I'm looking for a way to make it so my booster doesn't droop under high load. If I set it to 3.4v in the BIOS, it drops to 3.25 under any type of load. Does anyone have a suggestion?
Printable View
I'm looking for a way to make it so my booster doesn't droop under high load. If I set it to 3.4v in the BIOS, it drops to 3.25 under any type of load. Does anyone have a suggestion?
:D Yeah - got the same problem. I see only two ways - first to live with it or second to swich to DFI NF4 with the 4-V-stock Vdimm and burry Your booster.
:toast:
Is your PSU rails stable, that could cause a problem. A droop mod...mmmm...never heard of one for the booster but I wouldn't be surprised if there was one
jjcom
That high voltage drop could also hint to a lack of cooling. If the mosfets on the Booster get too hot, you also get voltage fluctuation.
If it's not cooling, then try another PSU.
BTW, what's your 5V rail at? Isn't the booster powered from the 5V rail? If so, then it should be above 5V all the time. :toast:
Guys, dont confuse the term "droop" with every possible mod out there. The Vdroop is a voltage controller specific thing that can be modified via adding a resistor. You can raise/lower the droop only in specific circumstances.
I havent looked up the data sheet for the controller on the Booster, so I dont know if Vdroop is a part of its regulation.
I would agree with what the guys above said. Most likely heat or psu related. I would check the heatsink to see how warm it is and look it over to make sure its making good contact with the fets under it. Also look over the rails on the psu to see if it is bouncing around. You should also try a search here or on google about using the booster and your motherboard. My understanding is that there has been many problems associated with this combination.
I'm running a PC Power and Cooling ATX 475-PFC. At idle (tested with multimeter), I get 5.19 v on the 5v rail and 12.04 on the 12v rail. At load they are 5.18 and 12.03. I persoanlly ripped the orignal aluminum heatsink off the MOSFETs and Actic Silver Alumina Epoxied 4X Vantec Iceberq BGA copper heatsinks on them, the I glued two 80mm fans together, and they are blowing DIRECTLY on the entire ram assembly. The Heatsinks get warm under load, but I could keep my finger there all day and be fine. I know they're making contact great, as I did it myself. It's not heat, and it's not the PSU, I'm certain of that.
There's no voltage controller on the booster. Actually, the booster is just a big adjustable resistor connected from +5V to Vdimm, no droop mod possible there.Quote:
Originally Posted by Hell-Fire
You could try running the +12v on the booster on a dedicated PSU.
This will eliminate most of the fluctuation I think.
;) :)Quote:
Originally Posted by Micutzu
Well hurry up then damnit!Quote:
Originally Posted by hipro5
Don't you think that it would be better to be a little more friendly here in the forums??? :stick:Quote:
Originally Posted by Stealth5325
Hope your next postings will look different then this one. :rolleyes:
No.....It's OK........I know what he means...... :DQuote:
Originally Posted by celemine1Gig
I also know what he means, but he shouldn't talk in such a way, especially to people whom he doesn't even know.Quote:
Originally Posted by hipro5
If he knows you personally, then no problem. But me for example, I know that you are hipro5, the great overclocker and engineer from Greece, but I don't really know you as a person and I highly respect your work. And that's the point why I wouldn't talk this way to you, and everyone else shouldn't do so either.
We shouold keep some respect, even in these forums here.
There are other ways to express ones thoughts, than the way, Stealth chose.
Just my :2cents: . If you don't mind that, then ignore my posting.
I'm with Celemine, I really like RESPECT, respect it's what makes us all strong and unite
You are right guys and thank you very much about your good words..........BUT we supposed to be FRIENDS in here no matter what age-gentar-country-e.t.c. we are.......So I think that friends could talk freely and comfortable in here.....That's my opinion........Respect is good but friendship IS BETTER don't you think...?....... ;) :toast:
We all could say : "Respect your friend"...... :D :toast:
That's better and congrats on your work, very good!!Quote:
Originally Posted by hipro5
While I don't appreciate Celemine injecting his negative thoughts about me in my thread, I seem to be in the minority here.
Hipro, although you've indicated that you understood, allow me to apologize for my choice of wording.
I am finding my current DDR voltage solution inadequate. You have what seems to be a better solution, and I gt overly exicted for it. Please know I meant no harm.
