Part Selection Guide
Some of the parts below are optional. For most of these parts, you simply leave them out if you don’t want them. If a jumper is required, it will be mentioned.
You may want to print out a copy of the schematic for reference while working through this guide.
Capacitors
C1, C2 (Class X and Y film)
Along with L1 these form the AC line filter. C1 must be a Class X capacitor and the C2s must be Class Y capacitors, for safety reasons. These are sometimes described as interference suppressor caps. Check the datasheet to be sure they are designed to be used across an AC power line, and that they have at least a 250 V rating, even if you expect only to be using it on 120 VAC systems.
These capacitors are optional. You can install just C1 and leave the C2s out, perhaps because you want no coupling to AC ground, or because you can find Class X caps but not any Class Y caps.
At this time, only the Panasonic ECQU series caps recommended in the part table have been tried.
Largest Part Size: 19 mm × 6 mm for C1, 15 mm × 5 mm for C2.
C3 (metallized film)
When the power is removed, the fields in the line filter choke and the transformer collapse, generating a voltage spike. Along with R2, this capacitor forms a snubber to defeat this spike so it doesn’t damage parts further down the line.
This board position is currently the same size as the 0.1 µF cap used in the AC line filter (C1); you can use the same cap here. It’s a good value for the purpose, but if you want to make the ideal snubber for your particular power supply, the article Calculating Optimum Snubbers by Jim Hagerman explains how to do it.
This part is optional.
Largest Part Size: 19 mm × 6 mm
C4 (film or ceramic)
These caps suppress any noise generated by the bridge diodes as they switch on and off. The exact value to use here is not terribly critical. Values recommended for this purpose range from 100pF to 0.01 µF, depending on who you ask. My view is that since it’s a high-frequency application, smaller is probably better. Also, a smaller cap is going to avoid passing lots of line-borne noise.
Schottky diodes don’t generate large amounts of noise, and they already have a pretty high amount of parasitic capacitance across them. These caps really only make sense if you’re using more generic silicon diodes.
Largest Part Size: 2.5 mm × 10 mm
C5 (electrolytic)
These are the main filter capacitors. You can use just one, but using more reduces the filter bank’s ESR and reduces pre-regulator ripple.
There are three different sets of overlapping footprints here. You can use up to four 10 mm or 12.5 mm diameter caps, up to three 16 mm or 18 mm diameter caps, or up to two 20 mm or 25 mm diameter caps. In certain enclosures and with smaller transformers, you can even use a single 35 mm diameter cap.
It’s important to be careful about the part height. The two standard enclosures have 20 mm and 35 mm above the board level for most of the board area, but the corner roundings inside the case reduce this by about 5 mm, which can affect your C5 choice. Testing this here, I see that 10 mm diameter caps can use the entire 20 mm height in the Hammond 1455L16 series enclosures, but the 12.5 mm diameter caps that fit in the same place run into the corner roundings, so you can only use 15 mm high caps. Unless you’re willing and able to experiment, you should take the lower height as a maximum.
The voltage tolerance should be at least twice the rated voltage output of your transformer. For instance, if you have a ±12 V transformer, you should use 50 V caps. The reason is, when the transformer is lightly loaded, its voltage will go up by as much as 40%. Also, the peak voltages put out by the rectifier bridge will be about 1.4× the RMS voltage. Taken together, 2× the voltage tolerance is required for safety.
You don’t want too much capacitance here. I’d keep the total of all the C5s to 5000 µF or lower, myself. I usually use around 2000 µF with fine results. Overly high capacitance values will require larger AC line fuses, which reduces their usefulness in protecting against problems. There are other bad effects resulting from overly high filter capacitances. Rely on the regulator to reduce ripple; excessive filtering capacitance isn’t going to make a big difference in the performance of the power supply as a whole.
Largest Part Size: See above; it’s complicated!
C6 (film)
This cap just lowers the output impedance of the unregulated supply, and bypasses the input pin of the preregulator. The default value is fine.
Largest Part Size: 2.5 mm × 10 mm
C7 (electrolytic)
This cap is the post-preregulator filter cap. It just provides a little extra bulk capacitance near the regulator, and bypasses any ripple let through by the preregulator.
You can probably go down to about 100 µF here, if that’s all you can get in your cap line of choice.
