Hantek CC-65 Current Clamp Probe – Part 3 – Signal Chain, Improving BW and Noise

This is the 3rd (and final) part of a series of posts on the Hantek CC-65 current clamp probe. In part 1 we went over the probe structure and circuit operation, and discussed possible issues and improvements. In part 2 we’ve started measurements and modifications of the probe, focusing on the power-supply (PS) and the sensor bias circuit. That allowed us to achieve lower noise on the supplies and sensor bias, as well as extend the circuit operation down to lower battery voltage.
In this post I’d like to go into modifications of the actual signal chain. This consists of the amplifier structure at the heart of the probe, but will also touch on the offset cancellation circuit. The main goal from my point of view is to both extend its bandwidth (BW) by at least an order of magnitude, and reduce the equivalent input noise density so that limited BW measurements can be made on lower amplitude signals.

Main Amplifier – Op-Amps

As we’ve covered in part 1, the dominant noise source in the circuit in its present state is the operational-amplifier (op-amp) U3. We’ve considered a few possible options for a replacement to this part, each with its pros and cons. For my set of requirements, I’ve decided to go with the OPA1652. It offers a few benefits, such as much lower noise density of ~4nV/rt(Hz) vs the 25nV/rt(Hz) of the 272C op-amp that is already there, and much wider BW and higher slew-rate (SR), all of which we want. It also has some “nice to have”  benefits such as lower offset, higher open loop-gain, better CMRR and PSRR, etc. There are some drawbacks obviously, such as the higher quiescent current. At room temperature the OPA1652 has 2mA typical (2.5mA max) per channel, while the 272C has only 0.7mA typical (1.6mA max). I chose to use OPA1652 for U4 as well, despite the ability to optimize it somewhat better with a slower yet lower power and lower offset part (more on that at the end of this post). Once you consider the total consumption it will add approximately 5mA to the current draw of the probe if we compare typical figures (this drops to 3.6mA if we look at maximum figures).
Using the ~11.5mA current I’ve measured in my probe this will increase to ~16.5mA. We will offset some of this as will be discussed later, but its still an increased current consumption. Thankfully, the modifications covered in part 2 of this series discussed extending the operation of this probe down to lower battery voltage. This compensates for the increase is current draw and extends battery life sufficiently to be on par with what it was originally.

For U1 and U6 we aren’t looking for anything fast, nor very low noise, as explained in part 1. This allows choice of op-amps which could potentially save power. My original aim was to use the OPA2196 which I’m familiar with from previous projects. However, it is in short supply for a while now, and lead times are in the range of months. Therefore, I chose a somewhat lesser spec TLV9102. Its a low power, low offset, rail-to-rail op-amp which is an excellent match for this application. Its noise performance isn’t great, but thankfully, this is of no consequence here. U6-1 is used as a comparator for battery monitoring, and U6-2/U1-1’s noises are greatly attenuated before reaching the signal path. U1-2’s noise is not only attenuated somewhat by the structure of the circuit, its also band limited quite significantly by the modifications done to the bias circuit in the previous post. Therefore, the TLV9102’s main downside is of no consequence in any of these locations.
The quiescent current of this op-amp is very low, just 115uA typical (150uA max) per amplifier. If we replace U1 0.7mA typical (1.6mA max) and U6 200uA typical (250uA max), we can save ~1.5mA typical (3.4mA if we use max figures). This will offset a substantial part of the increase in current draw we’ve caused by upgrading U3 and U4.

Main Amplifier – Resistors and capacitors

Next we need to turn our attention to the resistors used in the amplifier. The equivalent resistance at the input of each of the amplifiers of U3 is ~500-ohm, so combining the two we get 1K-ohm, or an equivalent noise of 4nV/rt(Hz). At this point this isn’t dominant, but close to that of U3-1 and U3-2 (~5.5nV/rt(Hz) combined). We can adjust the resistors at the input to gain somewhat in noise, but due to sensors mismatch we will hit a limit as to how much better we can make this.

