Power supplies in-cable capacitors, why people dislike them ?

[Grog6]

Yeah, if they do anything at all, it would be on the bleeding edge of the ratings for the power supply, where the extra capacitance is more effective because it's not in series with the cable resistance.

So, max current, max OC.

An 18awg cable 2 feet long is 8 milli ohms; 0.008 ohms.

Do you have a part number for those caps?

The lowest ESR I see for a cap that size is ~0.016 ohms; the cable inductance must be what they are trying to reduce.

They don't want the wire to make the power supply on the motherboard have issues at heavy currents, so they are removing the effect of the inductance of the wire.

You can help this by twisting each power wire with a ground wire; the inductance is proportional to the area of the loop between the supply and return wire.
So a smaller loop is better.

 

[Bill_Bright]

Bill, as I mentioned before, these capped cables only appear exclusively in high-end PSUs, and effectively do nothing to improve an already perfect ripple suppression.

And as I mentioned before,

I think the idea is to suppress any capacitance caused by the cables or introduced into the cables by surrounding high speed circuits - which can happen.

The point being, it is not just about suppressing ripple riding the output of the PSU, but also to suppress anomalies that may be introduced once the voltage leaves the PSU.

An 18awg cable 2 feet long is 8 milli ohms; 0.008 ohms.

Ummm, the internet is full of wire resistance calculators. As seen here, a 2 foot length of 18awg wire has .013 ohms of resistance. An even more accurate calculator here shows it is .0128 ohms.

But of course, without knowing all the variables (copper purity, stranded or solid, bends and curves in the wire, etc.) these are just estimates.

You can help this by twisting each power wire with a ground wire

To be most effective, there is an optimal twist count per inch too. I personally just avoid running cables parallel with other cables as much as possible.

 

[delshay]

I did some checking of capacitors not in-cable but on the CPU power socket itself. The result is surprising.

100uf ceramic 16v capacitor soldered to the back of the motherboard CPU socket reduce ripple & noise. Measured with a scope meter which I should get out more often. It seems putting a 22uf or higher ceramic capacitor here shows "big changes" on the scope meter. I obtained even better results with 220uf surface mount tantalum, lowering ripple even further. This whole week as from today is focused on PSU ripple & I will be using my scope meter from now on, but I have already found some fantastic results, but this is mostly changes around motherboard & devices that have 16v capacitors on the 3.3v & 5v rail.

EDIT: Tantalum capacitor 16v "Super Low ESR" from AVX soldered on back of CPU socket & PSU from Seasonic (1200w) Platinum, but Antec Truepower 550w PSU also shows the same improvements with the same capacitor's.

Tantalum Result: 45mv to 25mv ripple rock solid. Ceramic capacitors not that far behind 26-28mv not stable. Overclocked 85w CPU @idle.

EDIT2: It was this video that made me get out my scope metre which I rarely use.

 

[Bill_Bright]

Sorry, but I don't see the value for dredging up this thread that's been dormant for nearly 8 months for this. Sure, excessive ripple is a problem. But virtually every 1/2-way decent PSU suppresses ripple and noise well below the maximum allowed thresholds of 120 mV p-p for (12VDC) and 50mV p-p (for 5 and 3.3VDC) allowed by the ATX standard.

IMO If the motherboard needs further suppression, swap in a better PSU instead of modifying the board by adding caps to it.

 

[xtreemchaos]

i must of led a shelterd life because i cannot remember seeing any or just plain blind , are thay common ?, interesting none the less.

 

[delshay]

Sorry, but I don't see the value for dredging up this thread that's been dormant for nearly 8 months for this. Sure, excessive ripple is a problem. But virtually every 1/2-way decent PSU suppresses ripple and noise well below the maximum allowed thresholds of 120 mV p-p for (12VDC) and 50mV p-p (for 5 and 3.3VDC) allowed by the ATX standard.

IMO If the motherboard needs further suppression, swap in a better PSU instead of modifying the board by adding caps to it.

It's not the motherboard, it's the PSU. The reading was taken from two different PSU directly behind from the CPU socket on the soldered pins. I'm talking about the soldered contacts that hold the CPU socket in place. What are the chances of two different PSU showing almost identical results. Don't take my word for it, try it yourself on a motherboard you don't care about & post result. This is what I did, stick a 220uf surface mount tantalum on there & post the difference in ripple. The result is clear as daylight. It will show up on most settings you set the scope metre to. It's a massive difference on the scope.

 

[Bill_Bright]

It's not the motherboard, it's the PSU.

You said in your very first sentence, "CPU power socket". The power socket is on the motherboard.

Regardless, my points remain. I don't see the value in dredging up this dead thread. If such ripple and noise suppression is needed, get a better PSU.

 

[delshay]

You said in your very first sentence, "CPU power socket". The power socket is on the motherboard.

