Fullscreen HTML5Med Res (??MB)HTML4

By DAVID L. JONES The µCurrent . . . a precision current adaptor for multimeters You might not be aware of it but your digital multimeter is unable to make accurate current measurements in low-voltage circuits because of its “burden voltage”. This precision current adaptor solves that problem and greatly improves the measurement accuracy, as well. D ON’T MOST multimeters already have current measurement ranges? Well, of course they do. But most multimeters, be they a no-name $10 hardware store throwaway model or a $1000 highly-accurate brand-name meter, all suffer from two rather annoying issues with their current measurement ranges – burden voltage and reduced accuracy. The biggest problem with current measurement ranges is burden voltage. This is the voltage that the internal current shunt resistor drops as the circuit’s current passes through it. The burden voltage is typically specified in millivolts per Amps (mV/ A). The value will change for different current ranges, so you might have 1mV/A, 1mV/mA and 1mV/μA for example. Normally, you may not give burden voltage a second thought, as like many, you probably think it’s fairly insignificant in most applications. In fact, most people would be hard-pressed to tell you what the burden voltage of their particular multimeter actually is. It’s usually buried away in the user manual, if it’s mentioned at all. Next time you borrow a colleague’s meter, ask them what the burden voltage is, and watch their reaction! At small displayed currents, the 58  Silicon Chip burden voltage is usually not an issue but at larger displayed currents (relative to full-scale) the burden voltage can be very high, even in the order of several volts! This can often force you to use a higher current range (with a lower-value shunt resistor), with subsequent loss of resolution and (often) accuracy. You may in fact have encountered this many times, with your circuit either not working or “playing up” on too low a current range. That’s the burden voltage at work, starving your circuit of the voltage it needs to function correctly. You usually have no option but to reluctantly switch to a higher current range to lessen the effect. The problem can also be highlighted with the many 4½-digit or “10000 count” meters on the market. In theory, they allow you to get an extra digit of resolution over a 3½-digit meter. But you may now find yourself trying to measure, for example, 990.0μA on the 1mA range with a burden voltage of just under 1V. Can your circuit really handle a 1V drop? The burden voltage of a multimeter is determined primarily by the shunt resistor used for measurement. However, on the higher current ranges (mA & A) it also includes the protection fuse resistance and, to a much lesser extent, any switch and test lead contact resistance. Some manufacturers will specify it as a total or just the shunt resistor, or in many cases not mention it at all! Some meters will specify it as a maximum voltage drop only. For example, “300mV max”. In this case, to get the mV/A value, you simply divide that voltage by the full-scale range current. Current measurements with low supply rails The recent trend toward low-voltage microcontrollers and other silicon devices (some operating from as low as 1V or less!) has really highlighted the need for considering the burden voltage when measuring currents. 3.3V supplies have been widely used for a long time now and the trend is heading lower. A common task these days is to measure the accurate “sleep” and operating current of a microcontroller. Indeed, with the lower supply voltages of today’s battery-powered circuits, accurately measuring the supply current has become more critical. So the industry has changed but digital multimeters haven’t really kept up with the pace when it comes to acsiliconchip.com.au Table 1: Burden Voltages For Typical Multimeters Approx Cost($) Burden Voltage (mA range) Burden Voltage (µA range) Meterman 5XP (3.5-digit) $65 1V max 300mV max JayTech QM-1340 (4.5-digit) $99 5mV/mA 0.11mV/μA Meterman 30XR $120 4.6mV/mA 1mV/μA Protek 506 $175 1mV/mA 1mV/μA Meterman 37XR (10,000 count) $250 10mV/mA 1mV/μA B&K 390A (4000 count) $380 2V max 500mV max Fluke 77 series III (3.5-digit) $400 6 mV/mA N/A Fluke 77 series IV (6000 count) $425 2mV/mA N/A Fluke 79 series III (3.5-digit) $375 11mV/mA N/A Fluke 177/179 Series IV (6000 count) $430 2mV/mA N/A Fluke 27 $900 5.6mV/mA 0.5mV/μA