resurecting an old post may not be a good idea, but i actualy rememberd this topic, and last week i did see a real linear psu powered computer.
was not built for audio use, but for overclock.
actualy there is quite a gain, it does not surprise me at all.
it does indeed look quite "dumb" to have a big powerbrick next to a pc,
but on the otherhand nearly every single test shown improvement.
better overclock capacity, and even , better hdd performance to mention a few.
(probably less noise in data lines helps that hdd read/write figure)
one thing that was verry intresting is to see running temp going down with the linear psu.
this one is a bit spooky for me, i *might* have a clue why is that, but it could be verry wrong.
The local geek does plans to run as mutch of the fans and other motors as possible on an isolated supply, this is what i would not try.
The ps is verry ineffective, and uses linear regulators. without them the computer fails to start up according to him.
Even so start up procedure is complex.
The psu has dummy loads on all rails, after mains power on there is a secund of delay prior motherboard powerup, or else the ripple halts the system.
So it does not look like an easy project, but surely cool for overclocking.
I resurrected the thread to share what i seen, who knows it may be usefull for someone out there on the internetS.
was not built for audio use, but for overclock.
actualy there is quite a gain, it does not surprise me at all.
it does indeed look quite "dumb" to have a big powerbrick next to a pc,
but on the otherhand nearly every single test shown improvement.
better overclock capacity, and even , better hdd performance to mention a few.
(probably less noise in data lines helps that hdd read/write figure)
one thing that was verry intresting is to see running temp going down with the linear psu.
this one is a bit spooky for me, i *might* have a clue why is that, but it could be verry wrong.
The local geek does plans to run as mutch of the fans and other motors as possible on an isolated supply, this is what i would not try.
The ps is verry ineffective, and uses linear regulators. without them the computer fails to start up according to him.
Even so start up procedure is complex.
The psu has dummy loads on all rails, after mains power on there is a secund of delay prior motherboard powerup, or else the ripple halts the system.
So it does not look like an easy project, but surely cool for overclocking.
I resurrected the thread to share what i seen, who knows it may be usefull for someone out there on the internetS.
Arty,
Thank you for the post. Very insightfull that someone did make it work. Any idea how much power the psu was able to deliver?
I guess he did all rails linear. That's tough. I also worry about the response to sudden powerdraws.
I've been thinking about a hybrid version. Something like a beefy linear ~24V cap multiplier with one of the hv picopsu supplies for the motherboard only. Audio and light work pc only, no fancy graphics (maybe a hd 5545). Low power sata devices can be made to run from separate linear supplies. You could get away with less than 150W peak and still be using an i5 (no oc). I would go for something like 10A @~24v continuous. This can be done with a few components and a heat sink.
This setup will skip the HV switcher on the primary, imho the biggest sinner in the atx supply. The HV pico psus have very wide input range (necessary for car use) and are therefore quite immune to voltage sags. This relaxes the regulation requirement for the input. This is also where the capacitance multiplier comes in. No tight regulation, but it reduces the input ripple drastically and has a low voltage drop so less waste heat to get rid off.
Furthermore the current requirement for a 24V supply is half of that of a 12V direct one.
The only concern is the remaining switching noise of the pico psu will not spoil the effort. A good linear supply will have much less ripple.
More information, if you have it, will be appreciated!
Regards, Coen
Thank you for the post. Very insightfull that someone did make it work. Any idea how much power the psu was able to deliver?
I guess he did all rails linear. That's tough. I also worry about the response to sudden powerdraws.
I've been thinking about a hybrid version. Something like a beefy linear ~24V cap multiplier with one of the hv picopsu supplies for the motherboard only. Audio and light work pc only, no fancy graphics (maybe a hd 5545). Low power sata devices can be made to run from separate linear supplies. You could get away with less than 150W peak and still be using an i5 (no oc). I would go for something like 10A @~24v continuous. This can be done with a few components and a heat sink.
This setup will skip the HV switcher on the primary, imho the biggest sinner in the atx supply. The HV pico psus have very wide input range (necessary for car use) and are therefore quite immune to voltage sags. This relaxes the regulation requirement for the input. This is also where the capacitance multiplier comes in. No tight regulation, but it reduces the input ripple drastically and has a low voltage drop so less waste heat to get rid off.
Furthermore the current requirement for a 24V supply is half of that of a 12V direct one.
The only concern is the remaining switching noise of the pico psu will not spoil the effort. A good linear supply will have much less ripple.
More information, if you have it, will be appreciated!
Regards, Coen
why use switchers?
Guys as far as I know, the switchers are used so that the rectified ripple requires a smaller size capacitor for the smoothing. On the commercial frequencies you have a ripple of 100 or 120 hertz but after the switcher the frequency is a couple of kilohertz and you do not need a googal (10 power99) farads to smooth it. When we still had those VIC20 and Commodore64 Sinclairs and the like and those linears were scorching the furniture and they were only running 4Mhz. The first IBM p.c.'s that had linears you needed a forklift to shift them. A 500watt transformer weighs more than 50 Kilos the smoothing caps were 50X 100 000 microfarads! The rectifiers were huge on huge heatsiks. (About 40mm diameter) The processor? 50Mhz 8008 intel cpu if my mind does not play tricks on me!
Guys as far as I know, the switchers are used so that the rectified ripple requires a smaller size capacitor for the smoothing. On the commercial frequencies you have a ripple of 100 or 120 hertz but after the switcher the frequency is a couple of kilohertz and you do not need a googal (10 power99) farads to smooth it. When we still had those VIC20 and Commodore64 Sinclairs and the like and those linears were scorching the furniture and they were only running 4Mhz. The first IBM p.c.'s that had linears you needed a forklift to shift them. A 500watt transformer weighs more than 50 Kilos the smoothing caps were 50X 100 000 microfarads! The rectifiers were huge on huge heatsiks. (About 40mm diameter) The processor? 50Mhz 8008 intel cpu if my mind does not play tricks on me!
Hi,
Imho the main benefit of the switching is that you can make very high power- very very small transformers. That's why we can have a 1kW switching pc supply fitting in your 50-60Hz 500W power transformer. The ripple on the input cap, before the HV switcher, is still 100-120 Hz. But that is compensated for by the regulation around the power switching transformer. Indeed the output caps are very small.
Fortunately we have moved on on the silicium side. No more powersavvy ECL and big (bipolar?) transistors on the cpu dye. Rectifiers have become much faster with less lossess. Torodial transformers are also more space efficent, yet still a ATX compliant 250W supply will be huge.
Regards, Coen
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.