by Oleg Artamonov
09/18/2005 | 05:36 PM
Passively cooled computer power supplies, once limited exclusively to systems of the most hardcore enthusiasts that were ready to manually finish off off-the-shelf solutions, have become quite ordinary products with times. Today, many PSU makers think it necessary to have one or two fanless models in their product range.
<%BANNER[article]%>Of course, you cannot solve the problem of noise completely by simply using a silent PSU because there still remain the central processor, graphics card, mainboard and hard drive. But after the PSU has got silent, it is not a hard task to ensure proper cooling and noise insulation of all these components, too.
You can use a system case with low-speed 120mm fans (it would be unwise to assemble a computer without any active ventilation at all), put rubber noise-absorbing pads on the hard drives, choose a graphics card with passive cooling, and mount a copper CPU cooler with a low-speed fan (or even deploy a liquid-cooling system). In a system assembled with silence in mind, a fan-cooled PSU may become the loudest component, spoiling the result.
So today we offer you a detailed review of two fanless power supplies from FSP Group and Topower. Frankly speaking, only one of them proved to be truly fanless, as you will learn below.
The FSP Zen looks quite strange even when compared with other fanless PSUs which usually have large protruding heatsinks. The Zen is a neat box without any jutting details at all.
Most of the case is covered with small-meshed vent grids (it would be better to have a larger mesh from the technical point of view, I should note). The power supply is only cooled by means of convection: warm air from the system case comes in through the grid in the case of the power supply and then goes out through its rear panel.
You can’t see anything inside after you’ve removed the cover – the components of the PSU are hidden under the three large heatsinks. Note that they differ from heatsinks in fan-cooled PSUs: instead of small bars with thin and densely placed fins we have massive aluminum blocks with thick fins here.
The top parts of the heatsinks can be removed (all junctions are carefully covered with thermal paste) to reveal us the internal design of the PSU:
The power semiconductor elements are distributed among three heatsinks rather than between two, as usual. The PSU’s high-voltage elements (the active PFC device and the transistors of the main and standby regulators) reside on the first two heatsinks; the third heatsink carries the diode packs of the output low-voltage rectifiers. A separate perforated heatsink positioned perpendicularly to the others cools the diode bridge on the PSU output (this bridge doesn’t heat up much, so there’s no need to add anything to the top of this heatsink).
The 270µF 450V smoothing capacitor that stands after the PFC device is rated for a temperature of 105°C, while ordinary PSUs usually have 85°C capacitors on their output. This is a reasonable solution. The capacitor itself doesn’t heat up much, of course, but the proximity of a hot heatsink might tell negatively on its service life as there are no strong airflows inside the case. A rather large active PFC coil is located near the capacitor. The choke coils of the line filter can also be seen on the snapshot (between the diode-bridge heatsink and the rear panel of the PSU). The line filter is complete, so I have no complaints about it.
The next remarkable thing is the output diode packs. A majority of PSUs employ packs like Mospec S30D40C in large TO-247 cases. This model uses Fairchild YM3045N (MBRP3045N ) diode packs in TO-220 cases, but there are as many as twelve packs here, six on each side of the heatsink:
At first, before checking the marking, I thought the engineers used a synchronous rectifier. For people who are not well versed in electronics, a synchronous rectifier is a circuit in which MOSFETs replace diodes and are controlled in such a way that one transistor opens on arrival of a positive half-wave and the other opens on the arrival of a negative one. Thus, the transistors imitate the operation of ordinary diodes but feature a higher efficiency – the voltage drop on a diode is constant (about 0.7V for a silicon diode and about 0.5V for a Schottky diode), while the voltage drop on a transistor depends on its type. So, transistors with a minimal resistance in the open state make it possible to reduce the voltage drop considerably and, accordingly, to increase the rectifier’s efficiency.
The Zen, however, uses ordinary diode packs (each rated for an up to 30amp current, 45V voltage and 150°C temperature of the die) which are connected in parallel in twos to ensure the required load currents at high temperatures.