Fluke 80 series V (4.5-digit) $720 1.8mV/mA 0.1mV/μA Agilent U1251A (4.5-digit) $680 1mV/mA 0.1mV/μA Extech MM570 (500,000 count) $680 3.3mV/mA 0.15mV/μA Fluke 289 (50,000 count) $950 1.8mV/mA 0.1mV/μA Gossen MetraHit E-XTRA (60,000 count) $1700 300mV max 150mV max Fluke 8808A (5.5-digit) $1100 1mV/mA 1mV max Fluke 8846A (6.5-digit) $2100 500mV max 15mV max Keithley 197A Microvolt (5.5-digit) N/A 300mV max 300mV max Multimeter Model curate current measurement. You may think that multimeters are getting more “accurate” for less cost but that’s only part of the story. Let’s look at how the supply voltage can impact your current measurement or vice-versa, as the case may be: Let’s say you want to measure the supply current of a chip or circuit taking 200mA using a 4000-count meter on the 400mA range. This is a fairly common scenario and one you would think would be pretty easy for any multimeter to handle. But maybe not . . . A typical high-end “accurate” multimeter will have a “low” 1mV/mA burden voltage (about as low as it gets), so this means the meter will drop 200mV across its shunt resistor at 200mA. This represents an almost tolerable 4% (200mV/5V x 100) of a 5V supply voltage. This may not be a big deal if your supply voltage is spot on 5V, as your chip will get 4.8V and still be within spec. But what if it’s only 4.8V? Your chip or circuit will now be getting only 4.6V which may well be below its operating specifications. This already shows the limitation of the current range on a typical multi­ meter. But that’s without even considering how the circuit current can differ siliconchip.com.au when you lower the rail by 0.2V. Let’s now say you need to do the same thing on a modern circuit or chip with, say, a 1.2V power supply, ie, the voltage from a single NiMH cell. That same 200mV burden voltage is now a whopping 17% (200mV/1.2V x 100) of the supply voltage. Your circuit may now fail to function correctly and this is clearly not acceptable, not to mention inaccurate. Think this is only a problem with “cheap” meters? Well, think again. The Fluke 87-V, probably the most popular high-performance meter available, has a burden voltage of 1.8mV/mA (which is still pretty good). So the above numbers are even worse – a 360mV drop for a 200mA current. Sure, you can switch up a current range, using the 10A jack, with its burden voltage of say 10mV/A, giving you a very nice drop of only 2mV. But your display is now showing 0.200 or 0.20 instead of 200.0 – you’ve just lost a valuable digit or two of resolution. The higher 10A current range is likely to be much less accurate than the mA range too! Let’s now take a look at the quoted burden voltage of some typical multi­ meters – see Table 1. As shown, things can improve a bit with the more expensive meters, particularly on the μA ranges. But an expensive precision meter is by no means a guarantee of a low burden voltage. Even many topof-the-line bench meters can have unacceptable burden voltages for many applications. It should be noted that while some meters will have a fixed burden voltage for all mA ranges, others like the Meterman 30XR have individual April 2009  59 S1b nA nA A A mA C1 100nF R12 100 S1a mA 1 3 7 IC1 2 R9 100 6 4 + R3: 75k* CURRENT INPUT R2 10k* R1 0.01  0.5% R8 10 * VOLTAGE OUTPUT R11: 24k* –1.5V R5 1k* – + – C3 100nF * = 0.1% TOLERANCE S2b S2a 3V LITHIUM BATTERY (2032) +1.5V 1 IN IC3 OUT 2 1 R4 470 GND 3 C2 100nF R6 100k A LED1  3 R7 100k 5 IC2 4 R10 100 2 K IC1: MAX4239ASA+ IC2: LMV321AS5X IC3: TPS3809L30DBVR LED1 SC 2009 MICROCURRENT DMM ADAPTOR A K Fig.1: the circuit is based on IC1, a Maxim MAX4239 ultra-low offset/drift, low-noise precision amplifier. IC3 is a voltage monitor while voltage follower stage IC2 provides a virtual ground reference for the circuit. specifications for each range; ie, 2mA range = 100mV/mA, 20mA = 13mV/ mA and 200mA = 4.6mV/mA. Some popular and highly regarded meters like the Meterman 37XR and Fluke 79 are particularly bad on their mA range, an order of magnitude worse than some cheaper meters – so beware. Taking the above example again, the Meterman 37XR would drop a whopping 2V (10mV x 200) on its mA range for 200mA. This will not be much good when your supply voltage is only 3.3V, 5V or even 12V. And the 37XR is a relatively expensive 10000-count meter that is supposed to be capable of measuring 999.9mA on its 1A range – which it will try to do. But that would be a gigantic 10V drop which