Hantek CC-65 AC/DC Current Probe Teardown and Testing

[Bob Sava]

Indeed this is a typo error for the OPA195.

In first place U4 was an OPA2350 but the response wasn't great due to the poor capability of driving capacitive load.
C15 represent the C-Load and is needed for noise attenuation.

Good observations Mechatrommer the OPA2350 can't work at 9V but the regulator of the clamp is set at 7V.
I think we can replace all op amp with OPA2192. This will need some simulation and testing but it is possible.
Tell me if you're interested. 

If capacitive load on U4-1 is an issue - is resistor R18 different than 1k?

[Noy]

So whats best now?
Every opamp OPA2192 or mixed with OPA2350?

[newbieswe]

Does anyone have a parts list for the caps used in the modded schematic?
I'm about to order but don't know which size specs to choose.
If anyone have a complete parts list, that would help a lot.

[Noy]

All 0603 ones. The caps you need are all so small that you can choose X7R or They are only available as C0G.. So you can use >25V types..
All OPAMPS are MSOP..

[toli]

Glad to have found this thread. I've ordered a unit yesterday, and was looking for the schematic in the hope of extending the BW further.

Looking at the schematic posted on page 5 of this thread, some things seem a bit odd. Did anyone else trace out to circuit to see this schematic is correct?

Most of all the output difference amplifier around U4 seems strange, but a few other things can be extended in BW too.
1 - R18, which in theory should help isolate the amplifier from the capacitive load, is wrapped inside the feedback loop which means it doesn't isolate the amp from the capacitive load, it can actually make stability worse. I would expect R18 to be located after the feedback point (or in other words, for R17 and its added bypass cap to connect to the output of U4 instead of after R18).
2 - C15 seems very high in value. R18 + C15 will create a low-pass-filter with a pole at 1.6KHz with the values shown. Since its inside the feedback loop, the opamp will try to compensate for it, but with the limited GBW of the opamp it will only be able to do this for a limited BW (and even that only with small signal amplitude). Since the measurement results people have posted seem to show fairly high BW compared to this 1.6KHz, the value of C15 is probably quite a bit lower than this. Perhaps its 100pF and its a typo only?
3 - 22pF in parallel to R17 will add a pole at ~280KHz, and I'd expect to see a matching capacitor to bypass R20 for a typical difference amplifier structure. This could perhaps explain some of the uneven response seen at >100Khz.
4 - The first stage around U3 also has C5/C6 which are 8.1pF (8.2?) following the mod, so in parallel with 11Kohm its a pole at ~1.8MHz. That is assuming an opamp with high enough GBW. At the low current range the gain at this stage is probably ~X20 so we need an opamp with GBW on the order of 36MHz to get to this 1.8MHz limit in this case. In the PDF file attached a few posts later I see these caps are 15pF, this will limit the BW further to ~1MHz, so a GBW of ~20MHz should suffice in this case.

A few thoughts (which I can't do anything about until I get my unit ):
1 - My first guess would be that the value of C15 is off and its actually much lower in capacitance. Even then, it can probably be omitted completely after the mod. If we assume ~100pF load (cable + scope input) and R18 of 1K its a pole at 1.6MHz already, so adding additional capacitance will limit this further. R18 can probably be made smaller in value quite a bit (say 330R) to push out the pole its creating with the load capacitance, which would allow extended BW.
2 - The feedback for U4 will be taken (we can modify that fairly easily if it isn't the case) from output of U4 instead of after R18. In this case the 22pF in parallel to R17 can potentially be omitted completely, which would align R17 with R20 with no bypass and should result in a more even response at >100KHz and better CM rejection at the frequencies. If some capacitance is still needed there, we can scale down R16/17/19/20 to extend BW, which would make this stage less sensitive to parasitic capacitance if this is a problem. It will also reduce noise for this stage which can help in the high current range where first stage gain is reduced.
3 - This one is more of a trade-off. Currently the first stage seems to be operating with a gain of >X20 in the low current range (this is assuming the trimmer in the schematic is roughly in the center position), while the output stage has a gain of ~2.5 (for the low current range only, in the high current range the first stage gain is lower). This means the first stage needs to have a higher GBW as its more limiting in terms of BW. Moving more of the gain to the second stage can help extend BW further in the first stage which would improve overall BW. The down side here is that it will result in some noise degradation. A possible solution to this can be to move the gain switching to the second stage, and leave first stage with a constant gain of X5-X8 that will render noise of second stage non-dominant, but that's a bit more work.