And the last interesting feature of this PSU model is the abundance of toroid-core throttles on the output which are a sign of independent regulation of the output voltages. As I wrote in my previous reviews, in the classic PSU design the +3.3V voltage has a dedicated auxiliary regulator on a magnetic amplifier (of which a choke coil is the main component), but the +5V and +12V voltages are regulated together with a so-called group-regulation choke. This solution helps to make the power supply cheaper and simpler, but the result isn’t quite perfect. For example, if there’s a high load on the +5V rail and this voltage “bottoms out”, the +12V voltage grows up above the norm as a consequence of the group regulation. In power supplies with dedicated voltage regulation the main regulator is only watching over the +12V rail, while the +5V is served with an auxiliary regulator like the one installed on the +3.3V rail. As a result, the +12V and +5V voltages of the PSU become in fact independent of each other. You can tell such a PSU visually by noticing not two but three large choke coils on the output (by the way, there can even be only one – group regulation – choke in some cheap power supplies).
The power supply offers the following cables and connectors:
The cables are rather short and there is no 6-pin power connector for the graphics card, but otherwise, I have no complaints about the cables. The 24-pin connector with a detachable 4-pin part allows using this power supply with new as well as older mainboards (that have a 20-pin power connector on board).
This model complies with the ATX12V 2.0 standard – the low load capacity of the +5V and +3V rails is well compensated with the 22amp +12V rail (which is split in two with over-current limiters).
The most interesting test for a fanless power supply is checking it at a sustained full load, i.e. at 280 watts in this case (I loaded only the +5V, +3.3V and +12V lines). The temperature of the PSU heatsinks was steadily growing up for several hours and then stopped at 90°C for the hottest of them (the heatsink with the diode packs). The heatsinks of the high-voltage section were less hot, but the heatsink of the active PFC device may reach the same temperature at an 110V input voltage (thanks to the same active PFC the reviewed power supply supports input voltages from 90 to 264V without any switches).
This temperature may seem high, especially since the power supply was not installed in a system case but was lying on my desk (the temperature would be 10-15°C higher if the PSU were installed in a computer). On the other hand, even if the temperature of the cases of the diode packs becomes 125°C, they will still be able to work with an up to 15amp current (the maximum allowable current diminishes as the temperature of the case of a diode pack grows up; 150°C is the maximum temperature of the die of the pack, not of its case, and the temperature difference between the die and the case depends on the current. In other words, the temperature of the die will be 150°C at 125°C case temperature and 15amp current). The packs are connected in parallel, so we can assume the current for such a “pack made of packs” to be 22.5 amperes (it is not correct to double the currents at parallel connection. To ensure stable operation we must assume that each additional parallel-connected element increases the load capacity by only 50-70%). But any of the output currents of this PSU is below 20 amperes, so there will be no problems here.
So, the power supply can yield its maximum output power without problems. Let’s now check the stability of the output voltages depending on how the total load is distributed among them.
The diagram above shows you the loads at which all the main output voltages of the PSU are within the acceptable ranges, i.e. within ±5% of the nominal value. The X axis shows the load on the +12V rail, and the Y axis shows the combined load on the +5V and +3.3V rails (in each point the load current on the +3.3V rail is 50% of the load current on the +5V rail; thus, the load power of the +3.3V rail is about 25% of the total load power, like in a typical today’s computer). The deflection of the voltages from the nominal in percents is marked in different colors as explained in the legend in the top right corner of the diagram.
The FSP Zen behaves quite predictably for a PSU with dedicated voltage regulation: the +5V voltage deflects from the nominal by less than 1% and the +3.3V and +12V voltages by less than 2% at any allowable load. As you can see, the power supply maintains the required levels of the output voltages at any allowable load.
The only negligible drawback is that the PSU is unstable at highly misbalanced loads. For example, its protection circuitry wakes up when there’s a 250W load on the +12V rail and less than 7 watts on the +5V rail. That’s why the diagram is shifted a little away from the axes – I usually start with a load of 5W, but here the initial load is bigger. Anyway, so greatly misbalanced loads just can’t possibly occur in a real computer system.
The next parameter we must check out is high-frequency voltage ripple on the PSU output. It turned out that the result depended greatly on the load. For example, there was a strong ripple at high loads on the +5V rail. The oscillogram below was shot at a load of 280W on the whole power supply of which 100W fall on the +5V and 20W more on the +3.3V power rail:
The voltage ripple on the +5V power rail reached 75 millivolts – 50% above the maximum acceptable value. But if the load on the +5V is reduced to 70W, the PSU “calms down” immediately:
The ripple doesn’t exceed the acceptable limit of 50 millivolts. And if the whole load is transferred to the +12V rail, the ripple almost vanishes – this PSU is obviously designed with such working situations in mind, quite expectedly for an ATX12V 2.0 model.
