Flow Meters?

[jaydee]

I am looking for a good flow meter to base results upon.

So far I found these:

http://www.flowmeters.com/catalog_li...ion_set_id=110

http://www.sparlinginstruments.com/eggdelta.htm

One of these would be ideal BUT do I need such a nice one. Would a paddle wheel type still give a accurate reading? I know they slow flow rate but with the pump I am planing on using it shouldn't matter to much. And bing I plan on using multiple flow rates anyway.. So will a cheaper paddle wheel one be usable on a test bed?

[Deke_Kun]

I've got the innovatek flowmeter and I love it. Has gorgeous blue LEDs built into it, lovely effect, the only quam I have with it is that whilst it reports to a motherboard header, no-one seems to actually know WHAT it is reporting. Oh well...its still nice.

~Deke

[jaydee]

Originally posted by Deke_Kun
I've got the innovatek flowmeter and I love it. Has gorgeous blue LEDs built into it, lovely effect, the only quam I have with it is that whilst it reports to a motherboard header, no-one seems to actually know WHAT it is reporting. Oh well...its still nice.

~Deke

I definatly need function over looks as these results will be critical in calculation. Definatly will need to know what it is reporting.

[morphling1]

Why don't you look at something like this.
http://cgi.ebay.com/ws/eBayISAPI.dll...tem=3105026217
I bought something similar, but I'm still waiting to recieve it, it's fairly accurate, and simple to use also this kind of flow meter isn't to big of a flow killer.
Look at McmasterCarr for those. All those non mechanical flow meters are very expensive.

[jaydee]

Originally posted by morphling1
Why don't you look at something like this.
http://cgi.ebay.com/ws/eBayISAPI.dll...tem=3105026217
I bought something similar, but I'm still waiting to recieve it, it's fairly accurate, and simple to use also this kind of flow meter isn't to big of a flow killer.
Look at McmasterCarr for those. All those non mechanical flow meters are very expensive.

That one looks to only go up to 100GPH? The pump I plan on using is rated 1,100GPH+. Will be one of the one's on this site. http://www.aquatictech.com/pumps.html

After getting the pump installed it should be able to put more than 100GPH through the system. I have been serching and serching for flow meters and I found quite a few of them and they all seem pretty close to 2.5% to 3% accuracy. I am not opposed to paying $300, but for that price it better be better than a $40 one. Also not opossed to a paddle wheel one as the pump should easily handle the extra flow restriction. If I have to I will get a spendy one but if I don't then I will not.

[pHaestus]

If you want one of the bell-type flow meters then I have one hardly used.

http://www.procooling.com/~phaestus/flowmeter.jpg

Huge pressure drop and ~6% accuracy. If you are testing below 1.5GPM only or with an enormous pump then I guess they are "ok". You need to consider (IMO) exactly what you want to test and then purchase accordingly. BillA seems to be about an order of magnitude up from me in both accuracy and cost when it comes to testing gear. It works out ok as we have (somewhat) different questions of interest.

I have a GPI paddle wheel type (0.01GPM res and 3% accuracy or so), and to me it is the upper end of acceptable price ($40-80 on ebay). The important thing with all such used equipment is to pay the extra $10-15 for current calibration. The Sparling looks nice, but I seem to recall Bill having some issues with his and it ended up being cheaper to replace than to repair. Maybe he can comment.

If you check www.labx.com and www.ebay.com you can find flowmeters that just have a 4-20mA output. Couple those with an indicator and you may have another relatively cheap solution.

[jaydee]

Originally posted by pHaestus


You need to consider (IMO) exactly what you want to test and then purchase accordingly. BillA seems to be about an order of magnitude up from me in both accuracy and cost when it comes to testing gear. It works out ok as we have (somewhat) different questions of interest.

This is my problem I have yet to solve. I am not to sure what I really need to give decent results. My test bed will be similar in nature to BillA's but I am not really planning on being as spacific on some of the things he does nor quite as accurate.

How much will +- 3% really matter in the flow results? I plan on measuring from 50gph on up to whatever the pump will handle in the system. At 50gph that is only 1.5GPH +- off. Will that really make a substantial problem in results?

Now that you posted here I also have a question about the Fluke temp gauges. What would be a decent one to use and what type of probes would work for it. I found a bunch on ebay but I am not sure how to get the probes to fit where I want them to. :shrug:

This more accurate testing is something I really never thought all that usefull but I have to eat crow a bit and suck it up. I go off bitching about bad reviews but I intend on doing something about this time. My goal is to be able to do a much better review (not necassarily perfect by any means but better than DIGI DOC and onboard probes) than what has been out there lately. And also this might motivate me to purse my water block designing again as I will be able to test better.