You want to pay more attention to the voltage. This cap sees nearly the full unregulated voltage, so you should probably use the same voltage tolerance as for C5.
Largest Part Size: 10 mm diameter
C8, C9, C10, C11 (electrolytic)
The value of these caps is not critical, but changing these values shouldn’t change the performance of the supply greatly, either. The best reason to use different values is just that you can’t get the default value in a can that fits on the board. As long as you stay within a factor of about 2 of the recommended value, you should be fine.
C8 should only see the D3 voltage at most. C9 and C10 should only see about half the regulated voltage in normal operation, but in fault conditions they can see the full voltage. And C11 always sees the full regulated voltage. Since these caps have the same board footprint and the same value, to keep things simple, I recommend just making all four caps the same, with C11 setting the voltage rating. For instance, if the maximum voltage output is 30 V, you should use 35 V caps.
Largest Part Size: 8 mm diameter
C12 (film)
This cap slows the voltage rise at T1’s gate to prevent triggering due to short transients. The default value is fairly conservative. If you want the crowbar to trigger faster, you can use a smaller cap here, or even leave it out.
Largest Part Size: 2.5 mm × 7.5 mm
Diodes
D1
These diodes make up the AC rectifier bridge.
You can use any diode in a TO-220 or DO-201 package here. You can use smaller barrel diodes such as DO-41s, too. There are a great many types that will work; the parts table and the schematic give just a few ideas for you to consider.
Be sure to check the reverse voltage ratings as well as the forward ratings. Pick diodes with a voltage rating at least equal to the fully loaded voltage of the transformer. 25 V diodes with a ±12 V transformer is a good minimum. It won’t hurt performance to use higher voltage diodes, and it may provide a necessary safety margin.
D2
This LED sets the bias voltage for the constant current source. You can tune the regulator’s dropout voltage a bit by using LEDs with different forward voltage drops. But, unless your voltage situation is very tight, don’t worry too much about the exact value here. Jung recommends a 2 V green LED, but they also tried red ones (1.8 V) with good results. Some standard green and amber LEDs can have forward voltage drops as high as 2.2 V, which I also expect will work.
You’re on your own if you try more exotic LEDs, such as blue, white, or the newer pure green types, all of which have much higher forward voltage drops. This will raise the regulator’s dropout voltage, so I’d recommend against doing this.
You can adapt many different LED package types to work here, but it’s easiest to use a standard 3 mm LED.
D3
Unless you’re changing D4, too, I don’t see a good reason to use any part here other than the one recommended. If this diode doesn’t have the proper zener voltage, the regulator can fail to start correctly in some situations.
If you’re changing the stock design to get an output voltage below 6.9 V, you’ll probably have to change this part to either a lower voltage zener or possibly a small series diode, typically a 1N4148. Again, changing this part’s value can affect whether the regulator starts up correctly, so be prepared to experiment if you find yourself needing to change this part.
D4
This is a voltage reference used by the error amplifier to set the regulator’s output voltage. This reference voltage is multiplied by the error amp’s gain (set by R10, R11, and VSET) to get the regulator’s output voltage. The standard values give a gain range of 2.6 to 4.2; with the standard reference voltage of 6.9 V, the output voltage range is approximately 18 to 30 V, centered on 24 V.
We use such a high reference voltage because the recommended part is uncommonly quiet, with a typical wideband noise figure of just 7 µV. Voltage references with lower output voltages are readily available, but they’re all noisier. Therefore, the only good reason to change this power supply’s reference is to get an output voltage below 6.9 V. In that case, we recommend the LM385, which has the same pinout as the default LM329. It has far more wideband noise than the LM329, at 120 µV, but that can’t be helped if you need to get down below 6.9 V. You may also need to change the error amplifier to a low-voltage type, and will probably have to fiddle with D3’s value as well.
If you must get below 2.5 V, you will probably have to rig a nonstandard voltage reference to fit the LM329/LM385 pinout. Finding a replacement for U2 will also require some care. Also, remember that this is an “HDO” regulator: high drop out. To get 2 V at the output, you’ll have to feed it something like 7 V. Such a supply would be less than 30% efficient, but that can’t be helped if you want to use this design.
D5
This is the power indicator LED. If you don’t want a power indicator, jumper across this position and install R12 anyway. See that part’s description for the reason.
