Fullscreen HTML5Med Res (??MB)HTML4

High-Bandwidth Differential Probe This high-bandwidth, high-voltage differential probe is ideal for use with oscilloscopes, although it could have other uses. It has an internal rechargeable battery and fits in the same case as the Isolated Current Probe we published last month. It will be an invaluable addition to your test equipment arsenal! By Andrew Levido I f you ever work with high-voltage circuits, a differential probe is an indispensable piece of test equipment. In fact, they’re also useful with many low-voltage circuits; any time you want to monitor a differential voltage between two points in a circuit. This one can be built for a fraction the cost of a commercial device with similar performance and functions. The ground sides of most oscilloscope inputs are connected directly to mains Earth. This means you can only measure Earth-referenced signals – either those already referenced to Earth, or those that you can safely connect to Earth on one side for the purposes of the measurement. That generally includes truly floating circuits, such as battery-powered devices. Unfortunately, many signals in circuits such as switch-mode power supplies or motor controllers are referenced to voltages well above Earth potential. Connecting a scope to these using a standard probe would create a short from the circuit reference to mains Earth, via the probe ground lead and the ‘scope itself. This will potentially be catastrophic for your scope, the probe and your circuit. Even if your circuit is floating and you can safely Earth one point for testing, if you want to measure another voltage at the same time that’s referenced to a different point, you’re out of luck. That’s because if you Earth two different points in your circuit, you are adding a short circuit; usually not a great idea! A differential probe (or multiple probes) totally solves that problem. As an interesting and slightly terrifying aside, my very first oscilloscope, an Australian made BWD830 purchased in the early 1980s, actually has a “ground isolate” switch on the back panel that allows the user to open the mains Earth connection, allowing the scope common to float. Fig.1: a high-voltage differential probe is essential if you want to see signals that cannot be Earth referenced on your oscilloscope. In this example, three probes help to measure the phase-to-phase voltages of a variable speed drive. The scope display is a real capture made with the prototypes. 32 Silicon Chip Australia's electronics magazine siliconchip.com.au ● Maximum common-mode voltage: ±400V DC (280V RMS) ● Maximum differential-mode voltage: ±400V DC (280V RMS) ● Common-mode input impedance: 2MΩ || 2.5pF ● Differential-mode input impedance: 4MΩ || 2.5pF ● Attenuation ranges: 100:1, 10:1 ● Basic DC accuracy: better than 1% ● Bandwidth: >30MHz (x100), >25MHz (x10) ● CMRR: >100dB (DC-100Hz) ● Battery Life: >4 hours ● Charging time: <3 hours ● Charging socket: USB-C ● Input sockets: 4mm banana sockets, 20mm spacing ● Output socket: BNC This avoids the risk of blowing up the scope, but can allow the scope case and front panel terminals to rise to lethal voltage levels! Thankfully, this dangerous practice is a thing of the past. (And it still doesn’t help for monitoring multiple points referenced to different voltages anyway...) Fig.1 shows an example of where a differential probe is indispensable. Here, the three phase-to-phase PWM output waveforms from a variable speed drive (suspiciously similar to the one we published in the November & December 2024 issues) are displayed on three channels of an oscilloscope. None of the U, V or W phases can be safely Earthed, and the voltages involved are in the order of 400V peakto-peak. The differential probes provide 100:1 attenuation of the differential voltage and over 10,000:1 (100dB) attenuation of the common-mode voltage, allowing the phase-to-phase voltages to be measured safely. The waveforms shown on the scope are from a real screen capture made using three of these devices. A high-voltage differential probe translates the difference in voltage between its two high-impedance inputs into a voltage that you can safely connect to your oscilloscope’s Earthed input. The output is proportional to the difference in voltage between the positive and negative inputs. Any common-mode signal present on both inputs is almost entirely rejected. The differential probe is housed in a small plastic case measuring 82 × 65 × 28mm. The inputs are two shrouded 4mm banana jacks at one end, with a 20mm spacing. That is close enough to the 3/4-inch (19.05mm) standard for siliconchip.com.au dual-banana-plug accessories to fit. The BNC output, range switch and USB-C connector are at the other end of the case. The power