Low Power Isolated PSU

There are countless reasons that you may need a low-power supply, and particularly if it has to provide galvanic isolation. That means that there's no electrical connection between the powering circuitry and the powered circuit. One example is a MOSFET or hybrid relay, as described in Project 227 or the MOSFET Relays article. There are other applications too, and one of those is the subject of another project coming soon.

In some cases the supply will have to be isolated to handle mains potentials, but these aren't as common as basic 'air gaps' that may only have to withstand a nominal voltage of as little as a few hundred millivolts, up to 100V (AC or DC) or so. A common reason to introduce this 'air gap' (galvanic isolation) is to prevent ground loops and other troublesome issues when routing audio signals any distance. The same can apply with digital systems that require an isolated supply for the line driver section (for example, isolated RS-485/ RS-422 Data Interfaces).

There are many applications now that rely on galvanic isolation to separate high-current, high-voltage circuitry from control systems. Having the two interconnected is often very difficult, because ground currents (in particular) can cause serious errors, and a fault can be catastrophic. Electric vehicles (manned or otherwise) are a case in point, with IC manufacturers developing new devices all the time to isolate the high voltage circuitry from other parts of the system. Even at a comparatively low current (say 10A), a ground resistance of 0.1Ω will create a 1V voltage-drop, more than enough to introduce a large error into a measurement system. Galvanic isolation can eliminate this type of error completely.

One application for galvanic isolation is in electric vehicles. High current (with the likelihood of a great deal of noise) is used for the motor(s) and their controllers, but the interface back to the computer systems has to be isolated. A fault current shouldn't fry all of the processing systems, as most people would (quite rightly) be 'annoyed' - something of an understatement I expect. The need for isolated gate drivers isn't limited to very high voltage/ current systems though - it's becoming necessary in a multiplicity of new designs.

Audio applications are often susceptible to ground current, with the most common problem being hum at the mains frequency. This problem has existed for as long as people have been connecting mains-powered equipment together with signal leads. The use of balanced connections (initially with transformers) has long been a common way to exclude ground loops, but electronically balanced circuits can (and do) often fail to provide a complete solution. Transformers are still the best option, but good ones are expensive.

There are countless cheap isolation transformers for audio, but to be genuinely useful these can only be operated at unrealistically low signal levels. Their common-mode rejection is generally not wonderful, as they don't have an electrostatic shield. Performance can be enhanced by using an opamp circuit, so there are two separate common-mode rejection circuits, and you can have a decent output level. This can be done using transformers costing less than AU$2.00 each, compared to AU$100 or more for a quality transformer. If the signal level is kept low, the cheap transformers will have performance that's little different from the expensive ones.

Another use is for battery operated pedals, as used by musicians. These generally run from a 9V battery, and some use positive to chassis, while others use negative to chassis. You can always use two external supplies, but getting the wrong lead for a pedal can be a disaster, either killing the pedal, the power supply or both. A selection of completely isolated power supplies prevents any mishaps, because each outlet is isolated from the others. Using a separate power supply for each pedal is one method, but you could end up with a lot of power supplies.

Users may also want to generate a negative voltage when only a positive polarity is available. Because the supply is isolated, you never have to worry about which way it should be connected to avoid short circuits. Inverting (but non-isolating) DC-DC converters are available, but they aren't as flexible as an isolated type.

In most of the circuits shown here, the output uses a voltage-doubler rectifier. This provides more voltage, but reduces the available current. Voltage-doublers can be replaced by a bridge or full-wave rectifier (the latter requires a centre-tapped transformer secondary) and vice versa. A voltage-doubler generates roughly twice the voltage at half the current compared to a bridge rectifier.

