A7N8X-DLX 2.0 Dimm Regulator Circuit Failure
in [Asus]
|
From: Kyle on 4 Apr 2008 03:02 Anyone have tech info on the Vdimm regulator circuit for this mobo? My A7N8X-DLX board died, after a simple power down. Upon removal from the system, I finally discovered why: 2 swollen caps in the area of the Vdimm regulator circuit (looks like there are actually 2 regulator circuits in that vicinity). I replaced the swollen caps (820uF 6.3v caps), but board was still dead, and I believe one of the surface mount power transistors in the Vdimm circuit is bad based on some ohm checks. Anyone have any circuit info on this or tips/tricks/advice to offer please do! The Vdimm is about .6 volts right now, based on some voltage probing of the location identified by some volt mod sites, which explains why I get no post, but I DO get a bad betty "cpu not found" message if I remove the CPU and turn the board on, so it's not quite dead yet! FWIW, I dropped all the old stuff into another identical board and all the parts work, thankfully. Am merely interested in fixing my bad board, have soldering iron, will solder new parts on board. Also, I need a suggestion on where to get some replacement transistors/FETS (have not yet found substitutes for the bad component parts), and yes, I'm aware of Digikey, but $25 min order is required. -- Best regards, Kyle
From: Paul on 4 Apr 2008 04:10 Kyle wrote: > Anyone have tech info on the Vdimm regulator circuit for this mobo? > > My A7N8X-DLX board died, after a simple power down. Upon removal from > the system, I finally discovered why: 2 swollen caps in the area of > the Vdimm regulator circuit (looks like there are actually 2 regulator > circuits in that vicinity). I replaced the swollen caps (820uF 6.3v > caps), but board was still dead, and I believe one of the surface > mount power transistors in the Vdimm circuit is bad based on some ohm > checks. > > Anyone have any circuit info on this or tips/tricks/advice to offer > please do! > > The Vdimm is about .6 volts right now, based on some voltage probing > of the location identified by some volt mod sites, which explains why > I get no post, but I DO get a bad betty "cpu not found" message if I > remove the CPU and turn the board on, so it's not quite dead yet! > FWIW, I dropped all the old stuff into another identical board and all > the parts work, thankfully. Am merely interested in fixing my bad > board, have soldering iron, will solder new parts on board. Also, I > need a suggestion on where to get some replacement transistors/FETS > (have not yet found substitutes for the bad component parts), and yes, > I'm aware of Digikey, but $25 min order is required. > There would probably be two supplies in the area. You need something for the rail supply for the DIMMs. Also, you need something for the termination supply. For the thing that is the switching circuit, you'd start with the datasheet for the regulator chip. That will have info on how to select components for a design. Since you have an identical board, you can do some ohmmeter checks, and compare readings from the good board, to the bad board. You may be able to detect the failed component, by such a comparison. Or you may get lucky, and spot some heat damage to some component. Collateral damage is always fun to clean up, and many happy hours can be spent, trying to find a MOSFET that comes close to the bad one. The one I was working on, the first trick was the original part number. The number on the MOSFET was Kxxxx and I started looking for that, and wasn't having much success. I eventually discovered that the part number was 2SKxxxx and they'd decided to drop the leading two characters. And the smaller the components get, the more cryptic the marking scheme. This is an example of a DDR regulator. It includes a PWM regulator, with Top Gate and Bottom Gate drive for the switching part. That gives you 2.5V for DDR. It also has integrated, a linear regulator for VTT (sink/source termination current). Not the best example, but the first one I found. http://www.semtech.com/pc/downloadDocument.do?id=468 Here is another one, from Intersil. Intersil gives a few hints on component selection, and in this case, that starts on page 14. http://www.intersil.com/data/fn/fn9143.pdf So you'd want to start, by identifying the regulator chip in the vicinity of the Vdimm regulator. Paul
From: Kyle on 4 Apr 2008 12:27 "Paul" <nospam(a)needed.com> wrote in message news:ft4nau$ra0$1(a)aioe.org... | Kyle wrote: | > Anyone have tech info on the Vdimm regulator circuit for this mobo? | > | > My A7N8X-DLX board died, after a simple power down. Upon removal from | > the system, I finally discovered why: 2 swollen caps in the area of | > the Vdimm regulator circuit (looks like there are actually 2 regulator | > circuits in that vicinity). I replaced the swollen caps (820uF 6.3v | > caps), but board was still dead, and I believe one of the surface | > mount power transistors in the Vdimm circuit is bad based on some ohm | > checks. | > | > Anyone have any circuit info on this or tips/tricks/advice to offer | > please do! | > | > The Vdimm is about .6 volts right now, based on some voltage probing | > of the location identified by some volt mod sites, which explains why | > I get no post, but I DO get a bad betty "cpu not found" message if I | > remove the CPU and turn the board on, so it's not quite dead yet! | > FWIW, I dropped all the old stuff into another identical board and all | > the parts work, thankfully. Am merely interested in fixing my bad | > board, have soldering iron, will solder new parts on board. Also, I | > need a suggestion on where to get some replacement transistors/FETS | > (have not yet found substitutes for the bad component parts), and yes, | > I'm aware of Digikey, but $25 min order is