Typical mains power for mid-range PC?

The fact remains that your overly simplistic description of relevant ripple, is eclipsed by the actual ripple induced by factors you don't understand.

The steady state ripple at full rated output is not what crashes systems unless the PSU itself was operating beyond it's capable current output or simply didn't meet ATX specs at all.

You fail to see the larger picture here.

If we wanted to take your very narrow interpretation and claim the only "performance" variable we were meauring was full load ripple, then yes of course by definition that would be true.

Unfortunately a PSU doesn't really care, nor does a system, that you want to pose narrow definitions instead of all--encompassing ones of total power fluctuations as used for the intended purpose- in a computer, and by a reasonable extension- in a modern system running modern OS and multi--GHz parts which have LARGE changes in current consumption.

Nope.

You mean you can't think of any.

Here's one example- Cement load resistors to the filter caps so they run hotter, at lower ESR, but fail a lot sooner. I'm not going to point fingers but there's more than one PSU out there built like this.

I'm considering compliant supplies. Please tell us you're not trying to use non-compliant supplies.

Since we can assume you will claim you're not using non-compliant supplies, and that everyone else should be rejecting non-compliant supplies too, we can ignore them and move on to compliant supplies. Sheesh, talk about pedantic, do I have to spell out everything for you?

Believe it or not, educating yourself about the issue of dynamic loads would help.

Did you plan on getting on topic or is this just more confirmation that you're done?

Then you'd know the issues of dynamic loads.

So you feel the existence of one specified parameter makes any other parameters irrelevant? Apparently so since you suggest it in your argument.. Let's hope the person who did design these, didn't have the limited understanding of dynamic load induced ripple that you do. Ripple in an absolute sense is any and all deviations in voltage, particularly when they occur continuously from many parallel loads as in a computer.

They can be measured BTW, did you ever bother to do any testing? We can be sure you either didn't, or don't understand it. Let's break it down into overly simplified terms.

Hypothetical PSU outputting 20A RMS full load with very low,

100mA ripple with a constant load. What happens when we swap a dynamic load in place of the 20A fixed load? To keep the comparison fair, we'd still have 20A RMS, or would we use 20A peak? Next reduce that peak (or RMS), let's use 15A, and connect to the dynamic load that is a modern computer.

Either way, your ripple is significantly higher than 100mA. What does it mean? It is not the full, steady load ripple that mattered, it is easily eclipsed.

Seems a bit more likely you know someone or some company that makes these units if they exist at all.

LOL, that's mouthing off?

You need a reality check.

If you are checking the frame temp, yes that's reasonable. Not so reasonable with any other parts that do not attempt to provide an effective 'sink to the surface.

You apparently have no clue. Those involving logic will lock up, while others will burn junctions. Temp is important for most IC (actually ALL, but within a certain region...) and selectively ignoring some at high temp is a recipe for disaster.

Perhaps you don't "need" to know, "IF" you assume the design was sound and operating in the spec'd environment. That IS the general situation facing most computer users. We'd already gone beyond that though, when considering the basic subtopic of whether a SMART reported temp was applicable or there were other temps to consider.

Hardly. Anyone experienced with power ICs knows better. Some of those more sophisticated (or integrated) will have shutoff mechanisms but still no external access to the temp reading.

There is no heat sink. You are not on the same page here, we're talking about HDD PCBs.

Go ahead and extrapolate it then, surely you will be able to do this "easy" thing...

The industry has indeed taken surface temps. They have access to more data to put it in context such as the thermal conductivity of the package. You're making claims of things that do not allow you to reach the supposed goal of a higher precision of the important core temp, without more data... which goes back to what I'd already posted.

Known by the fab, yes. Also known by whoever has access to this data.

You conveniently failed to mention this (more important) consideration, and yet still allude to someone knowing it, but not that the data is availabe to the person taking the reading.

So ultimately we're back to square one, that the additional effort is not going to achieve what you claimed unless there was other data already availble, or more importantly, obtained.

Nope, just being practical.

Do you care to demonstrate to us that you aren't completely full of it yourself? Here's what you'll have to do, a most relevant and necessary test, as real-world as it gets:

We're talking about HDD PCB chips, so I will pick up this aforementioned Samsung drive and write out one of the chips off of it. You will have to provide the package thermals and calculate the core temp based upon what you presume to be the infinitely accurate surface temp (which has not been established either, as even the epoxy is an additional junction and conductance on it's own).

So here it is,

6AOG7W3 HA13645 Malaysia

Further observable; SMT, 10mm square, low profile

So instead of your vague claims, you'll have to actually find the missing data to do what you claim should be done, out of necessity or your entire argument collapses.

Again, you have not described a method of achieving the goal, only mentioned a vague idea that infrared is more accurate.