The heart of any isolated power supply is generally a transformer. There are photovoltaic isolators (PVIs) that you can use, but they have very limited output current, with most being less than 100µA. This isn't enough for any opamp circuit that's expected to drive a long cable (for example). PVIs are designed to turn on MOSFETs, and that's all they can be used for. Another class of isolator uses capacitive coupling (e.g. Si8752), and while these are much better than PVIs, they are still low current devices and are currently difficult to get from any supplier. The output voltage and current are limited, and if you need more than a few milliamps you're out of luck.

For an isolated supply, a realistic output current is around 50mA, which is enough for several opamps. The voltage should not be less than 10V, but you may be able to get away with 9V, depending upon your application (and the opamps - some won't work properly with less than 10V). The following drawing shows the general (highly idealised) approach for isolated converters. The transformer is assumed to have close to zero resistance. Reality will be different, and the two things that suffer most are output voltage and regulation. If the supply is to be used only with low current (perhaps 20mA or so) then simple circuitry can be used to get a satisfactory result.

The general principle is shown below. A squarewave inverter converts the incoming supply (12V in this case) to a high frequency squarewave, and the DC component is removed with a capacitor. The cap doesn't need to be a high value if the frequency is high enough, and if the frequency is over 100kHz, a 1µF cap will be sufficient. Its reactance at 100kHz is only 1.6Ω, so losses across it will generally be quite low. A more-or-less typical transformer will have a winding resistance of perhaps 5Ω or more, so the cap's reactance is negligible.

The transformer couples the ±6V waveform to the secondary rectifier. Voltage doublers are convenient, as they allow you to use a 1:1 transformer. In many cases, you don't need a 'bulk' capacitor at the output, as the two voltage doubler caps can be sized to ensure that the ripple voltage is within reasonable limits. Although they are shown as 1µF, you can use more. If space is at a premium, multilayer ceramic capacitors (MLCC) are available up to 10µF, usually in a surface mount (SMD) package. These have very good performance up to very high frequencies.

For transformers you have a number of choices. You can wind your own, which is somewhat tedious but quite satisfying once it's done. You can also use common-mode chokes, used for filtering interference at the input of a switchmode supply (SMPS), and these can be very effective. A typical example is seen in Fig. 8 (third device from the left). You need an inductance of at least 1mH or the switching frequency will be too high to obtain from cheap parts. Everything is a compromise, and the transformer generally involves the most trade-offs. Everything gets harder when the input DC voltage is reduced. 12V is usually a reasonable choice.

Another alternative is a cheap audio transformer. Something rated for 600:600 Ω is ideal, and these are readily available from multiple suppliers. However (and this is very important), the isolation voltage is minimal, and exceeding 50V or so between input and output would be ill advised. There are several ICs designed for low-power DC-DC converters, but they are fairly specialised (and require equally specialised transformers) and both may be hard to get. An example is shown in Fig. 3, using a MAX253 switching IC, specifically designed for the purpose.

Small isolated DC-DC converters are available from multiple suppliers, and one that I've used is the B1212S-1W, which not surprisingly is 12V input, 12V output and rated for 1W (83mA). They can be obtained from multiple manufacturers and suppliers, with prices ranging from less than AU$5.00 to AU$15.00, depending on brand and supplier. Brands you've never heard of are at the low end of the price scale (not surprisingly).

It's very hard to go past these for convenience, and they work well and use minimal PCB space (about 12x6mm for single output types). Mostly, if I need a floating supply these devices will be my first choice, but there's a lot to be said for building your own because you can learn so much from the exercise. You'll find that it's surprisingly hard to get more than 30mA or so with a DIY version. The first circuit (Fig. 2) uses a comparator as a free-running oscillator, with a nominal frequency of 350kHz. The transformer should have an inductance of not less than 1mH, but more is often better. 10mH is probably ideal.

There are many reasons you may want to do it yourself. The first is to get a better understanding of switching supplies in general, and DIY is generally the only real option for that. Theory is always good, but practice means that you really do get to know how everything interacts. The first circuit is more complex than any of the others, but the parts are all low-cost even if you have to buy the comparator. The device shown is a dual, but only one section is used (ground the two inputs of the unused section). You can use any comparator you like, but be aware that some have 'odd' pinouts.

