Changing voltages.

[Spork Schivago]

Re: Changing voltages.

ripple current is the current (at the rated frequency) that a capacitor can absorb / release.

So higher ripple current for a cap, the better?

[Spork Schivago]

Re: Changing voltages.

...The ATX spec states that for the 12v line you may have a maximum of 120mV ripple.
And it's 50mV for 5v and 3.3v (easy to remember since it's a division of the major voltage )

So start up the comp again on the shit PSU if you are not afraid to kill it, connect the GND to any negative and the probe to 12v.
Set the scope for maybe 1ms and half a volt per division and coupling to AC, if the screen isn't already filled with shit then zoom in and see how much you've got

I got the scope hooked up and set like you said. I hooked up the PSU to my tester instead of the motherboard. I hooked up my probe, the probe tip is connected to 12V, the GND is connected to one of the GND pins. I set it for 1ms and a half a volt per division (.5) and coupling to AC. It's really cool looking. I can see a straight line and then like a wave on the ocean move across it. Is that the ripple?

After a while, the line straightens out. The longer the PSU is on, the straighter the line gets. Now it's pretty smooth. So, I zoom in, by setting it to 20mv per division and it's like a saw blade! Instead of a nice smooth line, it's all woobly. The new power supply should have it nice and smooth.

[stj]

Re: Changing voltages.

that's the ripple, bigger caps would smooth that out.

put it on the motherboard so it actually has a load and it will look atleast 10 times worse!

[stj]

Re: Changing voltages.

So higher ripple current for a cap, the better?

yes

[ChaosLegionnaire]

Re: Changing voltages.

I count 4 coils next to the CPU, 4 caps total. Wouldn't that make it 4 phase?

nope. the coil by the atx12v connector does not count. it also has a different inductance from the other 3. its 3 phases. there are also 6 switching mosfets so that further reinforces the point its 3 phases.

it looks the the 3phase switcher only has 3 caps - cheap bastards!

actually, using this picture of the board from the manufacturer which is a clearer view of the entire board without the cpu hsf in the way, the cpu vrm input has 4 electrolytic caps and the output has 5 polymer caps.

how do I calculate that FSB base frequency?

the fsb in pentium 4 era till core 2 era chips are quad pumped or quad data rate (qdr) so u divide it by four to get the base fsb clock. i would have thought u'd known this by now... as i explained this in another post!

it's like a saw blade! Instead of a nice smooth line, it's all woobly.

that is the ripple. do u know how to read the peak to peak ripple ala jonnyguru style? how much is it? im dying to know, especially whether the ripple exceeds atx specs or not. or just post a pic of what is shown on the scope.

btw, a warning why u should NEVER use a junk psu with high output ripple: it kills hard drives. hard drives totally HATE ripple esp on the 5v line which is converted to 3v or 3.3v and is meant for the sensitive electronics on the drive. hard drives dont tap on the 3.3v rail of the sata power connector due to the existence of molex to sata converters. high ripple = bye bye hard drive aka hard drive died? its probably your power supply...

[dmill89]

Re: Changing voltages.

btw, a warning why u should NEVER use a junk psu with high output ripple: it kills hard drives. hard drives totally HATE ripple esp on the 5v line which is converted to 3v or 3.3v and is meant for the sensitive electronics on the drive. high ripple = bye bye hard drive aka hard drive died? its probably your power supply...

^This, there are "ok" cheap Power Supplies that are reasonably well built but with cheap caps, etc. and/or are overrated that are good for a couple years and eventually fail due to the aforementioned cheap components (and are often ok after a re-cap, and keeping the load realistic rather than going by the "label rating"), and then there is the pure crap that doesn't belong anywhere near a computer and will output ripple well outside of spec causing damage to components even as it appears to be "working fine" and possibly fail in a spectacular manner causing massive overvoltage and severe damage to components. The Okia is in the latter category and belongs in the nearest garbage receptacle (possibly after striping any usable parts).

The "420W" Okia I pictured earlier handled a 5V heavy (since this PSU is an old design, the 5V at least had a 40A schottky, while the 12V used "diodes on a bracket") ~150W load (resistors, I'd never put that piece of crap in a computer) for about 10 minutes before blowing up (obviously no over temp protection as it got very hot before blowing). I don't have an oscilloscope so I can't comment on ripple, but the voltages were pretty unstable.

Considering you can get a decent (though not spectacular) 350W-450W new PSU for $30-$40 (FSP, Antec BP/VP, low-end Seasonics, CWT-built units from Corsair, etc.) that may not have the best caps but will be many times better than that Okia, or a good NOS/used PSU for less than $20 shipped on ebay there is no reason to use a piece of crap power supply.