The resistors in the feedback network however don’t suffer from such constraints and could be made lower in value. At the moment the combination of R6+VR1 (in parallel to VR2 branch which I’ll neglect for the moment) is ~1K-ohm, I’d like to make its noise half the original value, so a nominal value of 250-ohm for the series combination sounds like a good number. I chose to use 180-ohm for R6 and leave the remainder to VR1. For a gain of 21 (the approximate gain of the 1st stage in its current form), R15 and R22 should be 2.5K-ohm each. However, with such a gain, the OPA1652 can reach ~900KHz BW, which is a tad shy of the 1MHz I’ve defined as the “target”. Therefore, I’ve decided to trade off some of the gain between stages, and have a gain of X15 in this stage, which results in 1.75K-ohm. A value of 1.8K-ohm is a close enough standard value, so we can adjust the previous resistor via VR1 to compensate. For similar reasons we will revise R21 to 4.7K-ohm, this should still allow sufficient margin for gain adjustment for all cases.

To compensate for this reduced gain in the first stage, the second stage gain needs to increase from ~X2.4 to ~X3.4. The resistors in the second stage are of lesser impact on noise due to the gain in first stage, but they do come into play somewhat in the case of the lower sensitivity range. Additionally, high resistances and high BW aren’t typically a good match. Therefore, I’d go ahead and reduce these too. I went with 1.2K-ohm/3.9K-ohm which gives a gain of X3.25, close enough to target.
R18 which is a PTC also has to go, as was discussed in part 1. It has limited use for protection, one might even argue it not doing anything in some cases as there is no explicit power clamp for the positive rail. It does however induce significant distortion, which means it must be placed inside the feedback loop which upsets stability, and sets a hard stop on max BW.  This is especially true for high capacitance loads such as long cables. R18’s presence in its current location also means that U4-1 needs to drive significantly increased amplitude at higher frequency to compensate for that, which would limit its useful amplitude at higher frequencies due to attenuation and slew rate. I therefore chose to replace it with a linear resistor, and take the feedback before that resistor. To make life easier for U4-1 driving a capacitive load, I chose a 220R resistor.

Next are the capacitors adjacent to the same amplifiers. C11 has to be removed, it hurting CMRR of the probe by unbalancing the sensors when it kicks in. This might be outside the BW of the stock probe, but not outside the BW we expect to achieve. Same for C8/C14, they will affect frequency response and we have no need for them. C5/C6 will have to be replaced with 68pF caps, preferably C0G/NP0 type. C15 should also be replaced with a lower value, or removed completely. I chose to use it as an additional pole just in case (at >X3 the probe BW), so once combined with the 220R at the output, along with the cable capacitance and possible load capacitance (say ~150pF combined), we need C15 to be 100pF. As before, a C0G/NP0 cap is probably best just in case, as they are very easy to get for such low values. We can add optional capacitors of 15pF in parallel with R17 and R20. They can help with stability if the PCB layout has significant capacitance at the inverting input of U4-1.

A small note regarding D2/D3. They have limited use in actual protection of the circuit, as mentioned previously. However, such small diodes typically have fairly low capacitance and therefore shouldn’t affect performance of the probe for the limited 1MHz point I’m looking at as a target. Therefore, I chose to leave them as is for the time being.

DC offset cancellation circuit

In part 1 of this series we’ve covered potential improvements to this circuit that cancels the offset of the probe. These included increasing the value of R23 to make the circuit less sensitive to voltage variation of the storage capacitor C4, reducing the value of R26 to make the offset adjustment finer once we replace the op-amps, replacing C4 with better capacitors, etc.