D6
This sets the voltage trigger level for the crowbar. You pick its zener voltage to be just a bit higher than either the desired output voltage or the recommended maximum supply limit for your error amp.
For instance, you might have your supply configured for 24 V out, with an error amp capable to accepting a 30 V supply, and so select a 1N4751A, a 30 V part. This would be a good choice if the downstream circuit can also take at least 30 V, so it only triggers right before there’s a danger of damage. If instead your 24 V output is right at or below the downstream circuit’s limit, you might change to a 1N4749A or 1N4750A.
Beware of setting the YJPS output too close to this zener voltage. Although the crowbar won’t trigger until another volt or so past the zener voltage, you don’t want the D6 path to be conducting in normal operation, as that will wreck the supply’s noise performance and make the crowbar touchy besides. You do want the crowbar to react quickly, but at high output voltages, you probably can’t reliably configure the supply to react to less than about 2 V of overvoltage. D6 will have a 5% zener voltage tolerance, so a part rated for 30 V could start conducting anywhere between 28.5 V and 31.5 V. You should thus use such a zener only with a supply configured for 28 V or lower, and only to power a circuit that can tolerate 32 V briefly. If these conditions don’t match your situation, you should pick a lower voltage for the YJPS and/or this zener.
A lot of people get mentally stuck when picking their power supply voltage, as if there were something special about the standard voltages; 5, 9, 12, 18, 24, 30 V.... These standards are arbitrary, and this is a DIY supply. Unless your downstream circuit is very picky about its supply voltage, you have the freedom to pick, say, 22 V if that works better.
Inductors
L1
L1 is a common-mode choke. Together with C1 and C2 this choke forms a CLC filter to reject line-borne high-frequency noise that would otherwise sail right through the power supply. The toroid and filter cap bank aren’t much impediment to high frequency noise. The regulator does well up to a point, but this filter takes over well above the audio band where the regulator’s performance drops off, up where digital electronics are putting out hash onto the AC line.
This filter is somewhat “bidirectional.” It’s most effective at blocking high-frequency noise from getting into your powered circuit through the power supply, but it can also prevent HF noise from getting out of the powered circuit onto the AC line through the power supply. Thus, if you’re using this to power a DAC or similar, consider this feature mandatory, for the benefit of your other audio components.
The board was designed with Panasonic’s ELF series chokes in mind. You can get the full line from Digi-Key, and a small subset from Farnell. The board will work with any of the 4-pin chokes with 10 mm × 13 mm pin spacing in this line, or the big 850 series with 6 pins in a 15 mm × 21 mm array. The full set of supported types — from smallest to largest — are the 16M, 290, 15N, 17N, 200, 450, 650, 21V and 850.
If you can’t get the Panasonic ELF series, look into the EV20, EV24, EV28, and RN214 series chokes from Schaffner, or the PLA10 and PLY10 series from Murata. I’ve seen these, variously, at Mouser, Newark, Rapid Electronics, and Farnell. I haven’t personally tried any of these; I’m just going by what they say in the datasheets.
When picking a choke, you first need to pick the current rating based on how much current you intend to draw from the supply. Since this part is on the primary side of the supply, the current is reduced by the transformer’s winding ratio. For instance, if your supply will have a constant 320 mA load and you have an 8:1 transformer, the primary side current is only 40 mA. The actual current rating should be much higher to provide a safety margin, and to withstand the power supply’s inrush current. 200 mA should be safe enough in this example.
Once you have the current rating, you can then pick the inductance. The higher the value the stronger the filtering effect. It’s really a question of how much you want to spend. A little ELF type 16M choke will work fine, but some may prefer to spring for a bigger choke.
This filter is optional. The Step-by-Step Assembly Guide explains how to bypass it.
Largest Part Size: 26 mm × 36 mm
TR1
The transformer is covered below.
Transistors
Q1
This transistor forms a constant current source along with D2, R5, and R6. Its purpose is to provide the initial base current to Q2 to get the circuit started, and to force the error amplifier into class A for better performance.
You can use any TO-92 PNP transistor here that uses the standard US pinout. Some other transistors use a mirrored variant of this pinout; you’d just have to mount them backwards relative to the outline on the board. Using a different type than the one recommended is likely to change the noise level a small amount, either for good or ill.
You can use the same type of transistor here as for Q3 and Q4.
