I will have a small update on that because I did some further testing. The M12D has an incredible high sb-efficiency with up to 76% which is really good compared to 40-ish% of many others.
Can someone explain the significance for the DC-to-DC configuration over what they or other PSU manufacturers normally do? I mean other than the odd efficiency curves, it doesn't really provide anything unique when compared to say, the Enermax.
I'm guessing here and haven't checked my "facts", but I'm guessing the DC-DC conversion helps because it allows the PSU manufacturer to concentrate most of their effort on providing a clean and efficient +12V rail from the AC supply, then tag the conversion circuitry onto the end of that to generate the less critical +5V and +3.3V rails.
The alternative is to have seperate circuitry for generating +12V, +5V, and +3.3V all individually from the AC supply.
I believe that's right - it's easier to do 12V DC to 3.3V/5V DC than it is to do 120V/230V AC to the same, so you do one good AC-to-DC conversion and then a couple simpler DC-to-DC conversions.
It's not necessarily easier, it's just that if you first convert to 12V alone you have maximized the amount of 12V current a single transformer of a given size can produce, then let that be the primary feedback voltage to determine switching pulse width. That typically also means you can have higher ratios of 12V to 5V or 3.3V current without going out of spec. It makes more and more sense to do so when a PSU is of high wattage as it's then expected to have so much more 12V current consumption as a % of total power used.
Maybe you could put a bug in Seasonic's ear about their "little" 120mm fan? Looking at the acoustic results, you can't help but wonder what a 140mm fan could have done for that curve. A 140mm might have shifted the whole curve down two or three decibels.
Bear in mind half of the fan is usually covered towards the front of the PSU, I'm really not sure a 140mm fan helps at all, because that's simply more airflow that's blocked by a baffle. This baffles me (hurr)... Seems like 120mm is perfectly adequete, and more to the point, if you so wish to change the fan yourself, at least 120mm is a standard size
140s are mainly interesting because they produce any given amount of airflow at a lower RPM then 120s. So rather then run it at the same RPM as the 120, and just block more airflow, they could run it at a lower RPM while producing the same airflow. Then maybe it would top out around 28DB instead of a 31DB, or idle at 15db instead of 17. Hopefully there wouldn't be a need to change the fan then.
First of all, in a design like this no they don't produce their airflow at a significantly lower RPM. You've fallen victim to dubious free air ratings, and ignored that lower quality fans tend to increase in noise level sooner due to wear.
Second, choosing 140mm fan means you can't pick among many of the good fan makes which the Sanyo Denkis are. Let's just leave a good fan alone instead of replacing it with something junky.
I'm going by fan testing from SPCR, not the manufacturer's fluff specs. From what they've seen, to hit x cfm rate, a 80mm is going to have to spin faster then a similarly built 92mm, while that 92mm is going to rev higher then a 120mm, and presumably any good 120mm is going to be rotating faster then any quality 140mm at the same real CFM.
And cheap fans are definitely not what I was suggesting.
Yes, it's me harping on about ripple and noise again :)
There appears to be no ripple in the traces you have shown. The "ripple" figures you quoted are actually noise figures. However, I suspect that the ripple is still there, just overwhelmed by the noise. A low-pass filter to cut off anything more than 1 MHz or so would possibly clean it up and allow ripple numbers to be be determined.
Also, the comparison of 12V1 and 12V2 is just a waste of space. They both come from the same place (they just have independent current limiting, that's all).
This is where I see a future for PSU tech. As more people want to create HTPC/Music Servers I think it would be really good to be able to buy PSUs that deliver high quality, stable, filtered, clean power with the minimum of RFI hash etc.
Units would only need to be around the 300-400w mark (could probably get away with 250w).
Would be nice to be able to get stripped down motherboards with traces designed to be as short as possible and ultra quality power circuits to go with them with.
You are entirely misunderstanding the issues. The power supplied to the system is not directly driving anything where the ripple matters. Voltages are stepped down on the mainboard itself which is a separate more local source of noise, IF that noise would've mattered. A decent audio card will have a linear regulation stage and the digital circuits don't care about a minor bit of ripple.
IOW, there is no reason to believe it would make a difference.
Also, stripped down motherboards with short traces also has no point. These are digital circuits while you are thinking in terms of crude analog audio where the traces need be short because the designer didn't properly shield or ground-plane the circuits. Within the analog audio circuits on boards, the traces ARE short, as much so as reasonably possible. Even then, it's drifting the wrong direction since we've had digital output for audio.