I didn’t perform my fan speed test due to obvious reasons, but I did measure the efficiency and power factors. FSP promises an efficiency of no less than 89% for this model and it is really so:
The efficiency is 89.3% at the maximum and this is an excellent performance. But I want to remind you once again that I perform my tests at 200V input voltage and if you connect the PSU to an 110V power grid, its efficiency is going to drop due to loss in the active PFC circuitry. It means FSP Group doesn’t tell you the whole truth – the efficiency of this PSU is really up to 89%, but only in 220V power grids.
The power factor is rather low (for a PSU with an active PFC device). It is a little above 0.95, while in theory active power factor correction allows for a power factor of 0.99. Yet, 0.95 is still good if compared with PSUs without PFC (0.65-0.7) and with PSUs with passive PFC (0.7-0.75).
I would like to tell you about one mistake users and testers often do as they put together the efficiency factor and the power factor of a power supply. These two parameters do not relate to each other in any way. I put them in one diagram for the sake of convenience only (as they can be shown well on a same-scale diagram and are also measured with the same tools). You can’t calculate the power factor basing on the efficiency whatever formulas you use. Moreover, the efficiency value is not used at all to calculate the power factor.
So, the FSP Zen is quite an appealing product. Despite the lack of any fans, it quite successfully works under the full load (but of course, the top of the system case right above the PSU will hardly be cool to the touch). This model can work at input voltages from 90V to 264V (you are going to appreciate this if the voltage of the power grid in your area lacks stability). Using dedicated regulation of the output voltages, the power supply keeps them stable at any allowable load, while the load capacity of its +12V rail should be sufficient for all midrange computer systems and even for some top-end configurations. Of course, it wouldn’t cope with a system with two GeForce 7800 graphics cards, but people who own such SLI configurations have other things to worry about than a noisy power supply…
As for drawbacks, the cables of this PSU are rather short, and there’s a big voltage ripple at high +5V loads. The latter thing doesn’t matter much for modern computers, however, as they mostly load the +12V power rail.
The text on the cardboard box the TOP-420NF power supply comes in reads “Fanless Enhanced Cooling Power Supply”. The words “enhanced cooling” actually mean an ordinary 80mm fan installed on the front panel of the PSU (this panel will be inside the system case after the installation). Why “fanless” then? The manufacturer claims the fan starts to work only at a 250W and higher load, but remains absolutely silent at other times. You can also turn the fan on with a button on the PSU case, but the fan will then be working at its full speed only.
Topower is not only known under its own brand, but also as the actual manufacturer of power supplies from OCZ and BeQuiet!. The TOP-420NF has the familiar features like a black glossy case, a graphics card power cable with an LC-filter on the connector and screening, and black heatsinks with small and dense ribbing:
Unlike with the FSP Zen, this model has one heatsink on the outside – its L-shaped internal part is put on the heatsink with the diode packs. Otherwise, the design of the PSU is somewhat closer to the classic samples than the Zen. The TOP-420NF seems to be an adaptation of an existing power supply rather than a product developed from scratch. The adaptation wasn’t too serious, though. For example, the employed heatsinks with numerous small ribs and slits are obviously supposed to be cooled with a fan. Even the external heatsink is shaped like that, while a heatsink with larger and less densely placed ribs would suit better for natural passive cooling.
Moreover, the ribs of the heatsinks are directed towards the inside of the power supply rather than to the outside which worsens the efficiency of passive cooling even more. The external heatsink also almost entirely covers the opening in the PSU rear panel, so the air from the fan (when it is turned on) goes out through the opening in the top of the case mostly. As a result, the power supply is driving hot air in a circle rather than exhausts it to the outside.
In my tests the fan turned on not at a 250W load, but after twenty minutes of working at 150W load when the temperature of the heatsink with the diode packs reached about 70°C. The fan is going to turn on even earlier in a real computer where it will be additionally heated up with warm air from the processor and the graphics card.