[DaneLaw]

Also got one of these CC65 and been positively surprised with this unit, my first Hantek product.
Purchased the 4 cheapest AC & DC current probes I could find in China at one of their pre-NY-sales and some of them could likely work great with a new Micsig scope I had just purchased at the start of 2020.. So I ended up with 4 budget AC/DC current probes, and saw it as cheap entry tools to learn from, and compare one to the other and get an idea of frequency-rating and the noise and all that jazz..
The total price' a tad over 100 bucks' and it were more or less the same ballpark f I should have purchased Fx a Hantek CC65 inside Europe from a regional seller with danish +25%VAT.

ETCR007AD 100KHz (was the most expensive at 45USD delivered it has a +/- wheel on the side, while the Hantek have a degauss-button)
Hantek CC65 35USD 20KHz
AllSun EM264 25USD ?? Hz  (dedicated zero button, like the Hantek CC65)
HoldPeak HP605A 13USD ?? Hz
All 4 were AC+DC

ETCR007 at the top waveform and CC65 at the bottom. (spotwelder)

The Hantek CC65 of the 4 AC/DC current probes' got way the cleanest signal.. the other AC/DC like 605A and the EM264 is primarily for DMM's and quite high amps.

Here a vid with the 13USD Holdpeak AC/DC 605A  https://i.imgur.com/yUrAeXz.mp4

and here the 20KHz Hantek CC65 below and the 100KHz ETCR007 at the top.. noisy to say the least.. can be seen when going full BW like I have on the CC65 at the bottom.. (load is a USB colorchanging RGB diode with like +/-150mA.) https://i.imgur.com/Cj2B5AJ.mp4

But its very likely Im doing something wrong (rookie) but so far the Hantek CC65 has been the one I been most satisfied with, and to an extent that I even wondered about getting the CC650 but the noise will be that higher, and most of my use-cases are in the lower amp-segment so decided to pass on the CC650 but CC65 have been a pleasant first purchase from the brand "Hantek".

[dcac]

1 - My first guess would be that the value of C15 is off and its actually much lower in capacitance. Even then, it can probably be omitted completely after the mod. If we assume ~100pF load (cable + scope input) and R18 of 1K its a pole at 1.6MHz already, so adding additional capacitance will limit this further. R18 can probably be made smaller in value quite a bit (say 330R) to push out the pole its creating with the load capacitance, which would allow extended BW.

Yeah I probably missed C15 when I measured the other caps - the 100nF value is just an arbitrary value from the first versions of the schematics. And I also notice neither R18 or C15 was included in the spice simulation I did here: https://www.eevblog.com/forum/testgear/hantek-cc-65-acdc-current-probe-teardown-and-testing/msg3158118/#msg3158118

Adding C15-R18 to the simulation it will actually cause some signal peaking as C15 'steals' part of the signal from the feedback loop. And question is still what exact value C15 is - I agree 100nF seems pretty big.

[dcac]

Measuring C15 + the output cable which sits in parallel - it seems to be about 3nF.

I used the same technique as I did before using a low level sine wave at 10mV RMS and measuring the -3dB point it forms as a RC network with a known resistor value.

Below is the simulation with R18-C15 added - it didn't make that much difference in the frequency response. Though there is peaking of about 3.1dB at 100KHz. The lower blue trace is with C15 set to 0nF.

[toli]

Thanks for the reply, 3nF sounds much closer to expected value as this will add a pole at around 50KHz.

