Building a ddr booster

[stealth17]

is there anyone that would have the knowledge to help me build a ddr booster?

what im looking for is Xtreme Clean power, and something that can adjust and display the voltage in thousandths.

i dont need vtt tracking or any of that stuff just a something basic to supply stable and clean power to my ram. i can solder it to the boards vdimm mosfet if possible to eliminate needing a pcd card in a memory slot.

if this all works nicely im considering making one for the vcore :woot:

i know a little about circuitry, ive been reading up on power stabilization and EMI. but i still think im ganna need a bit of help :p:

thanks! :toast:

[Micutzu]

You should try this out, it works just like the booster, but the voltage drops in load at least 10 times less, and the power delivered is also much cleaner, especially if the 1000uF capacitors are high quality.

http://www.crazypc.ro/forum/attachment.php?attachmentid=621

The "remote sense" and the "output" wires need to be connected together, but only at the soldering point on the motherboard, this way you'll have the best voltage stability. Watch out, the BD135 and especially the BD911 transistor gets very hot, and will need heatspreaders.
The output voltage will be adjustable from 0 to ~4.2V, with an output current that is more than enough for 2GB of TCCD or BH-5/UTT in any conditions. The maximum voltage depends on the stability of the +5V line, and is about 0.8V lower; if you have an adjustable PSU and you'll get the +5V rail up to let's say 5.5V, then the maximum voltage will be 5.5-0.8=4.7V.
It doesn't display the voltage though, you'll need a multimeter for that.

[persivore]

This (http://www.drunken-student.co.uk/tom/mods/DDR%20Booster/complicated/mso_Booster_complicated_CD.jpg) is the circuit that I used to provide regulation for the DDR booster. It can be simplified a lot by removing some of the components, but it might make the Vdimm supply slightly more unstable. It will provide a variable Vdimm supply of between 2.5 to 4.3v iirc, with a fluctuation of 0.01v with good grounding, or possibly better. You can use any general purpose MOSFET, as long as it is able to withstand the high currents RAM needs (a rating of about 15A should be suitable)

[stealth17]

You should try this out, it works just like the booster, but the voltage drops in load at least 10 times less, and the power delivered is also much cleaner, especially if the 1000uF capacitors are high quality.

http://www.crazypc.ro/forum/attachment.php?attachmentid=621

The "remote sense" and the "output" wires need to be connected together, but only at the soldering point on the motherboard, this way you'll have the best voltage stability. Watch out, the BD135 and especially the BD911 transistor gets very hot, and will need heatspreaders.
The output voltage will be adjustable from 0 to ~4.2V, with an output current that is more than enough for 2GB of TCCD or BH-5/UTT in any conditions. The maximum voltage depends on the stability of the +5V line, and is about 0.8V lower; if you have an adjustable PSU and you'll get the +5V rail up to let's say 5.5V, then the maximum voltage will be 5.5-0.8=4.7V.
It doesn't display the voltage though, you'll need a multimeter for that.

pic doesnt work :(

[Micutzu]

There you go.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=31748&stc=1

[stealth17]

There you go.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=31748&stc=1

looks nice. should i add any inductors to help clean the power more?

i cant read all of the lettering....what does C5 say? is that a diode?

thanks :D

[Micutzu]

I've also used a inductor but there was a load voltage drop of about 0.02V on it, so i decided to give up on it. C5 is a 22uF capacitor, connnected in paralel with D1 wich is a 1N4148 diode.

[stealth17]

I've also used a inductor but there was a load voltage drop of about 0.02V on it, so i decided to give up on it. C5 is a 22uF capacitor, connnected in paralel with D1 wich is a 1N4148 diode.

this is a linear design thus the heat correct?

[Micutzu]

Yes, that's a linear voltage regulator.

[stealth17]

anyone have a scematic for a switching design?

[persivore]

anyone have a scematic for a switching design?
If you can wait until next week, I'll have a go at drawing up a switched mode design, but its likely to be a lot more complicated than a linear supply (at a rough estimate I think I can probably do it with 4 opamps and a few resistors and capacitors and a MOSFET/transistor, unless you can find a dedicated chip to do it all for you)
IMO, a switching supply will be a bit overkill for the purpose. A simple linear regulator should provide similar performance, but with more heat output.