and charge LEDs are visible through the top of the case via two light pipes. Design goals When I set out on this project, I set myself a few design goals. I wanted a probe that could safely be used in mains-voltage projects like that described above. This means the device should be able to measure differential-mode signals of ±400V magnitude and withstand a similar level of common-mode voltage. This corresponds to an AC voltage of 280Vrms. We want to show these on a standard scope, so an attenuation of 100:1 (-40dB) would be appropriate to give a ±4V full-scale output signal. Sometimes, we will want to measure a low-voltage signal riding on a high common-mode voltage; for example, to examine the gate signals of the IGBTs in Fig.1. These signals would normally be within a ±40V range, so a 10:1 (-20dB) attenuation range was also one of my requirements. The common-mode rejection ratio (CMRR) at DC to mains frequency should be at least 100dB. This means a 400V common mode voltage would contribute less than 4mV at the output. The input impedance should be in the megohm range with fairly low parallel capacitance (say <10pF). I wanted an upper bandwidth limit as high as I could reasonably get, at least 25MHz, to get good representation of high-speed switching signals with fast rise-times. Bandwidth and rise time are related according to the Australia's electronics magazine approximation trise ≈ 0.35/BW, so a 25MHz bandwidth means the fastest rise time we will see is about 14ns, which should be short enough. I also wanted the unit to have the smallest form factor possible and include an internal rechargeable battery. My bench gets cluttered enough as it is without having bulky probes and their power cables added to the mix. More than three hours’ battery life and USB-C recharging was mandatory. Operating principles In principle, the concept of a differential probe is pretty straightforward: a matched pair of input attenuators followed by a classic three op-amp differential instrumentation amplifier will do the job. Fig.2 shows the bare bones of the circuit, along with differential and common-mode voltage sources we will discuss later. You can think of this circuit has having three sections: a dual input attenuator, a buffer stage and a difference amplifier stage. The overall differential-­mode gain of the circuit is given by multiplying the gains of each of these stages, which are given in the figure. We have to set the gains of each stage such that we respect the input common-mode voltage range of each op amp (voltages A+/A- and C+/C- in the figure) and their maximum output swings (voltages B+/B- and Vout). With ±5V power rails, it seemed fairly safe February 2025  33 Fig.2: the differential probe consists of two matched attenuators followed by a classic three-op-amp instrumentation amplifier. The latter has a buffer stage with a gain programmable via a single resistor (RG) and a difference amp stage with a fixed gain. to assume an input common-mode voltage of ±2V and an output swing of ±4V as a starting point. A division ration of 200:1 would give 2V at point A with a 400V input, leaving the rest of the circuit to provide a gain of 2 or 20 to achieve the overall target of 100:1 or 10:1 attenuation. As you can see from Fig.2, the voltage at any one of the inputs will actually be a combination of some common-mode voltage, Vcm, plus one half of the differential-mode voltage, Vdm. The maximum voltage of 400V at the inputs will therefore be made up of a combination of common-mode and differential-mode voltages. We can construct a graph (the yellow area in Fig.3) showing the allowable ranges of input voltage that keep the op amp voltage within the ±2V band. This input range is more than enough to measure signals likely to be encountered in a circuit powered by 230-240V AC. The area shown in pink is the combination of inputs that can be measured on the 10:1 attenuation range. In this case, the range is limited by the ±4V output swing of the op amps, rather than the input common-mode voltage. It is important to keep in mind that all of these are limits relate to the faithful reproduction of the input signal. The maximum voltage that the inputs can safely withstand is considerably higher, as we shall see. With the attenuator gain determined to be 1/200th, we can consider the gains for the other two stages. The buffer stage gain can be set by selecting a single resistor RG, so this is the obvious candidate for switchable part of the gain. We can’t put all the remaining gain in this stage, or we run the risk of exceeding the difference amplifier’s 34 Silicon Chip common-mode input voltage range. Thus, I chose to make the buffer stage gain switchable between 1 and 10 and set the difference amp stage gain to a fixed