The peaking is probably due to stability of the feedback network now that this pole is there. It should improve peaking if you move the feedback to the other side of R18 which is the opamp output.

[dcac]

Not quite sure what you mean about moving the feedback.

I any case the schematics with regard to R18 and C15 placement seems correct. I haven't opened my cc-65 again but i.e. loading the output with a 1K resistor the signal level doesn't drop - suggesting R18 is inside the feedback loop. And adding capacitance in parallel with the output (and in parallel with C15) causes additional peaking or in other words an increased level at increased frequency. This scenario is of course only valid until the opamp saturates.

[toli]

I meant moving R18 outside the feedback loop (connecting R17 at the left side of R18 instead of the right side) would fix the stability issue.
For 20KHz this isnt a big problem probably, but for extended BW this is needed obviously.

[dcac]

Oh I get it now, yeah for a higher bandwidth the current configuration is far from ideal.

[toli]

Thanks dcac for your help so far, appreciate it 
I will be sure to update here when I have a chance to play with my probe when it gets here in a few weeks time.

The parts used really makes you wonder why they aren't offering a higher spec unit for slightly more.
For instance, from a noise density standpoint, the amplifiers used there are the P272 if I understand correctly from the pictures online. These are 25nV/Hz, the resistors in the feedback loop of the first stage will give some extra noise but its not dominant compared to the opamps used. So if we take 2 such opamps for the instrumentation amp input we get ~35nV/rt(Hz). Lets round that up to 40nV/rt(Hz) with the feedback resistors + second stage + resistors infront of it which are ~1K in series to each sensor on each side. For the lower current range given the gain of both stages (x60 combined gain assumed) and translation to equivalent input signal this will result in ~24uA/rt(Hz). As a sanity check, assuming ~20KHz -3dB BW, single pole means noiseBW is ~30KHz. So 24uA/rt(Hz) * rt(30K) will give ~4.25mArms noise which seems reasonable from the info shared in this thread.
With the replacement opamps suggested here in this thread (OPA2350) the noise of the opamps is no longer dominant at 5nV/rt(Hz) per amplifier. So now the opamps and resistors will be more balanced, and the overall noise density will be ~X3 (10dB) lower than the original. If we modify the feedback network around U3 to use lower value resistors we can get this down further by another 2-3dB.
A similar argument could be made for the BW. These opamps have a GBW of >1MHz (the stock units), so its possible to get well over 20KHz from this unit. Splitting the gain evenly across the stages could achieve ~150KHz BW at the cost of noise (I didn't measure slew rate limitation though, so this might only suffice for limited amplitude). A better opamp would not cost much and will give higher BW with lower noise floor with minimal circuit modifications.

The main down-side to using higher spec opamps (BW and noise) would be the current consumption which is X2-X3 higher and will reduce battery life. But this one could have been easily solved in a higher price unit with an integrated battery of higher capacity. The higher integrated noise can also be circumvented by a BW limit switch, so that a much lower noise can be achieved for a limited BW.

All the noise numbers above are for high frequency. It does rise rapidly at low frequency, and at low frequency (<300Hz) the OPA2350 is actually noisier than the P272. As a possible alternative, I quite like the OPA2156. A quick look at it seems promising with lower noise across the band, sufficient GBW, sufficient max voltage rating, etc.

BTW, looking at the teardown photos, R18 doesn't seem like a resistor at all. Edit: yep, its a PTC indeed/EDIT. That would explain why its inside the feedback loop (to limit amplitude and timing distortion due to its non linearity) despite the fact it has an undesired effect on stability of this stage.

[dcac]

Nice job identifying the PTC - that seems rather sophisticated choice for a device at this price point - I think I paid about 32 GBP for it a couple of years ago. Not that PTCs are really expensive but that they actually bothered to add that extra protection as supposed to just having a resistor there.