[Micutzu]

I think a linear regulator is more appropriate here, since it generally offers cleaner voltage output with no switching noise.

[stealth17]

linear supply has cleaner power? hmmm.....switching is much more efficient though. i need something that can put out about 40 amps without breakin my balls. i just want good clean stable power...thanks for the help guys!

[Micutzu]

I do agree that switching is more efficient, but in this case it really isn't required. Plus, the schematic i gave you puts out 15A, and considering that a standard 512MB PC3200 stick needs ~1.5A in full load, you can guess you have a nice margin there for even 2GB of memory.

[stealth17]

I do agree that switching is more efficient, but in this case it really isn't required. Plus, the schematic i gave you puts out 15A, and considering that a standard 512MB PC3200 stick needs ~1.5A in full load, you can guess you have a nice margin there for even 2GB of memory.

opps i told oyu the wrong amperage.

http://www.xtremesystems.org/forums/showthread.php?p=633701#post633701

so i need about 12amp atm. will linear be enough to put out 15amps without getting hot? id like a switching so it doesnt put out as much heat if possible. is there any way to add some high uf caps (i have a bunch of 1200uf caps from an old p2 mobo i can use) to help stabilize the voltage more? i basically want a current that good and stable like a vcore current and use that for my vdimm :D

[Micutzu]

Well, if it's done right (thick wires, good soldering, quality parts) that schematic will have a voltage drop of about 2mV/A, that's more than 10 times less than the OCZ Booster and comparable with a good onboard RAM voltage supply.

[stealth17]

If you can wait until next week, I'll have a go at drawing up a switched mode design, but its likely to be a lot more complicated than a linear supply (at a rough estimate I think I can probably do it with 4 opamps and a few resistors and capacitors and a MOSFET/transistor, unless you can find a dedicated chip to do it all for you)
IMO, a switching supply will be a bit overkill for the purpose. A simple linear regulator should provide similar performance, but with more heat output.

ygpm

[SoddemFX]

I dont like linear, maybe just me :D Switching with good filtering would be the way to go in my opinion...

What about using the PWM from a PIC to implement a full single phase PWM?

Feed a N-channel enhancement mode MOSFET as the upper FET with an invertor feeding the lower N ench FET. Feedback referenced against a LM336, output fed through a potential divider to scale it the the ref diode range before going into a ADC channel to compentsate PWM duty cycle to desired voltage. Ouput to basic LC...

You could also have LCD voltage display (or even temp. with a LM335?), RS232 communication, whatever. The whole thing in one package with a minimal amount of extra components...

Ive got a shed load of LM336's, low ESR caps, ceramics, inductors, N-channel ench MOSFETS rated from 10A - 48A. I've been planning to build one of these for a while, im planning on using a 16F877, but you have 2 PWM and 8 ADC -why not do Vtt too?

Only problem i can think of is the ADC being 10-bit, resoloution of 4 mV. Nasty thoughts of it going unstable...

Maybe just use one of the buck regulator IC's instead and add extra control?

How about drawing up plans for an XS homebrew DDR booster?

Tom

[stealth17]

I dont like linear, maybe just me :D Switching with good filtering would be the way to go in my opinion...

What about using the PWM from a PIC to implement a full single phase PWM?

Feed a N-channel enhancement mode MOSFET as the upper FET with an invertor feeding the lower N ench FET. Feedback referenced against a LM336, output fed through a potential divider to scale it the the ref diode range before going into a ADC channel to compentsate PWM duty cycle to desired voltage. Ouput to basic LC...

You could also have LCD voltage display (or even temp. with a LM335?), RS232 communication, whatever. The whole thing in one package with a minimal amount of extra components...

Ive got a shed load of LM336's, low ESR caps, ceramics, inductors, N-channel ench MOSFETS rated from 10A - 48A. I've been planning to build one of these for a while, im planning on using a 16F877, but you have 2 PWM and 8 ADC -why not do Vtt too?

Only problem i can think of is the ADC being 10-bit, resoloution of 4 mV. Nasty thoughts of it going unstable...

Maybe just use one of the buck regulator IC's instead and add extra control?

How about drawing up plans for an XS homebrew DDR booster?