The comparator oscillator is capable of operating at 500kHz or more, but 350kHz is more than sufficient with a 1mH transformer (anything from 1mH to 10mH will work). The required load pull-up resistor (R6) is bootstrapped with R5 and C3. This allows the comparator's output to exceed the supply voltage, providing a useful increase in the output swing. Gaining an extra 2V at the isolated output is well worthwhile. Using Schottky diodes also gains a small but worthwhile increase, and with the values shown the output will be about 9V at 10mA (12.3V at 22mA with a 15V supply). This requires a transformer with low winding resistance, with no more than 10Ω resistance for the primary and secondary (5Ω for each). The oscillation frequency depends on many factors, but with the values given for R1, R2 & R3, it's roughly ...

To allow the use of a 1:1 transformer, the rectifier is a voltage-doubler. The output isn't regulated, so it will vary depending upon the load current. Following the circuit with a linear regulator IC (e.g. LM317 or 7809) will reduce noise and provide a clean regulated output. It's worth noting that the total cost for the supply is probably greater than that for a B1212S-1W, but it's one that you build yourself and you'll get to understand just how everything works. You don't get that if you buy one. The B1212S-1W is also much smaller than you can make the Fig. 1 circuit, even if you use SMD parts throughout.

However, if you build or select the transformer to have a very high isolation voltage (≥ 2,500V DC) then it may be suitable for isolation of mains voltages up to 230V. This is something that the small modules are not rated for, so a DIY version provides options that are otherwise unavailable. Bear in mind that your DIY circuit will not have any safety agency approval unless you use an approved transformer. Isolating anything to mains standards requires good knowledge of creepage and clearance distances and adhering to accepted practice for everything. This is never trivial!

The MAX253 IC is specifically designed for making isolated supplies, and at first look it's an appealing option. There are no support parts other than a bypass capacitor, transformer and output rectifier, and there's a range of transformers designed specifically to work with the IC. Reality sinks home when you look at the prices, as just the IC and a transformer can easily cost almost AU$20.00 (depending on supplier). In addition, the maximum input voltage is 6V (5V is recommended), so you need a substantial step-up if you need a 12V supply, and you need to include a 5V regulator to drive the IC. The output isn't regulated.

Both the MAX253 and suitable transformers are available in through-hole and SMD, so home construction (using Veroboard) is easy. The transformers are available with various ratios, allowing operation with either 5V or 3.3V supplies. If you need a higher voltage you're back to winding your own transformer, and you lose any approvals that apply to those designed specifically for the IC. This is an option that I didn't test, because I don't feel like paying a premium for a power supply that won't be of any use to me anyway.

There are other alternatives such as a 555 timer, but they have a significant output voltage sag when loaded, and are only suitable for perhaps 10mA or so output current. As shown in Project 95 (Low Power Negative Car Power Supply) you can also use an LM386 'power' amplifier IC. You'll probably get a bit more output voltage, but don't expect to get much output current, as they have limited dissipation. Unless used with a transformer, the Project 95 circuits are not isolated. Sometimes you can use an IC that's intended for a completely different application, such as ...

The UC384x series of ICs are designed for flyback SMPS, but the drive circuit has enough capability to allow them to drive a small transformer directly. Apart from decoupling and coupling caps, they only need a resistor and capacitor (R1, C1) to set the oscillator frequency. With 10k and 1nF, the oscillator will run at about 100kHz. D1 protects the internal transistors from damage due to negative voltages. These ICs are cheap (under AU$2.00), readily available, and are made by several manufacturers. The transformer is the same as for Fig. 1, with an inductance of between 2mH and 20mH.