That said, I would love to see your unit in production as soon as possible!
Hipro, cant you make a quick circuit with a small voltage controller to control the fet's on the booster so it will correct itself when the voltage is dropping under load?
I doubt that Hipro5 would give you any details on how to improve the booster, as he (as you might already know) wants to release his own kind of booster (the maximizer; much better than the booster anyway) soon. Helping others improve the product that his direct opponent is marketing, would be like shooting his own foot.Quote:
Originally Posted by Rub87
But perhaps you are lucky and he'll help you, although it would be kind of bad for him.
@Stealth5325:
You are a really funny guy. :D Me and negative thoughts :stick: :p:
I only pointed out, that we (I think I can speak for the majority of the community here) would like to keep a certain level of conversation which indicates a certain respect towards eachother. And as your last post partly showed, you are indeed able to hold that level. So, was it that hard to behave? :D :toast: Just kidding! :)
BTW, if the posting, that I was referring to when complaining, was meant to be funny or ironic, or wahtever, then just use smilies next time, so we all know the way it was meant to be. ;)
I've designed a circuit which should help reduce the droop on the Booster. I haven't got/seen a booster myself, so i'm not sure if it will work, but I think it should workQuote:
Originally Posted by Rub87
http://www.drunken-student.co.uk/tom/reg.JPG
The OPAMP should be powered from the 12v line on the booster.
BTW, I'm not trying to take any business away from Hipro, and I am sure that what he has planned will perform much better than this.
I'm not that good with electronics, but if I get this correct is q1 the 4 fets under the hs on such a booster (haven't removed them yet) and then we replace the stock 10k pot with a 100K one and add a little opamp..
The opamp needs to be powered by +12v..? and -Vz needs to be connected to gnd or -12v?
I didn't know that there were 4 FETs on the booster. It depends how they are connected if this will work, but I'll look into it further.
It should be ok to use the existing 10k pot thats on the booster.
The + leg should be connected to the +12v line and the - leg to ground.
They're al 4 feeded by +5v and switched parralel, The middle (from top view) leg is +5v, and the left one is the controll line, the left one I can't see where that's going as this is a multi layer pcb, but I suppose that it goes to Vdimm...
It should be fine to just disconnect the wiper (center pin) of the pot and connect it to the non inverting (+) input on the OPAMP and then connect the point that the center pin of the pot was connected to on the PCB to the output of the OPAMP.
Is something like this usable as opamp for this purpose?
It's the LM386N-3 version..
Any OPAMP will work fine.Quote:
Originally Posted by Rub87
Edit: you might have to connect pins 1 and 8 together to set the OPAMPS gain to 1 on the LM386
What is just the gain function?
I've just tried to breadboard the circuit with an LM386, and I can't get it to work properly.
The gain is the ammount that the input voltage is multiplied by to get the output voltage, and in this circuit it needs to be 1 (so that the voltage set by the pot=Vdimm). The LM386 has an inbuilt gain of 20, which needs to be disabled for it to work in this circuit, but I haven't worked out how do it yet, but I'll keep trying.
Edit: I can't get it to work with an LM386 as it appears to have internal connections to make it into an audio amplifier. :(
Atm I'm trying with an LM358, ripped from and dfi infinity...
An LM358 will work fine :). I've tried with an LM358 and a 741, and both worked ok.
Sorry for my paint skillz, but is it +- like I showed on this pic? (digicam is death, only got webcam)
So this is when the stock pot is removed...
That looks ok. The - input on the OPAMP should be going to the output of the MOSFETs. I've never used a DDR booster before, but that looks like it should work.
The pin were I connect the inv. input of the opamp is the output of the mosfets, not? The middle one comes straight from the +5v and the left one are connected by a very thin interconnect that goes to the middel leg of the pot its also splits to a little 10K smd resitor (R7) (don't know what's that for) so there is only one pin left as output..
The inverting input should be connected to the Vdimm output of the Booster (the 1 remaining pin, sounds like the pin on the right from your description).
The output of the OPAMP should be connected to the point that the middle leg of the pot was removed from (or you could try using the left leg of the MOSFETs, but it shouldn't matter what you use as they should be connected together).