Since then I’ve considered a few other possible improvements, and have made a few calculations/measurements to check the effect. One potential change is moving the point that C4/R269 connects to from Vg potential to half the sensor bias voltage. The logic behind it was fairly simple, the output of the sensors is at this DC point, and therefore this should be the center point for the DC offset. If not, there is an inherent imbalance from R23, which would change the offset with temperature unless it is zeroed out again. However, the gain at this path is fairly low, punching in the numbers ~1.1mV/C at the center point of the sensor from the diode temp-co, and the gain of the offset cancellation, this results in ~0.35mV (3.5mA) error at the output per degree C for the high sensitivity range. This seems like a small enough error to have no significant effect compared to other possible sources of error such as stray fields affecting measurements. However, some people might think its worth the modification, so I’ll discuss it briefly here with a few possible implementations.
Implementing this calls for a couple of changes, and there are a few different methods of doing this. One way would be to tap the point between R1/R2, which is at exactly that voltage. Another would be between D1/R7. Both of these I don’t like because they affect control voltages of U1-2, but both will probably work reasonably well. Another option would be to tap the center wiper of VR5 or VR6. This could affect gain somewhat (for VR5 this actually increases it as there is a positive feedback loop formed, but the low gain means it would remain stable), but nothing significant that we can’t adjust later (<1% error in gain). Another way would be to create a new voltage divider to tap for this. We will need to keep the resistors there small enough for reasonable fast settling, but high enough to minimize the power consumption. I chose 6K8 resistors so that the equivalent output resistance is almost equal to R24. R24 can (should) now be shorted, as the voltage divider we’ve drawn replaces its function of adding a zero in the transfer pole during offset calibration. At room temp, with ~0.95V over this voltage divider, it will draw 70uA, which seems negligible with the total current draw of the circuit as a whole. I’ve drawn this in a different color in the schematic below, to note that this one is really more of preference, and not something that is strictly advised.

Next, I turned my attention to the capacitors leakage. I’ve ordered a low-leakage capacitor (UKL1A681KPD), and observed its leakage current over a period of a few hours at different DC voltages from -0.2V, up to 3V, with a few steps. The leakage current always seemed to settle to very low values (on the order of 1-2nA after a few hours), which is completely insignificant in this application. However, there was something far more interesting to note, and that is the effect of changing the voltage over the capacitor and then moving back to the first value. It seems that the capacitor leakage during the first couple of hours is more of a “memory” effect than actual leakage, perhaps dielectric absorption is to blame here. This has a significant effect on the route we should take to minimize leakage, and that is to keep the voltage over the capacitor as close to 0V as possible, as this is where R269 will try to pull C4 when the probe is turned off. This meant that increasing the value of R23 might make us less sensitive to voltage variation over the capacitor, but it will also increase the leakage for the first few hours of operation after turn on since the capacitor will “remember” the 0V it had during power-off. Therefore, I’ve decided to leave R23 as is, and try to limit the voltage over the capacitor during operation. One small “improvement” can be reducing value of R26 to make adjustment via VR7 finer, while still meeting the complete spec of OPA1652 for offset voltage. I’ve used 390R for this part. Interestingly enough, it seems like the range covered by the original values doesn’t cover the complete spec of the 272C op-amp offset voltage.

While at it, I’ve measured the DC drift of my probe in its original state. I’ve measured -30mVDC over C4 in my unit when offset is zeroed out, so fairly close the 0VDC it will try to “settle” to when powered off. I was very impressed to see that with the probe standing still so that no external field changed its offset, the output remained at <0.1mVDC for >1 hour, at which point I stopped the measurement as this was already more than I’ve ever imagined it would hold. Because of that, I’ve decided to leave C4 capacitor in place. The replacement capacitor I’ve ordered might be better, but it might as well be worse considering the low change. What’s more interesting to note here is that R269/C4 form a time constant of 4700s, or ~1hours:20minutes. This means that after the measured time period, the voltage over the capacitor should have dropped by at least 50%, and the output should have drifted by 5mV if we assume all else is ideal. Therefore, it seems that my unit holds the voltage as stable as it does by nothing more than “luck”, perhaps the memory effect (or pure leakage current) of the capacitor is simply strong enough to counteract the 3nA current caused by R269 having 30mV over it. I will measure this once again after all other modifications are done, and will take action if I see a note-worthy drift of the output voltage.

Fig. 1. Initial Summary of Schematic Changes to CC-65 Probe Circuit 

Figure 1 shows a summary of the schematic changes to the probe circuit. I’ve used the schematic posted here, and marked on it the changes described in this blog. In red are the changes covered in part 2 for the PS and bias circuits. In blue are the changes I would consider as the “core” of the mods of this post to improve noise and BW. In green is the optional change to the offset circuit to improve thermal stability of the offset zeroing loop.