value of two times. That’s about it for the high-level design – a pair of matched 200:1 attenuators, a ×1/×10 switchable buffer stage and a ×2 difference amplifier. Now we just have to make it all work – and the devil is in the details, as they say. The attenuator The circuit diagram (Fig.4) shows the complete design. The attenuators have to withstand high voltages, have reasonably high input impedance and be very closely matched to maximise CMRR. The attenuators are identical, so I will focus this description on the positive side for simplicity. The resistors I have chosen for the Fig.3: the range of common-mode voltage and differential-mode voltages the probe can faithfully reproduce at the output. The yellow area is for the ×100 range and the pink area is for the ×10 range. It can safely tolerate much higher voltages without damage. Australia's electronics magazine upper leg of this divider are 1MW ±0.1% ¼W devices with a voltage rating of 700V. The maximum continuous voltage we can apply across each of these resistors is limited to 500V by power dissipation. With two resistors in series, the inputs can withstand a sustained voltage of 1kV (DC or AC RMS), giving a comfortable safety margin. The resistance value required in the bottom leg of the divider for 200:1 attenuation is 10,050W. In each half, this is made up of a 10kW resistor, a 10W resistor and half of 100W trimpot VR1. This trimpot is shared with the negative attenuator, allowing us to tweak the divisor ratios so that they are precisely equal, as necessary for maximum rejection of common-mode signals. With VR1 centred, the total resistance of each resistor string is 10,060W, not 10,050W as calculated. The extra 10W is necessary to compensate for the 10MW resistors, which are effectively in parallel with the lower leg of each divider. You can ignore trimpot VR2 in this calculation, since its value is much smaller than the error due to the 1% tolerance in the value of the 10MW resistors. The overall resistance of the lower leg of each divider is therefore 10,060W || 10MW = 10,050W. The purpose of VR2 is to allow us to inject up to ±5mV into one input to compensate for any op amp offset errors. We will discuss this further below. Diode pairs D1 & D2 protect the op amp inputs from overvoltage by limiting the voltage swing at the divider output to ±5.6V or thereabouts. That covers the DC performance of the attenuator, but we want the divider to operate properly up to 25MHz or more. We know that there will inevitably be some capacitance at the output of the divider. The protection diodes, for example, will contribute about 1.5pF each; the op amp input capacitance will be about the same. There will also be 3pF or 4pF of stray capacitance inherent in the layout. At 25MHz, this ~10pF of total capacitance will have an impedance of around 630W, reducing the divider ratio to something in the order of 1/3500. The incidental capacitance is moreor-less unavoidable, so we potentially have a real problem. The solution is to deliberately add some capacitance across the upper leg siliconchip.com.au of each divider to reduce its impedance by the same ratio and maintain the attenuation. With a 200:1 divider, we would need an upper leg capacitance 199 times lower than the ~10pF in the lower leg. Clearly, this is impractical. Instead, we put a small known value of capacitance across the upper leg and add more capacitance across the lower leg to compensate for it. I selected series pairs of 4.7±0.1pF 1kV NP0 capacitors for the upper legs, to match the high-voltage tolerance of the input resistors. Together, they amount to 2.35pF of capacitance in the upper leg of each divider, requiring 467pF of capacitance in the lower leg to compensate. This latter capacitance is made up of the ~10pF of incidental and stray capacitance we have already mentioned, plus the parallel combination of 390pF and 27pF fixed capacitors, plus VC1 (12-60pF). This combination gives us a range of capacitance adjustable from nominally 440pF to 490pF. It is useful to have a range to account for capacitor tolerances and other uncertainties. Moreover, the overall bandwidth we can ultimately achieve will be quite sensitive to perfect frequency compensation. The buffer stage The op amp we use for the buffer stage is critical. It must have high input impedances so as not to load the attenuator, and low bias currents since the input impedance is ~10kW. Thus, a FET input op amp is required. It must also have a high large-­signal bandwidth, and a common-mode input range of ±2V with ±5V supplies. I chose the ADA4817, which is expensive at around $15 each, but it fits the bill nicely. It has an input impedance of 500GW in parallel with 1.3pF and the input bias current is ±20pA. The large signal bandwidth