When looking closer at the Booster, there's nothing that regulates the Vbt so that it'll be always the half of Vdimm
Is it possible to disgn the same type of power supllie to keep the vbt half of vdimm?
Could this work out?
Wich type of fet, mosfet? can I use for this purpose?
I'll add some caps here and there...
An IRF640 or IRF630 would work. There aren't any Vtt/Vbt carrying pins on the Dimm slot itself (there's only Vref, which is also 1/2 of Vdimm, but that usually tracks properly on motherboards) so you might have to mod the motherboard itself.
I've done a similar mod on my NF7-s using an OPAMP and an IRF630 MOSFET to make Vbt/Vtt follow.
The LM358 is a dual OPAMP, so you could use the other half to control Vbt.
Quote:
Originally Posted by persivore
I know, thats why I came on the idea to fix the vbt/vtt also :p:
Problem is that the LM358 is kind'a small, this came of an dfi infinity right inder the agp slot.. It's quite vulnerable, but I don't have the equipement to make a pcb for it:/
I'm using the non-SMD/bigger version of the LM358 on my NF7-s, without a PCB. I just glued the OPAMP to the top of the new Vtt MOSFET. It doesn't look very tidy, but it works great!
Yeah, I'll think I'll get monday a bigger version plus some mosfets...Quote:
Originally Posted by persivore
Pffff,
I take a new nf7-s, put the still unmodded booster in, put the power on and the chip under dimm 3 went up in smoke, this pisses me off :frag:
Vref is dead on the board, vbt/vtt is still 1.4v..
It was while vbt/vtt was connected vref...
Now I disconnected vbt/vtt vref and removed the booster, would it be safe to reconnect vbt/vtt to vref as in the 3in1 mod?
I'm not sure how much current Vref draws, but I don't think that it will be much as there is only 1 pin for it on the Dimm pinout. Try doing the Vbt/Vtt to Vref mod, but monitor the voltage as you turn on the board to check that it doesn't go too far out of spec.Quote:
Originally Posted by Rub87
That's the problem, the vbt/vtt=vref mod was done when the vref regulator blew... But I don't knwo why, could it be due the booster?
What could have happened is that Vref followed Vdimm, but Vtt/Vbt didn't, which caused a voltage difference between the 2 regulators and that ended up killing something.
I don't how that could happen as there is nothing that is capable of producing a vbt/vtt that is the half of vref on the booster, goddammit
Vref will usually follow as 1/2 of vdimm. I think that its just produced with a voltage divider from Vdimm to ground, whereas Vtt/Vbt is produced via a regulator.
Yean, but why is vref death and vbt not?
Its on this picture the right regulater that went up in smoke..
I guess that the voltage which could supply the most current was the one that survived. Where are you reading Vtt/Vbt and Vref from?
Vref from the dimm and vbt/vtt from the capacitor...
I find it very strange that Vtt/Vbt has managed to survive, even with a burnt regulator, and Vref has died. Replacing the burnt component might help. Is the Vdimm supply still alive?
Sorry, you were right it Is the vbt/vtt thats dead, vref is working, switched wire, sorry..
Jep, vdimm is @ 2.9
If you want to replace the MOSFET with one of the same type, I think that someone over at OCforums managed to get one from the manufacturer (fairchild I think). It might be ok to just replace it with a general purpose MOSFET (eg IRF630) tho. I think that the chip might also have something to do with the Vdimm supply, but I've got the chip totally removed from my board, and it still works ok.
And then?
Made a vbt supllie with an opamp and mosfet? like you said?
I have no mosfets atm so it will be for monday or so, maybe I'll try tomorrow with some diodes from +3.3v...
Vref is 1.45v, vbt/vtt is dead, the two upper right pins of the burnt chip are vdimm, the 2 upper left ones are 0.3v, these 2 should be vbt/vtt..
With the death chip in place I measured 0v, now I removed it and measure 0.3v...
I just tried with a 470 ohm pot between vref and vbt, when I lower the resistance, vref drops quickly, I stopped @ 0.6v for vref, at that point, vref was vbt/vtt, both 0.6v, but this way I'll kill the vref supllie I think..
You could use an OPAMP, voltage divider and MOSFET to rebuild the Vbt supply. It would be good to try to get the board working using some diodes tho, so that you can check that it is just the Vbt supply that has died.