With regard to improving the probe I’ve made two mods to mine - I replaced C4 with a 1000uF/10V high quality cap - this have lower leakage and lower dielectric absorption so the zeroing process is and bit more distinct now and the larger value cap means less/slower drift. And then I lowered R24 to 1K to compensate for the increased charge time of C4 and to make zeroing slightly quicker again. But I never really measured the actual overall improvement except for the leakage in C4 which was significantly lower in the new cap and as it was an easy dropin mod in it went and then a 1.5K resistor in parallel with the 3.3K R24.

[toli]

Replacing the cap is probably a good idea actually since its the dominant leakage source at this point, only problem is getting a cap with low enough leakage. This can be a bit of a trial and error perhaps. Did you find one that fits and has a good spec on the leakage or did you pick one based on measurements?

[dcac]

It was just one I had in my caps box - I believe it's an Elna low ESR type so not really the specifically low leakage variants. I really only wanted to compare the leakage with the mounted 470uF/25V and that 1000uF/10V I found not only had double the capacitance but also considerable lower leakage and it was the same diameter 10mm but slightly taller 20mm vs 13.5mm but still no issue to fit it.

The drift now is less than 0.5mV after 5 minutes in high gain mode so often plenty of time to make the measurement but of course not stable enough for any logging purpose. And it's easier getting it to zero - with the original cap you had to hold the button for more than a couple of secs and then often press it again and again to really get it to zero - now it locks on much quicker.

[dcac]

Updated Schematics - just an updated value for C15 and that R18 is a PTC - also added possible types for the HALL sensors (suggested by Zhao).

C15 is set to 2.7nF - I measured it earlier to 3nF but forgot to subtract 120pF for my test coax cable. And the fixed output coax on CC-65 is probably 100-150pF and is also in parallel with C15.

[dcac]

I had a look at the noise performance and wanted to compare with what the spice simulation shows.

I measured CC-65 output noise in high gain mode to be about 2.5mV P-P but this is on a 100Mhz oscilloscope. My Fluke 45 shows it as about 0.5mV RMS. But I noticed how sensitive CC-65 is to EMI - just holding the probe in my hand I can get almost 3mV P-P 50Hz hum signal. Interestingly the front of the probe - or the actual clamp section - isn’t that sensitive to EMI so adding some shielding around the amplifier section should really be worthwhile if you plan to measure low level signals.

Simulating the noise in Micro-cap 12 I get about 133uV RMS output noise at 50KHz BW. This is with the Hall element simulated as generators with 400 ohms source resistance and with the TLC272 opamps.

If replace the opamps with TLC2272 which is an improved version that has lower noise at 9nV/sqrt(Hz) compared to 25nV/sqrt(Hz). I get 104uV RMS noise - so some improvement compared to 133uV RMS. But compared with the output noise I measured from CC-65 (roughly) 500uV RMS the overall improvement will probably not be noticeable as the Hall elements seems to contribute the majority of the noise them self.

And the simulated current consumption increased from 2.9mA to 3.3mA with the TLC2272

But keep in mind the measured noise is more a ball park value as the BW it represent isn’t well known. But from my previous experiences with Micro-cap noise simulations they seem to mimic the real world pretty well.

Below are RMS noise traces for TLC272 and TLC2272 - I only changed opamps at U3-1-2 as these stand for the majority of the amplification.

EDIT: There seem to be an error in the TLC272 simulation - there are four different spice models for these parts, it has to do with the supply voltage level and I don't think I selected correct model.

[toli]

The noise density of 25nV/rtHz even if we assume this is the only source of noise (that is U3-1 and U3-2), we still get 35nV/rtHz. For a -3dB BW of >20KHz, and single pole response the noise-BW is 30KHz. The gain of the first stage would be (R22+R15)/(R6+VR1) (I've neglected the path via VR2 here). The second stage is R17/R16. So overall gain assuming VR1 is about half its max value would be (18.6/1.06)*(26.1/11) which is >40. I wrote it down explicitly just so that its easy to check my math in case there's an error.
We can calculate the total RMS noise as:
Input noise density * gain * sqrt(BW), and we use only the white noise density which is somewhat optimistic but reasonable. So we get 35e-9 * 40 * sqrt(30000) = ~0.25mV. Since we have a transfer gain of 10mA/mV this is ~2.5mArms.
This all assumed only U3 noise is relevant, and that its a white noise source. So in practice this will probably be higher than this value.
C5/C6 will add a pole, but C14 seems to offset this partially. This will therefore result in lower noise-BW, but with limited benefit, perhaps 20% at most. So that's not too far out of your simulations, about 3-4dB difference.