Regardless, my points remain. I don't see the value in dredging up this dead thread. If such ripple and noise suppression is needed, get a better PSU.

Will do. I have Seasonic 850w & Silverstone SFX 800w PSU both Titanium class. CPU power socket can only be in one place, but work was done on soldered contact back of motherboard. What I tried to do is simulate in-cable capacitor without cutting my cables up. That was the idea behind this experiment. Does it make a difference.

 

[xtreemchaos]

yoohoo im here stop bickering a mo im waiting for some enlightenment and at my age every minute counts....
look what arrived 30min ago, isnt she lovely

 

[EarthDog]

I don't believe the efficacy of such a thing is in question... but its relative worth in doing so. Ripple from high quality PSUs which are recommended are well below specification. Adding something to the back of the board or inline from PSU, while it can likely improve things.......the question is to what end? In a vast majority of cases this will not yield another 100 MHz of overclocking, or yield anything tangible for the average user and daily overclocker.

Where there is value, perhaps, is with EXTREME overclockers using Dry Ice/LN2/LHe.... not for the average enthusiast.

It is largely irrelevant, these inline caps and and improvements, is the take away. Not worth the marketing we see for us.

 

[Vayra86]

When reading power supplies recommendations threads i often find people dislike PSUs with bulky cables that has capacitors in them but then they are there to further help with the ripple suppression for the power supply so clearly they have a legit function yet i often fine people dislike them, sometimes going as far to avoid PSU's with in-cable caps You see many manufactures highlighting in-cable caps as a helpful features that slightly increase your PSU's performance in ripple suppression yet the wise people recommend i dont get one ?

Can someone please elaborate why that is ?

Quite simply put, if they can't get the ripple suppression down to a good level within the PSU itself, their design is shit and its a PSU to avoid.

The extra suppression from cables to me sounds a whole lot like audio cables forged by elves in moonlight. Probably vegan approved too. I'm staying FAR away.

And yeah, that last 0,1% seeking the best results. In other words, not a market at all if it was restricted to just them. The rest just tags along for the ride, feeling special for buying hot air

I've never seen that, sorry.

Seems clunky to me; I'd design them in, if it were me.

I just know power supplies.

Yeah, this. The logic eludes me completely. And I don't even know power supplies, I just add 1 + 1...

 

[Bill_Bright]

I don't believe the efficacy of such a thing is in question... but its relative worth in doing so.

Exactly!

I see no advantage in doing this. It will not improve performance. It will not increase efficiency. It will introduce additional points of failure, and it surely will void warranties. And that's assuming the correct capacitor value can be determined, the PSU does not balk, and the system boots.

And let's say the optimal capacitor value can be determined. Can that same cap be used universally with every PSU/motherboard combination?

While ED may be right about EXTREME overclockers using extraordinary cooling alternatives, it is highly unlikely those exceptions to the exceptions would be using anything but the highest quality power supplies in the first place.

I'm not even sure it will give anyone bragging rights - other than they can say, "I did it and it didn't break the computer".

Excellent ripple and AC noise suppression is critical and certainly a worthwhile goal. But at some point there is just no returned value for the effort to squeeze an extra drop out of turnip. Its like trying to get an additional 2°C of cooling for your CPU when you are already sitting at 20°C. It will not improve performance, stability or longevity.

What I tried to do is simulate in-cable capacitor without cutting my cables up. That was the idea behind this experiment. Does it make a difference.

IF a capacitor was needed in the first place, then "in theory", maybe because there would be slight differences in the resistance values at that point in the circuit compared to a cap on the other side of the cable connector. But in practice, that slight difference would likely be negligible - assuming quality connectors and a quality "mechanical" connection.

if they can't get the ripple suppression down to a good level within the PSU itself, their design is shit and its a PSU to avoid.

If that is why they put the caps there, then I agree completely. But my guess (and reviews seem to back that up) is that some PSU makers put caps there simply as "marketing fodder" and not because the PSU needs them to meet ATX Form Factor minimum standards.

Maybe it wasn't a mistake to dredge up this old thread as even those of us who frequently disagree on things are clearly on the same page here. So that hopefully will dispel the myths marketing weenies are trying to impose on us.

Conclusions:
1. "Quality" PSUs don't need in-line caps.
2. If a PSU needs in-line caps, see #1.

 

[Grog6]

Forgive me if I've said this before, but a switching power supply is a Charge Bucket.

It uses an inductor or transformer to move charge thru a diode, where it gets stored in a set of capacitors.

The charge that comes from the inductor has a wide range of frequency content, centered at the switching frequency, but the waveform is nasty; take a look.

The more efficiently the Capacitor Bank can accept the charge in the time allotted, and absorb and store the particular frequency content, the more efficient it is.

When the capacitor bank starts to discharge, here comes another bucket of charge.