Bottom line- all you need is a good audio card with digital output. There will be zero difference which board or PSU you choose so long as they otherwise worked fine.
ROFL. I'm sure that Monster would be all over that market segment. One thousand dollar 400watt power supplies with gold plated connectors, EMI shielding, and "audiophile grade" filtered power.
Granted, clean power is what sets the good power supplies apart from the chaff, but as soon as you step into the music realm, the marketing machine rules supreme.
Indeed there could be an element of that. However, unlike hifi gear which only gets tested by listening. These components, as they are still PC based would be subject to decent technical testing as per websites such as these.
Extravagant claims would soon be put to the test. Unfortunately thats not something that is done quite nearly enough in the audio world.
Since you've been posting all these great articles on not only PSUs but also energy requirements and effeciency it's really made me look realistically at my system.
I picked up a kill-a-watt a few weeks ago and put it on the plugs for my desktop and my server and the numbers I saw completely re-alligned my expectations. Neither of those systems pull over 200W with anything I throw at them. The server isn't high end stuff but has lots of drives spinning and runs 24/7.
The desktop isn't a gaming monster but even plugged into the wall outlet for the UPS I maxed out at about 250W, and that's with the 19" CRT on the same outlet.
So thanks for the education, it's saved me money already. I needed to replace the server PSU and saved money by shopping in the right portion of the spectrum as well as shooting for max efficiency at the wattage I saw vice some bloated "recomendation".
If the 5v and 3.3v draw from the 12V rail(DC to DC) then what does the 40A rating per rail mean? Is this the rating after the draw from the other rails? What is the max amp draw from both 12V rails?
The reason for this question came from looking at the specs from "Antec Signature850". See here http://www.antec.com/usa/productDetails.php?lan=us...">http://www.antec.com/usa/productDetails.php?lan=us... This has rail 1&2 at 22A each and Rails 3&4 at 25A each, yet the total is 65A max.
The ripple measurements need a more severe judgment, since they are higher than, for example, Corsair's TX750 and worst of all, they have a very bad waveform, with very high frequency components.
Very bad waveform with very high frequency components? The "waveform" exists since it cannot perfectly make a flat line voltage.
Besides, the ATX12V 2.3 spec allows 120 mV ripple for all 12V lines and 50 mV ripple for the 3.3 and 5V lines. It is very well within spec, and consider that these ripple specs are stricter than previous versions of the ATX12V spec, you are going to be fine.
Waveform exists since we are looking and measuring it... and there are good and bad ripple and noise waveforms on the DC rails, just compare with Antec Signature or Corsair 750W or Enermax Revolution or... definitively a very bad waveform.
These waveforms are more like very high frequency noise covering ripple and this is not good, since on the PSU's DC rails we should see ripple and not the same noise level.
Actually it won't make a bit of difference in use. The day a little bit of ripple or it's frequency matters, will be the day we all start using linear PSU again.
Why does Seasonic insist on putting crappy ADDA fans in 150.00+ units? For the same price, they could at least source from Panasonic or Yate Loon.
That ticking sound in the "heavily undervoltaged fan" (nice heavily murderaged of the englished language, there) is incredibly annoying for those of us who haven't destroyed our hearing by playing our iPods at 90db every day for the last 5 years.
WTF? Yate Loon is the bottom of the barrel, ADDA is mid quality. If you want a yate loon on your case wall it's not so bad but those fail pretty frequently when placed horizontally in a PSU.
FWIW, that was another error with the speech-recognition. Dragon NaturallySpeaking does so well that sometimes I miss errors in dictation. Apparently, it thinks "undervoltage" is an allowable word, but I know I said "undervolted". Whatever.
In case you're wondering, Christoph is not a native English speaker, but he is fluent in at least English, French, and Chinese -- besides his native tongue of German. Feel free to critique his use of a second tongue; me, I'm happy to speak English, Danish, and some German. Anyway, I do pretty heavy editing of his text to clean up the English, but errors slip through on occasion. I was pressed for time on this one (had to run off to the airport) so I didn't give the final text a second proofread.
On page 3, about the cables, the speech-recognition software seems to have made another error
[/quote]Most of the cables are detachable and come in an extra bag Seasonic provides. The 24-pin ATX connector, 4-pin ATX12V and 8-pin EPS12V connectors, 260cm 6-pin PEG connectors...[/quote]
I think that should read two 60cm 6-pin PEG connectors.