[Spork Schivago]

Re: Changing voltages.

nope. the coil by the atx12v connector does not count. it also has a different inductance from the other 3. its 3 phases. there are also 6 switching mosfets so that further reinforces the point its 3 phases.

actually, using this picture of the board from the manufacturer which is a clearer view of the entire board without the cpu hsf in the way, the cpu vrm input has 4 electrolytic caps and the output has 5 polymer caps.

So can I always fall back on dividing the number of switching mosfets by two? I almost thought I remembered something about how some boards might not use those switching mosfets at all. Thanks for telling me the coil by the atx12v connector doesn't count.

the fsb in pentium 4 era till core 2 era chips are quad pumped or quad data rate (qdr) so u divide it by four to get the base fsb clock. i would have thought u'd known this by now... as i explained this in another post!

I have some issues with my brain and some days, it just doesn't work. The memory part of it I mean. I wish there was I way I could switch that part out with better memory! That'd be real nice. Come tomorrow, I'll more than likely remember. Seems today was one of those days. I had trouble remembering stuff all day.

that is the ripple. do u know how to read the peak to peak ripple ala jonnyguru style? how much is it? im dying to know, especially whether the ripple exceeds atx specs or not. or just post a pic of what is shown on the scope.

No, I do not know. I watched a video on it but they had a much, much, much nicer scope! I do have this option on my Hitachi 1065-A to take measurements. Just not sure how it works. Once I left the PSU on for a while, at .5 volts per division, the line appeared real straight. I was thinking maybe I didn't give my scope enough time to warm up or maybe somehow the ripple stopped?

btw, a warning why u should NEVER use a junk psu with high output ripple: it kills hard drives. hard drives totally HATE ripple esp on the 5v line which is converted to 3v or 3.3v and is meant for the sensitive electronics on the drive. hard drives dont tap on the 3.3v rail of the sata power connector due to the existence of molex to sata converters. high ripple = bye bye hard drive aka hard drive died? its probably your power supply...

My oscilloscope would be fine with high output ripple though, right? Maybe instead of having the PSU hooked up to my power supply tester I should hook it up to an older board of mine that I don't use. Stj mentioned I'd get a load by hooking it to the board, so I'll give that a shot tomorrow, instead of risking damaging the customers board.

[Spork Schivago]

Re: Changing voltages.

^This, there are "ok" cheap Power Supplies that are reasonably well built but with cheap caps, etc. and/or are overrated that are good for a couple years and eventually fail due to the aforementioned cheap components (and are often ok after a re-cap, and keeping the load realistic rather than going by the "label rating"), and then there is the pure crap that doesn't belong anywhere near a computer and will output ripple well outside of spec causing damage to components even as it appears to be "working fine" and possibly fail in a spectacular manner causing massive overvoltage and severe damage to components. The Okia is in the latter category and belongs in the nearest garbage receptacle (possibly after striping any usable parts).

The "420W" Okia I pictured earlier handled a 5V heavy (since this PSU is an old design, the 5V at least had a 40A schottky, while the 12V used "diodes on a bracket") ~150W load (resistors, I'd never put that piece of crap in a computer) for about 10 minutes before blowing up (obviously no over temp protection as it got very hot before blowing). I don't have an oscilloscope so I can't comment on ripple, but the voltages were pretty unstable.

Considering you can get a decent (though not spectacular) 350W-450W new PSU for $30-$40 (FSP, Antec BP/VP, low-end Seasonics, CWT-built units from Corsair, etc.) that may not have the best caps but will be many times better than that Okia, or a good NOS/used PSU for less than $20 shipped on ebay there is no reason to use a piece of crap power supply.

Yeah, I have the low-end Seasonics PSU ordered. I ordered it yesterday and have a tracking number already. For my personal board that I just bought (a Gigabyte one), I need to buy a new case. I didn't realize how damaging a crappy power supply could be. I'm so glad I posted here. For my system, I was hoping of going for a modular type system. I'll probably start a new thread when I'm ready and ask for suggestions on the case / PSU.

[Spork Schivago]

Re: Changing voltages.

For those of you guys that have a scope, how do you check the PSU lines with it? I mean, how do you physically hook the probe up to the line and the GND up to the GND? I think I have what's called a poor man's ground or something like that. It's a separate GND cable that comes off the probe and has alligator clips on it. I believe this type of GND can cause noise on the line. I've seen in a video a guy using a different tip that had a much shorter GND on it.