All the other voltages are fine...
Is your board also using a standalone vdimm supllie? Or only vbt/vtt..
How much amps does vdimm and vbt/vtt pull? So my mosfets are ably to handle the load..
But now my first concern is to know why the vbt/vtt died...
My other nf7-s has also the vref=vbt/vtt mod and is working for months 24/7 without any troubles...
on my board, I've got a Vdimm supply based around an OPAMP/MOSFET regulator fed from the 5v supply, giving 0-5v adjustable vdimm supply. I've also got the Vbt supply which follows Vdimm. The Vdimm MOSFET only gets slightly warm when I'm using 1*256mb BH-6 with 3.8v through it, but there is a fan/heatsink on the MOSFET. When I use my 1gb of TCCD with 2.6v in the board, the MOSFET and cabling to it gets VERY hot, even with a fan over the MOSFET sink, so its probably drawing about 10 to 15 amps. I'm using an IRF630 MOSFET for the vbt supply (which is more than adeqate), and an IRF640 MOSFET for the Vdimm (I would suggest using a higher rated MOSFET for the Vdimm, maybe a IRL2203N which is rated at about 80 amps at 100*c!)
Yeah 80 amps, thats the big stuff;)
I'll try to get to the shop tomorrow! I've a few broken psu's so I can sink the mosfets...
Then I'll buy the bigger version of the LM358 and make a nice vbt/vtt and vdimm supllie :toast:
Did you cut/removed the vdimm mosfets on your board?
Is there any kind of OCp or OVP on the board?
Btw, where did you connect the vdimm to you board? The vdimm cap? The mosfet or straight to the dimm's?
dubblepost
I removed the resistor that you solder to when you do the normal Vdimm mod, removed the 2 legs from the Vdimm MOSFET and connected the point that you normally solder to on the Vdimm mod to the gate (marked G on the PCB) via a 15 ohm resistor. This tricks the regulator into giving out whatever voltage you have set in the BIOS, but without it being connected to the RAM in any way. I don't know if you actually need to do this, but I did it incase you do.
I then connected the new Vdimm supply to where the other leg of the Vdimm MOSFET was. I left the Vdimm MOSFET in place on the board becasue it would have required too much heat to remove it.
I haven't found any problems with OVP or OCP on the board.
BTS240A
Is this also a possibility?
That should be ok.
Sweet,
I've a few out of car heating systems to control the electric fan speed, they're rated @ 58A continous and 200A pulse..
And, they're already attached to a big hs :banana:
So, back to the original post: have we come up with a mod that reduced the voltage drop under load? Someone want to sum it all up for us noobs?
Quote:
Originally Posted by Rub87
Unsolder the pot that's already on the booster, and connect the wires as shown in the picture, You can choose what potmeter you use, I'd say anything between 20k and 100k will work, then take an OPAMP, something like an LM358 for example, and connect the inverse imput to vdimm and the positive to the middel leg of your 20k-100k pot...
one side of the pot you connect to +5v the other to ground..
If someone can give me a hi res picture of the rear side of the booster then I'll mark the thing on it...
58A should be enough for RAM. The MOSFET I'm using is only rated to 16A iirc and it works ok (although it does get a bit hot!)Quote:
Originally Posted by Rub87
If someone can post some pictures of the front and back of the DDR booster, I'll write up a guide :)Quote:
Originally Posted by Stealth5325
The pot would need to be connected to the +5v supply, so that it gives a reference voltage at its wiper/middle pin of between 0-5v.
You're right, I was confusing with the "difference amplifier" (don't know the correct english name...
I'll delete the post, otherwise there will be more confusion..
wORKS LIKE A CHARM
Nice work! =)
Good to see it worked for you Rub87.
I finally got an OS onto my NF7-s yesterday to check how stable the Vdimm was under load, and I don't get any voltage fluctuaton read in MBM or on a DMM :slobber:. I need to get a scope to check what the voltage fluctution is really like tho.
