1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

1.04 WHR-HP-G54 rebooting?

Discussion in 'Tomato Firmware' started by RobNC, Feb 11, 2007.

  1. RobNC

    RobNC Network Guru Member

    Has anyone experienced issues with the device rebooting (for no apparent reason)? I have a Buffalo WHR-HP-G54 and had overclocked the device to 228MHz (from stock 200). Lots of ppl on the DD-WRT and other forums have said that this device is pretty stable at this frequency. I also increased the wireless TxPower to 50mW (from default of 20 or something like that).

    Wish I could get CIFS to work, then I'd perhaps be able to log everything (hopefully before it chokes). SSH wouldn't work, for some reason, using Firefox (had to use IE7 to copy/paste the id_rsa.pub key).
     
  2. andersonbc

    andersonbc LI Guru Member

    I don't overclock my WHR-HP-G54 but it couldn't be more stable.

    This router has a built-in amplifier, so pumping up the broadcom chipset won't do much (and probably creates noise). I keep mine set at 10 mW and get great range.
     
  3. azeari

    azeari LI Guru Member

    actually as long as its 84mw and below, you shouldn't get much more noise. here i can only guess the overclocked cpu's restarting the router. well, it could also be lack of memory due to running too many services though
     
  4. RobNC

    RobNC Network Guru Member

    more memory

    Dunno if this makes any difference, but I did clear all the nvram parameters relating to memory, and then I replaced the SMT TSOP SDRAM chips with 2 32MB (256Mbit) chips.

    The original chips were 64Mbit x 2 => 8MByte x 2 MIRA chips:

    P2V64S40ETP-G6 =>
    backside:
    Mira 64Mbit/SDRAM 4M*16
    P2V64S40
    54-pin
    RAM : mira p2v28s40btp [5409fa03-6]
    spec: http://www.deutron.com.tw/data_sheets/sdram/p2v28s_0btp11_07024.pdf

    P2 == mira DRAM
    V == LVTTL
    28 == density (128mbit)
    S == synchronous DRAM
    4 == x16 organization (4 banks - 16-bit)
    0 == random column
    B == 3rd gen
    TP == TSOP(II)
    G6 == 166MHz 6ns (3-3-3)

    but since the memory operates at 100MHz anyway, no real need to go faster (so thus I only used 143MHz parts as the replacements).

    I replaced with Micron 4Mx16 -> MT48LC16M16A2 (256Mbit) x 2
    to give 64MBytes but Tomato only recognizes 32MBytes. I checked the traces on the board, and the unused (with 16MB) trace for the higher address bits exist on the PCB (A12, which is pin 36 of the memory). The block diagrams for the Micron memory (pg 9 of 256Mbit datasheet) and the MIRA (page 3) are compatible. I did see a few times that the Buffalo would dip below 14MB free (which means more than 16MB would have been depleted).

    I also used as a guide: http://oleg.wl500g.info/sdram.html (which I am not sure if it is accurate but I merely removed those NVRAM parameters, commit, then rebooted and it then is supposed to auto-sense during boot).

    Something else it could be, is perhaps some kind of memory leak in the interface, if perhaps you leave the webpage to the router up too long and refreshing too often. I have found that on the QoS page, there were a lot of opened (unclosed) connections to the router (more than a few hundred) if I let some auto-update page (i.e., site survey) run for a day.
     
  5. nsumner

    nsumner LI Guru Member

    The HP has a built in amplifier which must be enabled on the router. It also does not work above 10mw of power.

    The firmware automatically sets the boardflag to enable the amplifier and the power to 10mw. I have compared 10mw to 150mw (where the amplifier will automatically get disabled), and get better results with 10mw.
     
  6. azeari

    azeari LI Guru Member

    if i remembered correctly, the broadcom chip should be able to recognise all 64mb of memory, then again tomato won't need that much ram anyway.

    well anyway, the unclosed connections could possibly be due to bad programming on the interface's ajax i guess. but it shouldn't make much of a difference anyway, since the default contrack settings is to expire open connections after 4hrs, and with 14 or more mb of free ram, it clearly isn't the issue of the router rebooting

    -shrugs- try lowering the clock speed back to 200mhz, or 216mhz and sees if it makes a difference
     
  7. RobNC

    RobNC Network Guru Member

    TxPwr, Clkfreq

    I did change the power to 10mW and sure enough, it looked similar to the settings for 63mW. I thought that the external amp was controlled via nvram variables but maybe there's another way to control it that is beyond nvram vars.