Adjustment and Measurements

I’ve implemented all the modifications marked in blue color in the figure above. Before turning on, make sure you clean the flux properly after all this soldering, then short C4 (important, as it can reach negative voltages at this point before we adjust everything).

I’ve then adjusted the trimmers. First, to make sure the two sensors as balanced as I can get them I’ve adjusted VR5/VR6 (they weren’t optimally balanced beforehand). To do this you need to rotate the probe (try all directions and see which one results in the maximal change of output voltage), and see how much the output voltage changes. Then adjust VR5 or VR6, and see if things improve or degrade, and continue accordingly. Repeat for the additional trimmer of the two.
Next I’ve adjusted the offset via VR4 (coarse) and VR3 (fine). Following adjustment of the gain. Starting with the 65A range (VR2), and then for the 6.5A range (VR1). Make sure you adjust at this order, as the 65A setting will affect the gain of the low current range too.
Finally, you can remove the short from C4, press the zero button, and adjust VR7 to get as close to 0VDC at the output as you can. This last step can be done with C4 still shorted, so if you prefer to make it a tad faster (no need to charge C4), you can leave C4 shorted while adjusting VR7, and only remove the short after it is done.

Frequency response

Next I’ve moved to some more interesting measurements, starting with frequency response. Measuring on my audio test setup, you can see results are better, as shown in figure 2. However, in this case, the setup itself is far more limiting than the probe. So we have to move to something more capable in terms of BW. There is a -1dB difference in gain between this plot and the one posted in part 2 for a similar test before modifications, this is due to proper adjustment of the probe gain, which wasn’t the case previously.

Fig. 2. Frequency Response Measured on Audio Testing Setup – Limited by Setup

To extend frequency response measurement BW, I’ve connected a load resistor to the end of a BNC cable driven from a 50MHz function generation, and the output of the probe to the scope. The output of the generator was fed into the scope as well, to monitor it, and trigger from it due to the lower SNR at the output of the current probe for such low currents (<50mArms).

The result didn’t match what I was expecting, the -3dB BW was far higher than I’ve estimated (~4MHz if I recall correctly), and had significant peaking (6dB for low current range, 10dB for high current range). This peaking had a zero in the transfer function at approximately 300KHz. After considering the possible sources, the only one that makes sense is the sensor/magnetic core it is fixed to. To compensate for this, there’s a need to add a pole at the same frequency. Doing so in the first stage isn’t possible as it follows by a zero close-by in the high-current range. Additionally, its beneficial to have gain as early in the signal chain as possible to maximize SNR. Therefore, I’ve modified the circuit and replaced the 15pF capacitors I’ve added earlier with 120pF capacitors, as shown in figure 3. If you don’t mind having a few dB of gain peaking, you can use a smaller value capacitor here instead of 120pF, and you will be rewarded with a higher -3dB BW point. It can easily be at a few MHz (3-4 or even higher perhaps), but from my point of a view, such significant gain peaking is un-acceptable as it will cause significant overshoot in the transient response. I prefer to have a probe which is slower than it could potentially be, than having one which creates significant overshoot where there is none.

Fig. 3. Revised Schematic

I’ve repeated the measurements once more, and this time the peaking was far lower at <1dB max. The -3dB point was at 1MHz for the high sensitivity (low current) range, and 2.5MHz for the other. The normalized frequency response for both ranges from 100KHz to 5MHz is plotted below in figure 4. Below 100KHz its flat, just as expected.