Q2
This is the regulator’s output pass transistor. Because it is directly connected to the load and controlled by the error amplifier, the type you use can greatly affect the performance of the circuit.
The recommended type is not the easiest thing to find, but it is known to perform very well in this circuit. It is likely that there are other equally good performers out there, but I do not have any suggestions as of this writing.
This part’s board footprint has room around it for an optional heat sink. Much of the voltage dropped in this circuit is across this part, so it can get quite hot without a heat sink. Unless you know for a fact that you don’t need it, use the recommended heat sink. (Most of the rest of the voltage drop is across the preregulator.)
Q3, Q4
These PNP transistors are being used as low-leakage diodes, for clamping the error amplifier’s inputs to safe values. This is necessary because the op-amp inputs are connected to the circuit output, where any number of common circuit faults could fry an unprotected op-amp. Some op-amps have such diode clamps acrosss their inputs already, so you could leave these transistors out in that case, though more protection can’t hurt.
You can use the same transistor type here as for Q1.
Resistors
You should only use 1% (or better) metal film resistors in this circuit. Lesser resistors go against the whole point of this power supply, which is high accuracy and low noise.
All of the resistors can be 1/4 W types. If you want to use the Vishay Dale RN/CMF series resistors, the RN55s are 1/4 W at the temperatures you should see in this power supply. They’re specified for lower wattage because their temperature range is so high.
R1
This is a bleeder resistor to discharge the line filter caps. Its value is not at all critical, but it should be high.
R2
Along with C3 this forms a snubber to suppress voltage spikes generated by the magnetics when the power is removed. See the C3 section for further details.
R3, R4
These resistors set the preregulator’s output voltage. You can of course change the values to get a different output voltage, but the whole point of this preregulator design is that it floats some fixed distance above the final regulated voltage. If you set it lower, the preregulator would put out too much ripple, and if you set it higher you’re just wasting power.
That said, you might need to burn power. For instance, you might be stuck with a transformer that puts out way more voltage than you want at the regulated output. Rather than drop most of it across the pass transistor, you could set these resistors to drop more of it across the preregulator, thus splitting the heat load across the two heat sinks. Obviously it’s best to pick your transformer so you don’t get into this situation, but that isn’t always practical.
R5
This resistor biases the CCS voltage reference, D2. It just needs some current to turn the LED on and allow it to achieve a stable voltage drop. You can change the regulator’s dropout voltage slightly by fiddling with this resistor’s value, but unless you have a really tight voltage situation, I wouldn’t bother.
The best reason to change this value is if you’re configuring the supply for a lower output voltage than the default. That implies a lower unregulated voltage, so you need a smaller resistor value here to put enough current through the LED. I recommend keeping at least 1 mA going through this LED in the worst case situation.
R6
This resistor sets the value of the regulator’s constant current source. The voltage across it is fixed at your D2’s Vf value minus the Vbe drop of Q1; a constant voltage across a fixed resistance gives a constant current.
With the default values, you get about 5.6 mA through here. This current gets multiplied by the hFE of the pass transistor to give the regulator’s maximum output current. With the minimum guaranteed values for the standard pass transistor, this puts the maximum output current at about 330 mA. You might think to lower this resistor’s value to get more output current, but beware: the error amp has to sink away some fraction of this current based on the load and supply variations to keep the output voltage stable. With no load and a high input voltage, the error amp might have to sink away nearly all of this current. The datasheet for the recommended op-amp claims it can sink 50 mA continuously at minimum, but this reduces its performance and may require heat sinking.
Also, keep in mind that the heat sinks in this supply are relatively small. Even splitting the voltage drop across them, you’re going to have a hard time keeping heat under control with higher current loads. You’ll have to do the math to figure out if more output current is practical.
Bottom line, leave this value alone unless you know what you’re doing.
R7
This resistor just performs a little buffering and prevents dead shorts on the error amp’s output. Stick with the default value.
R8, R9
These resistors are current limiters for various parts of the regulator. It’s best that R9 be at least 10× as high as R8. You can shift the two values around a bit if you need to, but I would keep them near the recommended values. For instance, going with 4.7 kΩ and 470 Ω because you have those values on hand would be reasonable, while going with 10 kΩ and 1 kΩ would be less so because the larger values would add noise. I would find it hard to justify saving a little hassle on parts sourcing at the cost of higher noise in a supply of this caliber.
