Last year I bought an MFJ-4125P 13.8V 22/25amp power supply. It's widely touted as being one of the smallest, lightest supplies avalable, and having good RFI properties. Since I travel a bit, I was looking for good portable power that could deliver current above typical battery levels, for times when I want to use full power rigs like my IC-7000 or K3 on the road. Almost all my radio equipment and power cabling is terminated with Anderson Powerpoles in the ARES-standard configuration, and the 4125P model has integrated Powerpoles.
Phil Salas (AD5X) liked the MFJ-4125, though the version he got didn't have Powerpoles, so he added some.
Eham Reviews of the 4125 are generally good in terms of power out, reliability, and RF noise, but fan noise is a big issue for many users. For example, AI7AZ's 2012 review notes the "truly outrageous amount of noise the fans in this little box put out." This worried me, but still the small size and good quality power were attractive.
On receipt, I plugged in the MFJ-4125P to try it out. The fan noise was truly loud, much too loud to be comfortable in any half-way quiet shack or room. On a recent trip, while operating my IC-7000 in one section of a basement, I tried putting the MFJ-4125P into a workbench drawer to muffle the noise - even that didn't do the trick, and I had to live with it.
I thought of selling off the PS, but the small size, my sense of adventure, and AD5X's confident modding convinced me that the downside risks weren't that great - maybe I could manage the noise problem somehow and get the satisfaction of homebrewing a solution.
High fan noise in this and several other power supplies has been noted and addressed before.
AD5X addressed it for the Samlex 1235). The stock Samlex 1235 includes a heat-sensing circuit that switches on the noisy fan at full speed only when the heatsink temperature gets above a threshold. Phil's popular mod involves adding a parallel fan circuit to run the fan constantly at slower speed, to keep the temperature down. A 100R resistor in the parallel circuit drops the fan voltage and lowers the speed. If the temperature rises anyway under load, the original heat-sensing circuit will close and the fan will run at its original higher speed with attendant noise. This provides a very nice backup. In practice, Phil found that with the constant low-speed fan the temperature rarely rose enough to trigger the original circuit, so fan noise was almost never an issue after the modification.
Joseph Marler (KQ1Q) replaced the 1235's noisy fan with a new ultra-quiet fan.
Unlike the Samlex 1235, the MFJ-4125 has no heat-sensing fan control - the fan runs continuously which makes the noise issue that much more problematic. Alan Bloom (N1AL) addressed the noise issue for the MFJ-4125 by adding his own heat-sensing circuit. This is somewhat involved, though the circuit, parts list, and a PC board are available. The fan speed becomes a function of the heat sink temperature. This was one approach I considered.
I first opened the 4125P to see what I could find out about the internals. (NOTE: Several folks have pointed out that one of the heatsinks carries 120V live, so that's an issue to be careful about.) I noticed several things:
In 2008 on the Elecraft reflector, John Gibson (NO8V) said, "On my MFJ-4125, over half of the fan exhaust openings on the back of the supply are blocked by the fan's motor housing...." This issue was also noted in 2011 on the QRZ General Technical forum. Figure 1 is a photo (from AD5X's MFJ-4125 Powerpoles mod article cited above, and used with his permission) of the fan opening and the housing "grate" that covers it, showing the level of obstruction:
In addition, I looked at how the cooling airflow was routed inside the PS. Intake holes extend all the way down both sides of the P/S (see Figures 3 and 4). This means that air can come in at any point on the sides, and it exits through the fan out the back. It seemed to me that a significant part of the airflow, then, could enter toward the rear of the PS, flow pretty directly to the fan and not even pass over either of the heatsinks. Overall, it seemed that the combination of airflow blockage at the rear opening and airflow bypassing the heatsinks probably reduced the net cooling effect of the fan operation significantly.
Since the net cooling effect is already compromised, I surmised that positive airflow improvements from clearing the fan opening and changing the airflow routing would counterbalace any negative effects of reducing the fan speed and airflow volume with a voltage-dropping resistor. This gave me a three-part approach.