What speech recognition s/w do you use Jarred? Do you use it just for large amounts of text entry or is useful for other things like quick posts like this. Does it make things quicker? Or does fixing the errors eat up a lot of time? I guess you just save time and don't fix those errors. haha j/k!
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
45 Comments
Back to Article
CEO Ballmer - Saturday, November 29, 2008 - link
I have one, works nicely!http://fakesteveballmer.blogspot.com">http://fakesteveballmer.blogspot.com
InterClaw - Friday, November 28, 2008 - link
How much power does the M12D consume while in standby mode (switch on) running on 230Vac?Christoph Katzer - Saturday, November 29, 2008 - link
I will have a small update on that because I did some further testing. The M12D has an incredible high sb-efficiency with up to 76% which is really good compared to 40-ish% of many others.daar - Wednesday, November 26, 2008 - link
Can someone explain the significance for the DC-to-DC configuration over what they or other PSU manufacturers normally do? I mean other than the odd efficiency curves, it doesn't really provide anything unique when compared to say, the Enermax.PrinceGaz - Wednesday, November 26, 2008 - link
I'm guessing here and haven't checked my "facts", but I'm guessing the DC-DC conversion helps because it allows the PSU manufacturer to concentrate most of their effort on providing a clean and efficient +12V rail from the AC supply, then tag the conversion circuitry onto the end of that to generate the less critical +5V and +3.3V rails.The alternative is to have seperate circuitry for generating +12V, +5V, and +3.3V all individually from the AC supply.
JarredWalton - Thursday, November 27, 2008 - link
I believe that's right - it's easier to do 12V DC to 3.3V/5V DC than it is to do 120V/230V AC to the same, so you do one good AC-to-DC conversion and then a couple simpler DC-to-DC conversions.mindless1 - Saturday, November 29, 2008 - link
It's not necessarily easier, it's just that if you first convert to 12V alone you have maximized the amount of 12V current a single transformer of a given size can produce, then let that be the primary feedback voltage to determine switching pulse width. That typically also means you can have higher ratios of 12V to 5V or 3.3V current without going out of spec. It makes more and more sense to do so when a PSU is of high wattage as it's then expected to have so much more 12V current consumption as a % of total power used.Mr Perfect - Wednesday, November 26, 2008 - link
Maybe you could put a bug in Seasonic's ear about their "little" 120mm fan? Looking at the acoustic results, you can't help but wonder what a 140mm fan could have done for that curve. A 140mm might have shifted the whole curve down two or three decibels.piroroadkill - Thursday, November 27, 2008 - link
Bear in mind half of the fan is usually covered towards the front of the PSU, I'm really not sure a 140mm fan helps at all, because that's simply more airflow that's blocked by a baffle. This baffles me (hurr)... Seems like 120mm is perfectly adequete, and more to the point, if you so wish to change the fan yourself, at least 120mm is a standard sizeMr Perfect - Thursday, November 27, 2008 - link
140s are mainly interesting because they produce any given amount of airflow at a lower RPM then 120s. So rather then run it at the same RPM as the 120, and just block more airflow, they could run it at a lower RPM while producing the same airflow. Then maybe it would top out around 28DB instead of a 31DB, or idle at 15db instead of 17. Hopefully there wouldn't be a need to change the fan then.mindless1 - Saturday, November 29, 2008 - link
First of all, in a design like this no they don't produce their airflow at a significantly lower RPM. You've fallen victim to dubious free air ratings, and ignored that lower quality fans tend to increase in noise level sooner due to wear.Second, choosing 140mm fan means you can't pick among many of the good fan makes which the Sanyo Denkis are. Let's just leave a good fan alone instead of replacing it with something junky.
Mr Perfect - Wednesday, December 3, 2008 - link
I'm going by fan testing from SPCR, not the manufacturer's fluff specs. From what they've seen, to hit x cfm rate, a 80mm is going to have to spin faster then a similarly built 92mm, while that 92mm is going to rev higher then a 120mm, and presumably any good 120mm is going to be rotating faster then any quality 140mm at the same real CFM.And cheap fans are definitely not what I was suggesting.
emboss - Wednesday, November 26, 2008 - link
Yes, it's me harping on about ripple and noise again :)There appears to be no ripple in the traces you have shown. The "ripple" figures you quoted are actually noise figures. However, I suspect that the ripple is still there, just overwhelmed by the noise. A low-pass filter to cut off anything more than 1 MHz or so would possibly clean it up and allow ripple numbers to be be determined.