R10, R11, VSET
These resistors set the regulator’s output voltage range. It’s a simple noninverting op-amp gain situation: the regulator’s output voltage is the reference voltage (Vref) from D4 multiplied by the error amp’s gain:
Since the reference voltage is 6.9 V by default, the stock values give approximately an 18 to 29 V output range, centered on ~22 V.
If you want a different output voltage, the simplest thing is to vary R10. If you leave R11 and VSET at their default values, and use a 6.9 V reference, the equation for R10 is:
This puts your desired output voltage in the middle of VSET’s adjustment range.
R12
This sets the current through the power indicator LED. In addition to telling you that the regulator is putting out power, it also keeps a small, steady draw on the power supply, which can improve performance in some cases. It also helps discharge the filter caps when the power is removed. If you don’t want a power LED, add R12 anyway, and just jumper across the LED.
You should select this resistor’s value to allow at least 5 mA through this path, even though that will make the LED rather bright. You might even go up to the LED’s maximum, typically 20 mA. This will ensure quick discharge of the filter caps.
R13
This limits the current through the crowbar’s zener to a safe level, and ensures that the gate of the crowbar’s SCR is tied to ground in the normal case, preventing accidental triggering.
The proper value here is not critical, but there is a correct range. If this resistor is too big, the zener’s turn-on behavior will be squishy, potentially delaying the crowbar turn-on too much to be useful. If it’s too small, the zener could be fried when it starts conducting, potentially fast enough that the crowbar doesn’t actually turn on, so that the overvolage goes right through to the downstream circuit.
Thyristors
T1
This SCR is part of the crowbar circuit, which protects against output overvoltage.
If you can’t find the parts recommended in the part list, you’re looking for an SCR designed for crowbar use. Other types of SCRs aren’t suitable.
You might be surprised by the high current rating chosen for this part. Although the YJPS is inherently current-limited to only a fraction of an amp, that only happens when the circuit is working correclty. If the crowbar triggers, the circuit is not working correctly, so we have to assume that current limiting is gone, and that the biggest capacitors on the board can all dump their charge down this path at once. The SCR simply cannot be allowed to fail in this situation, as it’s the last hope of preventing damage to the downstream circuit.
You might then wonder why there is so little space set aside around this part, scarcely enough for the smallest sort of heat sink. The reason for that is, this part only dissipates large amounts of power very briefly even in the worst case. Unless the SCR itself fails, the power through this device will eventually drop to under 1 W, well within the capabilities of a naked TO-220 package.
Integrated Circuits
U1
This is the preregulator. It reduces the amount of ripple that the next stage has to deal with. Although there are many different regulators that will fit the generic TO-220 3-pin adjustable regulator pinout here, only the LM317 is currently recommended. The LDO variants in this pinout will probably oscillate due to C7, and the higher-current variants are unlikely to have any benefit over the LM317.
If you want to spend money on a better regulator here, use the National Semiconductor LM317A, rather than the non-A variant or an off-brand clone.
U2
This op-amp is the error amplifier. Its job is to zero out regulation errors by adjusting the base current of Q2 so that the actual output voltage matches the gained-up reference voltage.
This component has the single greatest effect on the stability and performance of the power supply. A poor choice can be either unstable, or give higher noise than the circuit is capable of achieving. Unless you know what you’re doing or are willing to experiment, it’s best to stick with the tried-and-true here. At this time, only the AD825 is known to work in this particular implementation.
If you’re contemplating using a different chip just because it’s hard for you to get an AD825, keep in mind that generic chips have poorer performance. Our goal here is to make a high-performance power supply, so ask yourself whether it’s worth compromising its quality just to save a mail order.
If you want to use another type, it needs to be a standard single-channel SO-8 op-amp. The YJPS doesn’t make use of any features found on the 3 unstandardized pins of this package.
The best reason to switch error amps is when you need an output voltage outside the AD825’s recommended supply range, 10-30 V, where the chip doesn’t perform at its best. I’ve used an AD825 successfully down to about 5 V in other circuits, but I haven’t tried that in a YJPS. As for the other extreme, beware of using this chip for output voltages above ∼24 V unless you also install the crowbar circuit.
At low output voltage levels, you probably need to use a rail-to-rail type chip, else it won’t have enough output swing to provide good regulation. Walt Jung recommends the AD848 in his original articles on this design, but I haven’t tried it in a YJPS.