[BillA]

goddamn jd, you get hit by lightning ?

as pHaestus said, and I think you too now understand, the equip totally depends on what you wish to accomplish
- its not really so much 'what will work', for many things can be made to work, as how much accuracy is needed

so let me flip all the questions around and say, from my direct experience, what certain specific capabilities will enable one to do, and also what certain deficiencies will preclude one from doing

lets start with temperature: 2 principal applications:
- heat die/CPU/wb bp/hsf bp temps
- coolant temps

for heat die temps and such, 0.1°C resolution will work
BUT the instrument accuracy will probably be on the order of ±0.2 to 0.4°; this WILL play hell with one's calculations

-> do not think for an instant to 'ignore' accuracy (aka uncertainty), with sloppy thinking like "an instrument's error is always the 'same'", do read on this subject before buying anything - it is the same issue over and over

coolant temps are much more difficult because the increment being measured (that of interest) is much smaller, 0.01°C is necessary (and in fact insufficient if back-calculating power from a temp difference)
[I am now going from 0.01 to 0.001°C resolution for just this reason]

next: what sensors are appropriate for what resolution ?

for 0.1°C resolution with a TC only type T can be used (but yet again their accuracy is also ±0.2° or so; plays more hell with the calcs)
the Fluke 2190A is a good, and sometimes cheap, thermometer; but a multipoint selector will be needed also
(there are older models on eBay that can work, but the 2190A is better)

for 0.01°C resolution thermisters can be used, but they are rather special and rather expensive
-> and the indicators for 'matched' thermister reading ?
Very specialized, and pricey

a more common approach is with RTDs, but can only be 4-wire for this accuracy and require a good thermometer/ohmmeter
- note that the 24hr accuracy of a Fluke 2180A can be as low as ±0.02°C, everything calibrated of course

to go to 0.001°C can be as 'simple' as using a 6½ digit ohmmeter (the thermometers doing the conversion are VERY expensive)

for all of the above calibration is required
I have found it cheaper to buy the equipment and have the capability 'in house' so that it can be used as often as needed
- what's needed depends on the gear: calibrated mercury thermometer with TC, lab std resistors, lag bath, etc (its a chore)

call Omega and request their "Temperature" catalogue - tons of info (they have ss sheathed 'T' TCs in 0.040" dia)
they have a "Flow" catalogue also

flow meters are a problem which I think pHaestus has addressed
-> do remember: You can't make a silk purse form a sow's ear

the best low cost type is an industrial paddle wheel WITH a built in indicator
-> if the indicator is separate, how will you be able to calibrate it ?
answer: only by sending the whole thing off (big $$ !)

work with a pencil and see what the effect of 3% is (and is that 3%, or ±3% ?)

I can tell you this: repeatability is a B*TCH with 3% variation
you will not get the same results, close perhaps, but not the same
and the 'better' everything else gets, the worse the flow meter looks
this will drive you nuts !
[I have used 5 different types, form a Brooks rotameter to the mags - now have 4 - I use now.)

jd, glad to hear you wish to be a serious tester, feel free to contact me
(sometimes I have extra equip I've upgraded)

it will not be difficult to best what the 'review' sites do, for they have no equipment (and apparently less understanding)
- but I have totally rebuilt my test bench (The Ghetto Bench) 5 times and do assure you:
It will be much cheaper to get the 'right' equipment in the beginning.

good luck

[koslov]

Bill (or maybe pH can answer this), do you have any experience with the Digitec 5810 meters? I bought one from eBay, and I am having trouble locating the right thermistors and connectors. I bought two "micro-bead" thermistors, but I am not sure if they will work. Where can I find the probe connectors? Also, once I do this, where can I get it calibrated without spending more than I paid for the thing (<$20)? I don't need more than +/-1°C accuracy.

[pHaestus]

I have 3 working digitec 5810s (2 that Bill is intimately familiar with) and a non-working spare. I use YSI 400 and 700 series thermistors with them. Instuctions on calibration are in the manual; need decade resitors. Ask at the local engineering dept maybe?

[BillA]

nice thermometers, but a lot more than what you need
and may be costly to get 'running'

you will spend a LOT more than $20 for the dual lineraized thermister probes
and the manual cost me $35, google for it

calibration is pretty easy if you have (access to) the decade resistors

[koslov]

Here are the exact thermistors I have: http://cgi.ebay.com/ws/eBayISAPI.dll...tem=3105237277

Will these 'cut the mustard'?

I am willing to pay more than $20 for thermistors, I was just hoping calibration wouldn't be more than $20.

The cheapest YSI thermistor I can find is $50, please tell me they can be gotten cheaper. They had better come with connectors for that much.

I guess my budget would have to be under $100, for probes and calibration. If I can only get one probe, then I guess it'll do, but I would like to have 2.

edit: I just found some probes for $9.00! Please tell me these would work for CPU temp!

[jaydee]

Thanks for replying Bill A! Hit by lightning? You could say that. I have a bad learning curve (as I am sure you know by now). I can read and read and read the same material over and over but it will never click untill some time down the road when a spacific keyword, picture, or event happens then it all clicks together.