Tom

that sounds like what Im looking for. since vtt seems to be a problem even on the dfi nf4 boards, yes, lets implement vtt also.

Im not that good at drawing schematics (obviously lol), so I cant really help there :(

The vtt circuit will need 3 amps, Id say call it 5 to be safe and furture proof.

-Jordan

[goatman]

I dont like linear, maybe just me :D Switching with good filtering would be the way to go in my opinion...

What about using the PWM from a PIC to implement a full single phase PWM?

Feed a N-channel enhancement mode MOSFET as the upper FET with an invertor feeding the lower N ench FET. Feedback referenced against a LM336, output fed through a potential divider to scale it the the ref diode range before going into a ADC channel to compentsate PWM duty cycle to desired voltage. Ouput to basic LC...

You could also have LCD voltage display (or even temp. with a LM335?), RS232 communication, whatever. The whole thing in one package with a minimal amount of extra components...

Ive got a shed load of LM336's, low ESR caps, ceramics, inductors, N-channel ench MOSFETS rated from 10A - 48A. I've been planning to build one of these for a while, im planning on using a 16F877, but you have 2 PWM and 8 ADC -why not do Vtt too?

Only problem i can think of is the ADC being 10-bit, resoloution of 4 mV. Nasty thoughts of it going unstable...

Maybe just use one of the buck regulator IC's instead and add extra control?

How about drawing up plans for an XS homebrew DDR booster?

Tom
what about the electrical noize?
the ram and the rest of teh system will pick all that up!!
u'd need to encase the power supply

[SoddemFX]

goatman - What about electrical noise? I dont understand what you mean...

As long as you have good isolation on the PCB i dont think it would be an issue, you'll have the PIC main oscillator, RS232 and LED strobing but dont see why with good filtering and PCB design you couldnt avoid any probelms.

I've got to admit that ive never built anything like this before with such accurate limits on voltage output, closest ive done is motor and heater control. My analogue is a bit ropey so any help would be much appreciated :)

Stealth17 - The lowest FETs i have are 10A, so it would be about 15A Vtt, not like you need that much, just that i dont want to buy unnecessary stuff. Also TO220 is a nice size :D

Tom

[lyquochaokiet]

i search the web and finnally i found this , i seeen micutzu pic , but i dont know what is sense line in his pic ??

i wonder which pin of ddr module we feed the voltage ?

sorry , english is not my main language.

[stealth17]

i search the web and finnally i found this , i seeen micutzu pic , but i dont know what is sense line in his pic ??

i wonder which pin of ddr module we feed the voltage ?

sorry , english is not my main language.

dont use that micutzu pic....He made a new one that is better. Ill post it tonight....

[lyquochaokiet]

i'm waitting for urs pic ;) ,

but when we build that ddr booster module , we need a special pcb for that ? i was sacrifried a 256 ddr module for that ...... what did you use for booster pcb ?

[stealth17]

T = STP55NF03L
IC = TL431
C1, C3 = 3300uF/6.3V low-ESR
C2 = 100-470uF/16V low-ESR
R1 = 390ohm
R2 = 1.5Kohm
P = 1Kohm multiturn

Try to get high-quality parts, so the voltage stays as clean and stable as
possible. The mosfet (T) gets very hot, so you need to mount it on a
heatspreader; watch the isolation, it's metallic part of the case is
connected to the D pin and therefore directly to the PSU's +5V line. The
output wire should be as thick as possible, because that's the only thing
it can cause voltage drop. If well built, this circuit should provide zero
voltage drop and extremely clean power.

http://img484.imageshack.us/img484/8518/reg6xp.th.jpg (http://img484.imageshack.us/my.php?image=reg6xp.jpg)

To supply the voltage you find out where to solder the output of the booster onto the motherboard. You dont use a DIMM slot....

Make your own PCB, doesnt need to be anything special size or shape wise.

[lyquochaokiet]

so we need to solder the output to the output of mosfet on the mobo rite ? but then , we need to cut the default voltage line ? cos if not , if will feedback the default mosfet too...

[Micutzu]

Just solder the output on the mobo mosfet's output (drain). If the mobo is compatible with the OCZ booster, you don't need to do anything else.

[lyquochaokiet]

so wat happen when mobo do not support ocz booster ?

we need mod vtt or something else ?

