Best 80mm fans for silence?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Best 80mm fans for silence?
Going to need 4 of them in a case. What do you reccomend that is not too expensive but is not that loud.
These are quiet and cheap:
http://www.svc.com/suf-s82-eb-19.html
This fan is also quiet, but more expensive:
http://www.jab-tech.com/80mm-Panaflo-fa ... -2344.html
http://www.svc.com/suf-s82-eb-19.html
This fan is also quiet, but more expensive:
http://www.jab-tech.com/80mm-Panaflo-fa ... -2344.html
I recommend the 80mm Yate Loons. You pasted a UK link so I assume you are in the UK. You can buy Yate Loon fans from Silencio_777 on Ebay, who is an SPCR member:Going to need 4 of them in a case. What do you reccomend that is not too expensive but is not that loud.
Silencio's Ebay Shop
Are the Panaflo L1A's still the top 80mm fan? Seems to me they have been superceded.
-
Aleksi
- Friend of SPCR
- Posts: 889
- Joined: Sun Dec 05, 2004 11:34 pm
- Location: Finland -- Folding For SPCR
Nici, the reason they aren't constant in quality is that people are buying surplus and reject (banged in transit etc) Panaflos from shitty vendors. Most of the Japan made Panaflo go in this category these days. I've posted about this many times on many threads on the fan forums. If you get them fresh they are very very good. Period.
The new NMB-MAT 80L1A (yes it's Panaflo) is nice, but pretty hard to get as most of NMB-MAT's production goes to fulfilling big OEM orders. You can ask from Dorothy Bradbury in the UK, but wouldn't hold my breath.
Yate Loon 80mm is nice sleeve bearing fan, would personally take it over a Nexus simply because of the price.
Papst also has some good 80mm models around the 2000RPM range.
The new NMB-MAT 80L1A (yes it's Panaflo) is nice, but pretty hard to get as most of NMB-MAT's production goes to fulfilling big OEM orders. You can ask from Dorothy Bradbury in the UK, but wouldn't hold my breath.
Yate Loon 80mm is nice sleeve bearing fan, would personally take it over a Nexus simply because of the price.
Papst also has some good 80mm models around the 2000RPM range.
-
Krazy Kommando
- Posts: 274
- Joined: Sun Aug 14, 2005 10:12 pm
- Location: Sydney, Australia
regarding those hard to find panaflo's
recently someone imported a batch to Croatia for the first time and i ofcourse ordered one pronto. it was pricy, really pricy, around 12$ for the FBA08A12LBX model.
i installed it and was promptly disapointed. it was the noisies fan in the room even after putting it on 5v. its whining and you can hear the mechanism.
after inspecting the specs i noticed the Made in China line. damn! its a 4D23A73 batch and it goddamn awful, its noisier than the stock fan in the codegen PS that came with the case. i tossed it out and put back the Arctic Cooling Pro that u had and everything is back to normal.
so Chinaflo's suck and its really a gamble to buy a Panaflo unless you use SPCR's proven vendors
recently someone imported a batch to Croatia for the first time and i ofcourse ordered one pronto. it was pricy, really pricy, around 12$ for the FBA08A12LBX model.
i installed it and was promptly disapointed. it was the noisies fan in the room even after putting it on 5v. its whining and you can hear the mechanism.
after inspecting the specs i noticed the Made in China line. damn! its a 4D23A73 batch and it goddamn awful, its noisier than the stock fan in the codegen PS that came with the case. i tossed it out and put back the Arctic Cooling Pro that u had and everything is back to normal.
so Chinaflo's suck and its really a gamble to buy a Panaflo unless you use SPCR's proven vendors
FBA08A12L1BX...
o Never a stock item released to distributors
---- uses different PCB & older motor
---- uses non soft-edged switching on the chipset
o Only released to OEMs to an OEM specification
---- different bearing grade, less oil, cheaper specs
Any available are surplus only...
o Panasonic/Minebea do not keep supplies of OEM specs
---- no channel to buy them or store them (J-I-T)
o OEMs who ordered excess will dump on surplus market
---- graded -- variety of grades (new, returned, G damaged)
o OEMs who got ship/handling damaged get a refund
---- refunded -- OEM refunded by Panasonic/Minebea re shipping/etc
---- and sells -- OEM dumps on surplus market who grades them
Problem is that refunds are given without stock returned/binned,
so OEMs get 2 incomes - on refund & again on surplus sale.
Surplus market doesn't like being stiffed, but IT resellers just see
part number & price differential and see easy money, users don't.
Lots of EBM-Papst & generics in this category currently.
New fans have a production date window that slides through time...
o 80mm L1A -- current is 5D-5I -- with the emphasis at 5G currently
o 80mm L1A -- unavailable, demand has eaten supply deep into 2006
4D (April 2004) 80mm L...
o Production date never issued by (old) Panasonic, only to OEM
o The 4D will be based on 4A 4B which used an outsourced bearing
---- 100% ok for reliability & performance (TQM data improved)
---- 67% rejected by me for acoustics -- they went back, literally
o Panasonic USA & Osaka advised about the friction issue
---- all outsourcing was to end, 4F+ was the revised product
---- of those the 4G were particularly nice, 4H pretty good etc
There are continual revisions to the 80mm fans after this time.
