This page lists a couple of suggestions for improvements on the Newisys NA-1400 or it’s successor.

FANtilation problems

While examining the system I noticed some cooling problems which result in noise coming from the fan:

Inefficient harddisks cooling

A large open space between the front of the drive cages and the harddisks itself and a relatively large opening towards the systemboard compartment of the housing allows the fresh air to take the path of least resistance: Through the front of the drive cages, down into the systemboard compartment, over the systemboard into the double walled housing and out into the fan compartment. This leaves a relatively small amount of air traveling the longer, narrower route over the harddisks towards the SATA backplane where it enters the double walled housing. This results in little ventilation of the harddisks which requires the fan to use higher rpm’s....

Fan intake clearance

When the fan is at high rpm, trying to cool the harddisks down, the distance between the underside of the systemboard and the fan is so narrow (~1 cm) that the air is turbulently sucked in by the fan. In fact the distance is so narrow that the fan actually creates a ‘vacuum’ between the systemboard and the fan itself (To prove this: hold the intake side of a fan running at full speed close to a flat surface and you will feel the fan pulling itself towards this surface).

The vacuum and turbulent air flow produce lots of noise and greatly reduce the fan’s efficiency.

Fan intake environment

The fact that the underside of the systemboard is littered with small components is also a contributing factor to the noise (To prove this: take the systemboard and hold it horizontally in front of your mouth and then blow air over it, you will hear the turbulent air flow around the small resistors and capacitors).

Fan outtake clearance

The clearing between the bottom of the housing and the surface on which it stands is, again, only about 1 cm which also adds turbulence to the airflow and results in additional noise and a reduction of fan efficiency.

PWM fan control

To control the fan’s speed, Newisys uses a microcontroller (PIC16F819). This microcontroller has built-in PWM circuitry to switch the fan on and off at regular intervals called cycles. The amount of time that the fan is powered is called the duty time and the longer the duty time the faster the fan will turn. There is some nice background information here.

Without proper testing gear I guess the duty cycle of the fan control at low speeds is about 50% with a frequency of about 1Hz because I can hear the fan spinning up and down symetrically every second. These relative long periods of on/off time produces the rattling noises as described here. Of course I could be wrong in my conclusions, but the observations are consistent with the theory.

Proposed solution

The solutions for problems mentioned above are, in my non-expert opinion, quite simple:

Use a SATA backplane with rectangular holes centrered between two adjacent harddisks (i.e. use as little PCB for the backplane as possible, without losing structural strength ofcouse). Place a fan in the back of the housing behind the SATA backplane (make sure to keep the fan at a reasonable distance from the backplane in order to prevent intake turbulency. Also allow some of the air to enter the systemboard compartment in the front of the device, but keep it isolated as much as possible from the harddisks up to the backplane.

This solution draws air in at the front and expells it at the back using a near strait path through the housing. This results in a better airflow over the harddisks, a less turbulent outtake and also less turbulence when it has left the housing (The network- and power cables force the device to keep a certain distance from the wall). All these factors contribute to a quieter, more efficiently cooled system. The resulting housing would be a few centimeters deeper, but also a few centimeters less high.

I will try to create a drawing of the air circulation somewhere in the future. For now you will have to visualize it yourself :)

As the frequency for the simple brushed fan of the NA-1400 can easilly be 40 to 60 kHz, using a frequency of about 20 to 30 kHz would make the fan rotate much smoother and with less noise. The onboard microcontroller is also suitable for this solution (20 kHz at 8 bit resolution when running at 8 MHz).

System health monitoring

Most modern pc’s provide fan speed monitoring and can shutdown automatically in case of any failure. The Newisys NA-1400 doesn’t provide this mechanism. If the system fan of the NA-1400 fails this could lead to damage to the harddisks eventually causing loss of data. Generally speaking, Newisys would have made a good choice if they’d incorporated a proper fan failure protection scheme. It is a simple mechanism and wouldn’t have added much to the cost.

USB connectivity

When the NA-1400 is also to become a printer server for example it would be very convenient to have two USB ports on the back of the device. One or twoUSB ports on the front is good for quick and temporary connecting external storage devices, but not if devices are to be connected permanently.

The onboard Philips USB chip has four ports available, but only two are currently in use. Two USB ports on the back could easilly be added with only minimal additional costs.

CVM: the resource hungry blob

CVM, which is some king of Java Virtual Machine, is used to provide the userinterface. While this interface look quite good, has a lot of features it is also slow and resource hungry. When a users requests a simple page like the CPU usage, the CVM process huggs the CPU for several seconds between 50% to 90%. It also eats up lots of memory: 7%. CVM also causes much of the constant harddisks activity I noticed when the system should be totally inactive (When I killed the process the NA-1400 .

Because much less powerfull embedded systems (e.g. routers) also have good looking webinterfaces and lots of features, it is certainly possible to provide the NA-1400 with a more elegant solution.

 
doc/general/suggestions.txt · Last modified: 2006/06/16 23:04 by admin