extends to 200MHz, with 0.1dB gain flatness to 60MHz. The worst-case offset voltage is ±4mV (which is good for a FET input op amp), and the input common mode voltage range is -4.2V to +2.2V with ±5V rails. If I only required a gain of one for this stage, I could have simply wired IC1 and IC2 as non-inverting buffers. But since we need the option of a gain of 10, I had to close the feedback loop around each op amp with resistors. It is a good idea to choose a fairly low value for this resistor as it will form an RC low-pass filter with the op-amp’s input capacitance, the effect of which will be to increase the gain of the buffer as the frequency rises, causing unwanted ‘peaking’ in the frequency response. When the ×10 range is selected via S1, the parallel combination of the 110W and 220W resistors is switched in between the two buffer amplifiers’ inverting inputs. The resistance values were chosen to give this stage a gain of 10 in this configuration. Consistent with the attenuator, I used 0.1% tolerance resistors for gain-setting. Fig.4: the complete probe circuit. Power is provided by an 800mA Li-ion cell via a dual-rail DC-to-DC converter (REG5). The battery is charged via a USB Type-C connector (CON4) and IC4. siliconchip.com.au Australia's electronics magazine February 2025  35 You can see the input attenuator components arranged vertically outside the banana sockets near the top. The 510W resistors in series with the non-inverting inputs of IC1 and IC2 are critical to the stability of the circuit. High-speed op amps like the ADA4817 love to oscillate. One of the (many) things that can bring this on is extraneous capacitance on the inputs, and we have plenty given the compensation network we just discussed. The 510W resistors are ‘stopper’ resistors that isolate the op amp inputs from this capacitance. The 500GW input impedance and 1.5pF input capacitance mean that these resistors don’t otherwise affect the operation of the probe. Just as for the input divider, we add 10pF & 47pF frequency compensation capacitors to this gain stage. I did not bother with a variable capacitor here because the low impedance of the surrounding circuit makes it less sensitive to an error of a few picofarads one way or the other. Difference amplifier The requirements for the difference amplifier (IC3) are not quite as stringent as for the buffers, but we do need a high large-signal bandwidth and good output characteristics. The LMH6611 fits the bill. It has a large signal bandwidth of 85MHz and a gain-bandwidth product of 115MHz. The output swing with ±5V rails is ±4.5V into a 150W load and the output drive current is ±120mA. 36 Silicon Chip The LMH6611’s input common-­ mode voltage range is -5.2V to +3.8V, giving plenty of headroom. As a bonus, it is considerably cheaper than the ADA4817s. IC3 is set up as a difference amplifier with a fixed gain of two using low-value 0.1% tolerance resistors. The 10W resistor helps overall stability by providing a little bit of isolation between the LMH6611’s output and any load capacitance. This stage does not need frequency compensation due to the low gain and low impedances involved. My design calculations indicate that the end-to-end gain error of this circuit should be comfortably under 1% over the temperature range of 0-40°C, and nearer to half this at 25°C. However, the untrimmed offset error could be in the order of ±5mV on the ×100 range and ±45mV on the ×10 range. The big difference is due to the buffer stage amplifying the ADA4817’s offset when on the ×10 range. This is why it is necessary to add the offset trim. If we added the offset to the difference amplifier (where we would in ideal world), we would need a different offset trimpot for each range and an extra gang on the range switch to select the right one. The compromise I selected was to add the offset before the gain stage, meaning we can trim out most of IC1’s and IC2’s offset but may not be able to fully eliminate that from IC3. Since this is ±4mV at the output (0.1% error) in the worst case, I decided I could live with it. Power supply The ±5V power supply is derived from a single Li-ion 14500 (AA-sized) 800mAh cell via a TPS65133 dual-rail switching power supply (REG5). This chip accepts a 2.9-5V input and can source up to 250mA on each rail. It is Fig.5: use this overlay diagram to place the components. We recommend mounting the LCC-packaged DC-DC converter (REG5) and supporting components first. Once you have confirmed they are working, you can move on to the rest of the parts. Australia's electronics magazine 92% efficient at 100mA and requires only a couple of inductors and three capacitors to operate. The chip has an undervoltage lockout to protect the Li-ion cell from over-discharge. The TPS65133 