Country of origin is not a factor...
o Japan 4A 4B were in fact worse
---- Japan & China use the same bearing, it's made in the same plant
---- the outsourced part was to spec, but acoustics varied (no kidding)
o Karuizawa Japan is where all the Plant is made
---- the HQ R&D creates the plant, ships it out to the plants
---- TQM & control link to Arco Tower, Tokyo (HQ) & Karuizawa (R&D HQ)
o Minebea do not use outsourcing
---- the current migration is to use their unique BB process to make FDB
---- this will be also utilised in the cross-licensing with Seagate-Minebea
Majority of blue-chip & mid-sellers globally use...
o OEM spec -- usually cut-down spec
o Surplus spec -- new when it is in fact recycled parts or from other OEM runs
An example of OEM spec re 4710KL valueline VA1
o Valueline bearing
---- Valueline = Non-Shielded Non-Closed Open Dual Ball Bearing
-------- cheapest + shortest albeit very long life + fractionally highest noise
---- Standard = Non-Shielded Closed Dual Ball Bearing
-------- more expensive + longer life + fractionally lower noise level
---- Reference = Shielded Closed Dual Ball Bearing
-------- most expensive + longest life + fractionally lower noise level
---- Minebea are the world leader in ultra precision miniature bearings
---- now approved for aerospace, military, not just medical & automotive (E)
o Valueline oil/grease level
---- level varies 50-95%, in this instance 70%, Reference is 95/97/98%
o Valueline housing, wires, stator etc
---- lighter weight, less glass fibre, thinner wires, lighter stator
---- product weight = shipping cost and 120x38mm Panaflo way FAR too much
So OEM spec can vary according to what the OEM wants.
That is a good thing, you can get a higher grade tier-1 brand fan in place
of generics - it may not be the top one, but cheaper so supplied originally.
--
Dorothy Bradbury.
o Never a stock item released to distributors
---- uses different PCB & older motor
---- uses non soft-edged switching on the chipset
o Only released to OEMs to an OEM specification
---- different bearing grade, less oil, cheaper specs
Any available are surplus only...
o Panasonic/Minebea do not keep supplies of OEM specs
---- no channel to buy them or store them (J-I-T)
o OEMs who ordered excess will dump on surplus market
---- graded -- variety of grades (new, returned, G damaged)
o OEMs who got ship/handling damaged get a refund
---- refunded -- OEM refunded by Panasonic/Minebea re shipping/etc
---- and sells -- OEM dumps on surplus market who grades them
Problem is that refunds are given without stock returned/binned,
so OEMs get 2 incomes - on refund & again on surplus sale.
Surplus market doesn't like being stiffed, but IT resellers just see
part number & price differential and see easy money, users don't.
Lots of EBM-Papst & generics in this category currently.
New fans have a production date window that slides through time...
o 80mm L1A -- current is 5D-5I -- with the emphasis at 5G currently
o 80mm L1A -- unavailable, demand has eaten supply deep into 2006
4D (April 2004) 80mm L...
o Production date never issued by (old) Panasonic, only to OEM
o The 4D will be based on 4A 4B which used an outsourced bearing
---- 100% ok for reliability & performance (TQM data improved)
---- 67% rejected by me for acoustics -- they went back, literally
o Panasonic USA & Osaka advised about the friction issue
---- all outsourcing was to end, 4F+ was the revised product
---- of those the 4G were particularly nice, 4H pretty good etc
There are continual revisions to the 80mm fans after this time.
Country of origin is not a factor...
o Japan 4A 4B were in fact worse
---- Japan & China use the same bearing, it's made in the same plant
---- the outsourced part was to spec, but acoustics varied (no kidding)
o Karuizawa Japan is where all the Plant is made
---- the HQ R&D creates the plant, ships it out to the plants
---- TQM & control link to Arco Tower, Tokyo (HQ) & Karuizawa (R&D HQ)
o Minebea do not use outsourcing
---- the current migration is to use their unique BB process to make FDB
---- this will be also utilised in the cross-licensing with Seagate-Minebea
Majority of blue-chip & mid-sellers globally use...
o OEM spec -- usually cut-down spec
o Surplus spec -- new when it is in fact recycled parts or from other OEM runs
An example of OEM spec re 4710KL valueline VA1
o Valueline bearing
---- Valueline = Non-Shielded Non-Closed Open Dual Ball Bearing
-------- cheapest + shortest albeit very long life + fractionally highest noise
---- Standard = Non-Shielded Closed Dual Ball Bearing
-------- more expensive + longer life + fractionally lower noise level
---- Reference = Shielded Closed Dual Ball Bearing
-------- most expensive + longest life + fractionally lower noise level
---- Minebea are the world leader in ultra precision miniature bearings
---- now approved for aerospace, military, not just medical & automotive (E)
o Valueline oil/grease level
---- level varies 50-95%, in this instance 70%, Reference is 95/97/98%
o Valueline housing, wires, stator etc
---- lighter weight, less glass fibre, thinner wires, lighter stator
---- product weight = shipping cost and 120x38mm Panaflo way FAR too much
So OEM spec can vary according to what the OEM wants.
That is a good thing, you can get a higher grade tier-1 brand fan in place
of generics - it may not be the top one, but cheaper so supplied originally.
--
Dorothy Bradbury.
That should say...
Majority of blue-chip & mid-sellers globally use...
o OEM spec
---- usually cut-down spec as my example shows
---- this can involve irritating proprietarisation
o Surplus spec
---- new when it is in fact recycled parts
------- saleable return laptop, PC + saleable components of return laptop, PC
---- new when it is from other OEM runs
------- seller has high-price permanent-spec PCs & low-price variable-spec PCs
------- variable spec comes from acting as a supply/demand clearing house for parts
For example, someone has 50,000 graphics card or fan X available...
o Tier-1 buyers get the pick of the crop
---- they will get the as factory spec parts discounted as overorder etc
o Surplus sellers get the rest of the crop
---- could be the same, more usually it is not
---- can be components or box-damaged on Ebay
Classic example of the latter is...