    I also set freq back to 200MHz to see if that would make a diff. I don't know why it won't recognize 64MB. I wanted 64MB so I could run Xwrt if I wanted (which would also involve replacing the flash, which is possible provided that the density can be correctly read via bootloader). But I also wanted more than 16MB since tomato seems to only leave less than 1MB free (on my WRT54G v1.0 w/16MB).
     
  8. GeeTek

    GeeTek Guest

    There are 2 extra amplifiers in the HP. One is to amplify the transmit power, and is on all the time. The second extra amplifier is to boost the receiver, and can be turned on or off with the board flags.

    Edit - I would check your 5v power supply to be sure that it is producing a clean and regulated 5 volts.
     
  9. StevenG

    StevenG LI Guru Member

    Do you know what the boardflag is for the receive amp? I'd like to verify it's on in Tomato 1.06.

    Thanks.
     
  10. GeeTek

    GeeTek Guest

    0x1758 = Pre amp Off, 0x2758 = Pre-amp on

    Telnet to router. Telnet is a network command environment. Open a telnet session in windows by typing telnet 192.168.1.1 in a command window.
    The username is 'root', password 'admin'
    Enter the following commands

    nvram show | grep boardflags
    boardflags=0x1758


    if you have 0x1758, change it to 0x2758
    Use this code:

    nvram set boardflags=0x2758
    nvram commit
    reboot

    This setting appears to turn on a feature on the HP receive side, dramatically increasing receive capability.
    This feature is not saved with the GUI Backup, so must be redone after every instance of going to default.
     
  11. StevenG

    StevenG LI Guru Member

    Thanks.

    Mine is already set to 2758. Are you sure that's the receive amp though? I thought, from everything I've read (and admittedly, there's a lot of confusion) that that boardflag (0x1758 or 0x2758) is the transmit amp. Perhaps the transmit amp is always on, and this is for the receive amp? Is there another setting for the transmit amp?

    Thanks again for replying, didn't mean to hijack the thread.
     
  12. GeeTek

    GeeTek Guest

    I know at very least for a fact that switching those numbers back and forth will definitely toggle a 15 Db boost in receive sensativity. I just saw on the forum that those instructions were copied from that the newer firmware releases also take these settings into account when adjusting the TX power. I know that the older versions of DD wrt do not take the board flag setting into account because I have fried more than one HP model with RF settings in excess of 35 Mw. As far as I can tell, the TX amp is always engaged, and cannot be turned off. How Tomato or other newer firmware adjusts the TX drive is really speculation, and is badly in need of testing with a watt-meter or service monitor. It seems that Tomato 1.04 will still allow the RF to be cranked up to very high levels, but there is also a lot of beer in the fridge........
     
  13. RobNC

    RobNC Network Guru Member

    what is actual power?

    Heck I just realized I bought a spectrum analyzer for work (Agilent MXA) and I could actually measure the TxPower. The only two problems here are the cabling and the fact that our lab uses wifi as well. Also, I guess I'd have to set the SA into peak-hold mode and measure the TxPower by scanning very slowly.
     
  14. GeeTek

    GeeTek Guest

    That right there qualifies you for a nick-name of "Jesus Christ" (Unless of course you would take that as an insult). Anyway, all you need is a 2' patch cable with an RP-SMA connector on one end, and a TNC or N connector (Or whatever connector the spec-an needs) on the other end. The other wi-fi traffic in the area will not be a factor. In the Tomato firmware select antenna "B" for TX and RX (external antenna port). Then set the spec-an for center frequency of CH 6 Wi-Fi (I'll find the freq real quick if need be) and a bandwidth of 10 MHz / Division. The spec-an should be calibrated in DBm which readily converts to MW. Just turn the radio on (CH 6) and you will immediately get a reflection on the scope indicating the power output. Then play with the MW settings in the firmware to see what they do to the output, or tell me where you are and I'll fly over and help ! Don't worry about the RX sens of the radio. That part is real easy to figure out with field tests, and would be hard to do with the machine.
     
  15. RobNC

    RobNC Network Guru Member

    RP-SMA connector

    That's the only issue I can see. We've never needed or used a RP-SMA connector on one end. I could solder one up, I suppose, to some semi-rigid cable, if I could find some. Most high-end SA's (like the one I use) uses 50-ohm N-type. Only problem here would be to clone my config to another router, as I'm currently using my WHR-HP-G54 for VoIP, and my S.O. can't stand to be out of touch with my MIL :)

    Anyway, the SA is pretty nice; 25MHz simultaneous analysis b/w, 20001 points per sweep, etc. In fact, there are a bunch of tests that are pre-built for 802.11b/g type tests (for actually demodulating the signal!). Just too bad that I can't use those functions in my real job as there are no decoding packages for it yet (can't disclose due to NDAs etc.).