required. | > | | There would probably be two supplies in the area. You need something | for the rail supply for the DIMMs. Also, you need something for the | termination supply. | | For the thing that is the switching circuit, you'd start with the | datasheet for the regulator chip. That will have info on how to select | components for a design. | | Since you have an identical board, you can do some ohmmeter checks, | and compare readings from the good board, to the bad board. You | may be able to detect the failed component, by such a comparison. | Or you may get lucky, and spot some heat damage to some | component. Collateral damage is always fun to clean up, and many | happy hours can be spent, trying to find a MOSFET that comes close | to the bad one. | | The one I was working on, the first trick was the original part | number. The number on the MOSFET was Kxxxx and I started looking | for that, and wasn't having much success. I eventually discovered | that the part number was 2SKxxxx and they'd decided to drop the | leading two characters. And the smaller the components get, the | more cryptic the marking scheme. | | This is an example of a DDR regulator. It includes a PWM regulator, | with Top Gate and Bottom Gate drive for the switching part. That | gives you 2.5V for DDR. It also has integrated, a linear regulator | for VTT (sink/source termination current). Not the best example, | but the first one I found. | | http://www.semtech.com/pc/downloadDocument.do?id=468 | | Here is another one, from Intersil. Intersil gives a few hints on | component selection, and in this case, that starts on page 14. | | http://www.intersil.com/data/fn/fn9143.pdf | | So you'd want to start, by identifying the regulator chip in the | vicinity of the Vdimm regulator. | | Paul Thanks Paul, you did not have to go to the effort to find the sample circuits as I have done such before (fixed my 350W Antec supply recently using sample circuits and reverse engineering my circuit), and your effort is appreciated. The other regulator circuit in that area of the board looks to be providing 5v, so I think it is ok, tho I'm not certain as to its normal output voltage. Does that sound right for the second voltage? No visual heat damage appears on the devices. I can measure resistors easily, it's a matter of knowing where the resistor resides in a circuit that determines, of course, what value one should measure, and having the circuit schematic is critical to know if the resistors are correct in value. I suppose a schematic is unavailable, as is typical with mobos. As to fixing switched mode circuits, I have a scope, variac and dvm and am able to fix most anything. Agreed, finding the collateral damage is more challenging, but in this case, the circuits are fairly simple, and typically the damage done by a bad cap is normally found in the switching transistors, tho I agree, other components can also get burned. My problems all started when I initially started to swap out the PS from my system, I observed careful static grounding techniques, PS swap was in fact cancelled once I got it out and reviewed the specs sticker, I reinstalled same PS (the PS was allegedly a 450w, and I was moving it to another machine which now has a new vid card that requires lots of power, but as it turns out, my 350W Antec had better specs for combined output and 12v rail current capability than the 450w PS, not unusual), then machine would not post. PS outputs measure fine, PS is not bad as it is running my "other spare" board now. Funny how a simple operation such as this results in disaster. When machine would not post, I began inspecting things. Found the CMOS battery very low, changed the battery, but had trouble with clearing CMOS as the machine was setup to automatically turn on when power was applied and this action/feature would not "clear" or reset when I cleared CMOS, oddly. So the mystery is what caused the vdimm circuit to fail, machine ran w/o problem prior to the PS removal/reinstall, so I doubt the vdimm circuit was bad beforehand. I'm pretty darn certain I turned PS off (switch on back) before removal and install, and only 1 stick of memory (512meg) is not a load likely to overheat the caps in the regulator circuit. Frustrated, I even swapped BIOS chips between my boards (found out the custom FSB frequency of this system is in the BIOS, no wonder these nforce2 boards have failures with BIOS ROM since FSB info is stored in the flash ROM when custom FSB's are setup). The machine is never unplugged, and powered up 24/7, so a bad CMOS battery would not be discovered until a power down event, such as PS swap, occurred. Oh well, just another day gone bad in the computer ownership world. The voltage on the Vcore circuit seems to track the type of CPU I install, an old Athlon 1400 showed 1.75v, and a newer Applrebred Duron 1600 showed 1.5x volts, so the Vcore circuit seems to setup correctly. Vdimm regulator circuit appears to be a RT9202 chip, found here: http://www.richtek.com/www/Docs/DS9202-08.pdf . I wonder if the Vdimm controller somehow got programmed up wrong because of CMOS battery failure? Guess I'll be probing around some to find the answers. -- Best regards, Kyle