Fig. 4. Modified Frequency Response Following Circuit Changes

The higher BW for the low sensitivity range is due to the lower gain of the amplifier, which in turn results in increased BW. I’m not really interested in the frequency response of this range, since at such high currents I doubt I will ever measure high frequencies. The high sensitivity range is the one I’m interested in, and it meets my target spec of 1MHz -3dB BW. The 1dB peak is acceptable for my needs. It can be flattened better by replacing the 120pF capacitors with 150pF ones. This will place the pole closer to the zero, and result in a flatter frequency response.
If we’d like to push this further, we can obviously go back to the circuit, and extend BW further by adding a resistor in series with the 120pF we’ve added, and we will get to a few MHz with no trouble, potentially even leveling off the peaking completely by replacing it with an additional 27pF  in parallel to that combination. However, I have no need for such BW, and even 1MHz is more than I need from this probe. The 1MHz target was more of a “nice to have round number” from my point of view.
With that being said, I might still reconsider and revisit this part if my needs ever change.

Transient Response

I wanted to estimate the performance with a transient response as well as the frequency response, to get more insight as to what it can actually do. As a test, I’ve constructed a simple switching circuit and observed its current pulse. The “reference” is voltage measurement directly over a 0.1-ohm resistor in the circuit. To connect the clamp I’ve constructed a metal “bridge” over the board so that I can wrap the probe around it. There were local decoupling capacitors on board. The slope was limited by a series resistor to the switching MOSFET in this case. The current pulse is with an amplitude of 4A, and extends approximately 20us every 10ms.

As a reference I’ve measured this with the probe before modifications took place. The response can be seen below in figures 5-6. The purple pulse is the drive signal to the switch, the red signal is the “reference” via shunt resistor, the green is the response of the CC-65 probe before modifications. Two different time bases are shows for better visualization.

Fig. 5. Transient Response, Pre-Modifications, 5us/div
Fig. 6. Transient Response, Pre-Modifications, 200ns/div

Next, I’ve repeated the same measurement once the probe circuit was modified. This is shown in figures 7-8. Same color scheme is used as previously.

Fig. 7. Transient Response, Post-Modifications, 5us/div
Fig. 8. Transient Response, Post-Modifications, 200ns/div

We can clearly see this is now much faster, and can almost keep track of the slope of the reference signal.
I should note that this is by no means an accurate or well controlled test setup. There are multiple wires connecting to the ground node where the current is probed (from the PS, from the function generator, from the scope), each at a different position and angle relative to the current probe, all in close proximity. So there is without a doubt some bias in the measurement so that even an infinitely fast probe with ruler flat frequency response might show a different result to the “reference” signal. With that being said, the probe output now looks much closer to the reference than what it was with the original circuit, and the transient pulse response helps demonstrate just how fast the probe is now compared to what it was.

Noise Density

Our probe is now much faster, so despite improvement to its noise density, the total integrated noise over the complete BW of the probe will be somewhat higher than previously. However, as long as we are interested in slow signals, we can limit the BW to <1MHz of the probe. This is something that is inherent to an audio measurement setup such as the one I’m using, so it can be a good demonstration and comparison of the two. Figure 9, was demonstrated in part 2, and is included here for completeness. This is the noise density at the output of the probe with no input signal present. Figure 10 repeats this measurement with the modified probe circuit.

Fig. 9. Probe Output Noise (6A Range), Following Part 2 Modifications Only
Fig. 10. Noise Spectrum Following Modifications

We can clearly see the noise density at these (<100KHz) frequencies dropped. Flicker noise now dominates to a higher frequency, this is because thermal noise was reduced more than flicker noise with the parts I chose to use. However, it still surprised me the flicker dominated area extends as high as it does, I’ve expected it to flatten earlier based on the spec of the parts used. Overall, integrated noise up to 100KHz -3dB BW (~140KHz NBW) is now 150uVrms instead of 310uVrms it was before modifications. Add this to the fact we have much wider integration BW now (140KHz NBW vs ~36KHz NBW previously), and this shows just how much we’ve improved the noise.

To try and visualize this better, I’ve measured a 3mArms 1KHz sine wave with the probe before and after mods. I’ve used a 192KSPS configuration so this is far “worse” for the probe after mods since more of its noise can enter into the measurement, while before mods it was too slow to affect this. The results are shown below in figures 11-12. As a bonus, I’ve repeated the measurement after the mods with a 48KSPS configuration, which lowers BW to almost that of the probe before mods (a tad lower actually), shown in figure 13.