[Micutzu]

so wat happen when mobo do not support ocz booster ?

Pretty much you'll have a nice disco smoke generator :D
You don't need to mod anything else, vtt wil track automatically on most mobos.

[lyquochaokiet]

....omg , how about i remove default fet ouput (just the output) and feed moded to that line,i will try it

and about 55FN03L (rate at 30v/55A) , in my country (i'm living in vietnam now :) ) it seem i cant buy that chip.so i found NEC k3296(from my dead mobo) , it's rate at 20v /35A , is that enought for the circuit ?

[Micutzu]

That NEC should work with no problems. Watch out, it will need a heatspreader.

[Vassili]

dont use that micutzu pic....He made a new one that is better. Ill post it tonight....
Where do you get the picture from, can you give us a link?

Does anyone have a picture of there home made booster?

[Micutzu]

The circuit is made by me, and i've sent the schematic to stealth17 to test it out. The only sample i've built is used by sangboy here (http://www.xtremesystems.org/forums/showthread.php?t=77056).

[Johnny Bravo]

I have built and installed the circuit onto a P4GD1 to compensate for it's questiojnable stability. running 2 sticks of BH-5 at 3.5v maximum voltage flucation was 0.005V. Cheers Micutzu :toast:

Pics to follow

[Micutzu]

I'm glad you're happy with it :)

[Johnny Bravo]

As promised pics:-

Solder point on a P4GD1
http://img365.imageshack.us/img365/6655/vdimmsolderpoint6xf.jpg

Shots of the top side and underneath of the board itself:-
http://img365.imageshack.us/img365/3465/vdimmpowerboardtop7ig.jpg
http://img365.imageshack.us/img365/6650/vdimmpowerboardbottom6sm.jpg

[Taro28]

hi!
very good job Micutzu!
I have Asrock dual sata2 and a twinmos UTT memos but this motherboard only it gives 3.6V with 3.3 Vdimm mod and psu 3.3rail mod. I need more voltage for memos and I think that this circuit is very useful. the problem is that I don't find the STP55NF03L mosfet. the more seems it that I find is IRFZ48 (http://futurlec.com/Transistors/IRFZ48.shtml) I wanted to know if this model is equivalent and if I can use this for make a DDR booster.

greetings, Gaizka

[Micutzu]

The IRFZ48 is just as good.

[Lastviking]

Before i do something stupid, will this also work?

http://www.elfa.se/pdf/71/07106271.pdf

[Micutzu]

No, that won't work. Don't you have a dead mobo, any mosfet from the Vcore regulator should do.

[Lastviking]

No, that won't work. Don't you have a dead mobo, any mosfet from the Vcore regulator should do.

Naa i have not have any luck to kill any mainborad yet that didnt take RMA :stick:

hmm do someone of this works?

http://www.elfa.se/elfa-bin/dyndok.pl?lang=se&dok=2441.htm

[Micutzu]

That isn't good either, but here (http://www.elfa.se/elfa-bin/setpage.pl?http://www.elfa.se/elfa-bin/dyndok.pl?lang=se&dok=2441.htm) you can find what you need ... alot of those are fitted to this application, they even have STP40NF03L to STP80NF03L, wich are the same type with what i've used, but with different current ratings, from 40 to 80 Amps.

[Lastviking]

That isn't good either, but here (http://www.elfa.se/elfa-bin/setpage.pl?http://www.elfa.se/elfa-bin/dyndok.pl?lang=se&dok=2441.htm) you can find what you need ... alot of those are fitted to this application, they even have STP40NF03L to STP80NF03L, wich are the same type with what i've used, but with different current ratings, from 40 to 80 Amps.

oh thanks found them now.

http://www.elfa.se/elfa-bin/dyndok.pl?lang=se&dok=2458.htm

Are the correct link :)

hmm do you know if boosters works on dfi lanparty nf4 and where do i connect it?