From: Paul on 4 Apr 2008 17:45 Kyle wrote: > > Thanks Paul, you did not have to go to the effort to find the sample > circuits as I have done such before (fixed my 350W Antec supply > recently using sample circuits and reverse engineering my circuit), > and your effort is appreciated. The other regulator circuit in that > area of the board looks to be providing 5v, so I think it is ok, tho > I'm not certain as to its normal output voltage. Does that sound > right for the second voltage? > > No visual heat damage appears on the devices. I can measure resistors > easily, it's a matter of knowing where the resistor resides in a > circuit that determines, of course, what value one should measure, and > having the circuit schematic is critical to know if the resistors are > correct in value. I suppose a schematic is unavailable, as is typical > with mobos. > > As to fixing switched mode circuits, I have a scope, variac and dvm > and am able to fix most anything. Agreed, finding the collateral > damage is more challenging, but in this case, the circuits are fairly > simple, and typically the damage done by a bad cap is normally found > in the switching transistors, tho I agree, other components can also > get burned. > > My problems all started when I initially started to swap out the PS > from my system, I observed careful static grounding techniques, PS > swap was in fact cancelled once I got it out and reviewed the specs > sticker, I reinstalled same PS (the PS was allegedly a 450w, and I was > moving it to another machine which now has a new vid card that > requires lots of power, but as it turns out, my 350W Antec had better > specs for combined output and 12v rail current capability than the > 450w PS, not unusual), then machine would not post. PS outputs > measure fine, PS is not bad as it is running my "other spare" board > now. Funny how a simple operation such as this results in disaster. > > When machine would not post, I began inspecting things. Found the > CMOS battery very low, changed the battery, but had trouble with > clearing CMOS as the machine was setup to automatically turn on when > power was applied and this action/feature would not "clear" or reset > when I cleared CMOS, oddly. So the mystery is what caused the vdimm > circuit to fail, machine ran w/o problem prior to the PS > removal/reinstall, so I doubt the vdimm circuit was bad beforehand. > I'm pretty darn certain I turned PS off (switch on back) before > removal and install, and only 1 stick of memory (512meg) is not a load > likely to overheat the caps in the regulator circuit. Frustrated, I > even swapped BIOS chips between my boards (found out the custom FSB > frequency of this system is in the BIOS, no wonder these nforce2 > boards have failures with BIOS ROM since FSB info is stored in the > flash ROM when custom FSB's are setup). The machine is never > unplugged, and powered up 24/7, so a bad CMOS battery would not be > discovered until a power down event, such as PS swap, occurred. Oh > well, just another day gone bad in the computer ownership world. > > The voltage on the Vcore circuit seems to track the type of CPU I > install, an old Athlon 1400 showed 1.75v, and a newer Applrebred Duron > 1600 showed 1.5x volts, so the Vcore circuit seems to setup correctly. > Vdimm regulator circuit appears to be a RT9202 chip, found here: > http://www.richtek.com/www/Docs/DS9202-08.pdf . > > I wonder if the Vdimm controller somehow got programmed up wrong > because of CMOS battery failure? > > Guess I'll be probing around some to find the answers. I was thinking, if a MOSFET fails short, you may be able to detect that with an ohmmeter. By comparison to your powered down, but still functional motherboard. Maybe. There is a bunch of stuff, like banks of capacitors, that the meter won't be able to charge very easily. The 0.6V is about a diode drop, but that may not have any significance. These are specs from a DDR datasheet. It appears VTT is 1/2 the supply voltage. On some boards, that can even be done with a five pin linear regulator chip. The regulator requirements on that one are special, as the regulator has to be able to sink some amps or source some amps, on a cycle by cycle basis. There are usually a bunch of chip caps on that rail, to smooth out the cycle by cycle high frequency variations, and give the regulator a chance to respond to the changing current. The termination type on the memory bus is SSTL (stub series termination), and the termination current that results is what VTT is handling. Supply voltage VDD +2.5 to +2.7 V I/O supply voltage VDDQ +2.5 to +2.7 V I/O reference voltage VREF = 0.49 � VDDQ to 0.51 � VDDQ I/O termination voltage (system) VTT = VREF - 0.04 to VREF + 0.04 More info here. http://en.wikipedia.org/wiki/Stub_Series_Terminated_Logic http://www.jedec.org/download/search/jesd8-9b.pdf (fig.3 page 14) With regard to the RT9202, it is a pretty "dumb" chip, and doesn't take a VID code, like a larger device might. The feedback ratio of R2 and R3 appears to set the voltage. That could be modified by a GPIO pin and resistor (if you wanted a second voltage value like 2.7V). The GPIO pin would be tristated (open circuit) when the board starts, causing the regulator to start at 2.5V. The BIOS doesn't need DRAM, to start computing. The BIOS code eventually gets to the part, where it sets the DDR voltage, if that voltage is programmable via a GPIO trick. They would enable the GPIO and do whatever is necessary, once the CMOS setting for Vdimm is consulted. If the CMOS is corrupt, the CMOS checksum will be wrong, and defaults will be loaded in their place. The CMOS checksum is not a strong check, so a bogus set of values could in theory be accepted. But the Vdimm probably has limits as to how far it can be programmed anyway. Asus are not DFI after all - they don't generally design in good "burn potential" :-) http://www.richtek.com/www/Docs/DS9202-08.pdf And even when a BIOS shows four values for Vdimm, the actual circuit may only have two values. Just because the BIOS shows them, doesn't mean the implementation matches the interface. The RT9202 has overcurrent protection, so it may be able to protect against some fault modes. It looks like the OC amplifier is looking for the voltage developed between OC and Phase. HTH, Paul
|
Pages: 1 Prev: Windows XP Pro OS Next: flashing the BIOS for the A7N8x e |