o Companies now will only refund if it is returned
---- to stop people a) getting $ from refund and b) another $ from surplus sale
---- which in turn stiffs the company re manufacturing scheduling to brand etc
o Some companies can't do this
---- re some industrial products (although try it with MSFT COA or Panasonic Whisper
HVAC or Toughbook, they will find you and write you nasty letters, then break legs)
o Some companies employ "pallet it up & sell it cheap"
---- people take 10 broken & re-arrange to create 7 working
---- sold on Ebay etc, can be fine, can be not fine, priced accordingly (hopefully)
For some products this works fine
o Unitary replacement of parts until working
o PCs in that respect are simply like LEGO, just MSFT ties you up in glue, roll on OS/X
For some components it can get silly
o Relabelling into hot applications where they fail
o Tyan found out to its cost earlier this year with server fans
---- there were 3 specs on a China Component Distributor Website
---- English Western spec was drivel, Traditional Mandarin was correct
---- one spec broke the Fan Laws derived from CFD equations & laws of physics by 47%
A note re country of origin...
o Some components involve zero labour
---- raw materials in
---- total automation of all component parts - raw in, finished out
---- total automation of all assembly - finished product out
---- labour is skeleton plant maintenance staff
o So producing to gain low labour cost is immaterial
---- there is no labour cost on a Minebea bearing nor on the fans using them
---- so there is no benefit in producing in low labour cost areas
o Producing in Asia is done for such items because of integration locale
---- you site the plant next to those plants building the products using them
---- simple as that
Hence German (re VW) & Japan (Panasonic, Mitsubishi, Toyota) still build all the
plant equipment in their own German & Japan R&D HQ - then ship it out, commission, run.
All data per car VIN or product code is straight back to the HQ re realtime control.
This changes of course for formerly China State Owned makers...
o China company -- choose non-precision bearings -- sleeve
---- bearing cost is 12p
-------- hence fan price independent of size, bearing cost dominates
---- automation is minimal
---- labour cost IS a major part of their production
o Japan/German company -- choose super-precision bearings
---- bearing is 400-920p (Minebea, Japan), 420-810p (EBM, EU)
-------- hence fan price independent of size, bearing cost dominates
---- automation is absolute, via a very unique process
---- labour cost is simply not a factor
Japan/German use a different approach:
o Vast economy of scale from capital plant & expensive production process
o Huge production plant cost eliminated by economy of scale
---- you become the world leader in ultra-precision miniature bearing
---- you use a fan as loss-leader to drive economy-of-scale of bearings
---- with 62% of the world market, 3,720,000,000 bearings per year it works fine
o No-one can undercut you on bearing re #1
---- plus you can refuse to supply (Minebea-Seagate did not supply WD Raptor
Typical Japanese approach, you sell at cost initially and let economy of scale
with the intention of total domination provide profitability & barrier to entry.
--
DB.
Majority of blue-chip & mid-sellers globally use...
o OEM spec
---- usually cut-down spec as my example shows
---- this can involve irritating proprietarisation
o Surplus spec
---- new when it is in fact recycled parts
------- saleable return laptop, PC + saleable components of return laptop, PC
---- new when it is from other OEM runs
------- seller has high-price permanent-spec PCs & low-price variable-spec PCs
------- variable spec comes from acting as a supply/demand clearing house for parts
For example, someone has 50,000 graphics card or fan X available...
o Tier-1 buyers get the pick of the crop
---- they will get the as factory spec parts discounted as overorder etc
o Surplus sellers get the rest of the crop
---- could be the same, more usually it is not
---- can be components or box-damaged on Ebay
Classic example of the latter is...
o Companies now will only refund if it is returned
---- to stop people a) getting $ from refund and b) another $ from surplus sale
---- which in turn stiffs the company re manufacturing scheduling to brand etc
o Some companies can't do this
---- re some industrial products (although try it with MSFT COA or Panasonic Whisper
HVAC or Toughbook, they will find you and write you nasty letters, then break legs)
o Some companies employ "pallet it up & sell it cheap"
---- people take 10 broken & re-arrange to create 7 working
---- sold on Ebay etc, can be fine, can be not fine, priced accordingly (hopefully)
For some products this works fine
o Unitary replacement of parts until working
o PCs in that respect are simply like LEGO, just MSFT ties you up in glue, roll on OS/X
For some components it can get silly
o Relabelling into hot applications where they fail
o Tyan found out to its cost earlier this year with server fans
---- there were 3 specs on a China Component Distributor Website
---- English Western spec was drivel, Traditional Mandarin was correct
---- one spec broke the Fan Laws derived from CFD equations & laws of physics by 47%
A note re country of origin...
o Some components involve zero labour
---- raw materials in
---- total automation of all component parts - raw in, finished out
---- total automation of all assembly - finished product out
---- labour is skeleton plant maintenance staff
o So producing to gain low labour cost is immaterial
---- there is no labour cost on a Minebea bearing nor on the fans using them
---- so there is no benefit in producing in low labour cost areas
o Producing in Asia is done for such items because of integration locale
---- you site the plant next to those plants building the products using them
---- simple as that
Hence German (re VW) & Japan (Panasonic, Mitsubishi, Toyota) still build all the
plant equipment in their own German & Japan R&D HQ - then ship it out, commission, run.
All data per car VIN or product code is straight back to the HQ re realtime control.
This changes of course for formerly China State Owned makers...
o China company -- choose non-precision bearings -- sleeve
---- bearing cost is 12p
-------- hence fan price independent of size, bearing cost dominates
---- automation is minimal
---- labour cost IS a major part of their production
o Japan/German company -- choose super-precision bearings
---- bearing is 400-920p (Minebea, Japan), 420-810p (EBM, EU)
-------- hence fan price independent of size, bearing cost dominates
---- automation is absolute, via a very unique process
---- labour cost is simply not a factor
Japan/German use a different approach:
o Vast economy of scale from capital plant & expensive production process
o Huge production plant cost eliminated by economy of scale
---- you become the world leader in ultra-precision miniature bearing
---- you use a fan as loss-leader to drive economy-of-scale of bearings
---- with 62% of the world market, 3,720,000,000 bearings per year it works fine
o No-one can undercut you on bearing re #1
---- plus you can refuse to supply (Minebea-Seagate did not supply WD Raptor
Typical Japanese approach, you sell at cost initially and let economy of scale
with the intention of total domination provide profitability & barrier to entry.