Some image in this thread

http://www.xtremesystems.org/forums/showthread.php?t=84492

[Micutzu]

Even if i haven't tested yet i'm sure it works with the DFI no problem. I'd suggest connecting directly to the memory slot pins (3.3V strap style), it gives the best power delivery. Here's the pinout for the DDR module, you need to connect to VDD or VDDQ pins.


http://img491.imageshack.us/img491/2051/dimmnbsppinoutnbspddr15od.th.jpg (http://img491.imageshack.us/my.php?image=dimmnbsppinoutnbspddr15od.jpg)

[Lastviking]

Micutzu:

What do you think about this,do it work good enogh? or are it much beter to connect it derectly to the dimms?

http://www.overclocking-team-switzerland.com/thread.php?threadid=4

I will need atleat 4.0v+ on my 2x512 and atleast 300mhz+ 2-2-2-5 :)

And one more thing 1k pot on max value give lowest output to the dimms?

[Micutzu]

As i said before, connecting directly to the slot pins gives the best power delivery, but that should be good enough. Use wires as thick as possible and be sure the soldering points are very strong. Also, to get the absolute best from this circuit i would recommend this little mod:

http://img295.imageshack.us/img295/7378/reg6xp0uj.th.jpg (http://img295.imageshack.us/my.php?image=reg6xp0uj.jpg)

You connect the output to the mobo (either on a Vdimm cap or directly to the pins of the slot) and the FB wire to another VDD/Vddq pin of the DDR slot. The circuit uses the FB input to measure the voltage at the consumer (memory sticks in this case) and regulates the output accordingly, so this input needs to be connected as close to the dimms as possible, to read the voltage as accurately as possible.

WARNING! If you use this version of the circuit you NEED to connect the FB pin, otherwise the output voltage will always be 4.15V.

Edit: to get a voltage higher than 4.15V you need a lower value for R2; for example, R2 = 1.2Kohm will give you a range from 2.5V up to 4.58V :D.

[Lastviking]

As i said before, connecting directly to the slot pins gives the best power delivery, but that should be good enough. Use wires as thick as possible and be sure the soldering points are very strong. Also, to get the absolute best from this circuit i would recommend this little mod:

http://img295.imageshack.us/img295/7378/reg6xp0uj.th.jpg (http://img295.imageshack.us/my.php?image=reg6xp0uj.jpg)

You connect the output to the mobo (either on a Vdimm cap or directly to the pins of the slot) and the FB wire to another VDD/Vddq pin of the DDR slot. The circuit uses the FB input to measure the voltage at the consumer (memory sticks in this case) and regulates the output accordingly, so this input needs to be connected as close to the dimms as possible, to read the voltage as accurately as possible.

WARNING! If you use this version of the circuit you NEED to connect the FB pin, otherwise the output voltage will always be 4.15V.

Edit: to get a voltage higher than 4.15V you need a lower value for R2; for example, R2 = 1.2Kohm will give you a range from 2.5V up to 4.58V :D.

If i can change my +5 line to 5.5v+ then the vdimm will be more than 4.15v ? or will that trick make some instability on the booster?

[Micutzu]

No, raising the 5V line won't help you.

[Lastviking]

Hmm do the booster need to be connected to the mainboard to work or will it work if i only plug in it in the molex and check voltage with a multimeter?

[Micutzu]

You can power the booster from a molex, but make sure it's the only thing connected to that wire because it will draw a significant amount of power and might corrupt your HDD for example if it's conncted to the same wire from the PSU.

[Lastviking]

You can power the booster from a molex, but make sure it's the only thing connected to that wire because it will draw a significant amount of power and might corrupt your HDD for example if it's conncted to the same wire from the PSU.

K , but think i got a problem with my booster its only give out 0.5v

[Johnny Bravo]

As i said before, connecting directly to the slot pins gives the best power delivery, but that should be good enough. Use wires as thick as possible and be sure the soldering points are very strong. Also, to get the absolute best from this circuit i would recommend this little mod:

http://img295.imageshack.us/img295/7378/reg6xp0uj.th.jpg (http://img295.imageshack.us/my.php?image=reg6xp0uj.jpg)

You connect the output to the mobo (either on a Vdimm cap or directly to the pins of the slot) and the FB wire to another VDD/Vddq pin of the DDR slot. The circuit uses the FB input to measure the voltage at the consumer (memory sticks in this case) and regulates the output accordingly, so this input needs to be connected as close to the dimms as possible, to read the voltage as accurately as possible.

WARNING! If you use this version of the circuit you NEED to connect the FB pin, otherwise the output voltage will always be 4.15V.