The external heatsink is covered with a protecting grid which is not very necessary in fact. The heatsink is not hotter than 60°C even at the maximum load, so you won’t have a chance to scorch your fingers.
Otherwise, the power supply presents nothing particularly interesting. It is a typical design on a TL494 PWM controller (located on a separate card), without any power factor correction or additional regulation of the output voltages.
The PSU offers you the following cables:
The mainboard’s power cables are sheathed into plaited pipes, the graphics card cable – into a flexible plastic tube (this cable has additional screening, which is not however connected to ground). The wires in the other cables are twisted like in a twisted pair. So, the cables of the TOP-420NF are overall better than with the above-described FSP Zen – they are longer and have more connectors. Of course, you can use adapters and splitters, but it’s handier to do without them.
This model formally belongs to the ATX12V 1.3 standard, despite the 24-pin mainboard’s connector, but this version of the standard does not describe power supplies with wattage higher than 300W, so I can only say that the TOP-420NF surpasses the requirements of the standard in every parameter. On the other hand, the PSU is obviously intended for high loads on the +5V rail which is not as important for modern computer configurations as the +12V rail, and the +12V rail of the TOP-420NF has the same acceptable load as the one of the considerably less powerful FSP Zen.
The cross-load characteristics of this PSU don’t look that beautiful. First, the +5V voltage is too high. It will be at about 5.2-5.3V in modern computers where the load on this power rail seldom exceeds 30-40W. Second, the +12V and +3.3V voltages deviate rather far from the norm, too. If you compare the TOP-420NF with other, similarly designed PSUs, it will appear an ordinary, average product, of course, but it doesn’t look appealing at all against the ideal diagrams from the FSP Zen power supply with its dedicated voltage regulation.
The voltage ripple under the maximum load was rather strong, but not above the normal range: 45 millivolts on both +5V and +12V rails, the acceptable maximums being 50 and 120 millivolts, respectively.
As I said above, the fan of our sample of the PSU began to rotate after the unit had worked for 15 minutes under a 150W load. The speed of the fan was steadily increasing from the initial 1100rpm as the load was growing up:
Topower claims the noise from the fan doesn’t exceed 22dB and is not heard against the rest of the computer’s noises. Alas, it is not quite true: at the maximum speed of 2560rpm the stream of air is not loud, but quite audible. A Yate Loon D80SH-12 fan on sleeve bearings is employed here.
The efficiency of this PSU reached 82% at the maximum which is worse than that of the FSP Zen. The power factor is about 0.65-0.68, like in other PSUs without power factor correction circuitry.
So, it would be an overstatement to regard the TOP-420NF as a fanless power supply. It is an ordinary PSU with the fan speed control set up in such a way that the fan starts to work only at a certain temperature of the heatsinks (about 70°C). The external heatsink helps somewhat, but I think the efficiency of passive cooling would be better if more massive heatsinks, specially designed for fanless operation, were employed. On the other hand, if there were fewer obstacles on the path of the air stream (the outer panel of the PSU has very small vent openings because of the external fan), active cooling could be made more efficient, too, and a slower fan might be used then.
On the other hand, if you view the TOP-420NF as an ordinary, but low-noise power supply, much fewer complaints remain. The PSU is really silent under small loads. At higher loads the noise from its fan isn’t too loud and many users won’t notice it at all. The electrical parameters of this model are average. From this point of view, however, the price of the PSU (above $100) may look somewhat too high.
On the whole, I can’t say anything bad about either of the tested power supplies. Products from FSP Group and Topower feature a highest quality of manufacture, and the particular models I have reviewed here easily meet their specifications.
The FPS Zen was obviously designed as a fanless power supply from the start, while the TOP-420NF model from Topower is a redesign of an ordinary actively cooled PSU. As a result, the Zen can really be regarded as an absolutely silent PSU whereas the TOP-420NF is quiet, but not completely silent.
Moreover, the FSP Zen is better of the two in other parameters: modern computers don’t need a high load capacity of the +5V rail, while the load capacity of the +12V rail in the Zen is theoretically no worse than in the TOP-420NF and is even better in practice due to a smaller voltage ripple and a higher stability of the output voltages. Thus, the FSP Zen seems to be a better choice as a power supply for a noiseless computer.