The MAX253 and UC3845 ICs are not the only solutions of course. There are many other ICs designed for this application, because it's a common requirement in system design. The examples shown are examples, and a web search will find other solutions. Some become quite complex, which limits their usefulness. When a designer wants/ needs an isolated supply, the logical approach is to use the simplest circuit that does the job. Sometimes a more complex version may appear to be cheaper (e.g. lower BoM costs), but assembly and test time has to be considered for production systems.

Another device designed specifically to provide a low-current isolated voltage supply is the TI UCC25800-Q1. This is an 8-pin SMD chip, intended to drive a small transformer. It's designed specifically to provide MOSFET/ IGBT gate current, and can operate from supplies up to 35V (40V is the absolute maximum). It doesn't require many external parts, and appears to be easy to use. I'm unsure why the datasheet extends to 47 pages, but it is very comprehensive.

It's intended to be used in a resonant LLC (inductor-inductor-capacitor) converter and is designed to provide very low EMI emissions. In a simple design, the transformer would be driven directly from the output, via a suitable capacitor. It can operate at up to 1.2MHz to keep the transformer size to the minimum, but of course it must still provide insulation suitable for the voltage differential between the low voltage and high voltage sides (e.g. 230V AC for mains isolation).

The output current capability is higher than most of the others described, with the internal MOSFETs able to switch 600mA. The IC itself is tiny - only 3 x 3mm, with eight pins. Mounting it using hand soldering would be a 'challenge', to put it mildly. No circuit is provided because it's unlikely that anyone will build a PSU using this IC. It's not a cheap IC unless you buy thousands. The 1-off price at the time of writing is almost AU$10 each, depending on supplier.

Also from TI is the DCH01 Series of DC-DC converters. These are 19.5mm x 10mm x 8mm (L x H x D) modules, available with several output voltages. The input voltage is 5V, and the total output power is 1W. They are available with single or dual outputs, at 5V, 12V and 15V. They are rated for an isolation voltage of 3kVDC (1 minute) and are configured in a 7-pin SIP (single inline pin) package. Note the wide spacing (comparatively speaking) between the input (left two pins) and the output. This provides sufficient creepage and clearance distances between the input and output.

The converter IC isn't specified, but operation will be identical to the others described. Depending on supplier, expect one of these to cost no more than AU$20 or so. It may be necessary to add input and output filtering if noise is an issue (or to ensure EMC compliance). The output is unregulated, in common with most similar parts.

The fact that there are so many DC-DC converters available, along with suitable ICs and transformers shows that there is a definite need for galvanically isolated power supplies for a wide range of applications. This can be expected to increase within the next few years. Transformer coupling is the most reliable method of providing isolation, and with modern ferrite materials and high-speed switching, the transformer size can be minimised.

Capacitive coupling isn't included in this article, other than as a simplified example shown next. It's too hard to get a usable current, and it's a requirement to keep the coupling caps low value to prevent common-mode AC from creating interference. Even if you use a 1MHz oscillator, the usable current is less than 10mA with sensible-sized coupling capacitors (i.e. no more than 22nF).

Capacitive coupling may be used internally with the Si8751/2 MOSFET gate drive IC, but if so the caps must be very low value. The internal workings aren't disclosed in the datasheet.

The example shown next uses 22nF caps, which will provide very basic isolation, but only for a few volts difference between input and output voltages. This does provide galvanic isolation, but there are many limitations and it's not a recommended circuit. It works, but probably won't be satisfactory in anything but the most basic application. Magnetic coupling (a transformer) works so much better that I wouldn't consider using capacitive coupling in a 'real' circuit.

The example in Fig. 5 is just that - an example. To keep the capacitance low, the frequency must be high, and the simulation I used had a generator frequency of ~1MHz. The current is still very low, so expect no more than around 5mA. It's limited further by the output current from the 4584/ 40106 hex Schmitt inverter, and even two gates in parallel can't supply much current. You can get more current using a higher frequency or increasing the value of C1 and C2, but it's still low unless high-current buffers are used to drive the coupling caps. C1 and C2 have to be a low value to prevent 'disturbances' between the controller ground and the isolated output from causing problems. The 2.2nF caps have an impedance of less than 75Ω each at 1MHz, but the current remains limited. Higher current starts making unrealistic demands on the drive circuit.