To clear the opening, I removed the original fan by removing the four large screws that hold it in. Being as small as it is, the PS interior is quite crowded and the fan fits tightly. Lifting it out requires a bit of care (see Figure 5). Once the four attachment screws are removed (see Figure 1), the fan itself has to be twisted slightly to lift it out over heatsinks. I also disconnected the fan's power connector from the front of the main PC board, with a pair of longnose pliers. With the fan out, I got out my nibbling tools and nibbled away all of the rear panel that serves as the fan grating, being careful to leave enough metal around the fan mounting holes for a secure re-mount with the original screws. I filed down burrs and carefully vacuumed out any metal filings from the inside of the PS. Then I re-mounted the fan and installed a new 60mm wire fan grille (Mouser Part #562-08147, $.68). This gave finger - and travel-packing - protection with negligible airflow obstruction. See Figure 2 below for the result.
I thought about drilling some additional airholes in the top of the case, but decided not to do that yet. Instead, I simply used black electrical tape to mask off the rearmost six rows of intake holes on each side of the housing top, to stop the rearmost airflow. (See Figures 3 and 4 below) This forces more air to enter forward, and flow over the heatsinks en route to the fan. (Heatsinks are shown in Figure 5 below).
|Figure 3 (click for larger version)||Figure 4 (click for larger version)|
Finally I added a 68R 2W resistor in series with the red (+) fan power cable. I soldered the resistor inline in the cable, covering the joints and short resistor leads with heatshrink tubing for insulation. I re-seated the fan's power cable onto it's original pins, routing the cable through the interior of the PS carefully, and reinstalled the fan itself. See Figures 5 and 6 below for the result.
|Figure 5 (click for larger version)||Figure 6 (click for larger version)|
Why 68R 2W? AD5X and others used 100R resistors. To experiment, with the fan still mounted in the PS I connected a 300R pot in series with the fan supply. I varied this pot until I found the appropriate (low enough) noise threshold, and measured the pot's resistance at that point. It seemed that there was a definite "articulation point" in the noise level with the resistance at about 63R-65R - lower than that and the fan noise was objectionable. Above that the noise was quite low and the airflow still significant. At 100R the fan was definitely inaudible, but the airflow seemed unacceptably low. Computing the voltage drop and power dissipation (with a safety factor), I needed a 1W resistor. I had access to 68R 2W resistors - they fit the bill nicely so I used one.
With these modifications, the PS is quiet enough for desktop operation in my shack. Most of my shack power is under the desk, so there the PS is inaudible over the general ambient room noise. Traveling, I don't expect to have any more noise issues with this power supply.
There is now some risk that with the fan never operating at full capacity there could be destructive levels of heat buildup in the power supply. Others don't report this problem with similar approaches (resistors to lower fan speeds). I'm betting the added improvements in airflow offset this prospect. In addition, my operating characteristics play a role here. I'm 99% a CW operator. Powerwise, CW is an intermittent mode: due the the periodic nature of CW codes, average CW power load is typically computed at around 40% of peak when transmitting. Moreover, most operating time (> 50%) is spent listening, so the over a span of time the average CW power consumption is something like <20% of peak. SSB also has average-much-lower-than-peak power characteristics. When I use continuous digital modes such as PSK31, it is at much reduced power levels consistent with the lower power requirments of those modes for reliable communications. I almost never operate continuous, high power modes like AM, FM or RTTY with the radios that would use this power supply. But we'll have to see, as I haven't yet done any significant tests of heat production/performance under load.
The only complaint that remains about the MFJ-4125P is the fact that the power cable is hard-wired. A standard IEC modular connector would make it much easier to pack this PS for travel. I considered replacing the original power cord with an IEC module - though the fit would be tight, with the right choice of IEC module there does seem to be space. I'm leaving that mod for the future.