the meter itself cannot even handle, so it’s limited to a nominal 400mA with a 4V drop on that range. Crazy huh? By now you should understand that burden voltage can be a real hidden problem lurking in your meter. What is your meter rated at? Accuracy And the second problem we men60  Silicon Chip tioned? That would be one of accuracy or lack of it. Most multimeters have a much poorer accuracy specification for current than for the DC voltage ranges or the “Basic DC Accuracy” as it’s called. The Meterman 37XR, for example, is quite an accurate meter at ±0.1% (+5 counts) on DC volts and is sold and marketed as such. But its current accuracy is a not so impressive ±0.5% (+10 counts) on DC current and ±1.5% for the 10A range. An even better example is the Fluke 27, with ±0.1% (+1 count) DC volts accuracy and ±0.75% (+2 counts) mA/μA DC current accuracy. Other multimeters are very similar, with a factor of five or more between the DC volts and DC current accuracy being quite typical. This issue applies to the AC voltage vs AC current ranges as well. Some meters can actually have very poor AC current accuracy and/or reduced AC frequency response compared to their AC millivolt range. Take the Fluke 27 again as an example. Its ACV accuracy is ±0.5% (+3 counts) to 2kHz but the AC current range is considerably worse at ±1.5% (+2 counts) to 1kHz. μCurrent adaptor is the solution You guessed it, the project presented here presents a neat solution to these issues. The “μCurrent” (pronounced “micro current”) is a simple yet accurate professional grade precision amplified current adapter for multi­ meters. It provides up to a 100-fold reduction in burden voltage for a given current range! An additional feature is a nanoamp (nA) current range. This gives any cheap 3.5-digit multimeter the ability to resolve 0.1nA (100pA). On a 4.5-digit multimeter it will resolve 0.01nA (10pA). And this comes with an excellent accuracy of <0.2%. In most cases, μCurrent is also able to improve your meter’s current range accuracy by using your meter’s more accurate mV DC voltage range to display the DC or AC current. (Yes, yes, we know that AC current is a tautology but what else can you call it?) For AC, the frequency response extends up to 10kHz although the siliconchip.com.au circuit’s THD (total harmonic distortion) increases substantially above 2kHz. This is still a very respectable AC response range, surpassing that of many digital multimeters on current and voltage ranges. Typical accuracy of the μCurrent itself is better than 0.2% on the μA and nA ranges, and 0.5% or better on the mA range. Unfortunately, it is not easy to obtain a 0.1% precision shunt resistor for the mA range, as the 10milliohm value is too low. The burden voltage of the μCurrent is a fixed 10μV/µA and 10μV/nA on the lower ranges. It varies on the mA range due to the switch resistance but 70μV/ mA is a nominal upper figure. These figures are unmatched by almost any meter on the market. So, for example, at a full scale of say 1000μA, that’s a maximum burden voltage of only 10mV. So measuring the current rail of a 1.2V logic supply with full-scale resolution would give you a worst case drop of around 0.8%, a fairly insignificant figure. The output voltage in mV is directly proportional to the input current, so you can simply read the current value from your multimeter’s mV DC range. The μCurrent thus effectively eliminates burden voltage by making it insignificant in all but the most extreme applications. How it works A current adapter is basically just a shunt resistor with an amplifier. But there are a few extra neat features to the μCurrent design to make it as professional and handy as possible, as we’ll see. The full circuit is shown in Fig.1. The heart of the design is IC1, a Maxim MAX4239. This is a special “ultra-low offset/drift, low noise precision amplifier”. As the name suggests, it’s a pretty high-spec device. The key figure in this application is its nearzero offset voltage. It’s not just “low offset” like many precision op amps; this one has almost no practical offset voltage at all. It is typically 0.1μV, with a maximum figure of 2.5μV over the entire temperature range. This class of op amps is known as an “auto-zero” (or “chopper”) amplifier. Maxim is a bit hush-hush on the actual internal workings of their particular device, saying only that “these characteristics are achieved through siliconchip.com.au VosB Vin+ VOUT AB Vin– VnB