Also, the comparison of 12V1 and 12V2 is just a waste of space. They both come from the same place (they just have independent current limiting, that's all).
jabber - Wednesday, November 26, 2008 - link
This is where I see a future for PSU tech. As more people want to create HTPC/Music Servers I think it would be really good to be able to buy PSUs that deliver high quality, stable, filtered, clean power with the minimum of RFI hash etc.Units would only need to be around the 300-400w mark (could probably get away with 250w).
Would be nice to be able to get stripped down motherboards with traces designed to be as short as possible and ultra quality power circuits to go with them with.
mindless1 - Saturday, November 29, 2008 - link
You are entirely misunderstanding the issues. The power supplied to the system is not directly driving anything where the ripple matters. Voltages are stepped down on the mainboard itself which is a separate more local source of noise, IF that noise would've mattered. A decent audio card will have a linear regulation stage and the digital circuits don't care about a minor bit of ripple.IOW, there is no reason to believe it would make a difference.
Also, stripped down motherboards with short traces also has no point. These are digital circuits while you are thinking in terms of crude analog audio where the traces need be short because the designer didn't properly shield or ground-plane the circuits. Within the analog audio circuits on boards, the traces ARE short, as much so as reasonably possible. Even then, it's drifting the wrong direction since we've had digital output for audio.
Bottom line- all you need is a good audio card with digital output. There will be zero difference which board or PSU you choose so long as they otherwise worked fine.
Mr Perfect - Wednesday, November 26, 2008 - link
ROFL. I'm sure that Monster would be all over that market segment. One thousand dollar 400watt power supplies with gold plated connectors, EMI shielding, and "audiophile grade" filtered power.Granted, clean power is what sets the good power supplies apart from the chaff, but as soon as you step into the music realm, the marketing machine rules supreme.
jabber - Thursday, November 27, 2008 - link
Indeed there could be an element of that. However, unlike hifi gear which only gets tested by listening. These components, as they are still PC based would be subject to decent technical testing as per websites such as these.Extravagant claims would soon be put to the test. Unfortunately thats not something that is done quite nearly enough in the audio world.
djc208 - Wednesday, November 26, 2008 - link
Since you've been posting all these great articles on not only PSUs but also energy requirements and effeciency it's really made me look realistically at my system.I picked up a kill-a-watt a few weeks ago and put it on the plugs for my desktop and my server and the numbers I saw completely re-alligned my expectations. Neither of those systems pull over 200W with anything I throw at them. The server isn't high end stuff but has lots of drives spinning and runs 24/7.
The desktop isn't a gaming monster but even plugged into the wall outlet for the UPS I maxed out at about 250W, and that's with the 19" CRT on the same outlet.
So thanks for the education, it's saved me money already. I needed to replace the server PSU and saved money by shopping in the right portion of the spectrum as well as shooting for max efficiency at the wattage I saw vice some bloated "recomendation".
computerfarmer - Wednesday, November 26, 2008 - link
If the 5v and 3.3v draw from the 12V rail(DC to DC) then what does the 40A rating per rail mean? Is this the rating after the draw from the other rails? What is the max amp draw from both 12V rails?The reason for this question came from looking at the specs from "Antec Signature850". See here http://www.antec.com/usa/productDetails.php?lan=us...">http://www.antec.com/usa/productDetails.php?lan=us... This has rail 1&2 at 22A each and Rails 3&4 at 25A each, yet the total is 65A max.