The time between buckets gives rise to the Ripple, and adding caps will help. But at some point, you will go above the level the feedback relationship of the switcher controller can handle, and it will blow transistors.

You can change some resistors in that feedback chain, add bigger mosfets, better diodes, etc. and you will always make it better, as long as you don't run over any of the datasheet limits.

As many an entry level EE has learned, if you pull the output of a 12v switcher to 13V with another power supply, one or both will blow up, due to loss of feedback.

 

[EarthDog]

Just reading through the thread...I don't think the concept/result is a gimmick, but the tangible improvements and mentions in marketing do smell a lot more like a gimmick than are proven useful. People see a cleaner o'scope it and say 'that has to be better', do they not? And while on paper/technically speaking it is better (lowering ripple), but in the end, that yields nothing tangible for the end users. So, in essence, I agree with @silentbogo here that its benefit for the end user just isn't there. I'd have to imagine adding these inline costs a few more pennies for each one as well as the labor/machine time for stringing them inline which all costs us a few more dollars in the end. A pittance when talking $150+ PSUs, surely, but still. $144.99 is better than $149.99 for the exact. same. result (though the same could be said about other things RE a PSU too, but, i'll stick to the inline cap discussion).


EDIT: How are PSU's reviewed? BThe boxed cables are used, right? If they use the cables, then the caps are already there in the testing result. It is what it is and while there is improvement in that 'leg' of the power from the wall to the component, does it really improve ANYTHING when it is already part of the whole end product and result? What do informed users think when another unit performs better without them than one with...? I suppose for ripple, it's better, lol.

 

[xtreemchaos]

ok guys i guess you dont what to tell me, thats ok.

 

[ne6togadno]

ok guys i guess you dont what to tell me, thats ok.

both cable at the right end of the rows (pcie for gpus)
note that there is something under shrinking hose at the end where connectors to gpu are.
the cable in this area is very stiff and hard to bend. there is the capacitor.

not every psu has it but those that have it there is a portion of the cable that is very hard to bend (it is where capacitor is located).
this particular set is from corsair ax860i.

iirc your focus dont have capacitors in cables.

edit:
here is an example of cable w/o capacitors

 

[xtreemchaos]

thank you so very much mate, psu"s have never been my strong point, i said ive never come across this type before but maybe i have and not noticed.
thanks again

thanks for that, im i right that the PCi-e are sleved if thay have capacitors ? if so i must of used that type at sometime.

 

[Bill_Bright]

You can change some resistors in that feedback chain, add bigger mosfets, better diodes, etc. and you will always make it better, as long as you don't run over any of the datasheet limits.

Ummm, "always make it better"? I can't agree with this.

Starting with the latter first, datasheets are just published specs "claimed" by the manufacturer that may or may not be factual. The information on those datasheets may be incorrect due to honest mistakes. Or the companies may intentionally publish false information to deceive consumers or even go so far as the modify their products to "cheat" the system. Remember the Volkswagen cheating scandal?

So in reality, datasheets are irrelevant. What is relevant is whether or not a device meets or exceeds industry standards - as verified by rigorous, unbiased testing. And in this case, for standards I mean the ATX Form Factor standards, UL, FCC, EU/CE and other standards of the country or region the product is sold in.

As far as the ability to "always make it better", well that is just not true either for it assumes all products have poor or less than optimal designs and/or use inferior components in the first place. That is not always the case - at least not in the "practical" sense. "IN THEORY" engineers can make perfect designs. In practice, Man has yet perfected the production of 100% "pure" raw material and flawless, precision, always consistent manufacturing techniques.

But those EEs know this so what do they do? They add additional (and often highly sophisticated) "intelligent" filtering circuits that auto-adjust to ensure the output waveforms are exceptionally clean of unwanted ripple and other anomalies. One cannot simply swap out resisters, add mosfets or "better" diodes to "always" make it better. Reality just does not work that way.

 

[delshay]

"always make it better"

Of coarse you can make it better by selecting higher grade component's & do you own internal testing to insure it meets all standards. They will not always (if any) be picking the best top of the line component's.

If it's for your own personal use, then it does not need to meet any standard whatsoever if you are modifying it for yourself.

 

[Bill_Bright]

Of coarse you can make it better by selecting higher grade component's & do you own internal testing to insure it meets all standards.

Nope. It does not automatically work that way. That's why "always" is just a pipe-dream. Higher grade/top of the line components mean nothing if they are not the right electrical specification for the job. Nor do those higher grade components automatically imply better ripple suppression, for example. The engineering/design plays a significant roll - and not just of the circuit in question, but of the entire device. For example, shielding and input waveform filtering can affect the waveform and regulation of the output.

The better PSU makers like Seasonic and EVGA (or their OEM suppliers) are not hiring "entry level" EEs or "backyard technicians" to design and build their products.