φB VosA AA φA VnA φA φB C M1 C M2 EXTERNAL FEEDBACK Auto-Zero Phase A: Null amplifier nulls its own offset Fig.2(a): how a basic auto-zero amplifier works. In the first phase, the main amplifier (AB) is offset with the voltage stored on capacitor CM2. The nulling amplifier (AA) measures its own offset voltage and stores it on capacitor CM1. VosB Vin+ VOUT AB Vin– VnB φB VosA AA φA VnA φA C M1 φB C M2 EXTERNAL FEEDBACK Auto-Zero Phase B: Null amplifier nulls the main amplifier offset Fig.2(b): in the second phase, the nulling amplifier (AA) measures the input difference voltage on AB and stores this value on capacitor CM2, ready for the next cycle. a patented auto-zeroing technique that samples and cancels the input offset and noise of the amplifier. The pseudo-random clock frequency varies from 10kHz to 15kHz, reducing intermodulation distortion present in chopper-stabilized amplifiers”. However, we can get a good idea of how a basic auto-zero amplifier works by referring to Fig.2. An auto-zero amplifier is basically the combination of a normal op amp (AB) with a “nulling” op amp (AA) that continually corrects for the DC offset voltage of the main amplifier. The device is driven by an internal clock that drives a 2-phase offset process. In the first phase, in Fig.2(a), the main amplifier (AB) is offset with the voltage stored on capacitor CM2. The nulling amplifier (AA) measures its own offset voltage and stores it on capacitor CM1. In the second phase, in Fig.2(b), the nulling amplifier (AA) measures the input difference voltage on AB and stores this value on capacitor CM2, ready for the next cycle. This process continually eliminates the offset voltage of the main amplifier. A side benefit of this is that it also eliminates typical op amp 1/f noise, as the low frequency is treated as a slowly varying input offset voltage and hence gets cancelled out. The pseudo-random clock used in the MAX4239 also helps to reduce the effects of intermodulation distortion as April 2009  61 TPS3809 VDD R1 RESET LOGIC + TIMER R2 RESET GND OSCILLATOR 1.137V REFERENCE VOLTAGE Fig.3: inside the TPS3809L30 Supply Voltage Supervisor. Fig.4: the discharge curves for the 3V lithium battery specified (CR2032), using a number of different loads. AC signals approach half the chopping frequency (10-15KHz). This remarkable DC performance allows the μCurrent to have insignificant output offset error. As a result, it will display 0V output for a zero current input. It is also quite a low power device, drawing around 600μA with a supply voltage specified down to 2.7V. This makes it ideal for operation from a single 3V lithium battery. The MAX4239 also has a companion device, the MAX4238. The only difference is that the MAX4239 is a high bandwidth “decompensated” version of the MAX4238. The MAX4239 requires a minimum gain of 10 which we have in this circuit, so it’s better to use the higher bandwidth device. If you want to use the MAX4238 then that is possible without any circuit changes, only the bandwidth and other AC performance measurements will differ. A fixed gain of 100 is defined by 62  Silicon Chip precision resistors R5 and R3+R11. These are 0.1% resistors with negligible temperature drift. The 100Ω resistor R9 at the output of IC1 ensures stability. This value will be low enough to ensure error-free operation with multimeters having greater than 100kΩ input impedance. If for some reason your meter is lower than this, than you’ll have to lower the value of R9 appropriately. Current ranges There are three current ranges that are defined by the shunt resistor on each range, together with the gain of IC1. R2 (10kΩ 0.1%) is the shunt resistor for the nA range and is permanently connected across the input terminals. It gives a burden voltage of 10µV/nA (1nA x 10kΩ). The other shunt resistors R1 and R8 are disconnected in the nA range. R2 is permanently connected, ie, not switched, to ensure that the input is not left open-circuit. R8 (10Ω 0.1%) is switched in parallel with R2 in the µA range by S1b which gives a burden voltage of 10μV/ μA (1μA x 10Ω). R2 contributes a small error of less than 0.1% in this case. It can be ignored. R1 (10mΩ 0.5%) is switched in parallel with R2 in the mA range by S1b which gives a (resistor) burden voltage of 10μV/mA (1mA x 10mΩ). Because R1 is such a low value, the solder joints and the copper tracks of the PC board can contribute large errors, so a special purpose-designed “shunt” resistor is used. This is a 4-terminal device that includes the 10mΩ resistor and two “sense” terminals connected directly across the resistor on the substrate. This eliminates any errors caused by solder joint or copper track resistance. However, because the 10mΩ shunt resistor is such a small value compared with the resistance of the range switch, the switch itself will dominate the actual total burden voltage. The switch contact resistance is rated at 70mΩ maximum, so the actual burden voltage on the mA range will vary from unit to unit and will change with time, but can be taken as a nominal 70μV/mA. The maximum current in the mA range is a nominal 300mA, as this is the contact rating of the switch. But in practice it can be higher than this. You will notice that the virtual ground is connected to the sense side of R1. This means that the sense currents for R2 and R8 also flow through this terminal but these currents are negligible and so they have virtually no effect. The switch contacts of S1a select which shunt resistor voltage gets fed through to op amp IC1. Power supply Any current adapter must be able to handle both positive and negative inputs and so a dual-polarity power supply is required. In a battery-powered device, this can be achieved in one of three ways. The first way is by using two or more series batteries to a middle “0V” tap. This method is convenient but takes more space, there are more batteries to replace and you can get uneven current drain from the batteries, thus making true low-battery detection more difficult. The second way is by using a single siliconchip.com.au Specifications Three current ranges: (1) ±0-300mA (70μV/mA burden voltage typical) (2) ±0-1000µA (10μV/uA burden voltage) (3) ±0-1000nA (10μV/nA burden voltage) Output Voltage Units: 1mV/mA; 1mV/μA & 1mV/nA Resolution (nA range): 100pA (3.5-digit meter), 10pA (4.5-digit meter) Accuracy (typical): <0.2% on μA and nA ranges, <0.5% on mA range Output Offset Voltage: negligible on 4.5-digit meter Bandwidth: 2kHz nominal (±0.1dB) Temperature Drift: insignificant over normal ambient range Noise: < -90dBV THD: < -60dB Battery: CR2032 lithium coin cell Battery Life: >200 hours (LED OFF); >50 hours (LED ON) Connection: 4mm banana, screw terminal inputs, standard 19mm spacing battery supply and generating a negative supply using a switched capacitor inverter. This is convenient for low current applications but it generates noise and requires filtering. Also, using a 3V lithium battery means a total power supply voltage from 5.4V to over 6V. But our MAX4239 can only handle a maximum 5.5V supply voltage, so extra diodes would be required. The third method involves a “virtual ground” split supply circuit and this is the technique used in the μCurrent circuit. In effect, the two 100kΩ resistors comprise a voltage divider and this is buffered by op amp IC2 which is connected as a unity gain voltage follower to provide a low impedance output. However, the output impedance is increased by the series 100Ω resistor which has been included to ensure output stability. The output from the 100Ω resistor (R10) is now the “virtual ground” reference for the rest of the circuit. This ensures that IC1 has a ±1.5V supply from the battery and the input current shunt resistors can now sense current in either direction. IC2 is an LMV321 general-purpose, low-power, low-voltage op amp (essentially a low-voltage version of the venerable LM351). The total current drain for this portion of the circuit is about 145μA. Low battery detection To ensure that what you read on your multimeter is accurate, it is imsiliconchip.com.au portant to know if the battery voltage is low and thus possibly affecting the measurement. IC3, a Texas Instruments TPS3809L30 Supply Voltage Supervisor, does this job accurately in a single chip. It contains a precision resistor divider, a voltage reference and an output circuit with timer (Fig.3). If the input voltage on the VDD pin drops below 2.64V then the Reset-bar output will go low. In our application, Reset-bar will be high and thus the BATT LED will be on if the battery voltage is above 2.64V. Conveniently, this is about the “end point” for a 3V lithium coin cell. The discharge diagram for the lithium battery, using a number of different loads, is shown in Fig.4. By using the same type of 2-pole 3-position switch used for the current range selection, we are able to get a very handy “battery check” mode