How is anyone supposed to understand these specs?
valdir - Wednesday, November 26, 2008 - link
The ripple measurements need a more severe judgment, since they are higher than, for example, Corsair's TX750 and worst of all, they have a very bad waveform, with very high frequency components.sprockkets - Wednesday, November 26, 2008 - link
Very bad waveform with very high frequency components? The "waveform" exists since it cannot perfectly make a flat line voltage.Besides, the ATX12V 2.3 spec allows 120 mV ripple for all 12V lines and 50 mV ripple for the 3.3 and 5V lines. It is very well within spec, and consider that these ripple specs are stricter than previous versions of the ATX12V spec, you are going to be fine.
valdir - Wednesday, November 26, 2008 - link
Waveform exists since we are looking and measuring it... and there are good and bad ripple and noise waveforms on the DC rails, just compare with Antec Signature or Corsair 750W or Enermax Revolution or... definitively a very bad waveform.These waveforms are more like very high frequency noise covering ripple and this is not good, since on the PSU's DC rails we should see ripple and not the same noise level.
mindless1 - Saturday, November 29, 2008 - link
Actually it won't make a bit of difference in use. The day a little bit of ripple or it's frequency matters, will be the day we all start using linear PSU again.fri2219 - Wednesday, November 26, 2008 - link
Why does Seasonic insist on putting crappy ADDA fans in 150.00+ units? For the same price, they could at least source from Panasonic or Yate Loon.That ticking sound in the "heavily undervoltaged fan" (nice heavily murderaged of the englished language, there) is incredibly annoying for those of us who haven't destroyed our hearing by playing our iPods at 90db every day for the last 5 years.
mindless1 - Saturday, November 29, 2008 - link
WTF? Yate Loon is the bottom of the barrel, ADDA is mid quality. If you want a yate loon on your case wall it's not so bad but those fail pretty frequently when placed horizontally in a PSU.sprockkets - Wednesday, November 26, 2008 - link
Perhaps you can re-read page 1 where it says it uses a Sanyo Denki fan.And, what is "murderaged" spelling hypocrite?
Slash3 - Wednesday, November 26, 2008 - link
Wooshed.JarredWalton - Wednesday, November 26, 2008 - link
FWIW, that was another error with the speech-recognition. Dragon NaturallySpeaking does so well that sometimes I miss errors in dictation. Apparently, it thinks "undervoltage" is an allowable word, but I know I said "undervolted". Whatever.In case you're wondering, Christoph is not a native English speaker, but he is fluent in at least English, French, and Chinese -- besides his native tongue of German. Feel free to critique his use of a second tongue; me, I'm happy to speak English, Danish, and some German. Anyway, I do pretty heavy editing of his text to clean up the English, but errors slip through on occasion. I was pressed for time on this one (had to run off to the airport) so I didn't give the final text a second proofread.
RallyMaster - Wednesday, November 26, 2008 - link
"That's not entirely wrong, as PSUs are one of the components most people only think about when their old unit sales, or when building a new system."What does "old unit sales" even mean? Are you sure you're not saying "old unit fails?"
Spoelie - Wednesday, November 26, 2008 - link
Page 1"The PSUs are also supposed to have very tight voltage regulation in the future only Japanese manufactured capacitors."
I'm not quite sure what to make of that sentence.
JarredWalton - Wednesday, November 26, 2008 - link
I have no idea... I think I said "and they feature" perhaps? Heh. Go Dragon!Slash3 - Wednesday, November 26, 2008 - link
Argh! The fingers, use them! :) It's probably faster, too, after correcting two dozen mis-transcribes.JarredWalton - Thursday, November 27, 2008 - link
I would... but carpal tunnel issues make it a problem.BikeDude - Saturday, November 29, 2008 - link
Sorry to hear about your carpal tunnel issues.But I have always thought that such issues stems from mouse usage. So at work I use my left hand for moving the pointer, and at home the right hand.
Using a keyboard should not normally be much cause for concern?
JarredWalton - Wednesday, November 26, 2008 - link
Blame the editor (me) and my speech-recognition. Should have been "fails".PrinceGaz - Wednesday, November 26, 2008 - link
On page 3, about the cables, the speech-recognition software seems to have made another error[/quote]Most of the cables are detachable and come in an extra bag Seasonic provides. The 24-pin ATX connector, 4-pin ATX12V and 8-pin EPS12V connectors, 260cm 6-pin PEG connectors...[/quote]
I think that should read two 60cm 6-pin PEG connectors.
Zoomer - Thursday, November 27, 2008 - link
I thought someone was supposed to proof-read these. ;)feraltoad - Wednesday, November 26, 2008 - link
What speech recognition s/w do you use Jarred? Do you use it just for large amounts of text entry or is useful for other things like quick posts like this. Does it make things quicker? Or does fixing the errors eat up a lot of time? I guess you just save time and don't fix those errors. haha j/k!feraltoad - Wednesday, November 26, 2008 - link
Oh. U use Dragon. Well, other questions stand!