My probe has a hook on it that I clip on a wire. I had to strip the 12V and GND on the PSU 24-pin connector to hook it up. Do they make different tips where I can just stick them down in the pin holes? What do you guys use?

[keeney123]

Re: Changing voltages.

I'm a bit confused by what ChaosLegionnaire said though. The RAM is running at 800MHz but it has a base frequency of 400MHz. The CPU's FSB is running at 800MHz...but because it's a dual core, would the based frequency be 400MHz? Wouldn't that give a 1:1 FSB to RAM ratio?

So there are many different configurations that have been developed. By that I mean the hardware layout and the software that runs it. The memory cores from different processors are set out in different ways. The frequency of the clock may or may not be directly connected to the Front side Bus and the Ram depending on the architecture. This is why the hardware specs of each system are important in understanding how it is running. The speed of the computer as a whole is also how the software is written. When I use to shop for a computer I look at the nano second of the processor it's latency, the forward bus speed, the ram speed and amount allowed, what graphic processor it had and whether it was important to me. Although I have not done an deep study of this subject because it is of little concern to me now. What I do know is the faster your processor runs and operation the quicker it is. The more memory inside of the processor the better it is. The greater the number of bits and address the processor is capable of handling the better. The faster the forward side bus the better. And the faster the Ram the better. There is good and bad in the newer processors that run asynchronously. Personally I have seen old single processors that ran just as fast as the newer multi-core processors. The 32 nano second processor is about as fast as they have gotten. The bottleneck use to be the hard-drive, but the solid state drives have lessen that problem. I would imagine to improve the speed of data a completely new technology will need to be developed. Perhaps a better understand of photons and how to control them will lead to revolutionary idea I do not know?

[stj]

Re: Changing voltages.

For those of you guys that have a scope, how do you check the PSU lines with it? I mean, how do you physically hook the probe up to the line and the GND up to the GND? I think I have what's called a poor man's ground or something like that. It's a separate GND cable that comes off the probe and has alligator clips on it. I believe this type of GND can cause noise on the line. I've seen in a video a guy using a different tip that had a much shorter GND on it.

My probe has a hook on it that I clip on a wire. I had to strip the 12V and GND on the PSU 24-pin connector to hook it up. Do they make different tips where I can just stick them down in the pin holes? What do you guys use?

the grounding clip is usually 6inches or less, he idea is to get the ground near the signal for less noise.

connect your probe by sticking it into the back of the psu conector while it's on the motherboard.

[stj]

Re: Changing voltages.

Once I left the PSU on for a while, at .5 volts per division, the line appeared real straight. I was thinking maybe I didn't give my scope enough time to warm up or maybe somehow the ripple stopped?

the caps in the psu warmed up.
you need to be very wary of this phenomenon, caps have lower esr when they are warm.

if you desolder a cap to test it - let it cool down first!

[Spork Schivago]

Re: Changing voltages.

So there are many different configurations that have been developed. By that I mean the hardware layout and the software that runs it. The memory cores from different processors are set out in different ways. The frequency of the clock may or may not be directly connected to the Front side Bus and the Ram depending on the architecture. This is why the hardware specs of each system are important in understanding how it is running. The speed of the computer as a whole is also how the software is written. When I use to shop for a computer I look at the nano second of the processor it's latency, the forward bus speed, the ram speed and amount allowed, what graphic processor it had and whether it was important to me. Although I have not done an deep study of this subject because it is of little concern to me now. What I do know is the faster your processor runs and operation the quicker it is. The more memory inside of the processor the better it is. The greater the number of bits and address the processor is capable of handling the better. The faster the forward side bus the better. And the faster the Ram the better. There is good and bad in the newer processors that run asynchronously. Personally I have seen old single processors that ran just as fast as the newer multi-core processors. The 32 nano second processor is about as fast as they have gotten. The bottleneck use to be the hard-drive, but the solid state drives have lessen that problem. I would imagine to improve the speed of data a completely new technology will need to be developed. Perhaps a better understand of photons and how to control them will lead to revolutionary idea I do not know?

They have those new M.2 drives and if the interface on the board is implemented properly, they're supposed to be really fast. Didn't Intel find away to do away with the FSB? There was a socket, can't remember which one. Something like LGA 1366 that didn't use a FSB. The socket 2011 doesn't have one. My limited understanding of it was the CPU "talks" directly to the RAM and vice versa.