--
DB.
OT: I wasted money on yet another poor Panaflo/NMB-MAT 92mm sample. Still way too loud compared to other quiet fans, maybe someone can shed some light on why these fans are so noisy? The model:
NMB-MAT 5E16A73-1BX, MODEL FBA90A12L, DC12V 0.15A, MADE IN CHINA
Considering all the good praise I hear regarding Panaflo bearing-quality, etc., I keep buying these fans hoping to find a good sample and/or supplier. I got this fan just a week ago from Jab-tech.
NMB-MAT 5E16A73-1BX, MODEL FBA90A12L, DC12V 0.15A, MADE IN CHINA
Considering all the good praise I hear regarding Panaflo bearing-quality, etc., I keep buying these fans hoping to find a good sample and/or supplier. I got this fan just a week ago from Jab-tech.
The L1BX's (which is what this fan is) has never had a very good reputation on SPCR:I wasted money on yet another poor Panaflo/NMB-MAT 92mm sample. Still way too loud compared to other quiet fans, maybe someone can shed some light on why these fans are so noisy? The model:
NMB-MAT 5E16A73-1BX, MODEL FBA90A12L, DC12V 0.15A, MADE IN CHINA
Thread 1
Thread 2I just purchased an L1ABX to replace the stock fan on my zalman 9500. With the fanmate 2, the panaflo operates around 800rpm at the lowest setting...turning it up half way is where it gets bad. There's horrible bearing noise. I purchased the panaflo to remove this specific source of noise. If I operate the panaflo at the same rpm as the zalman was at it's lowest, it's noticeably, and annoyingly louder.I bought from jab-tech, they said it was a panaflo, and had a picture of a panaflo, and listed the specs of a panaflo, but sent me an NMB-MAT. There are two sets of numbers:
SE16A73 -1BX
FBA09A12L
Thread 3If you need something with more airflow than a nexus, consider an evercool instead. I've had TWO of the L1BX from Panaflow, they were buzzy at all levels.
Thread 4 (Dorothy)I got three of the M3 80mm Chinaflo M1BX's from Digi-Key a few weeks ago. Their code is "4J".
After running them for 24+ hrs to break them in, here's what I found:
One is excellent, very quiet at 5V and nearly no clicking.
One is pretty decent, with a slight hum and slight clicking at 5V.
One is pretty bad, with very noticeable clicking even at 5V.
So, FWIW, my three M3's seem to behave just like run-of-the-mill Chinaflos: some are good, some aren't.
Thread 5About signals:
o TACHO SIGNAL - only exists on 1BX option code fans
---- L1BX is not available (I can get it, but you'll wait 26-38wks)
---- L1BX has a different PCB & Motor-IC than 1A & may be noisier
---- yes I've been sampling & testing, 1A still wins out
I personally don't see any reason to persevere with Panaflo's. The quality is so variable, and the competition so much better (both in terms of noise and quality control) they don't even figure in my decision-making process when purchasing new fans. Specifically, Nexus fans are better and more expensive, Yate Loon fans are better and less expensive. What have the Panaflo's got to offer that you can't get from a more consistent, better value rival?i must say i was thouroughly disapointed with the panaflo 92mm (L1BX), it makes a very anoying ticking sound that is clearly audible at all voltages, and it's nothing like the very quiet 80mm L1A. btw, it was not a bad speciment, i tried a replacement and it was exactely the same.
...i feel the same way
i bought a 92mm panaflo last week to replace my zalman 92mm on my flower and its noisy at all voltages with the clicking
the 80mm is fine
Jaganath: The Panaflos don't seem to hold a candle noise-wise, but they supposedly are very long life. I know Nexus and Yate Loon fans (among others) are quieter, but I'd like a second fan choice for long-life/high-temp applications. So far, it seems Akasa's fans may be the only non-sleeve bearing models that are reasonably quiet.
FBA09A12L1BX...
o Manufacturers data -- 42.7cfm, 27dB(A)
---- at 12V operation, 22oC, 65% RH
o Testing to
---- ANSI S12.11-1987 method on all DC fan motors
---- Testing also to JIS C 9603:1988, DIN 45635 Appendix A of Part 38
---- Testing also to BS848 1966 Part 2 & ISO R266-1975
---- ISO 7779 is used to evaluate designs to engineer in tonal acceptability
Every fan has a particular characteristic...
o Venturi design, aerofoil design, specific Pressure-v-Airflow, motor torque, phase
---- fans actually vary from best to worst by only 10-19%
-------- unsurprising as the laws of physics have not been suspended
---- fans are tailored to suit specific markets
-------- Panaflo 40mm U sounds noisy but produces twice the pressure of any other
---- fan makers thus target specific P-Q characteristic segments
-------- laws of physics apply in that there is a tradeoff
o Bearing characteristics are much more predictable
---- sleeve offer 20,000-26,000hrs (Dynaeon to Acoustifan)
---- ultraprecision ball from Minebea & Papst far more
---- Tier-1 L10 uses -15% rpm as failure, others inflate L10 by -30% rpm
Ok, so fans vary - you have to tweak each particular usage...
o Vary distance of a fan from object being cooled
---- set a heatsink fan to different distances from a heatsink
---- over a range of distances temperature is near constant, noise varies wildly
---- the P-Q curve of the fan interacts with the P-Q curve of the heatsink
---- this is easy to modify - add washers accordingly to adjust separation
o Vary the distance of a fan intake from an object
---- ideally a fan of 25mm should have 25mm of clear space before the intake
---- if not the case, air intake is not laminar, boundary layer separation & noise result
---- this is difficult to modify - fitting 15/20/25mm fans can find an optimum
For those that have not done it, vary the separation of your fan from the
CPU heatsink. You may find you can improve on the default setting - but pay
particular attention to the intake side of the fan relative to any obstruction.