Also see the discussions about low voltage for D3 and D4.
There are some chips that will let you get above 30 V out. If you really push it, you might be able to achieve something in the neighborhood of 48 V with this design, using a high-voltage op-amp like the OPA604. Again, I haven’t tried this myself.
You should only experiement with different op-amps in this design if you have access to an oscilloscope. Like any other feedback circuit, it’s possible to make a YJPS oscillate, but unlike many other such circuits, it’s often difficult to tell when this is happening without looking at the output on a scope.
Configuring the Crowbar
T1, D6, R13 and C12 form a type of “crowbar” circuit. Crowbars are ways to prevent incorrect voltages from causing circuit damage by shorting the power supply. The image is of a person taking a real crowbar and dropping it across a device’s power terminals. Ohm’s Law tells us that it would take near-infinite current to prevent near-zero ohms from causing an immediate drop in voltage, so the voltage drops, hopefully quickly, to a safe level.
Other circuits on this site use a much simpler crowbar, a simple diode across the power inputs of a circuit, oriented so that it only conducts if the supplies are hooked up backwards. The one in the YJPS is a different sort, meant to protect against overvoltage.
As long as the YJPS’s output voltage is below D6’s zener voltage, its crowbar is inactive. Above this zener voltage, D6 begins to conduct, but nothing else can happen until the YJPS’s output voltage is at least this zener voltage plus T1’s gate threshold, typically around 1 V. If the YJPS’s output rises past this gate threshold plus the zener voltage, still nothing happens until C12 charges up to the gate threshold level. Thus C12 prevents short transients from triggering the crowbar. This triggering time is a function of the amount of overvoltage, the higher the faster, just as we want.
When the SCR is triggered, it clamps, shorting the YJPS’s output voltage to ground. T1 will stay triggered as long as there’s power to the circuit, even if the output voltage drops back below D6’s zener voltage.
The best case is that F1 blows, protecting everything else.
If F1 doesn’t blow, the SCR will try to drag Q2’s emitter down to ∼1 V. The worst case voltage across Q2 is therefore nearly the entire preregulated voltage level. It can’t stay there, though, because this high load will be causing the transformer secondary voltage to drop, eventually causing the preregulator to drop out of regulation. At that point, Q2’s collector voltage will start dropping just as fast as the unregulated voltage. If this voltage drop happens fast enough, Q2’s heat sink can be enough to protect it while these many watts of power go through it. In that case, the power supply’s output will eventually sag down to the SCR’s level, and sit there sulking until you unplug it.
It’s possible for neither F1 to blow, nor for the output to sag fast enough to protect Q2. If Q2 blows, it will either fail open, “fixing” the problem, or it will fail short, endangering U1, which isn’t any better protected than Q2 was and which sees more voltage. Thus U1 is likely to follow suit, either failing open (another “fix”) or closed, in which case the YJPS output becomes the unregulated voltage, which must continue to drop, eventually getting down to the SCR’s minimum.
The SCR itself should be unkillable, as you should be using a part here capable of handling the worst case surge currents until the output voltage drops to the SCR’s minimum. At that point, you have the transformer’s entire output current dropped through just a volt or so, which the SCR should be able to handle “forever;” long enough, certainly, for someone to notice the problem and power everything down.
All of this tells us that there are only two ways for the downstream circuit to see an overvoltage fault. One is the catastrophic case where F1 doesn’t blow and both Q2 and U1 fail short quickly enough that the transformer’s secondary hasn’t yet dropped below a safe level. The other is that the SCR is too small to clamp the overvoltage long enough to avoid dying itself. You avoid the primary risk simply by sizing F1 correctly. You avoid the secondary risks by not running Q2, T1 and U1 so near their limits that they can’t tolerate a failure.
Choosing a Transformer
There’s a lot of detail here, but I urge you not to gloss over this section. Choosing the wrong transformer is the single most common mistake I see people make when building linear power supplies.
It’s common in DIY when selecting part values to choose something that’s way overspecified to avoid getting something that’s too small to do the job without doing the detailed engineering to find exactly the right value. For most areas of DIY electronics, this tendency to overspecify is harmless and may give some useful safety margin. Transformer selection in linear power supplies is not one of those areas. Overspecifying the transformer usually results in a power supply that overheats because the regulator is being made to drop more voltage than is necessary to maintain good regulation. On overheating, the supply either just quits, or it shorts the regulator, supplying the full unregulated voltage to the downstream circuit, potentially destroying it, too.