The voltage-doubler is recommended and a (if desired) you can add a 12V zener diode in parallel with the output. This will cause the IC to draw more current, but CMOS is generally fairly tolerant of overloads. You'll need to experiment if you need a stable output voltage. Remember that capacitor coupling goes both ways (as does transformer coupling), so any disturbance on the secondary side will be passed back via C2 and C3. The difference between capacitive and magnetic coupling is that a transformer will only pass a signal that appears across the secondary winding (differential mode), but capacitors will couple differential and common-mode signals equally well.

If a significant voltage exists between the input and output supplies, there's a risk that the output stages of the 4584/ 40106 inverters may be damaged if an external voltage is applied to the 'isolated' side very quickly. If that's likely, you need to use extra diodes to protect the CMOS IC. As it's only an example, the extra protection isn't shown. By all means experiment, but don't expect it to be wonderful (and don't expect any support if you can't get it to work).

All of the messing around with separate components is eliminated by using a modular DC-DC converter, such as the B1212S-1W mentioned above. These are by far the easiest way to get an isolated supply, but they are not intended to provide mains voltage isolation. Mostly this should not be an issue at all. The pinout shown below is 'industry standard'. Regardless of manufacturer, a device indicated as B1212 (or any other voltage variant) will have the same pinout. The input and output are isolated, so the output can be used for a negative supply, 'stacked' onto the existing input supply (getting +24 for a 12V version) or used as an independent supply for isolated circuitry.

It's not possible at present to get anything else that's as compact, and while the output may not be regulated that's generally not a problem. Some are regulated, but you're looking at using a 'name brand' with full specifications available, and the regulation isn't especially good (±5% no load to full load is about average). You'd expect that a particular model number would have the same specs regardless of manufacturer, but that's not necessarily the case with these. However, for most applications you'd probably use the cheapest, but reliability will be unknown. One maker (MORNSUN) offers a 3 year warranty, regulated outputs and very detailed datasheets.

In some respects it's cheating, but when an ideal part exists that does what you need, then it would be silly to go to any more trouble. Of course you don't learn anything from the exercise, but if you're building a project you need something that works, doesn't cost much and does what you require. These modules are available with several different input and output voltages, and there are some that provide a dual-polarity output. If you need ±12V on a floating supply, you can get it easily.

Input Voltage	Output Voltage	Output Current	Efficiency (Typ.)
3.3 V	3.3 V	303 mA	72 %
	5 V	200 mA	76 %
	12 V	84 mA	80 %
5 V 12 V	3.3 V	303 mA	72 %
	5 V	200 mA	76 %
	9 V	111 mA	80 %
	12 V	84 mA	80 %
	15 V	67 mA	80 %
	24 V	42 mA	80 %

Table 1 - Abridged Bxxxx 1W Module Voltages

I've only shown 3.3V, 5V and 12V inputs, but the modules are available with 15V and 24V inputs as well, with 12V being (by far) the most common. Not every combination is available, but there are few omissions. The efficiency figure is 'typical' and is only relevant at full load. Although manufacturers and/ or suppliers may list all the combinations, that doesn't mean you can get them. Some are very popular and therefore readily available, but less common versions may be difficult or impossible to obtain.

You can tell instantly that these modules are not intended to provide mains isolation, as the pins are at 2.45mm (0.1") spacings. To achieve acceptable creepage and clearance, they'd need to have a minimum of 5mm between the input and output, but this is not the case. If you don't know what to look for, it's easy to make an assumption that results in an unsafe circuit.