Fig. 11. 3mArms 1KHz Sine-Wave Measurement, Pre-Modifications
Fig. 12. 3mArms 1KHz Sine-Wave Measurement, Post-Modifications
Fig. 13. 3mArms 1KHz Sine-Wave Measurement, Post-Modifications, Limited BW

We can clearly see the improvement. Now, I know what you are thinking, this is 3mA on a 65A clamp meter (on the 6.5A range). True, this is a very low signal, but this is exactly why it was selected. The RMS noise of the meter before mods was ~3mArms according to measurement, which is why I chose this specific number. This effectively gives us a SNR of 1 before mods to the probe.

Additional Notes

As was mentioned earlier, the choice of higher grade parts, comes at the cost of power consumption. After modifications, I’ve measured the current consumption of the probe at 15.9mA, which is a 4.5mA increase compared with the 11.4mA I’ve measured before any modifications. Based on typical figures from datasheets, this should be closer to 3.5mA, but this can obviously vary between samples, and in my case this summed up to 4.5mA difference.
As mentioned earlier though, this is somewhat offset by the ability to work down to 6V battery voltage, instead of >7V before any mods took place. I’ve verified the low-battery indicator, and it worked great, turning on at 6.05V, and staying bright for far below 6V which I would consider the point at which the battery should probably be changed.

Moving to the DC offset zeroing circuit, I didn’t replace the capacitor used there due to reasons explained earlier, nor did I implement the thermal tracking for this circuit (in green in the schematics above). I’ve monitored the DC output voltage of the probe, and I can indeed see the drift clearly. It is now at a rate of 1mV per ~30 minutes (at high sensitivity range). This is already very low leakage, and sufficient stability for practical use as it still has some sensitivity to surrounding fields so I don’t expect leaving this logging for very long periods of time without zeroing it occasionally.
[Edit August 12, 2021]: Decided to give the replacement capacitor (UKL1A681KPD) a try despite being happy with the original capacitor for C4. Right after installing it, it seemed to be worse than stock capacitor at 1mV per ~5 minutes drift. However, after leaving it on with the zero button pressed for 20 minutes (in case this is still dielectric absorption we are seeing here), this improved to 1mV per ~20minutes. After an additional 20 minutes with the zero button pressed, this improved further to 1mV per ~25 minutes (figure 14).
So the two capacitors both seem to be more than good enough for this use, with their leakage not being the dominant limiting factor for this. Given its a brand name capacitor it might hold up better over the years, I chose to stick with the replacement capacitor. I don’t like the fact there is so “much” of an effect from this phenomena of capacitor “memory”, as the offset changes between ranges and therefore the capacitor will leak more if you switch a range and zero it again. However, given the implementation of the circuit, this is something I’ll have to live with. Thankfully, as will be shown in a few lines, the “6.5A range” can actually do much more than 6.5A, so in practice I might use it for 99.9% of my needs. [/Edit]
I know some people consider reusing VR7 to set a constant zero value bypassing the sampling switch and feedback loop controlling it. I’d recommend not to take this path, as the offset can change with time, temperature, and selected range (for the 65A range offset of U4-1 is much more dominant than at the low current range). Additionally, since you will most likely have some residual sensitivity to surrounding fields, this will once more be something this method cannot compensate for.

Fig. 14. Probe Output DC Offset vs. Time. Low Current Range

Moving to frequency response, as noted earlier, there is some peaking (1dB max in my measurements) in the transfer function. You can flatten it by increasing the 120pF caps we’ve added to 150pF instead. This will reduce the -3dB BW to ~800KHz, so there is a trade-off between the two, and you can choose your preferred configuration according to your needs. Just as a comparison, I’ve repeated the transient measurement with an increased capacitor, to be able to show the difference in results. Figure 8 was with a 120pF, and figure 15 is with a 150pF (actually measured at 140pF) capacitor. You can see the difference in the transient behavior of the output. For example, the value to which it drops before settling.
On the same subject, it is possible to improve this further, by both moving this pole, and adding a zero at higher frequency to compensate for the drop as we approach the 1MHz mark. This will require adding another resistor in series with this 150pF capacitor, and preferably splitting it into 2 capacitors, so that some of it remains active to guarantee stability at higher frequency.
Alternatively, you can choose the easier way of getting higher BW by reducing this 120pF cap, at the cost of increased peaking in the frequency response. This would be an easy way to extend BW, at the cost of this increased peaking which I consider to be too much. Both methods will get you to a few MHz -3dB BW (3-4 or potentially even higher if you choose the parts right).
I didn’t implement any additional BW extension in my probe, as this will have a downside of significantly extending the BW on the low-sensitivity range due to the much higher BW achieved at this gain setting in the first amplification stage. This in practice has no benefit for signal content, at least not for my use case, but will extend noise BW, as well as potential coupling from the surrounding which can now be observed at increased frequencies.