If the L1BX acoustics do not suit, consider the L1A...
o L1BX uses a different PCB, chipset & switching to L1A fans
---- Tacho is not an appropriate sensor for FDB
---- indeed Dell themselves published a paper & patent on this issue
o L1A is acoustically preferred by some in light of this
---- Locked Rotor is the appropriate sensor for FDB
---- projective failure methods give meaningless data for FDB
Remember a DC if supplied by pure DC would not rotate...
o Rotation occurs because voltage is switched around the coils
---- so inducing rotation from interaction with an annular permanent magnet
o Rotation can be effected by complex switching
---- microprocessor manages switching timing, fall off & rise in voltage
---- microprocessor manages the phase frequency of the rotation
---- +ve -- achieves more torque (better real world P-Q curve) + lower noise
---- -ve -- costs more + minimum input voltage + can dislike PWM solutions
o Rotation can be effected by simple switching
---- transistor & capacitor simply induce rotation
---- +ve -- super cheap (ideal for generic) + low noise + PWM ok + lowest voltage
---- -ve -- doesn't manage bearing wear + doesn't optimise P-Q curve re performance
Simply, 92L1BX uses different phase frequency & other attributes to the L1A.
o Has a different acoustic signature whilst other characteristics are the same
---- some may prefer it, some may prefer others - like EBM-Papst growl
o Despite that the 1BX in 92mm has led to a full range of L1BX H1BX U1BX
---- it does well in bulk airflow applications, direct performance improvement
---- it often allows the replacement of 3 fans by 2 fans, direct cost saving
Despite that the 1BX in 92mm has led to the release of L1BX H1BX & U1BX to both
the PC, medical & industrial markets - it does well in bulk airflow applications.
Additionally it often allows the replacement with 3 fans with 2 fans re cost.
If the 12V operation does not suit, consider 6-12V use...
o Fans may be operated over a range of voltages
---- 12V fan -- typical range is 6-12V
---- 12V fan -- with Zener or Resistor may have a range of 7-12V
---- 24V fan -- relabelled to 12V has a range of say 10.2-12V
o Running a fan at 50% is predicted via the Fan Laws
---- fan laws are derived from CFD formuli & laws of physics
---- Noise_n = Noise_k + 50 * LOG (speed_n / speed_k)
---- Airflow_n = Airflow_k * (speed-n / speed-k)
o FBA09A12L = 2100rpm
---- At 9V the fan would be turning at 1575rpm
---- from applying the Fan Laws...
------ airflow @ 1575rpm = 42.7 * (1575/2100) = 32cfm (free air)
------ noise @ 1575rpm = approximately 20.8dB(A) (free air)
o Fan laws have two limitations
---- data input must be accurate to published standards
---- figures lose efficacy below ~20dB(A) as other sounds dominate
Fan laws require accurate data...
o Fans of same size should follow a largely straight line
---- some do not - they are arithmetically/marketing derived
---- some do not - utilise sample variation in measuring
-------- take +10% rpm sample to measure RPM
-------- take -10% rpm sample to measure dB(A)
o Fan data must include P-Q fan characteristic curve
---- HPQ DELL IBM etc publish Flotherm CFD model & validation papers
-------- create 2D FNM model of a product - refine re hotspots/shadowing
-------- create 3D CFD model of a product & input fan P-Q curve
---- that is where P-Q curve data must hit tarmac
-------- virtual product model is run, refined & goes for prototyping
-------- prototype is empirically tested to confirm model & fan P-Q data
---- P-Q data is verified in a published paper & used by Flotherm
-------- I can do this as I use it to design coloco rack cabinets
-------- so I nor Dell have any problem with the performance data
o Best & worst fans only vary by 6-19%
---- unsurprising - Laws Of Physics have not yet been suspended
-------- marketing often likes to believe that they have
---- fan P-Q characteristics present an envelope bound by laws of physics
-------- slotted housings, blade-count, dual-fans, motors, chipsets work within it
---- fan makers thus target specific product usage
-------- tier-1 actually often design the products to use their fans
---- eg, Panaflo 40mm U is noisy - but offers double the pressure of any other
-------- so you can see how you can tradeoff one characteristic for another
If you take the entire popular of fans and plot them...
o You can see they fall on a band which in %age terms is quite narrow
---- relabelled fans known as "IT fans" in the industry often are amusing outliers
---- bring up underlying fan characteristics from the maker/integrator shows true data
o Examining intra-population you see some clear segmentation
---- high resistance product = axial fans offering high pressure, ok flow, high noise
---- low resistance product = axial fan offering low pressure, high flow, low noise
Finding the underlying data is sometimes difficult...
o DF1209SL-3 is listed on 7 China Component Distributor sites
---- DF1209SL-3 - 92x25mm -- 25.2cfm, 27dB(A), 2200rpm, Sleeve
---- DF1209SM-1 - 92x25mm -- 32.4cfm, 33dB(A), 2600rpm, Sleeve
---- bearing life -- 20,000hrs @ 25oC
o DF1209SL-3 listed on Centenary Electronics has two sets of data
---- one -- 92x25mm -- 27cfm, 19.2dB(A), 1500rpm, Sleeve
-------- 1500rpm fan with operating voltage 10.2-13.8V
-------- indicating a voltage reduced fan (12V & Zener/Resistor or 24V)
---- two -- data as provided by the component distributor sites
-------- 92x25mm -- 25.2cfm, 27dB(A), 2200rpm, Sleeve, 20,000hrs @ 25oC
o A quick check via the Fan Laws
---- DF1209SL-3 v Alternative data
------ Noise at 1500rpm = 27dB(A) + 50*LOG (1500/2200) = 18.7dB(A)
------ Noise claimed at 1500rpm is 19.2dB(A), so somewhat validated
------ Airflow is not confirmed by the same laws
---- DF1209SL-3 v DF1209SM-1 data
------ again contradictory information is obtained
o Conversely checking their other fans...