Edit: to get a voltage higher than 4.15V you need a lower value for R2; for example, R2 = 1.2Kohm will give you a range from 2.5V up to 4.58V :D.


Out of interest what resistors do you vary to get lower voltages, we're talking in the region of 1.2-1.7v and 2.4-3.3v ?

[Micutzu]

Unfortunately this schematic cannot be modified to obtain a voltage lower than 2.5V, it's a limitation of the TL431 IC. The only option is to insert different power rectifiers in series with the output and their voltage drop will substract from the output value. 2X512MB BH-5 running at 250Mhz 3.4V draws approximatively 7.5A in average under full load BTW.

[Johnny Bravo]

Do you have any alternative circuits for such an application Micutzu?

[[cTx]Philosophy]

So this is what them aliens were saying when they abduscted me..
std43-to mosfet29 header, burgerkingme, then solder it to the wd-40...

I think I got it now, thnx guys...

[Micutzu]

Do you have any alternative circuits for such an application Micutzu?

You can try this (http://www.xtremesystems.org/forums/showpost.php?p=890138&postcount=2), but it's more complex and has a little worse performance. It's adjustable from 0V to 4.2V.

[bouc]

The IRFZ48 is just as good.

Hi all
would the "N" version of IRFZ48N do the job http://www.meditronik.com.pl/doc/a0-a9999/irfz48n.pdf ?
thx a lot :D

[Micutzu]

Hi all
would the "N" version of IRFZ48N do the job http://www.meditronik.com.pl/doc/a0-a9999/irfz48n.pdf ?
thx a lot :D

Yes, sure, it'll do just fine.

[bouc]

ok, can find some on ebay for 1,10 USD each :)
thx ;)

[Obelisk79]

I need a mod like this for my new set of BH5, since I feel more comfortable utilizing a regulated voltage than the 5v rail on my PSU directly with a pot. I plan on trying the original mod that allows full adjustment from 0v-4.2v but my question is, while looking at the schematics I don't quite understand what components K1-K6 are and what they are for. Also do those terminate to ground? Been a couple years since i read schematics so I'm rusty, are they inductors for compensating any fluctuations in current/voltage?

Man all this electronics talk has me breaking out my old electronics manuals... *sigh* why does technology have to be so addictive?

[Micutzu]

K1 ... K6 are ground connections. Unless you need voltages under 2.5V, the second schematic is better: it has improved thermal stability, simpler design and offers cleaner power troughout the whole the voltage range.

[bouc]

Is the same method available for vcore?
to stabilize it, and controll it?

[Micutzu]

No, external vcore regulation would require a much complex schematic; it's usually better to mod the already existing vcore regulation circuit on the motherboard by improving the LC filters' capacity and replacing the current switches (MOSFETS) with larger/better ones.

[bouc]

ok thx
any vcore regulation circuit mod for asrock 939dual sata2 then? :) :): )

[stealth17]

ok thx
any vcore regulation circuit mod for asrock 939dual sata2 then? :) :): )
Plz make a new thread for that.

[bouc]

ok, excuse me

[stealth17]

Your excused :nuts:

Don't mean to be a d!ck, just wanted to keep this thread on topic :up:

[bouc]

what's the mini voltage with the second pic? (without the sense on the dimm) ?

I saw in another thread someone that would need a 2V regulator with big Amp (~10A) cause the regulator of his 7900GT memory fried, could this be use insteed?
(in the graphx mod section, sticky topic 7900GT Voltmod-2)

thx 4 him

[Micutzu]

First schematic provides basically anything from 0 to 4.3V but with poor regulation at both ends of the scale, the second schematic from 2.5V to 4.15V with strong regulation all trough the range. For replacing the memory regulator on the 7900GT i would recomend rebuilding the original circuit, or using another switching regulator, as that is not such a critical application to require the clean power of a linear regulator (video memory erroring > artifacts, but system memory erroring > system crash).

[stealth17]

First schematic provides basically anything from 0 to 4.3V but with poor regulation at both ends of the scale, the second schematic from 2.5V to 4.15V with strong regulation all trough the range. For replacing the memory regulator on the 7900GT i would recomend rebuilding the original circuit, or using another switching regulator, as that is not such a critical application to require the clean power of a linear regulator (video memory erroring > artifacts, but system memory erroring > system crash).
I agree with Micutzu. The linear design would create a lot more heat too which is unnecessary and not worth the extra heat dump.

