRQ's. As an added bonus, it saved a bunch on hardware costs. At first, the 286 vintage processors had a difficult time doing the character decoding and processing. It was a gutless processor and couldn't keep up. This gave the Winmodems an undeservedly bad reputation.
However, as processors and memory started to improve, less and less of the processors daily life was spend dealing with serial data in and out of the modem. I think it was the 486DX25 vintage where the CPU load from a UART type modem, was equal to the CPU load presented by a Winmodem. As CPU's became even faster, the percentage CPU load for the Winmodem was drastically reduced to almost negligible.
Eventually, the same problem appeared with wireless devices. Why ship expensive hardware when there's a handy CPU that can do the same work for free? While the original wireless PCMCIA cards had chips that did most of the MAC layer crunching, the later and cheaper cards, along with MiniPCI and USB wireless just had the CPU do the MAC layer stuff (retrans, error control, PAD, power management, ad nausium). Most wireless routers contain a dedicated MAC chip to do this job, but the smaller MiniPCI and USB devices do it in the driver.
Does the monstrous 80Mbyte (compressed) Intel Proset download give you a clue as to where the real work is done?
In the PCMCIA and PC Card area, it does vary. Some have MAC chips, some do it in the driver. However, in the USB and MiniPCI area, all of them do the MAC layer stuff in software/drivers. Of course, routers and access points do it in a dedicated MAC chip
Oops. Forgot a few details. The MAC chip functions aren't totally unloaded onto the CPU.
Almost all current chipsets have a MAC processor chip. For example, the AR5213 "Multiprotocol MAC/baseband processor":
chips do much of the MAC layer processing and store the various radio timing and protocol generation settings. However, they don't do everything, such as the actual bridging. In addition, some of the functions often found in the RF section, such as the A/D and DAC's for modulation and demodulation (i.e. baseband processing), are in the MAC chip.
Also, WPA2, 802.11i, and AES encryption are sometimes implemented in either software or in hardware. The processing load for crunching
3DES or AES encryption and decryption is rather substantial. I'm amazed that it can even be done in software. Initially, it had to be done in software because there were no chipsets that offered AES hardware encryption. However, the current chipsets all do AES in hardware.
Incidentally, here's a new item in the AR5000x chipset. It has a "Wake-on-Theft" feature.
Atheros is also introducing a new Wake-on-Theft capability. This function can alert network administrators if a mobile device such as a laptop or PDA equipped with the new Atheros chipsets is removed from a company?s facilities without authorization. The Wake-on-Theft function can also wake-up and raise an alarm on the mobile device, even if the device is powered off.
That's the chipset used by the Ubiquiti MiniPCI card:
Heh, good thing the output is configurable. As I said, I'm less concerned about battery savings since this beast is already a power hog. I end up using it tethered to a power output 80% of the time.
Don't have a picture but there's a pair of them on 1" square flexible circuit boards up near the top edge of the lid. Their location 'sucks less' than that toshiba, that's for sure. I just had to replace them as a U.fl connector gotten ripped off the old one during a cleaning (wife spilled coffee on me and the laptop). The coax does run through the hinge but it's done in a pretty reasonable fashion. I suppose it might be worth considering use of some other type of antennae internally. But I've not seen much in the way of replacements, let alone reviews discussing them.
But of course!
I'd wonder how to determine which ones do or don't?
No idea. Yes, that eff'ing driver is HUGE. I've no idea how much of it actually gets installed or not. I just updated the wife's laptop last night. The device manager shows it as a 2200bg card but I have no idea if it's a mini-PCI card or embedded. The 10.1.3 driver does appear to have more configurability than the 9.x series. One new option being power output. Cranking it up does give better wifi signal in an area of the house that's been marginal up until now.
There really doesn't seem to be much coverage online of which cards are better performers thatn others, and why. I suppose it's no surprise given the vertitable avalanche of new devices being shipped every day. But still, I'd like to find some better reviews that focus on more of these sorts of details.
Where else online are these issues discussed besides in this newsgroup?
On Fri, 30 Jun 2006 21:03:42 -0400, "Bill Kearney" wrote in :
The problem is that more power won't necessarily help, since that doesn't help the receive side. In Jeff's colorful language (when referring to more transmit power at an access point):
This is commonly known as an alligator, which is an animal with a big mouth and small ears. The xmit amplified access point can be heard over a much larger area than it can hear the replies from the clients. Unless the client radios have a similar power amplifier, the system become asymmetrical, with more range in one direction than the other.