---- DF1208SL 80x25mm -- 21.5cfm, 23dB(A), 2200rpm
---- more consistency is observed
So care has to be taken - as well as loading the Mandarin character set
Indeed going through all the relabelled underlying fan makers data
I find the same thing, from Dynaeon to Yate Loon - they end up with
contradictory data. Some sleeve bearing can be quieter than high
quality fans as they use thick oil, double-sprung to mask the motor
pole axial thrust and so on. However basic laws of physics still apply.
The same applies to all the relabelled fans...
o A fan is an optimisation within the laws of physics
---- you can create very high pressure, good flow & low noise
---- you can create good pressure, very high flow & high noise
---- fan engineering will merely tweak things a little better - *at a cost*
o A fan in terms of longevity varies more widely
---- tier-1 offer lowest cost over their lifetime - *at a cost*
---- generic offer very low cost for consumer applications
Transistor switching of cheap generics can offer slightly lower noise,
but particularly lower starting voltage over microprocessor solutions.
Transistor switching does not allow optimisation of the P-Q curve by
optimising motor torque, soft-edging switching & phase control. One
reason why Panaflo show so much noise in the fan return line is all
the harmonics from the complex switching performed for performance.
Taking the FBA09A12L1A at 2100rpm 27dB(A) at 1500rpm it would give
30.5cfm and 20dB(A), from a recent Dell Flotherm study of the H1BX
those figures match perfectly. However if you have a *high* airflow
resistance application the Papst 3412 (and it's replacement due soon)
will offer better airflow despite the figures appearing inferior.
One is designed for quite low-resistance, bulk-airflow (case), the
other is more suited to quite high-resistance, pressure (heatsink).
As a rule the tier-1 & all fan makers don't directly "position a product
directly over a competitor" - and tier-1 in particular actually design
the likely products first, then design the fan range to suit that range.
Same with jet engines - the makers design the plane before the maker,
you thus create a range of likely market needs and products to suit.
You can create a fan with very good free-air figures, yet which does
somewhat poorly under load. You can also design a fan with very poor
free-air figures (side-vented, augmented flow), but which when matched
to a 2mm H2O resistance heatsink will outperform a conventional fan.
Problem is, if your heatsink doesn't match you don't get the win
So as a rule the tailored fan offerings are tailored to OEMs, so as
to avoid an unsuitable match - as well as providing broad use fans.
Sometimes it screws up, I think laptop designs need to begin again.
As for consistency, NMB-MAT RB have the highest consistency of all,
far above Panaflo FB & EBM-Papst. That is a key selling point, Papst
sell conversely on higher temperature capability (although Panaflo will
make a fan covering -40oC to +117oC for whatever odd reason).
Panaflo FB consistency will vary according to source, as will Papst.
A vast number of UK, EU, USA, Canada sellers are moving surplus
stock - easy for me to check, difficult for others. The reason is simple...
o Sell a new EBM-Papst or NMB-MAT
---- you will make 9% on a sale or 30% on conventional Papst pricing
o Sell a surplus fan
---- you will make 300% on a sale
---- so you need to sell 30x as many NMB-MAT to match - uneconomic
o Sell a relabelled fan
---- you will make 300-600% on a sale
---- you see why the "IT fans" migrate in this direction
o Sell OEM fans
---- provide tier-1 performance at lower pricing
---- Zalman PSUs adopt this approach with bottom of range 4710KL
---- going back to when Zalman rebadged Papst vs generics, amusingly
Conversely you can see why EMC etc move in the other.
o As yet Dell, IBM or HPQ high-end servers aren't fitting generics
o Despite the remarkable cost & benefits touted by affiliate marketing
o A quick verification of data filed with UL shows why
o A quick verfication with 3D CFD model & empirical validation shows why
Selling surplus & up-selling a relabelled is not a bad strategy for them.
Many IT retailers in the UK only survive through fan selling and the
affiliate reviews, advertising and such like. Margin is 5x a m/board.
The solution is to CAD Coloco to Servers, proving a revenue stream
there as well as providing a channel into which to sell tier-1 products.
However, people do have their preferences and where acoustics are
concerned that is not directly predictable. Comes down to whether
one is building a server with operational cost of downtime or whether
one is building a HTPC or w/cooling system, various fans, various uses.
--
DB.
o Manufacturers data -- 42.7cfm, 27dB(A)
---- at 12V operation, 22oC, 65% RH
o Testing to
---- ANSI S12.11-1987 method on all DC fan motors
---- Testing also to JIS C 9603:1988, DIN 45635 Appendix A of Part 38
---- Testing also to BS848 1966 Part 2 & ISO R266-1975
---- ISO 7779 is used to evaluate designs to engineer in tonal acceptability
Every fan has a particular characteristic...
o Venturi design, aerofoil design, specific Pressure-v-Airflow, motor torque, phase
---- fans actually vary from best to worst by only 10-19%
-------- unsurprising as the laws of physics have not been suspended
---- fans are tailored to suit specific markets
-------- Panaflo 40mm U sounds noisy but produces twice the pressure of any other
---- fan makers thus target specific P-Q characteristic segments
-------- laws of physics apply in that there is a tradeoff
o Bearing characteristics are much more predictable
---- sleeve offer 20,000-26,000hrs (Dynaeon to Acoustifan)
---- ultraprecision ball from Minebea & Papst far more
---- Tier-1 L10 uses -15% rpm as failure, others inflate L10 by -30% rpm
Ok, so fans vary - you have to tweak each particular usage...
o Vary distance of a fan from object being cooled
---- set a heatsink fan to different distances from a heatsink
---- over a range of distances temperature is near constant, noise varies wildly
---- the P-Q curve of the fan interacts with the P-Q curve of the heatsink
---- this is easy to modify - add washers accordingly to adjust separation
o Vary the distance of a fan intake from an object
---- ideally a fan of 25mm should have 25mm of clear space before the intake
---- if not the case, air intake is not laminar, boundary layer separation & noise result
---- this is difficult to modify - fitting 15/20/25mm fans can find an optimum
For those that have not done it, vary the separation of your fan from the
CPU heatsink. You may find you can improve on the default setting - but pay
particular attention to the intake side of the fan relative to any obstruction.