Fig. 15. Transient Response, Post-Modifications, 200ns/div, 150pF Bypass Caps Option

The current clamp is rated at 65A, which is why most people refer to the high-sensitivity range as 6.5A range. However, in practice, you can push it further. I didn’t check the limit before the mods, but I did stretch it after the mods. I’ve measured with my linear lab PS, and 3 turns of wire, and was able to measure (with error that is within spec of the probe) 18A in either polarity. It can probably be stretched further, but at this point I maxed out the PS output current, and couldn’t fit additional windings of the wire I was using into the core of the probe. I expect it to be fairly close to the limit, perhaps 25% more or so before we start seeing clipping of the output amplifier.

Previously, I’ve mentioned the possibility to optimize the circuit better by using a different op-amp for U4. If you consider the parameters we are interested in, they are:
– Maximum output swing, especially on the negative supply side due to lower negative supply amplitude. This would allow to operate with high current on the high sensitivity range.
– Sufficient BW, which for the modifications we did so far can be met with a GBW of just a few MHz.
– Sufficient slew rate. For a 10Arms signal at 1MHz on the high sensitivity range we will need slew rate of ~9V/us.
– Low offset. For U4-2 this is clear, as it will allow finer adjustment via VR7, perhaps even eliminating it completely from the trimming steps. For U4-1 this is more important actually, as this would mean the equivalent offset of both ranges will be closer, because the 2nd stage of amplification will have lower offset, and it will therefore not become a substantial part of the total offset as we move to the higher current range. This is desired due to reasons discussed earlier, where its in our interest to operate C4 with as constant of a voltage over it as possible, due to capacitor memory effect.
– While meeting all of these we’d like something with low noise and minimum power consumption.
I’ve used the OPA1652 for U4, just as I did for U3. It was easy, and it a fairly good match for this application. More importantly, it was in stock, which with current supply times is a significant advantage. However, it can be optimized better with other parts, and it wasn’t my first choice here. Potential candidates include the OPA2189 and OPA2197. As a comparison, the OPA2197 has an output swing which is even closer to the rails than the OPA1652, its GBW is lower but still sufficient at 10MHz, the slew rate is double that of the OPA1652 at 20V/us, and the offset is very low at 100uV max (25uV typical). What’s even better is that it has a quiescent current of about 50% less than the OPA1652. Its noise is a tad higher, but nothing substantial for this application. I think it is an excellent match for this application, and this is something I might do in my probe too if the OPA2197 ever gets back in stock.

Finally, if you modify your probe in any significant way, I recommend you verify you probe’s frequency or transient response if you can. This would give you additional confidence in the performance of the probe after you’ve made any modifications to it. This is of value in my opinion since its possible to make mistakes when modifying your own circuit. Additionally, there’s always the possibility of your probe behaving differently than mine, in which case even doing exactly the same things I did, might result in somewhat different behavior.


This series of posts finally came to an end. I want to note once more the complete group of EEVblog forum members who have participated in the discussion about this probe. The schematic I constantly reference and use to build upon is taken from that discussion.

I hope the information presented here will be of help to readers who are interested in this probe, and I do hope the amount of technical informational included would be of help. Its always difficult to find the right balance between going in depth to cover as much as possible, and only scratching the surface for it to be lighter to read and follow.

Leave a Reply

Your email address will not be published. Required fields are marked *