generates very little noise as far as switching converters go, but to be safe, I added an LC filter (10μH/220μF) between each output of the switcher and the analog circuitry. A green LED (LED2) across the power rails provides user indication that the power is on. The cell is charged from a USB-C power-only connector via a MAX1555 charger chip (IC4). This charges the cell at around 280mA, which is enough to charge an empty cell in under three hours. A yellow LED (LED1) lights when the MAX1555 is charging the Li-ion cell and extinguishes when it is fully charged. The charging voltage comes from USB-C connector CON4. Resettable PTC fuse PTC1 and transient voltage suppressor diode TVS3 protect against reverse-polarity circuits and overvoltage conditions. The two 5.1kW resistors pull the USB power delivery control channel lines down to passively signal to the source to supply 5V. The power is switched via a second set of contacts on range switch S1. In the Off/Charge position, the cell is connected to the charger and isolated from the rest of the circuit. In either the 100:1 or 10:1 position, the battery is connected to the switcher and isolated from the charger. PCB design High voltages can be present at the banana sockets’ exposed conductors and either end of the first resistor and capacitor of each attenuator. Those components are on the PCB outside the banana sockets. The track clearances PCB follow IPC2221-B standard B4 for boards with solder masks below 3050m altitude. So you must build this on a commercially-­ made PCB with a solder mask. During assembly, you should apply a conformal coating over the top half of the board once all components other than trimpots/trimcaps have been fitted. That will allow it to resist arcing even under extreme conditions (eg, very high humidity). These coatings are available in spray cans (see the parts list), are easy to apply and can be soldered through, although they should be reapplied later if you do that. Construction All the components mount on a small double-sided PCB coded 9015-D that measures 56.5 × 82.5mm. Most are through-hole or hand-solder-friendly surface-mount types. The only really tricky device is REG5, the TPS65133 switch-mode regulator. Unfortunately, all the useful power chips like it seem to only be available in tiny ‘leadless’ packages. During construction, refer to the PCB overlay diagram (Fig.5) to see which components mount where and with what orientations. Because of REG5’s package, we recommend assembling and testing the power supply first. REG5 has a thermal pad underneath the chip, so reflow (either hot air or IR) is the only realistic option to mount it. The best way I have found to do this is to use solder paste. Apply a small smear of it to all the pads. Don’t worry if a little gets between pads as it will ball up under surface tension when reflowed. Place the chip carefully, using the screen-printed lines as a guide. Make sure the orientation is correct. Heat the chip and the surrounding board with hot air until the solder melts, including that on the thermal pad. I use tweezers to hold the chip in place until I feel the surface tension of the solder ‘pull’ it into place. You can tell the thermal pad solder has melted if the chip re-aligns itself if you nudge it very slightly out of siliconchip.com.au Parts List – High-Bandwidth Differential Probe 1 double-sided PCB coded 9015-D, with solder mask, 56.5 × 82.5mm 1 Hammond 1593LBK 92 × 66mm case [element14 4437858] 1 adhesive panel label, 55 × 80mm 1 14500-size 800mAh Li-ion cell with PCB pins (BAT1) [Altronics S4981] 1 red PCB-mount banana socket (CON1) [Cal Test CT3151SP-2] 1 black PCB-mount banana socket (CON2) [Cal Test CT3151SP-0] 1 PCB-mount BNC socket (CON3) [Molex 73100-0105] 1 USB-C power only socket (CON4) [Molex 217175-0001] 2 4.7μH 1.1A M2520/1008 shielded ferrite inductors (L1, L2) [Würth 74404024047] 2 10μH 350mA M2012/0805 shielded ferrite inductors (L3, L4) [TDK MLZ2012M100WT000] 1 0.75A 24V M3226/1210 PTC polyfuse (PTC1) [Littelfuse 1210L075/24PR] 1 right-angle DP3T PCB-mount slide switch (S1) [E-Switch EG2310] 1 top-adjust 100W 3296-style multi-turn trimpot (VR1) [Altronics R2370A] 1 top-adjust 10kW 3296-style multi-turn trimpot (VR2) [Altronics R2382A] 2 0.6in (15.24mm) convex light pipes [Dialight 51513020600F] 2 No.4 × 6mm self-tapping screws 4 small self-adhesive rubber feet 1 can of conformal coating [Jaycar NA1610, Altronics T3175] Semiconductors 2 ADA4817-1ARDZ-R7 410MHz precision op amp, SOIC-8-EP (IC1, IC2) 1 LMH6611MK/NOPB 135MHz precision op amp, TSOT-23-6 (IC3) 1 MAX1555EZK-T Li-ion battery charger, TSOT-23-5 (IC4) 1 TPS65133DPDR dual DC-DC converter, WSON-12 (REG5) 1 SMBJ5.0C 5V transient voltage suppressor, DO-214AA (TVS3) 1 SMD M2012/0805 yellow LED (LED1) 1 