I have a scope, I'll check it in the weekend... don't have time due school :mad:
Now I tried with the big BTS240A mosfet, prob is, when I connect the gate with the drain @ 12v, the voltage drop is 4v:/ Probaly I'll use the IRF640, but when trying @ 12v with a delta FFB80 the mosfet went verry warm after 5secs without hs so I think It'll be a little light for 1 gig BH-5 @ 3.5v
The BTS240A should have worked. I can't see anything in the datasheet to suggest why it wouldn't work. I've just got a small, low ariflow fan blowing over the MOSFET heatsink.
Did you connect both the gate and drain to +12v?
Jep, voltage drop is way to big to use it powered by +5V,
When powering the mosfet by +12V then it will work...
Now I tried it on my nf7, so only the vbt/vtt supllie, when I use 2*20K as voltage splitters and and LM358 it doesn't work when the OPAMP is powered by +5v, when powering the OPAMP with +12v and the mosfet (IRF640) with +5V its works great...
I'll fix tomorrow the Vdimm supllie, but I don't see why the OPAMP doesn't work when powered by +5V, then vbt/vtt output at the mosfet source only gets around 1.4v when the input voltage @ the non inv input is 1.65v... :confused:
The OPAMP needs to be connected to +12v so that it gives an output voltage of between 0-12v (actually it will be closer to 2-10v due to OPAMP limitations, but that doesn't really matter).
MOSFETs start to come into conduction when the gate pin is at about 2v, and about 10v will turn it fully on. With a 5v supply to the OPAMP is probably only able to supply a max of 3.5v onto the gate of the OPAMP, which is only enough to bring the MOSFET into partial conduction, which is why it works with a 12v supply to the MOSFET and not a 5v supply.
With the Vdimm set to 3.8v on my setup the OPAMP outputs about 7.5v onto the gate.
If I overvolt my 5V rail will that increase the voltage to the memory?
If you are using a DDR booster, I think that increasing either the 5v or 12v rails will increase the Vdimm, but I also think that increasing the voltages screws with the voltage readout on the DDR booster as well.Quote:
Originally Posted by HiJon89
Whatsn the max voltage you can go now? I should measure the voltage with a multimeter to start, normally is you connect gare and drain to 5v, (like when to pot is turned to max) you'll get 5V- voltagedrop of the mosfets as vdimm...
How much vdimm you have now? measure with MM if you can? If your pot is already turned to max, then you can the gate and drain on the backside of you booster, I think this will give a little more, but if you still want more unsolder the +5V pin of the pot and connect it to +12v, this way, you'll be able to let the fets conduct soem more...
When turned all the way to the right it says 3.5V idle 3.4V load. This is with 12V rail at 12.1 and 5V rail at 5.1.
Can you measure what voltage you get at the gate pin of the mosfets?
Was that question for HiJon89 or me? If it was for me, I get about 7.5v on the gate, with 3.7v Vdimm, but it varies with the ammount of load on the Vdimm supply. If the question was for HiJon89 please ignore this post :p:Quote:
Originally Posted by Rub87
It's impossible to get 7.5v on the gate as the booster only gets feed by +5 and -5v... there is no 12v line to the booster so it can be max +5v...
Question was for HiJon89 :D
The gates are fed trough the pot from the +12V line ;)
Crap, now I see again, You're right, I verlooked the +12v
Hmm,
Installed windows, flashed the D25 merlin bios..
Actually I get pretty much fluctuation in Vdimm, goes when @ 3.5v in bios it goes to 3.38 sometimes when running sandra membench, 5v line drops fron 5.07 to 5.01, prolly that's the cause of the huge Vdimm drop, any ideas to make it more stable? The good news is: when running 1 gig BH-5 @ 3.7v, the IRF640 gets warm, but not hot.. (used a Hell of a Heatsink) :toast: I need the cable of my camcorder, or my digicam should revive so I can post post some pics!!
How long is the cable between the gate pin and the OPAMP? I've found that I get the most stable voltages with a very short wire between the 2 points.
Also, what is the gate voltage at? It should vary depending on the load on the regulator, and check that the reference voltage from the pot isn't varying (you could put a small capacitor between the non-inverting input and ground to help stabalise the reference voltage and provide "soft start" at power on. 0.1uf should be ok.
The wire between the OPAMP and gate is around 30cm...
The reference voltage is 5v, so it varies quite a bit...