[Spork Schivago]

Re: Changing voltages.

the caps in the psu warmed up.
you need to be very wary of this phenomenon, caps have lower esr when they are warm.

if you desolder a cap to test it - let it cool down first!

Will do. My wife and I aren't feeling too hot now. I don't know if we caught some bug or what. I'm going to go to bed. Depending on how I feel tomorrow, I might not be on the PC much.

[keeney123]

Re: Changing voltages.

So there are many different configurations that have been developed. By that I mean the hardware layout and the software that runs it. The memory cores from different processors are set out in different ways. The frequency of the clock may or may not be directly connected to the Front side Bus and the Ram depending on the architecture. This is why the hardware specs of each system are important in understanding how it is running. The speed of the computer as a whole is also how the software is written. When I use to shop for a computer I look at the nano second of the processor it's latency, the forward bus speed, the ram speed and amount allowed, what graphic processor it had and whether it was important to me. Although I have not done an deep study of this subject because it is of little concern to me now. What I do know is the faster your processor runs and operation the quicker it is. The more memory inside of the processor the better it is. The greater the number of bits and address the processor is capable of handling the better. The faster the forward side bus the better. And the faster the Ram the better. There is good and bad in the newer processors that run asynchronously. Personally I have seen old single processors that ran just as fast as the newer multi-core processors. The 32 nano second processor is about as fast as they have gotten. The bottleneck use to be the hard-drive, but the solid state drives have lessen that problem. I would imagine to improve the speed of data a completely new technology will need to be developed. Perhaps a better understand of photons and how to control them will lead to revolutionary idea I do not know?

Let me correct the above the 32 nano second should read 0.32 nano second. I know they have got down to 0.25 nano seconds and even a little lower. I think this is about far as they can push the performance of one clock cycle which would equal one clock cycle of a machine cycle.

[keeney123]

Re: Changing voltages.

Well if you have a scope I'm not sure why we are having this discussion with all of the rest of the FUD about RAM clockspeed related to how many CPU cores which is just completely bananas.

The ATX spec states that for the 12v line you may have a maximum of 120mV ripple.
And it's 50mV for 5v and 3.3v (easy to remember since it's a division of the major voltage )

So start up the comp again on the shit PSU if you are not afraid to kill it, connect the GND to any negative and the probe to 12v.
Set the scope for maybe 1ms and half a volt per division and coupling to AC, if the screen isn't already filled with shit then zoom in and see how much you've got

Is that yellow or red bananas? I think what would be useful is, if you get the time, to start a thread explain speed in a modern day computer. Something that is very basic as the audience goes from novice to experts and maybe give reference for further study for those who may want to learn more in detail.

[Spork Schivago]

Re: Changing voltages.

Is that yellow or red bananas? I think what would be useful is, if you get the time, to start a thread explain speed in a modern day computer. Something that is very basic as the audience goes from novice to experts and maybe give reference for further study for those who may want to learn more in detail.

I think that's a great idea! I love BadCaps.net because I can come here for just about any problem I have, whether it's related to PCs, TVs, DVD players, electronic theory, whatever. It'd be real nice if there was some thread that explained how the hardware communicated together, what the various terms meant, etc. Things in the IT world can change real quick like. It'd be wonderful if there was a place here to get all the up-to-date info on the various changes and stuff.

[keeney123]

Re: Changing voltages.

I think that's a great idea! I love BadCaps.net because I can come here for just about any problem I have, whether it's related to PCs, TVs, DVD players, electronic theory, whatever. It'd be real nice if there was some thread that explained how the hardware communicated together, what the various terms meant, etc. Things in the IT world can change real quick like. It'd be wonderful if there was a place here to get all the up-to-date info on the various changes and stuff.

Yea, I think it should start with the latest stuff and go backwards because people would be interested in now. The backward stuff could have references to other articles for learning. It would be hard for one person to do it all. That is why I suggested just a basic understanding of today computers and reference article, books and the like for the readers to do further study. If this was done by someone very knowledgeable their references would be accurate. Of course a person would have to have the time to do this.

[diif]

Re: Changing voltages.

They have those new M.2 drives and if the interface on the board is implemented properly, they're supposed to be really fast. Didn't Intel find away to do away with the FSB? There was a socket, can't remember which one. Something like LGA 1366 that didn't use a FSB. The socket 2011 doesn't have one. My limited understanding of it was the CPU "talks" directly to the RAM and vice versa.

The FSB was replaced by QPI - QuickPath Interconnect on the 1366 socket.
The memory controller was moved to the CPU.

[stj]

Re: Changing voltages.

so they ripped off amd!