If the L1BX acoustics do not suit, consider the L1A...
o L1BX uses a different PCB, chipset & switching to L1A fans
---- Tacho is not an appropriate sensor for FDB
---- indeed Dell themselves published a paper & patent on this issue
o L1A is acoustically preferred by some in light of this
---- Locked Rotor is the appropriate sensor for FDB
---- projective failure methods give meaningless data for FDB
Remember a DC if supplied by pure DC would not rotate...
o Rotation occurs because voltage is switched around the coils
---- so inducing rotation from interaction with an annular permanent magnet
o Rotation can be effected by complex switching
---- microprocessor manages switching timing, fall off & rise in voltage
---- microprocessor manages the phase frequency of the rotation
---- +ve -- achieves more torque (better real world P-Q curve) + lower noise
---- -ve -- costs more + minimum input voltage + can dislike PWM solutions
o Rotation can be effected by simple switching
---- transistor & capacitor simply induce rotation
---- +ve -- super cheap (ideal for generic) + low noise + PWM ok + lowest voltage
---- -ve -- doesn't manage bearing wear + doesn't optimise P-Q curve re performance
Simply, 92L1BX uses different phase frequency & other attributes to the L1A.
o Has a different acoustic signature whilst other characteristics are the same
---- some may prefer it, some may prefer others - like EBM-Papst growl
o Despite that the 1BX in 92mm has led to a full range of L1BX H1BX U1BX
---- it does well in bulk airflow applications, direct performance improvement
---- it often allows the replacement of 3 fans by 2 fans, direct cost saving
Despite that the 1BX in 92mm has led to the release of L1BX H1BX & U1BX to both
the PC, medical & industrial markets - it does well in bulk airflow applications.
Additionally it often allows the replacement with 3 fans with 2 fans re cost.
If the 12V operation does not suit, consider 6-12V use...
o Fans may be operated over a range of voltages
---- 12V fan -- typical range is 6-12V
---- 12V fan -- with Zener or Resistor may have a range of 7-12V
---- 24V fan -- relabelled to 12V has a range of say 10.2-12V
o Running a fan at 50% is predicted via the Fan Laws
---- fan laws are derived from CFD formuli & laws of physics
---- Noise_n = Noise_k + 50 * LOG (speed_n / speed_k)
---- Airflow_n = Airflow_k * (speed-n / speed-k)
o FBA09A12L = 2100rpm
---- At 9V the fan would be turning at 1575rpm
---- from applying the Fan Laws...
------ airflow @ 1575rpm = 42.7 * (1575/2100) = 32cfm (free air)
------ noise @ 1575rpm = approximately 20.8dB(A) (free air)
o Fan laws have two limitations
---- data input must be accurate to published standards
---- figures lose efficacy below ~20dB(A) as other sounds dominate
Fan laws require accurate data...
o Fans of same size should follow a largely straight line
---- some do not - they are arithmetically/marketing derived
---- some do not - utilise sample variation in measuring
-------- take +10% rpm sample to measure RPM
-------- take -10% rpm sample to measure dB(A)
o Fan data must include P-Q fan characteristic curve
---- HPQ DELL IBM etc publish Flotherm CFD model & validation papers
-------- create 2D FNM model of a product - refine re hotspots/shadowing
-------- create 3D CFD model of a product & input fan P-Q curve
---- that is where P-Q curve data must hit tarmac
-------- virtual product model is run, refined & goes for prototyping
-------- prototype is empirically tested to confirm model & fan P-Q data
---- P-Q data is verified in a published paper & used by Flotherm
-------- I can do this as I use it to design coloco rack cabinets
-------- so I nor Dell have any problem with the performance data
o Best & worst fans only vary by 6-19%
---- unsurprising - Laws Of Physics have not yet been suspended
-------- marketing often likes to believe that they have
---- fan P-Q characteristics present an envelope bound by laws of physics
-------- slotted housings, blade-count, dual-fans, motors, chipsets work within it
---- fan makers thus target specific product usage
-------- tier-1 actually often design the products to use their fans
---- eg, Panaflo 40mm U is noisy - but offers double the pressure of any other
-------- so you can see how you can tradeoff one characteristic for another
If you take the entire popular of fans and plot them...
o You can see they fall on a band which in %age terms is quite narrow
---- relabelled fans known as "IT fans" in the industry often are amusing outliers
---- bring up underlying fan characteristics from the maker/integrator shows true data
o Examining intra-population you see some clear segmentation
---- high resistance product = axial fans offering high pressure, ok flow, high noise
---- low resistance product = axial fan offering low pressure, high flow, low noise
Finding the underlying data is sometimes difficult...
o DF1209SL-3 is listed on 7 China Component Distributor sites
---- DF1209SL-3 - 92x25mm -- 25.2cfm, 27dB(A), 2200rpm, Sleeve
---- DF1209SM-1 - 92x25mm -- 32.4cfm, 33dB(A), 2600rpm, Sleeve
---- bearing life -- 20,000hrs @ 25oC
o DF1209SL-3 listed on Centenary Electronics has two sets of data
---- one -- 92x25mm -- 27cfm, 19.2dB(A), 1500rpm, Sleeve
-------- 1500rpm fan with operating voltage 10.2-13.8V
-------- indicating a voltage reduced fan (12V & Zener/Resistor or 24V)
---- two -- data as provided by the component distributor sites
-------- 92x25mm -- 25.2cfm, 27dB(A), 2200rpm, Sleeve, 20,000hrs @ 25oC
o A quick check via the Fan Laws
---- DF1209SL-3 v Alternative data
------ Noise at 1500rpm = 27dB(A) + 50*LOG (1500/2200) = 18.7dB(A)
------ Noise claimed at 1500rpm is 19.2dB(A), so somewhat validated
------ Airflow is not confirmed by the same laws
---- DF1209SL-3 v DF1209SM-1 data
------ again contradictory information is obtained
o Conversely checking their other fans...