between the ON and OFF modes, to switch in IC3 to light the LED. You can keep using the μCurrent in this mode with the LED ON if desired but it does use more battery power. The in-built timer will take about 0.2s to light the LED, so it’s possible to move the power switch through the BATT CHECK mode and not have the LED light if you are quick enough. Output voltage range The MAX4239 is capable of swinging its output fairly close to the supply rails. Given that the power supply will be at least ±1.35V for a working battery, this means that the output voltage Parts List 1 μCurrent double-sided screenprinted PC board, 79 x 50mm 1 UB5 plastic box, 83 x 53 x 28mm 1 CR2032 3V lithium cell 1 1060TR CR2032 SMD battery holder 2 miniature 3-position PCmount slide switches, C&K JS203011AQN 1 4mm black banana jack 1 4mm red banana jack 1 4mm black binding post 1 4mm red binding post Semiconductors 1 MAX4239ASA+ SO8 op amp (IC1) 1 LMV321AS5X SOT23-5 op amp (IC2) 1 TPS3809L30DBVR SOT23 voltage monitor (IC3) 1 LTST-C230GKT 1206 reverse green LED Capacitors 3 100nF 0805 capacitors Resistors 2 100kΩ 1% 0805 1 75kΩ 0.1% 0805 1 24kΩ 0.1% 0805 1 10kΩ 0.1% 0805 1 1kΩ 0.1% 0805 1 470Ω 1% 0805 3 100Ω 1% 0805 1 10Ω 0.1% 0805 1 LVK12R010DER 10mΩ 0.5% 1206 (current sense) Where To Buy This design is copyright to the author. Both kits and fully-built units are available from the author at: www.alternatezone.com/electronics/ucurrent can approach this figure within a few millivolts. Normally though, the μCurrent will be used with your multimeter’s mV range which will be typically up to a maximum of 999.99mV for a 10000-count meter. So there is some headroom left if you want to push it higher for any reason. Output units The output units are scaled by the shunt resistors and gain of IC1 to be precisely 1mV per range unit. So the April 2009  63 VOLTAGE OUTPUT 101 104 1mV/mA (10m ) 102 CR2032 BATTERY IN SMD HOLDER 101 104 IC1 104 104 1mV/nA (10k ) 471 471 103 R010 1mV/ A (10 ) + S1 + CURRENT INPUT – (FRONT PANEL SIDE) – IC2 104 104 IC3 102 101 CURRENT INPUT 100 uCurrent BATT OK – 104 753 – OFF 243 753 ON & BATT CHECK 104 104 243 ON K LED1 101 104 101 S2 A 103 100 – R010 – 101 VOLTAGE OUTPUT + + + (REAR/COPPER SIDE) Fig.5: install the parts on the PC board as shown here. You will need a soldering iron with a small chisel-point tip to solder the SMD devices to the board, along with a pair of fine-pointed tweeters and some fine solder. as measured with an Audio Precision analyser with a 1V output level on the μA range. There is little performance difference between the ranges. The nominal bandwidth is 2kHz, as the THD starts to increase exponentially after this. This figure is quite sufficient as most meters have a response 1kHz on AC current ranges. Overloads The top of the PC board forms the front panel and is attached to a UB5-size utility case. output will be 1mV per mA, 1mV per μA or 1mV per nA. This makes it easy and logical to directly read on your multimeter’s mV range. So if you read 100mV on your meter, that equates to 100mA, 100µA, or 100nA, depending on the range you have selected. AC performance The AC performance is shown in the accompanying screen shots (Figs.6 & 7) 64  Silicon Chip Fuses have been omitted from the design to ensure as low a total burden voltage as possible. Therefore you must be careful to ensure that the input is not connected directly across a supply voltage capable of providing a current that exceeds the selected range. Failure to take care here can result in a blown shunt resistor. Connectors The connectors are standard 4mm banana plugs, with standard 19mm spacing. This allows the use of various types of adapters if required. The screw-terminal type connectors are used for the current input, which is convenient for connecting to existing wiring without test leads. The top screw part can be completely removed to enable some short “shrouded” banana plug test leads to fit. Construction Apart from the connectors and min- iature slide switches, the entire design uses surface-mount components. This was done in order to give a professional look and to reduce cost and size by using a standard UB5 utility box. The double-sided PC board is used as the lid and front panel of the box. Its red solder mask on the topside provides a very elegant and durable appearance. The shield plane on the top layer is connected to VGND. All the SMDs are relatively large 0805, SO and SOT packages, so soldering is