I have never argued that it wasn't. Read more carefully, my argument was that it is not in itself useful to have the most accurate surface temp reading possible when we don't care what the surface temp is, instead of the core temp.

At that point you then made vague claims but nothing useful- we need the actual data, not a bs notion.

So you're done now.

You still seem to be completely missing the point.

We are:

- Taking the temp of a working hard drive to determine proper conditions.

- Not destructively testing it.

- Not voiding warranty on it, this is not a manufacturer QA test, it's a subsystem functional test.

- Not looking to spend the time or money to achieve the most accurate surface temp measurement possible UNTIL AFTER we already have the other necessary data to USE this more accurate temp, IF it weren't for the issues written above.

kony wrote

Wrong. Its a reasonable assumption with a device that only consumes 10W or so in total, and doesnt bother to have a heatsink on any of the ics on the logic card, that as long as the surface temps are reasonable for power ics, particularly the main driver ic, that there isnt any problem with the cooling of the device when the manufacturer doesnt mandate forced air cooling.

You can keep chanting that pathetic little mantra till the cows come home, child. Changes absolutely nothing at all, ever.

Have fun explaining why the hard drive manufacturer didnt bother to include a heatsink for those ics or to mandate forced air cooling either.

Mindless pig ignorant silly stuff.

What its actually about is the total power consumed by the particular ic, child.

Is that right ? You know it all with the design of a hard drive and the manufacturer knows nothing eh ?

Yeah, right.

And is the case with the hard drives being discussed too, child.

Nope, not on that later discussion of whether the surface temps of the ics is all that matters. The ics on a hard drive logic card dont consume enough power to need internal thermal diodes so the core temp can be monitored, child.

Fraid so.

Anyone with experience with power ics knows that the starting point is the power dissipated in that particular ic and that that isnt high enough with hard drive ics to need any monitoring of internal thermal diodes, child.

Doesnt happen with the ics being discussed, child.

Have you ever used usenet before?

I have to wonder because you can't manage to reply to posts without breaking them up.

Do concerntrate on not accidentally hitting that "send" button until you're done.

Now, if you're using epoxy, you have once again demonstrated that you either weren't paying attention to the thread or have no grasp what the purpose is of a temp reading.

The purpose was to demonstrate that the chips were hotter than a SMART polled report, suggested. Then you went off on a simpleton version of "I Know a better way to take a temp", without a consideration if it was useful to do it, any gain or you'd just be wasting parts.

- Warranty voided

- Time and expense

- Data still not applicable unless you also have the package specs.

You made a simplistic suggestion but failed to see all the other issues. So a better thermal junction will be closer to the core temp... no kidding! The remaining issue is still the same, we'd have to know the package specs to make it worthwhile.

No you're not, because I'm not done laughing at how you can't even keep one single post together without fragmenting it.

If there were a gradual thermal rise like a properly operating chip warming up from power-on, yes.

If the chip was already fairly hot in it's running state and it was a "runaway", as you wrote, no.

I never claimed otherwise, this is obvious.

What you don't get, is what is _ACTUALLY_ being done.

We're taking a REAL hard drive, that REALLY needs to be working when you're done playing with it, and REALLY needs to be in the same original condition when the measurements are concluded.

This isn't random dissimilar thing, it's not an argument about overgeneralized laboratory chip temps, it is quite specifically the topic that led up to our argument.

You suggest a means that takes great time, expense, equipment, and then it was all a waste because you don't have the rest of the data, only your supposed accurate enough temp that must then be extrapolated still.

I think you're a one trick pony, you've told us everything you know and have never had to apply it to real world situations.

Oh?

Is it not a fact that when all is said and done, you won't know the actual core temp either but will have spent a great deal of time, possibly money, possibly destroyed the part itself or at least voided the warranty, and then then still have no clue about what we were doing in the first place?

Damn you are slow. Focus real hard and maybe, just maybe you will see that the testing methodology is that which is applicable to the situation, it is not wise to always suggest the most elaborate and accurate methods possible unless there is a real expectation of gain from doing so.

In case it still hasn't sunk in yet: Any simple Google search is likely to turn up detailed examples of using infrared and thermal interface materials to achieve a better junction with the probe. This is such a rudimentary aspect that it never needed mentioned, instead we can move on to the inherant issues of whether any method has a realizable gain or detriment, and the other issues involved in testing. You don't seem to have that much forethought though, are still suck on step one.

Do you always act childish when you run out of argument?

I c>On Sat, 29 Jul 2006 22:49:33 -0700, "Ed Light"

So it's fair to say you are clueless about HDD temperature measurement, that a random idea about taking a temp in a different scenario is non-applicable here and you lack the data to get the job done even using the method YOU SUGGEST. Hell, you can't even reply to a normal usenet post without severe fragmentation of the replies.

LOL.

Could you have failed any more completely?