    I'm guessing that the same frequencies are used for 802.11g as for .b, and from this: http://www.qsl.net/n9zia/dsss-channels.html (shows 802.11b). Weird thing about 802.11g is it's supposed to be OFDM which means multiple sub-carriers modulated at the center frequency. Ideally, DSSS would be better, if implemented with multiple rake receivers for multipath correlation (but those are $$). I haven't been able to get a good hit from google describing the min/max bits/bin for the OFDM modulation for 802.11g, as well as cyclic prefix overhead, RS overhead, etc. Wikipedia is quite lacking there, perhaps due to fee-based standards body?
     
  16. GeeTek

    GeeTek Guest

    For determining the RF output level you relly needn't be concerned with modulaion schemes or how to de-mod/de-code them. It is all digital spread spectrum, and just a plain vertical deflection reading will suffice nicely. I am reaching on this one, but I belive the "Basic" rate setting in the "Advanced Wireless" setting is going to be the width of the modulation envelope as long as the radio remains un-synchronized with any clients. This would be either 1 or 2 MB in width, for both the "B" or "G" protocol. Just powering the radio up will induce the normal SSID broadcast, and should provide a fairly constant RF output.
     
  17. RobNC

    RobNC Network Guru Member

    another approach

    or another approach, would be to get the data sheets on the RF amp, then measure if it is enabled or not. I am guessing here; it is possible that an RF switch is used to either bypass the RF amp or to send the signal to the RF amp. Then a simple voltmeter could be used to confirm/deny the actual RF settings.
     
  18. GeeTek

    GeeTek Guest

    Let me guess. You work for Uncle Sam. He buys you an $85,000 piece of test equipment, sends you, a bright, intelligent young man to school for 3 years to learn how to use it, gives you a nice solid understanding of RF electronics and RF modulation schemes, a faraday cage, and a $2,986 dollar soldering station with hot air vacuum assisted de-soldering capability, and then runs out of budget when it comes time to buy the $40 universal RF adaptor Kit to allow you to actually use all those goodies. The only access point you could commodeer is from the top of a locker that the last nerd before you left behind and it was urgently pressed into "official" VOIP use, and cannot be dis-connected on pain of a court martial, since all the pots lines are bugged. Somehow, I'm not the least bit suprised.
     
  19. GeeTek

    GeeTek Guest

    How do you propose coupling and rectifying a milli-watt 2.4 GHz signal into a simple volt-meter with a 2:1 or better SWR ? If you do have an answer for that, what formula would you apply to deduce the RF level from the voltage you detect ? That is why I got so excited when you said you had a rectum-spannalyser !
     
  20. StevenG

    StevenG LI Guru Member

    Interesting. I read those threads on dd-wrt too. That's where I thought the boardflag controlled the transmit amp. Could still be some confusion, but I understand what you're saying. Either way, we are in agreement that there is only one boardflag to verify. Thanks for replying.
     
  21. RobNC

    RobNC Network Guru Member

    Not even close. Nice try, though! I was quite amused at your imagination.

    The problem here is that I can't justify purchasing something that will NOT be used for my job.
     
  22. RobNC

    RobNC Network Guru Member

    I think you're taking this waaay too seriously. Think more like an engineer with "stone knives and bear skins" (if I can steal an excerpt from Star Trek). If it has an amp, I am proposing measuring it's "output enable" signal to see if/when it is turned on. Presumably, it wouldn't make too much sense to have an RF amp powered and not use it, so there must be some sort of pass-through mode. And presumably it's off-the-shelf (to make it cheap), so data sheets should be readily available, I would assume. I am only using some information I've gathered about non-HP versions and how the nvram settings don't do anything on the other versions.

    If no pass-through mode, then, there must be an RF switch (pin) or two switches, that allows the antenna to either receive the signal from the BCM SoC internal amp, or from the on-board amplifier.

    Don't worry so much about the power level, just first determine what settings will enable the on-board amp. Isn't that what was asked here? After that is verified, then the amplitude must be measured. That will have to be measured from an SA.

    [Edited after reading GeeTek's post above]

    So if you're saying that the amp is always on (which conflicts with some posts about the boardflags; adding to my confusion), then my test doesn't make sense. I thought that the WHR-HP-G54 and WHR-G54 only differed by the HP having an on-board amp. Some posts have said that you can't use the SoC amp and the external amp at the same time, but who knows. Maybe we should ask Broadcom (let me know when you get done laughing).
     
  23. GeeTek

    GeeTek Guest

  24. RobNC

    RobNC Network Guru Member

Share This Page