---- DF1208SL 80x25mm -- 21.5cfm, 23dB(A), 2200rpm
---- more consistency is observed
So care has to be taken - as well as loading the Mandarin character set
Indeed going through all the relabelled underlying fan makers data
I find the same thing, from Dynaeon to Yate Loon - they end up with
contradictory data. Some sleeve bearing can be quieter than high
quality fans as they use thick oil, double-sprung to mask the motor
pole axial thrust and so on. However basic laws of physics still apply.
The same applies to all the relabelled fans...
o A fan is an optimisation within the laws of physics
---- you can create very high pressure, good flow & low noise
---- you can create good pressure, very high flow & high noise
---- fan engineering will merely tweak things a little better - *at a cost*
o A fan in terms of longevity varies more widely
---- tier-1 offer lowest cost over their lifetime - *at a cost*
---- generic offer very low cost for consumer applications
Transistor switching of cheap generics can offer slightly lower noise,
but particularly lower starting voltage over microprocessor solutions.
Transistor switching does not allow optimisation of the P-Q curve by
optimising motor torque, soft-edging switching & phase control. One
reason why Panaflo show so much noise in the fan return line is all
the harmonics from the complex switching performed for performance.
Taking the FBA09A12L1A at 2100rpm 27dB(A) at 1500rpm it would give
30.5cfm and 20dB(A), from a recent Dell Flotherm study of the H1BX
those figures match perfectly. However if you have a *high* airflow
resistance application the Papst 3412 (and it's replacement due soon)
will offer better airflow despite the figures appearing inferior.
One is designed for quite low-resistance, bulk-airflow (case), the
other is more suited to quite high-resistance, pressure (heatsink).
As a rule the tier-1 & all fan makers don't directly "position a product
directly over a competitor" - and tier-1 in particular actually design
the likely products first, then design the fan range to suit that range.
Same with jet engines - the makers design the plane before the maker,
you thus create a range of likely market needs and products to suit.
You can create a fan with very good free-air figures, yet which does
somewhat poorly under load. You can also design a fan with very poor
free-air figures (side-vented, augmented flow), but which when matched
to a 2mm H2O resistance heatsink will outperform a conventional fan.
Problem is, if your heatsink doesn't match you don't get the win
So as a rule the tailored fan offerings are tailored to OEMs, so as
to avoid an unsuitable match - as well as providing broad use fans.
Sometimes it screws up, I think laptop designs need to begin again.
As for consistency, NMB-MAT RB have the highest consistency of all,
far above Panaflo FB & EBM-Papst. That is a key selling point, Papst
sell conversely on higher temperature capability (although Panaflo will
make a fan covering -40oC to +117oC for whatever odd reason).
Panaflo FB consistency will vary according to source, as will Papst.
A vast number of UK, EU, USA, Canada sellers are moving surplus
stock - easy for me to check, difficult for others. The reason is simple...
o Sell a new EBM-Papst or NMB-MAT
---- you will make 9% on a sale or 30% on conventional Papst pricing
o Sell a surplus fan
---- you will make 300% on a sale
---- so you need to sell 30x as many NMB-MAT to match - uneconomic
o Sell a relabelled fan
---- you will make 300-600% on a sale
---- you see why the "IT fans" migrate in this direction
o Sell OEM fans
---- provide tier-1 performance at lower pricing
---- Zalman PSUs adopt this approach with bottom of range 4710KL
---- going back to when Zalman rebadged Papst vs generics, amusingly
Conversely you can see why EMC etc move in the other.
o As yet Dell, IBM or HPQ high-end servers aren't fitting generics
o Despite the remarkable cost & benefits touted by affiliate marketing
o A quick verification of data filed with UL shows why
o A quick verfication with 3D CFD model & empirical validation shows why
Selling surplus & up-selling a relabelled is not a bad strategy for them.
Many IT retailers in the UK only survive through fan selling and the
affiliate reviews, advertising and such like. Margin is 5x a m/board.
The solution is to CAD Coloco to Servers, proving a revenue stream
there as well as providing a channel into which to sell tier-1 products.
However, people do have their preferences and where acoustics are
concerned that is not directly predictable. Comes down to whether
one is building a server with operational cost of downtime or whether
one is building a HTPC or w/cooling system, various fans, various uses.
--
DB.
DB: Wow, thanks for the thorough reply. So the L1A should be quieter? That would seem reasonable. I've tried a few 80mm L1As and they had an entirely different motor noise than this 92mm 1BX. The 80mm L1A's motor sounded muted, whereas this 1BX sounds very raspy. It's definitely not a situation of airflow impendance--when I evaluate fans for noise, I test them in a free-flow environment sitting on a piece of soft foam.
-
Felger Carbon
- Posts: 2049
- Joined: Thu Dec 15, 2005 11:06 am
- Location: Klamath Falls, OR
Have you tried the Panaflo 92mm 'M' model? 2450RPM at 12V, but with no BX and hence soft edge-switching like the 80L. If that can be adjusted down to the RPM you want, it would be interesting what you thought of the noise level.GHz wrote:DB: Wow, thanks for the thorough reply. So the L1A should be quieter? That would seem reasonable. I've tried a few 80mm L1As and they had an entirely different motor noise than this 92mm 1BX. The 80mm L1A's motor sounded muted, whereas this 1BX sounds very raspy.