Wednesday, March 16, 2016

How to fix EMI (electromagnetic interference) noise generated by the computer case

I recently bought this new case made by Spire, model SP1071B. It was very cheap, only ~15 euros (clearance sale in a local shop).

Didn't have time to install it until about a week ago. After getting everything inside and started using it for some time, I noticed this really strange static-like sound in the headphones. It seemed to happen every time there was hard-disk activity. If there was music on, the sound was almost unnoticeable, but when nothing was playing I could clearly hear it. The sound itself resembled coil whine or a floppy disk unit reading something.

But this was nothing compared to the noise that was generated when copying files to / from USB sticks. Or the one made when using an USB Wi-Fi card. Oddly enough, with the old case everything worked fine, there was no interference at all, so I decided to investigate this a bit and find a fix.

I opened up the front of the case and got out the PCB that hosts the USB and audio connectors. The circuit is really simple, just some connectors on a board, nothing more. Following the traces I could see that even if the audio and USB ports have their own wires for the ground, on the board they were connected, forming a common ground.

All the ground wires are connected together

At the time, nothing seemed out of place so I went and dismantled the old case and looked at its front panel PCB. It looked almost the same, with the exception that the ground wires were completely independent. I thought that it would be a good idea to do this on the new PCB as well, so I separated the circuits and lo and behold, the crackling noise is completely gone!

The red parts are the audio ground, the orange the USB ground and the jumper wires green

To be sure nothing else could inject signals in the audio circuit (because long wires like the ones from headphones tend to work as antennas), I also passed the AAFP wire through a small ferrite ring harvested from a broken motherboard.

My guess is that somehow joining up the audio ground with the USB ground creates a ground loop, in which current from the USB port leaks in the audio circuitry creating the EMI. The generated noises seem to be correlated to the power drawn by the connected usb device.

Later I found out that tying the grounds together is a very common practice among case manufacturers. Maybe we should send some e-mails to them showing the issue?

If you have the same problem with your case, you can see the steps required to fix it by watching the video below:



Thursday, March 3, 2016

The bad capacitor saga continues! Recapping the FSP Hexa HE-500

Last week I got this older computer PSU from the local flea market to replace another one which was retired. Luckily, the old one went out without taking any other components with it, so all that was needed to have a fully working computer was to get a decent PSU.

The "new" PSU is a Fortron (FSP) Hexa HE-500 model. I thought that it would be a good idea to go with FSP since it's a well-known brand, respected by many people, and one of the biggest OEMs in the world. I've had laptop chargers made by them that worked almost daily since ~2006 without any problems at all.

But things seem to have changed and the build quality suffered quite a bit in the last years. As expected the Hexa was full of dust so before mounting it I decided to clean it up a bit. Vacuuming didn't work too well so the unit needed to be opened and brushed clean. Upon taking the screws off and cleaning most of the dust off, I was greeted by this:

CapXon capacitors...

Yes, a nightmarish cocktail of CapXon, OST and Teapo capacitors. CapXons are really bad quality (I have replaced many of them inside LCD monitors). TEAPOs are on the mediocre side now especially because they don't take heat too well. Too bad, since they had some good series in the late 90's.

The lower voltage CapXons tend to bulge and leak while the high voltage ones exhibit some sort of internal corrosion that eats away the internal wires that go to the snap-in connectors. This is really bad since it leaves the current unfiltered and it can make the the active PFC circuit to go up in smoke.

There is a post on the badcaps forum that explains this issue in depth:
http://www.badcaps.net/forum/showthread.php?t=8168

Another thing that has been hypothesised is that bad main capacitors make the active PFC coil burst up in flames. This is one of the problems that plagued many FSP models (like the Bluestorm II) - there are many posts on tech forums about it.

To fix all these potential issues, I decided to replace all the capacitors inside the unit with something better. The only high quality caps available here were Nichicon and Panasonic so I went with a mix of those. Before starting work I drew a capacitor map to know the location and type of each and every one of them. You can see it below (sorry for the bad drawing :P):

FSP Hexa HE-500 capacitor map

Then I made a table which contains all the information about the old capacitors. The specifications were extracted from the datasheets released by CapXon / Teapo / OST. Luckily, they were easy to find. In case you have a similar unit, you can use this table to choose capacitors with similar characteristics. Word of advice: make sure to also look at the size, see why below :P.

FSP HE-500 complete electrolytic capacitor list

The first to be replaced was the high voltage cap from the primary. Can you spot the difference? The original Capxon had 270 uF and a 420 volts @ 85 °C rating, a bit on the edge for an active PFC design. The replacement Nichicon fairs much better, with 390 uF and 450 volts @ 105 °C rating. One odd thing is that although the CapXon had smaller capacitance, voltage and temperature rating, the datasheet listed a ripple current rating 25% better than the Nichicon. Yeah right CapXon...


Quite the difference between the two.





The good thing is that there was enough space to fit the beast between the heatsinks. The bad thing is that  it required moving the small PWM control isolation transformer on a separate daughterboard which I bolted to the heatsink that also holds the switching transistors:

Making way for the capacitor. A resistor, a transformer and a coil had to be relocated.
The daughterboard that now holds the PWM isolator transformer

The daughterboard connected to the mainboard using a ribbon cable
It all turned out well in the end, nothing was mounted in reverse and the voltages are in spec.

Everything put back to where it belongs

After working on this unit I can safely throw out a mini-review:

Pluses:

  • Made by a well known company
  • Nice looking, black metal case
  • Pretty silent
  • Delivers the power as advertised
  • Good cross-loading performance for a group regulated design
  • Good build quality for such a low price

Minuses:

  • Horrendous choice of capacitors
  • Bad material for the PCB: it warped because of the heat, the traces lifted even after using a soldering iron with temperature control
  • No insulation under the PCB, combined with the warping it can lead to disaster
  • The heatsink from the secondary gets quite hot, maybe better diodes would have fixed that (it uses 2 x SBR30A60CT)
  • The Yate Loon fan used a low quality lubricant that seems to be acidic: it eroded the sleeve bearing and became something that looks like a rust-colored, semi-solid wax
  • The distance between the SATA / Molex connectors is too small, I needed a Molex to SATA adapter to connect both the HDD and DVD writer.


Would I get one again? Probably, if the price was right. But in all honesty after having to replace each capacitor, I'd rather save up some money and get a more expensive model, maybe another FSP - but Japanese capacitors are a must.

Thanks for reading!