Need a new quieter fan to HD IO..

[martthie_08]

Suad, thanks!
pulled one of the HD IO out of the rack today to take some pictures.

Here is a link to the full resolution images: AVID_HD_IO.zip

[Suad Cokljat]

Originally Posted by martthie_08

Suad, thanks!
pulled one of the HD IO out of the rack today to take some pictures.

In comparison to 192, HD I/O has PSU divided into 2 PCBAs and mainboard relocated to front panel metal plate. Fan connector and regulator are in the upper right corner of mainboard (picture attached). I can barely see even on maximum zoom, but I think fan voltage regulator is marked U6 (same as on 192) and this little blue component that is barely visible on picture I suppose is thermistor (picture attached). If your HD I/O is still out of rack, it would be good if you take some detailed pics of that part.

Here is my suggestion how to proceed:

1. Detect what are sources of heat inside device

- make HD I/O connected to HDX card and working with Pro Tools running. Put top lid on but remove screws for easy lid removal.

- after 1 hour of operation, remove top lid and touch all relevant components and mark your observation how hot are the components (rate them as cool - warm - hot - very hot). Observe converters, FPGAs, voltage regulators (especially ones with heatsink) and PSU MOSFETs and heatsinks. I suggest to turn device OFF when touching anything on PSU.

2. Make some performance measurements of the stock fan (do this at the same time while doing step 1)

- make fan voltage graph by measuring fan voltage in 1 hour period every 5 minutes (start when device is completely cool). Attached graph for 192 with stock fan and thermistor

- during 1 hour temperature testing, with lid on, observe noise and airflow from the fan hole. There is no strict procedure for that. Use your senses. I test airflow with thin piece of paper...

3. Install new fan and observe performance with stock thermistor

- install new silent fan (I use Noctua A8-FLX). Avoid fans with less than 2000RPM. Don't use PWM fans. Modify fan connector to fit HD I/O header and leave RPM wire unconnected.

- observe fan performance during 1 hour operation as in step 1 (start with device cooled down). Does fan spin during first 5 minutes of operation?

- observe temperature of hot components. Does it getting even hotter?

4. Replace thermistor if needed - if your fan does not spin fast enough and HD I/O internals get much hotter than with stock fan

- remove stock thermistor and measure it's value on room temperature (192 has 10-12K)

- try thermistors of lower value - 6.8K , 4.7K...

- repeat tests and find sweet spot with best noise/cooling ratio. Keep in mind temperatures will be higher in crowded rack with limited airflow.


For 192 with Noctua A8-FLX I use 4.7K thermistor.

Cheers!

S

[martthie_08]

thanks Suad,
unfortunately my HD I/O is not easily accessible, it took me 2 hours to get those pictures yesterday :) so it might take a while before I get back to that..

In the past I've replaced HD I/O fans with Noctua A8 (think it was FLX) and they do provide decent air circulation and are a bit more quiet than the original fan.

Though I didn't do any measurements and do not have the equipment for that, so this is just a gut feeling and there is still a risk if you are using the Noctua in a not well ventilated rack or in rooms with higher temperatures.

[David Clementson]

Saud, pretty much everything you have said is spot-on. Nice work.

If memory serves, the fan voltage can be calculated as follows:

V(fan) = 1.25 + (25000 / Rt) [volts]

where Rt is the resistance of the thermistor R19. The nominal 25C thermistor resistance is 10k. There is also a clamp that will limit the fan voltage to a little less than 12V regardless of temperature.

You can calculate a new nominal value of the thermistor if you need a different fan voltage, but it will also change the slope of voltage vs. temperature curve. You also need to consider fan stall, and the fact that the stall voltage will generally increase as the fan ages. The start-up circuit will get the fan spinning after powerup, but there is no stall monitoring after that.

BTW, the reason a PWM fan controller wasn't used was because when we tried it, we found that PWM current artifacts would appear in the audio noise floor and in the clock jitter residue. The analog fan controller eliminated that, although the pulsating current due to the fan's internal commutation was still a cause of concern for noise. The large LC filter (the big inductor and two big caps) minimizes the conducted noise path, however there is also a magnetic coupling mechanism that needed to be considered. We positioned the fan to minimize the magnetic coupling. This is also why we split the power supply in two pieces.

As you point out correctly, it was not a simple puzzle to solve. But at least the HD I/O didn't have all that Meitner Class-A discrete stuff the original 192 had, so it was comparatively easier to cool.

Also, the fan cost was no object: we would have gladly paid whatever for a fan that was quieter while still providing sufficient cooling for the unit to meet spec at the maximum rated ambient temperature. We auditioned many, many fans and did lots of fan listening tests before picking the one we did. But fan technology has improved, and I am sure there are better candidates today. And of course fans do age, so replacement is a valid maintenance procedure in my opinion.

DC

[tlindaas]

Originally Posted by David Clementson

And of course fans do age, so replacement is a valid maintenance procedure in my opinion.

DC

David, can you recommend a quiet replacement fan for the HD I/O?

Will it be enough with a 55 cmh (32 cfm) quiet fan (e.g. the Noctua NF-A8 PWM) without PWM controller so working at full speed the whole time? Instead of the loud original fan (CoFan) that ramps up somewhat as the temperature rises.

How important is the ambient temperature? Will it help if the room temperature is kept below say 25℃/77℉?

See attached data for the Noctua fans.

[SHIRK]

Pulled all of my HDi/o's (analog16x16) and 192 (analog 8x16) from the rack last month to replace a few of the i/o level trim pots that had stopped working. While I had them out I did a careful and thorough compressed air cleaning of all cards, fans and connections. After 9 years since I was last in there, there was a lot of dust. I mean a lot! The SilenX fans (EFX-08-15 Effizio 80x25mm 15dBA 32CFM PC Computer Case Fans) that I installed back in 2013 in all the units are still working great. I've had my 192 i/o since 2003 I think. It's now only used for Multi-track headphone sends to the Private Q as well as an AES i/o for my Latte, but I find it very satisfying to have gotten 19 years of mileage out of that box. Hope to get some more before I do my next upgrade.

[the.engineer]

I too have fallen into this trap recently. Here’s an up close shot of the sub-board with U6 on it.

[the.engineer]

For anyone still interested - the 12k Resistor mod works very well (I would still recommend changing the thermistor) but for now here is a close up of U6 in an HD I/O with a 12k resistor strapped across it.

Please note that a previous image posted by Suad Cokljat on 7-12-2021 is incorrect. The text is correct but the graphic overlay is incorrect.

My Noctua NF-A8's are now spinning happily (and still quietly) around 9-10V.

Best wishes,
James

[redmount]

Hello,
I am about to get a fourth I/O 16x16analog and was wondering if interfaces #SN could be of any use to determine the year these interfaces were manufactured and if some are more "desirable" than others when it comes to quality of parts like fans ?
(except for havng the opportunity to fire the unit and listen to how noisy it is )
Thanks

[martthie_08]

Originally Posted by the.engineer

Please note that a previous image posted by Suad Cokljat on 7-12-2021 is incorrect. The text is correct but the graphic overlay is incorrect.

James, are you really certain about this? To me it seems awkward to add a resistor across the outer pins of U6, that would put a resistance across the Vin and Adjust pins...

afaik the output voltage of the LM317 is adjusted by changing the resistance between the Vout and Adjust pins.
meaning that I believe Suad's image from 2021 is in fact correct.

Disclaimer: I am not an electrical engineer!