In this case I got bitched at because of a comment I made about a review ( http://www.amdmb.com/article-display.php?ArticleID=216 ) of a pump that showed nothing except it's ability to pump water into a container (in an attemp to test the flow).

And then the other posted review of the IceBurg Kit ( http://forums.procooling.com/vbb/sho...&threadid=5440 ) showing numbers that are not possible by the best of systems let alone a budget kit. I complained about it to the managment of the site but nothing has been done about it which is discouraging as it is one of my favorite sites. Hate to think people will buy this kit thinking it will outperform the best systems out there because of the numbers given. 9C over ambient?

So being not to many people seem to want to step up and start doing it better and just bitch, I felt it necassary to do this. Not to mention I would like to be able to test my own blocks better.

Once I get a better idea of what I am doing I will post a new thread of proposed parts. i have a couple heater cartriges coming to make the die simulator. Once they get here I will start a better parts list and.

[BillA]

the die simulator is a chore all by itself
I'm on rev.8 to my original
(which is really not so bad considering the wholesale replacement over and over of everything else)

my present list is here

koslov
I'm certain you need 700 series for the 0.01 resolution,
no idea about the 400s
(I sent the manual with the meters)

nor do I know what the effect of using 400s might be on calibration
- the 700s have 2 thermisters, one to 'correct' the other, and use a 3-wire cable and plug
(all included from YSI @ $120 or so, ea)

[koslov]

Bill or pH, could you recommend a specific probe on one of these pages?
http://www.coleparmer.com/catalog/pr...par=6207&cat=1
http://www.thomassci.com/productblock.do?id=4537

There are so many, I have no idea which would be better for this application.:shrug:

[BillA]

to my knowledge, takes a 700 series
THIS page
http://www.coleparmer.com/catalog/pr...=531&sku=&sel=

pHaestus perhaps can speak to the 400s

[pHaestus]

Just skimmed manual; indeed the 400 series is not mentioned. It isnt as thick as I had thought; maybe one lunch break this week I can scan it.

[koslov]

Thank you pH, I'd be happy to reimburse you for your time. Also, can you recommend a specific probe at one of those links? There are so many, I wouldn't know which one to choose. I would like to get one that fits under the WB to mount on the CPU package, hopefully one of those fits the bill.

[BillA]

noooooooooooooo
read those dimensions, never happen

that is an excellent dual probe (3 actually) thermometer for reading coolant temps;
what I did, and now pHaestus is doing with it

never gonna work for CPU temps due to the large probe size

sorry

[koslov]

I didn't even notice that none of those probes are under 1mm (die height). The smallest is 5/64" or 2mm!

Perhaps I could drill out a small channel for the probe to sit in in the base of the WB... not an option for a testing platform, but since I won't be changing WBs frequently, would you recommend this? I want accurate temps mostly so I can see the results of tweaking the other parts of my system, including frequency and voltage as well as the cooling system.

Oh, but wouldn't that be a shame if AMD changed Athlon die dimensions again, then I would be screwed!

Then again, I could always approximate by mounting the probe on the WB base as close to the die as possible. Would this yield feasible results?

Ugh, I'm discouraged now. Is there any other way I can get 0.1°C accuracy for under $100?

jaydee: Sorry for interjecting my little problem into your thread, but it seemed like a good time to bring it up, as long as we are discussing testing equip.

[jaydee]

Originally posted by koslov

jaydee: Sorry for interjecting my little problem into your thread, but it seemed like a good time to bring it up, as long as we are discussing testing equip.

No problem. Interested in this outcome myself.

[BillA]

no, you are as close as you can get for $100

you can get 0.1°C resolution with the Fluke and TCs described above
-->> but that IS NOT 0.1°C accuracy, a quite different beast indeed

to get 0.1°C accuracy it is held that you need a minimum of a 4-fold greater measurement capability
-> 10-fold is the norm (hence 0.01°C resolution)

I just now received an e-mail from a friend of mine, a (past ?) Sr. Editor of The Journal of Heat Transfer and who does CFD on a 30 tetraflop computer (guess where ?), part of which is copied below:

this is wrt a "Roundup" review of the Atlantis to go up on OCers tomorrow -

"Observation: start putting uncertainty bounds on all data plots. You have it on some, propagate it to all. As a modeler I have to know what is the range of error within the results, i.e. how near to the results do I have to be.

You start putting them there (you would never get published in a serious journal without them) then you can ask others for bounds on their results."

do you see the 'problem' ?
without the definition of uncertainty (what we more casually call accuracy), the numbers are - or may be ?? - junk

one can blithely say "oh, for what we're doing it doesn't matter", and if you believe that then stick with a digidoc and don't sweat reality

zeros are the most expensive digits

google on 'measurement uncertainty'
get 'EAL-22' and 'EAL-R2-S1'