It’s difficult to remove a PC mount transformer from the board once it’s soldered down without damaging the board. You want to be very sure you’ve got the right one from the start. Be sure you understand what the estimator is telling you before you purchase a transformer.
Part 1: Estimating your way to the right VA rating
In an effort to help people avoid the most egregiously wrong choices, I created the Power Supply Parameter Estimator. This will give you an idea of how the power supply will behave before you build it. It cannot do your thinking for you, but it can point out many common errors.
When using the estimator, give it 5 V for the regulator dropout voltage, to account for both the preregulator voltage drop and the minimum drop across the pass transistor. For the heat sink thermal resistance, it’s usually best to use the thermal resistance for just one of the heat sinks, and calculate with the θJC of the pass transistor. This ignores the heat in the preregulator and pretends that the pass transistor is carrying more load than it really is, which gives you a conservative result. If you have a serious heat problem, you could adjust the drop across the preregulator to split the heat load between the two heat sinks. In this case, you’d use half the thermal resistance of a single heat sink in your calculation, but double the thermal resistance number for the heat sink compound and add the θJC values for the pass transistor and the preregulator.
Naturally, you do have to avoid getting a transformer that’s too small, as well. There’s a simple rule of thumb for that: the load current times the load voltage should be no more than about 80% of the VA rating of the transformer. Exceed this threshold and you’re likely to “saturate” the transformer, causing it to perform badly, even overheat.
Part 2: Check the voltage limits
If you have never built a linear power supply or your only experience is with simple 3-terminal regulator based supplies (e.g. TREAD), you might be surprised by some of the voltage limits in this design, so pay attention.
The first limit you run into is that for the error amplifier’s power supply. This IC is powered from the regulated output, so in the most optimistic case, its supply limit puts a limit on how much output voltage you can expect to get from the supply. The default AD825 has a 30 V recommended maximum supply limit. Beware of approaching this limit, however. If the preregulator and regulator don’t always work flawlessly, they will put more voltage across the op-amp’s supply pins, potentially damaging it. See the U2 and crowbar sections for more on this topic.
Another important voltage limit to keep in mind is 50 V, that being the Vceo limit of Q1, which can see the full unregulated supply voltage in some situations. Since there is a possibility of an inductive spike getting through to Q1 on power-off, you shouldn’t come very close to this value when choosing a transformer.
Part 3: Selecting a part
The YJPS board is designed to hold one of the Amveco (a.k.a. Talema) 5 through 25 VA board-mount toroidal transformers. Digi-Key’s part numbers for these are 7002x through 7006x. The best way to find the part number you need is to do a search for “amveco pc toroid,” then narrow the results by voltage and VA rating.
Beware that only the 5 VA units will fit in the Hammond 1455L16 series inclosures. The others all require the taller 1455N16 series.
You can use any other toroidal transformer you like. The pads on the PCB are labeled so you can match up the color coding scheme used by your transformer with the proper pad on the board. If you get something other than a dual primary dual secondary type transformer, a little study of the transformer wiring diagram and the PCB will let you see how to avoid using the jumpers needed for the Amveco PC mount transformers. Since I haven’t used any such transformers with a YJPS myself, I can’t offer any detailed advice. If you can’t figure out the wiring on your own, either stick with the recommended units, or post to one of the DIY forums to see if someone will work out the correct wiring scheme for you.
One of the advances in this board over the STEPS is that it lets you wire the transformer’s secondaries in series or parallel with jumpers on the board. This lets you trade off current for voltage, or vice versa. For instance, if you choose a 12 VA transformer with dual 12 VAC secondaries, you can wire the secondaries in series for 24 VAC at 0.5 A, or in parallel for 12 VAC at 1 A. It’s 12 VA either way. This can open up options you wouldn’t have with a similarly configured STEPS.
Choosing an Enclosure
The YJPS board is designed to fit several of the enclosures in the the Hammond 1455 line, particularly the 1455L16 and 1455N16 types. These two basic case designs have the same footprint, so the YJPS board fills out the entire length and width of the case. The difference between them is height, with the “L” being about half the height of the “N”, so you have to be much more careful about part selection. I try to give multiple part choices above for the taller parts, with at least one short enough to fit in the “L” version, but check the sizes to be sure. There is just barely 20 mm above the board level in the “L” version, and more like 35 mm above the board in the “N” version, not counting the corner roundings, which subtract about 5 mm. This corner loss affects your C5 choices.