If it's for your own personal use, then it does not need to meet any standard whatsoever if you are modifying it for yourself.

Ummm, also incorrect. Those standards are not just for safety. The ATX Form Factor standard also establishes standards for regulation and voltages tolerances, for example. If your modified PSU does not comply with those ATX regulation or tolerance standards, your modified PSU may no longer be "compatible" with ATX compliant motherboards, or graphics cards.

 

[delshay]

Nope. It does not automatically work that way. That's why "always" is just a pipe-dream. Higher grade/top of the line components mean nothing if they are not the right electrical specification for the job. Nor do those higher grade components automatically imply better ripple suppression, for example. The engineering/design plays a significant roll - and not just of the circuit in question, but of the entire device. For example, shielding and input waveform filtering can affect the waveform and regulation of the output.

The better PSU makers like Seasonic and EVGA (or their OEM suppliers) are not hiring "entry level" EEs or "backyard technicians" to design and build their products.

Ummm, also incorrect. Those standards are not just for safety. The ATX Form Factor standard also establishes standards for regulation and voltages tolerances, for example. If your modified PSU does not comply with those ATX regulation or tolerance standards, your modified PSU may no longer be "compatible" with ATX compliant motherboards, or graphics cards.

Are you saying that if I go into the Seasonic PSU, upgrade the MOSFETS part number to higher temperature MOSFETS from the same family part number invalidates everything you are posting here.

You need a good answer for this one.

 

[Bill_Bright]

LOL

I am saying you can cherry pick "what if" scenarios that may work. But that does not IN ANY WAY make your claim of "always" being anything more than a pipe dream. Are you suggesting you are smarter than the EEs and their CAD/CAE systems at Seasonic? As seen via the link in my sig, I've got multiple electronics degrees and certs and I would not pretend to be better ATX PSU engineers than them.

But to your one-off anecdotal example, if the current MOSFETS are running comfortably within their normal operating temperature range, then yeah, your claim is once again invalidated. And for sure, I would expect a very reputable PSU maker like Seasonic to use MOSFETS that are already fully capable of functioning properly when exposed to the expected (plus some) temperatures those supplies are marketed for. And for sure, the better testing facilities put those supplies in 40°C or even 50°C hot-boxes to test just for that - and guess what? These better supplies do indeed meet or exceed those specs already.

 

[delshay]

LOL

I am saying you can cherry pick "what if" scenarios that may work. But that does not IN ANY WAY make your claim of "always" being anything more than a pipe dream. Are you suggesting you are smarter than the EEs and their CAD/CAE systems at Seasonic? As seen via the link in my sig, I've got multiple electronics degrees and certs and I would not pretend to be better ATX PSU engineers than them.

But to your one-off anecdotal example, if the current MOSFETS are running comfortably within their normal operating temperature range, then yeah, your claim is once again invalidated. And for sure, I would expect a very reputable PSU maker like Seasonic to use MOSFETS that are already fully capable of functioning properly when exposed to the expected (plus some) temperatures those supplies are marketed for. And for sure, the better testing facilities put those supplies in 40°C or even 50°C hot-boxes to test just for that - and guess what? These better supplies do indeed meet or exceed those specs already.

I'm not talking about that. What I am saying is, if I change the MOSFETS to higher temperature ones from the same family MOSFETS assuming no electrical, timings have change, just the working temperature range, invalidate what you said. You said yes.

You don't want me to post what I think of that.

Just for the record I can change the MOSFETS if I choose to do so.

 

[jsfitz54]

Is there some reason the cable caps cannot be placed on the end closest to the power supply?

 

[Bill_Bright]

I'm not talking about that. What I am saying is, if I change the MOSFETS to higher temperature ones from the same family MOSFETS assuming no electrical, timings have change, just the working temperature range, invalidate what you said. You said yes.

No it doesn't! The initial claim was, you can change resistors, add bigger mosfets, better diodes, etc. and "always" make it better". It is simply inaccurate to claim it will "always" make it better. You presenting one possible example does not prove it will "always" happen.

And again, when the existing MOSFETS are already fully capable of dealing with the current temperature characteristics of the circuit and environment (as they would be with a quality PSU), you adding higher temperature MOSFETS does NOT make it better - regardless how many times you say it, or how strongly you think it does.

You don't want me to post what I think of that.

I don't care what you think of that. The Laws of Physics don't change based on what you think might happen.

Just for the record I can change the MOSFETS if I choose to do so.

Nobody said you couldn't.

Is there some reason the cable caps cannot be placed on the end closest to the power supply?

Resistance plays a role in how capacitance and inductance affects a circuit. And cable length affects resistance. So the location of the caps on a quality cable may not matter significantly, but it could affect the value of the caps needed. I note some PSU makers put the caps in the middle of the cables because they simply look better there. The caps don't get in the way of the connectors and can be hidden away more easily.