pretty easy using a basic iron. Refer to the March 2008 issue of SILICON CHIP for a detailed article on how to solder surface-mount components, if you are new to this. There are a few things that make SMD hand-soldering much easier: a small chisel point tip (not conical), fine multi-core solder (0.56mm or better) and a pair of fine-pointed tweezers. Start with the three IC packages, making sure each one is mounted with the correct polarity. Follow these with the resistors and capacitors, taking care not to damage the precision resistors with excess heat. Applying a small amount of solder to one pad first makes it easy to “reflow” the component into place while you solder the other end. Next, solder in the LED. This is a special “bottom emitter” LED which is effectively soldered in upside down, siliconchip.com.au with the light coming through a hole in the board. Be sure to match the polarity to the silkscreen. Next, solder the battery holder into place, ensuring the correct polarity. Apply the iron and then solder to the topside of the flat pin instead of the pad for this part. The solder should then reflow easily to the pad underneath. Now turn the board over and install the two miniature slide switches, again ensuring correct orientation. If you have the vertical switches, then the side with the metal indent should face to the outside edge of the board. Side mount switches should have the switch lever towards the middle of the board. Ensure that the switches are flush with the board and straight, then tack one pin down first. Check that everything is OK before soldering the rest. Finally, install the banana connectors. Unscrew them completely first, removing all nuts, washers and solder tags. Install them on the topside with just the plastic spacers touching the topside of the PC board. Next, put the solder tag on the bottom side and solder it only to the smaller adjacent solder pad, then place the washer and screws on top and tighten. Feel free to add a thread-locker and/or glue if desired. Fig.6: this Audio Precision spectrum plot shows the residual noise of the μCurrent Adaptor circuit. Testing Testing is fairly straightforward. You will need a power supply, some suitable resistors and your multimeter. Insert the battery and switch to BATT CHECK mode. The LED should light within 0.2s. Switch to ON mode and the LED should turn off. Measure the DC voltage from the negative output connector (VGND) to first one then the other side of the battery in order to check the split supply system. You should get approximately ±1.5V and both values should match closely. Next, connect the Voltage Output terminal to your multimeter and set the multimeter to its mV DC range. With nothing connected, you should get a reading of zero on all three current ranges. The next step is to select a resistor for each range to give you a decent current level, eg, around half the meter’s full scale. For example, for a 5V supply, use a 47Ω 1W resistor (106mA), a 47kΩ resistor (106μA) and five 10MΩ siliconchip.com.au Fig.7: although largely of academic interest, this Audio Precision plot shows the THD vs frequency of the μCurrent Adaptor at a signal level of 1V. resistors in series (100nA). That done, connect the test resistor in series with the supply and the Current Input terminals. Ensure that you have the correct range selected before switching on your supply voltage – you don’t want to blow any shunt resistors! Your meter should read approximately 106mV (mA), 106mV (µA) and 100mV (nA) for the values mentioned. You can double-check your values by measuring the actual resistor values and supply voltage and calculating the current if desired. If these currents match, then your μCurrent is ready for operation, as the calibration is inherent within the precision 0.1% components used. The output value should not differ between BATT CHECK and ON modes. It might be handy to check the battery current also. It should be around 0.7mA with the LED off and around 3mA with the LED on. Don’t forget to switch off when you are finished measuring. The last step simply involves screwing the PC board onto the box. With typical infrequent use, the battery should last many years. That’s all there is to it. You now have a precision current measurement tool ready for those more demanding applications. We hope this article has got you thinking about the impact burden voltage can potentially have on current SC measurements. April 2009  65