[bouc]

Hi
can I hook this circuit on the 20 pin plug between the psu and the mobo?
or do I have to use a molex?
thx

[Micutzu]

Use a molex.

[bouc]

ok
thx

[burnout21]

can someone post a pic of one they made? plz

[bouc]

look in the thread, there are pics already ;)

[stealth17]

As promised pics:-

Solder point on a P4GD1


Shots of the top side and underneath of the board itself:

What program did you make the layout with?

[Death909]

Any circuits for stabilising Vtt? ECM2's thread kinda died off, mostly because he hasn't visited XS for a year and two months.

I found this in another thread:
http://xtremesystems.org/forums/attachment.php?attachmentid=34115&d=1121305069

My minimal knowledge in this field suggests that this circuit would not work (because if Vref = Vtt then VrefIn is going to be 0v, atleast without an offset equal to Vref). The graphs in that thread suggest that the diffamp used has a gain of 10, with the calculation Vout = 10(Vref - Vtt) + Vref. I couldn't figure out how to do that with a single opamp / diffamp, but with a 'passive averager' and a gain of 20 on the opamp / diffamp the calculation Vout = (20(Vref - Vtt) + Vmem)/2 becomes do able.

I cannot figure out how a Vtt augmentation signal would work. EMC2's seems to output up to 3A at 1.86v to keep Vtt stable at about 1.425v.

[stealth17]

Any circuits for stabilising Vtt? ECM2's thread kinda died off, mostly because he hasn't visited XS for a year and two months.

I found this in another thread:
http://xtremesystems.org/forums/attachment.php?attachmentid=34115&d=1121305069

My minimal knowledge in this field suggests that this circuit would not work (because if Vref = Vtt then VrefIn is going to be 0v, atleast without an offset equal to Vref). The graphs in that thread suggest that the diffamp used has a gain of 10, with the calculation Vout = 10(Vref - Vtt) + Vref. I couldn't figure out how to do that with a single opamp / diffamp, but with a 'passive averager' and a gain of 20 on the opamp / diffamp the calculation Vout = (20(Vref - Vtt) + Vmem)/2 becomes do able.

I cannot figure out how a Vtt augmentation signal would work. EMC2's seems to output up to 3A at 1.86v to keep Vtt stable at about 1.425v.

Well I think the best way to do it is to find another precision shunt regulator to replace the TL431 that will work down to around 1v.

There are a couple different ways of doing the reference now. I think to best suit our needs here a voltage regulator that is capable of reducing >4v down to >1v. The good thing about the reference is that it will be drawing only a few hundred milliamps at most, so a voltage regulator for the application should be easy to come by.

The voltage regulator should rely on two resistors to depict the output voltage. If you put a variable resistor on the second resistor then you can adjust the VTT voltage above or below the desired level. Sometimes running it a little higher then exactly half is beneficial, thus the variable resistor will allow for this if we want.

I think what you were trying to do is cut the vDDR in half after the voltage regulator. That will work but imho it is not a good idea since you would be cutting the voltage down twice and allowing for more chance of noise or "dirty" power.

On a side note I have found a dead motherboard with a RT9173 that I can use. I may see what I can do with it since I have an extra in hand :D

[Death909]

The RT9173A DDR termination regulator outputs "dirty power" to Vtt pretty much by default. EMC2's board was getting 90mv droops during Memorytest86+'s test 8 (with Vref = 1.4v). The RT9173A tracks Vref (pin 4). By altering Vref (or just seperating pin 4 from its pad and providing your own input) you can alter Vtt. By using a diffamp with a gain greater than 1 we can force the RT9173A to provide cleaner Vtt.

As for the augmentation signal, the circuit just needs to provide 3A at a voltage equal to the output of the diffamp. The output voltage range should be atleast .9v to 2.4v (Vref +/- .4v), but 1.2v to 2.2v should be fine.

(My RAM, OCZ VX, fails a Prime95 blend run after 2 minutes at 215mhz. It used to be able to do 255mhz just last year, but thats when the motherboard was new. Since its the same with another pair of VX, stabilising Vtt and Vmem is a priority.)