The gate voltage is between 6v and 8.5v, I'll ad some capitance soon ;)
The gate cable might be too long. I'm using a 5cm piece of sheilded audio cable on my regulator. The gate pin is very susceptible to noise.
You could also try taking the feeback between the Vdimm output and the inverting input of the OPAMP from a point connected to Vdimm on the underside of the Dimm sockets, which will help to compensate for any voltage loss in the cable between the output of the MOSFET and the Dimm slots.
Shortned the gate wires, added some 10µF to the wiper wire...
Soldered the feebback lines where the original vbt/vtt mosfet was...
Now the voltage drops from 3.52 to 3.48-3.49 when applying load.. If I want it perfect, my 5v don't may fluctuate or I should need a perfect stable external suplly for the pot that isn't load dependent...
You could try regulating the 12v line down to 5v using a zener regulator to provide a stable 5v supply for the pot.
http://drunken-student.co.uk/tom/mods/5v1reg.JPG
Ideed, I think there must be lying some 4.7v zener's somewhere over here.. but where...
I don't get it anymore, It's not even primestable @ 220 dcddr :(
Was it prime stable before you re-built the Vdimm/Vtt regulator?
my booster droops pretty bad also.
without the booster
p4c800e-d bh-5 256x2
no load- 2.81V
load- 2.71V
with booster
no load- 2.95
load- 2.80
using prime 95 torture test for load.
antec 550 true control. all rails very stable.
that .1 seems like a huge droop for this thing.
Nope, 240 wasn't stable, didn't test 220..Quote:
Originally Posted by persivore
As I still didn't found my usb clable for camcorder, I made a movie and captured a few pics and captured them via i-DV firewire cable and then captured some pics in MM .. enjoy the crappy stuff:D
Btw, 215 seems primestable...
I remember that I was having similar problems to you when I first built my regulator, but I can't remember how I solved them. I think that I found that putting a capacitor and resistor in parallel with each other into the feedback loop helped, but I'm not 100% sure.
Try putting a 1k resistor in parallel with a 100Nf capacitor and put that between the non-inverting pin for the Vdimm supply OPAMP and the Vdimm supply output on the MOSFET.
I'll try to think of some more ideas later and try breadboarding some of them.
Quote:
Originally Posted by persivore
Don't have such small caps, any source for such small ones out of old hw?:/ ..
What problem do you mean? that it isn't stable above 215 fsb? or the minor fluctuation between load and idle? When I've time I'll check the generated current with a scope? Do ypu have some info about how to use one? my dad has one but I don't know how to use on a effecient way...
You could try finding a 1uf capacitor on an old motherboard. I just looked at one thats next to me and its got a few of them on. Just replace the 1k resistor with a 100ohm resistor. The exact values of the componens shouldn't matter too much tho tbh.
You might also be able to fund a 100Nf capacitor inside an old PC power supply (could also be marked as 0.1Uf).
I was having the problem of the fluctuation between load and idle. I think that it might have been up to 0.3v when I had the problem tho.
With the scope you can measure the voltage fluctuation. Connect the ground lead to a good ground point on the motherboard, close to the Dimm slots. Connect the + lead, via a capacitor (value shouldn't be too important as long as it isn't too big) to the Vdimm supply. The capacitor should act to block the DC voltage, and just leave the AC 'ripple'/voltage fluctuation to go to the scope.
Allright, I'll give it I try somewhere in the next few days... The voltage fluctuation between load and idle is only 0.03v and the prime errors comes after 2-3min or so...
But when I check it with the scope it'll become more clear, btw, what Vdd are you running?
I don't use my NF7-s much any more, but I've had it prime stable up to 245MHz FSB with 2*512 Tccd with CPC on, 2.6v. I was running 1.7v Vdd, but anything over that didn't benefit me. I couldn't get it 3d stable above 250 with 2 double sided sticks of RAM, but I can bench at 260 with 1 stick in single channel.
I can bench sandra op to 255 dcddr, cpc on, get around 3750, due sucky 2ghz cpu speed...
Anything aboive that result in reboot while test...
I've an infinity to, but riped of the LM358..
It died due some problems with vcore after some bad ocp mods with those 3 little smd resistors...
Is it also possible to work out a vcore supllie the same way? Problem is, there is pretty much damaga around the vcore area :)