SMD M2012/0805 green LED (LED2) 2 BAV99 dual series signal diodes, SOT-23 (D1, D2) Capacitors (all SMD M2012/0805 size 50V NP0/C0G ceramic unless noted) 2 220μF 10V solid tantalum, SMC case 5 10μF 16V X7R 8 100nF X7R 2 390pF 1 47pF 2 27pF 2 10pF 4 4.7pF 1kV 2 6mm diameter 12-60pF variable capacitors (VC1, VC2) [EW GKG60015] Resistors (all SMD M2012/0805 size ±1% ⅛W unless noted) 2 10MW 4 1MW ±0.1% M3216/1206 size ¼W 700V [Vishay TNPV12061M00BEEN] 2 10kW ±0.1% 10ppm [element14 1140912] 2 5.1kW 1 1.8kW 3 510W 2 360W ±0.1% 25ppm [Panasonic ERA-6AEB361V] 2 330W ±0.1% 25ppm [Panasonic ERA-6AEB331V] 1 220W ±0.1% 25ppm [Panasonic ERA-6AEB221V] 2 180W ±0.1% 25ppm [Panasonic ERA-6AEB181V] 1 110W ±0.1% 25ppm [Panasonic ERA-6AEB111V] 3 10W position. Once it cools down, you can remove any excess solder or obvious shorts with solder wick around the edges (adding a bit of flux paste [not solder paste] makes the wick work better). Then clean up the flux residue with isopropyl alcohol. Australia's electronics magazine Next, fit the four inductors, L1– L4, the four capacitors around REG5 and the two large tantalum capacitors in the upper-left corner of the PCB according to the overlay. You are then ready to test the power supply. Solder a couple of lengths of fine February 2025  37 Fig.6: drill the enclosure end panels and top according to this diagram. The slots can be formed by drilling a pair of holes inside the perimeter and using a craft knife and files to open them up to the required dimensions. hookup wire to the board to power it externally. The easiest place to connect the negative supply is the through-hole for the battery negative terminal (the single hole on the righthand side of the board). The best place to connect the positive supply is the bottom right-hand through-hole in the group of six where the switch will later be mounted. These locations are marked by small triangles on the PCB silk screen overlay. Connect an external power supply set to deliver 4V with a current limit of 100mA and switch it on. The current draw should be negligible, and you should be able to measure 5V across both of the large tantalum capacitors. If there is a problem, switch off, check your work and, if necessary, reflow REG5 again. If all is well, you can remove the wires and proceed with mounting all the other parts, leaving the battery till the very last. IC1 and IC2 also have thermal pads on the bottom, so these will have to be reflowed too. However, they are SOIC-8 packages so are much easier to solder than REG5. Remember to apply the conformal coating we mentioned earlier on both sides of the board above the battery location before soldering the trimpots and trimcaps. Reapply it on the underside after soldering those components so their joints are covered. Immediately after you mount the battery, screw the board into the case bottom. This will help prevent accidental shorts under the board. The 38 Silicon Chip energy density of the Li-ion cell is such that accidental shorts can easily burn out tracks or cause other damage. Case preparation Drill the enclosure end plates and top case according to Fig.6. The slots can be most easily made by drilling a couple of holes inside the outline and finishing with a craft knife and small files. You will need to remove the two plastic bosses on the inside of the top case where the banana jacks are located – you can just snip them out with a pair of side cutters. The label (Fig.7) is simply glued to the front panel with some adhesive. I printed mine on glossy photo paper and covered it in transparent adhesive film for protection. Consistent with oscilloscope probes and commercial probes of this kind, the label describes the 100:1 and 10:1 attenuation ranges as ×100 and ×10, respectively. This refers to the multiplication factor you need to apply to the ‘scope’s vertical scale. For example, a 1V/division on the scope represents 100V/division on the ×100 range and 10V/ division on the ×10 range. Once it has been applied, punch out the holes for the light pipes and push them in from the front. They can be secured with a drop or two of The assembled PCB before it was installed in the case. Australia's electronics magazine cyanoacrylate adhesive (superglue) on the back side. Assemble everything except the top case and you are ready for calibration. Testing and calibration Start the calibration process by fully charging the battery. Connect a USB-C power supply to the probe and make sure the switch is in the off/charge position. The yellow LED should light, indicating the battery is charging. When full charge is reached, the LED will go out. This may take two or three hours if the battery is nearly discharged. Once charged, remove the USB cable and power the unit on by selecting either the ×100 or ×10 range and recheck that the power supplies are at ±5V as before. The green LED should be lit. The first step in calibration is to zero out the offset correction. We need to do this to make sure it does not impact the setting of the CMRR trim in the next step. Switch the probe to the 100:1 range and adjust VR2 until the voltage at its wiper is as close to zero as you can get it. You can clip your voltmeter’s negative lead to the GND test point and read the wiper voltage on the bottom end of the vertically orientated capacitor immediately below VR1, marked by a small square on the PCB overlay. You should be able to adjust the voltage to within a few millivolts either side of zero. Anything under ±10mV is fine. Now we need to adjust the CMRR trim. Set your bench power supply to the highest (safe) voltage you can get. For example, connect two channels of a dual 30V supply in series for 60V. Connect the positive lead of the power supply to both of the probe inputs (shorted together) and the negative lead to the GND test point. Switch the probe to the 100:1 range. Use your meter to measure the voltage between the mid-points of the voltage dividers while you adjust VR1. The suggested probe points are marked by small circles on the PCB overlay, immediately to the left of D1 and the right of D2. Adjust VR1 for a reading as close to zero as you can get at these points. You should be able to get a reading below ±20µV. With a 60V input, a reading below ±20µV implies a CMRR of 130dB. But you can probably do better than that with a good meter and some patience. Now you can set the offset voltage trim. Remove the power supply but keep the two inputs shorted. Measure the output voltage at the BNC connector with respect to the ground test point. Trim VR2 to get the output close to zero on both ranges. This may require a little backwards and forwards between ranges and the acceptance of some compromise (for reasons described above). For example, the best I could do was -1.1mV on the 100:1 range and +1.5mV on the 10:1 range. You should be able to get to within ±10mV of zero on both ranges simultaneously. The final step is to trim the frequency compensation. You will need siliconchip.com.au a function generator and an oscilloscope. The function generator should be set to deliver a 1kHz square wave at the highest amplitude you can manage. Connect the differential probe to the scope using a BNC-to-BNC cable and make sure the scope’s bandwidth limit is disabled. To set up the positive divider, connect the function generator’s output to the positive input of the probe and its common to the ground test point. Also connect the probe’s negative terminal to the ground test point. Switch the probe to the 100:1 range. Set up your scope to get a stable display of the square wave output of the differential probe and adjust compensation trimmer VC1 for optimum compensation, just as you would for an oscilloscope probe. The correct compensation is achieved when the rising edge of the square wave shows no overshoot or undershoot, as shown in Fig.8. Use a non-metallic tool to make this adjustment. It’s better to err very slightly on the side of over-compensation (a small amount of overshoot) if you are unsure, as this will maximise the probe’s bandwidth. Repeat the whole process for the negative divider, connecting the function generator output to the negative input of the differential probe and the probe’s positive terminal to the ground test point. This time, tweak VC2 for optimum compensation. Using it Screw the lid on and your probe is ready to use. I added four small self-­ adhesive rubber feet to the bottom of the case to prevent it from sliding around too much on the bench. Always take special care when you are using the probe with high voltage circuits. Make all connections – including that from the probe to the scope – before powering up any circuit under test. Never disconnect any high-voltage differential probe from the scope while the test circuit is powered on. If you do, the BNC connector on the probe can float to high voltages. There is no isolation barrier in these devices. Not much current can flow due to the high impedance of the probe, but you can still get a shock. Always use quality test leads with shrouded banana plugs for high-­ voltage connections, and check everything twice before powering it up. SC Australia's electronics magazine Fig.7: this label artwork can be downloaded from the Silicon Chip website as a PDF. For details on how we make front panels see siliconchip. com.au/Help/FrontPanels Undercompensated Correct Compensation Overcompensated Fig.8: correct compensation is achieved when the square wave’s leading edge shows no undercompensation droop or overcompensation overshoot. February 2025  39