Figure 8 - B1212 DC-DC Converter (Left) And Three Transformers

Fig. 8 shows (in order) a B1212 module, a small transformer I built to test a couple of the circuits (12.6mH @ 100kHz), a common-mode choke 'transformer' (20mH @ 100kHz) and an example of a commercial transformer intended for use with Ethernet. The inductance of the Ethernet transformer is very low and it can't handle high voltage, but it is rated for 2kV isolation. There are three separate transformers in the single SMD case, and acceptable performance is obtained with all three in series (a total of about 225µH).

It's immediately apparent that the B1212 (or any other voltage) is the most compact. In fact, it uses less PCB real estate than the Ethernet transformer alone, and it's only twice the height. Nothing else comes close, so while you may wish to experiment with the other ideas shown, in a project that doesn't need mains voltage isolation there's no contest. Not only is it smaller than the other circuits, but it will cost less as well.

PCB mount DC-DC converter modules are made by Mornsun, Murata, Recom, Traco, CUI, Texas Instruments and several other manufacturers. Some are interchangeable, others not. The are available as through-hole and SMD, but SMD versions are generally the same size as through-hole types. The art of miniaturisation only goes so far, and the transformer is always the largest single component. You need to check carefully, because some have little (or minimal) output filtering, and many require additional filtering to pass EMC testing. This only applies to commercial products - it isn't a requirement for DIY projects.

Some are less than AU$4.00 each, and even cheaper if you buy more. This makes trying to build your own a rather pointless exercise, but it's still good to make one just for the sake of doing so. You don't learn anything by soldering a part to a PCB (unless you reverse the supply polarity and blow it up of course). :-)

Like many ESP projects and articles, the circuits shown here are to provide ideas. All the DIY versions are capable of acceptable performance, but none could be classified as 'optimum'. In particular, the output current is very limited, but it's still enough to power an opamp or two. Despite any misgivings you may have, there's rarely a requirement for close-to-perfect regulation, and most of the time even supply noise (including [high-frequency] ripple) isn't an issue. Of course that depends on what you're trying to achieve, but most circuitry using ICs is surprisingly tolerant of the power supply. There are now many ICs designed specifically to provide isolated supplies, which is testament to the requirement for galvanic isolation in so many applications.

If your circuit is safety critical (isolating 230V AC for example), you cannot use Veroboard, and to ensure that it remains safe it should be 'hi-pot' (high potential) tested. This is normally the sole domain of accredited test labs, and they will charge you dearly for the service. You can use a 1kV 'Megger' (insulation tester) to get a rough idea of its safety, but if you design something that is sold to the public, it may require lab testing. If someone is killed or injured because of your work, you will be held responsible.

There aren't many audio circuits that need a floating (and isolated) power supply. They're more common in measurement and monitoring systems, particularly where high voltage and/ or currents are involved. It's often far easier to use an isolated supply than to have to try to eliminate noise or errors introduced when high and low current ground connections are shared. Despite the other examples, there is only one solution I'd consider for a project, and that's a ready-made module. When you consider that you can get these units for under AU$5.00 each, it would be folly to use anything else.

The idea to create a project for this came about because I wanted to use a cheap audio transformer for an isolated audio signal, but still have excellent response and low distortion. Because the cheap transformers use a very small core, the signal level must be kept low, so an opamp stage was needed after the transformer. Since the transformer is used specifically to get galvanic isolation, it was obvious that the power supply had to be isolated as well. This will be published when I've finished testing the circuits.

These circuits can be used anywhere that basic galvanic isolation is needed, but I strongly recommend that you avoid using any of them for mains isolation. As noted already, regulatory approvals may be a requirement for anything that uses mains power, and your safety and that of others must never be jeopardised. This is doubly true for medical devices, whether used in a medical facility or not. If mains isolation is a requirement for your project, then use only approved components across the isolation barrier.

There aren't many applicable references, because the circuits are largely conceptual rather than project ideas in their own right. The references shown are for specific parts from various suppliers.

Isolated Low-Power DC-DC Supplies