Keep heat sink selection in mind when choosing your case. The heat sink the board was designed to use comes in several standard lengths, 12, 25, 38 and 50 mm. Notice that two of these are just a little bit higher than the space above the board in the standard Hammond cases. If you’ve checked your configuration and have found that heat isn’t a big problem, you can just use the next size down that does fit. If your configuration is so close to not working that you need every last bit of drop in thermal resistance, you’ll have to find or make something nonstandard. The heat sink I’ve chosen for this board design is uncommonly compact, so there really isn’t a lot of space for “creativity” here. Perhaps the cleanest option is to trim one of the standard heat sinks to fit. A band saw would make this easy, though you could perhaps make do with a Dremel or a hack saw. Be sure to deburr the trimmed heat sink so you don’t short something out with aluminum flakes or whiskers.
The Hammond 1455 line has several standard options. One is whether the body of the enclosure is black or clear anodized; if black, just append BK to the clear version’s part number. Another is whether it uses aluminum end panels, or plastic, with the aluminum having 1601 in the part number, and plastic having 1602. Either will work for the YJPS, with the choice having more to do with esthetics and ease of working than RF interference blocking, as the YJPS shouldn’t be greatly susceptible to RFI if you stick to known-good error amps. The metal panel version will help conduct heat out of the case better, though, perhaps enough to matter.
The Hammond 1455s are rather expensive enclosures, so if you just want utility, almost anything can be made to work. Take some time to think about heat, however. I would insist on adding vents over the heat sinks for plastic or wood cases; it’s easier to justify having no vents with a metal case, as it can get rid of heat by conduction.
Another popular option is to get an enclosure big enough to hold the power supply and the circuit it’s powering. If you do this, be sure to allow for some room between the two circuits so you can put some distance between the AC and unregulated DC parts of the YJPS board and any sensitive parts of the circuit you’re powering. Just because we’re using a toroidal transformer doesn’t mean we can’t couple AC line hum over into the other circuit. It’s just harder, is all.
Miscellaneous Hardware
The board has room set aside in the corner of the board between the line filter and the transformer for a Qualtek 723W power input module, if you use the taller 1455N16 cases. This module has a fuse holder built in, and it’s very nearly as cheap as a plain IEC socket. An alternative to this part is the Schurter 6200 series; it’s very similar in design to the Qualtek socket. These sockets, though relatively compact, won’t fit in the shorter 1455L16 cases unless you get rid of the AC line filter, in which case you’re missing out on a lot of the value of the YJPS. Rather than sacrifice the line filter, I’d go with a captive AC line cord if I wanted to use the shorter enclosure. You could then either use the board-mounted fuse clips, or one of the more compact sorts of panel-mount fuse holders.
If you’re not going to use the remote-sensing feature of the board, you can get a DC power cord with the barrel connector molded already on it, rather than have to build your own. Kobiconn has several cords of this type, which you can get from Mouser; it’s a lot more convenient than building your own power cable. You should also get a strain relief for that cable.
If you’re going to build your own DC output cable, such as because you want remote sensing, you’ll need to carefully think out which cable, connectors and strain reliefs you need. When the power supply can’t live in the same enclosure as the circuit it’s powering, I’m a fan of the 2-wire remote sense compromise configuration. This requires 4 wires, but only a 2-conductor connector to the powered device. I like to use Canare Star-Quad cable, which fits tightly into the opening in the Switchcraft 760 type DC barrel plugs. This gives you four stout but flexible conductors for the remote sensing, without requring additional strain relief at the barrel plug end. You’ll need a strain relief bushing for power-supply end of the cable.
You’ll notice several holes around the board labeled TP1 through TP8. These are test points, useful for testing the power supply. I like wire loop type test points, so I can grab onto them with my meter’s grabber leads. Keystone type 5005 through 5009 wire loop test points work well here. You can also use the pins cut from SIP pin strips, or make loops from resistor lead cuttings, or leave the holes empty to form a crude DMM probe “socket.”
This space intentionally left blank. :)