[stealth17]

The RT9173A DDR termination regulator outputs "dirty power" to Vtt pretty much by default. EMC2's board was getting 90mv droops during Memorytest86+'s test 8 (with Vref = 1.4v). The RT9173A tracks Vref (pin 4). By altering Vref (or just seperating pin 4 from its pad and providing your own input) you can alter Vtt. By using a diffamp with a gain greater than 1 we can force the RT9173A to provide cleaner Vtt.

As for the augmentation signal, the circuit just needs to provide 3A at a voltage equal to the output of the diffamp. The output voltage range should be atleast .9v to 2.4v (Vref +/- .4v), but 1.2v to 2.2v should be fine.

(My RAM, OCZ VX, fails a Prime95 blend run after 2 minutes at 215mhz. It used to be able to do 255mhz just last year, but thats when the motherboard was new. Since its the same with another pair of VX, stabilising Vtt and Vmem is a priority.)

Well the RT9173A was just getting over loaded but if we make our own circuit for it then we could still use it for our reference voltage.

As for the shunt regulator, check out the NCP100SNT1 (http://www.google.com/url?sa=t&ct=res&cd=3&url=http%3A%2F%2Fwww.chipcatalog.com%2FDatasheet%2 F89CE47F09312A93B4AFAC39D37518C97.htm&ei=aQ5rRcaUHpfOpwKHhp3lDA&usg=__bzatEtOet7QfLE-kygJNd8_hO5E=&sig2=seYcQiehFAlaJH24Q6fIJw). I can't say for sure if it will fit right into the second/third schematic Micutzu made so someone would have to confirm this. It is good from 0.7v-6.0v.

Let me know what you think about that.

[Death909]

Well the RT9173A was just getting over loaded but if we make our own circuit for it then we could still use it for our reference voltage.

As for the shunt regulator, check out the NCP100SNT1 (http://www.google.com/url?sa=t&ct=res&cd=3&url=http%3A%2F%2Fwww.chipcatalog.com%2FDatasheet%2 F89CE47F09312A93B4AFAC39D37518C97.htm&ei=aQ5rRcaUHpfOpwKHhp3lDA&usg=__bzatEtOet7QfLE-kygJNd8_hO5E=&sig2=seYcQiehFAlaJH24Q6fIJw). I can't say for sure if it will fit right into the second/third schematic Micutzu made so someone would have to confirm this. It is good from 0.7v-6.0v.

Let me know what you think about that.

The RT9173A is supposedly below spec. Not only that but it seems to perform poorly pretty much by itself, possibly because a Vmem higher or equal to 3.3v is out of spec for the RT9173A (Vmem < 3.3v is the voltage range the RT9173A is looking for).

Anyway, a circuit to induce greater negative feedback on the RT9173A is simple. We simply supply our own Vref to the RT9173A. A difference or differential amplifier (amplifies the difference between both inputs) with its output connected to pin 4 would work.

Heres an altered version of EMC2's little diagram that should work:
http://img214.imageshack.us/my.php?image=rt9173anf1sn3.png
Note that R1 = R2. Altering the resistance of both resistors, as long as they are both equal, will alter the current. The RT9173A datasheet doesn't mention the current draw from Vref, but considering its an input it probably won't use much more than a mA.

However, the difference in current between the output of the amplifier and Vmem might make a difference in VrefIn. I don't think it will, since Millman's theorem takes into account voltage and V/R = I (current), but I'd still like someone with more knowledge on this subject to confirm it.

Instead of using a differential amplifier, any opamp could be used. The circuit would need to be modified to adjust the gain however (and I'm unsure on how to do that).

[AsAsIn8eR]

What do you use to make the PCB?OR what do you use for lettering.I use a OHP marker and draw the lines after I made holes with a drill machine.And put it in the coroding solution and the one markered stays.Is this what you use?Or you have some better technology?

[stealth17]

What do you use to make the PCB?OR what do you use for lettering.I use a OHP marker and draw the lines after I made holes with a drill machine.And put it in the coroding solution and the one markered stays.Is this what you use?Or you have some better technology?

I'm going to make a guide for it soon. I'll link you when I finish. Should be about 2 weeks.