The TLC2272 does look significantly better than the original unit with good power consumption too. Another good option might be the OPA2196. It has lower noise (15nV/rtHz wide band noise), similar BW (2.5MHz GBW), lower offset (100uV max), can swing closer to the supply rails (might allow extending output range further in the sensitive range), and lower power consumption (140uA typical per channel). If extending BW to 100KHz is sufficient, it might be an excellent match for this application.
I've also given this some more thought in the mean-time while waiting for my unit to get here. My aim is to extend BW among other things, and I want to see how far I can effectively stretch it with minimal effort. Therefore I'm willing to pay for it with some power. My current plan is to go with the OPA1652 for U3, which has a few benefits in addition to noise density. As far as noise is concerned though, it should reduce noise density by about 10dB for the overall transfer function if my math is correct, this should be easy enough to measure on my typical audio measurement setup that covers this frequency range. I have a few other modifications planned, I will get to it when I get my probe and order the replacement parts for it. I will be sure to update here as well as post about it with some background on my blog like I typically do for such things.

Regarding sensitivity to EMI, I wonder at what frequency is it most sensitive? Perhaps at a few KHz and above? Looking at the schematic, the possible problem with this structure as I see it is that it ignores the reason we have the 2 sensors to begin with. By bypassing R13 with C11, the input signal to the instrumentation amplifier is no longer an average of the two sensors like it is at low frequency. This means we no longer attenuate ambient magnetic field by averaging the two sensors in reverse polarity. Therefore, if it was me, I'd remove this bypass capacitor and modify the amplifier transfer function to get around it if needed.

[dcac]

In my previous post - there seem to be an error in the TLC272 simulation - there are four different spice models for these parts, it has to do with the supply voltage level and I don't think I selected correct model.

I'll post new simulations when I figured out which model is relevant for the CC65 circuit.

EDIT:
I think this should now be correct. The TLC272 error was quite substantial - simulated RMS noise is now 256uV compared to 133uV previously. Still though compared to the overall noise from the CC65 it would not make that much difference.

So Simulated RMS noise at 50KHz BW:
TLC272 = 256uV
TLC2272 = 108uV
OPA1652 = 67uV

Simulated current draw:
TLC272 = 2.9mA
TLC2272 = 3.3mA
OPA1652 = 5.0mA

Below are the traces for RMS noise and the Output frequency response - interestingly OPA1652 has slightly better reach even with the same capacitor values around the U3 lowpass filter.

TLC272 = Purple trace
TLC2272 = Blue trace
OPA1652 = Red trace

[toli]

Thanks, that now matches my calculations much better

BTW, don't forget the current consumption of the sensor bias. This can be a few mA per sensor (the spread is quite wide and sensitive to temperature) and we have 2 of them.

[dcac]

Yeah it was your calculation that made me check the simulation again.

The simulation only includes active elements U3-1-2 and U4-1 to do the Noise and Frequency response, so none of the Hall, DC control or Regulator current is accounted for.

[dcac]

I measure CC-65 total current draw to 10mA at 9.6V. So upgrading U3 really shouldn't make that much difference.

A rechargeable battery really is preferable in any case - else it will kinda ruin your day if you forgot to switch the probe off.

[toli]

My probe arrived. I'm yet to make any significant measurements on it, and no mods yet (will need to order parts first). However, I've probed inside of it for a few minutes just to fill some gaps I've had.
First, as far as the LDO stability is concerned, I don't see any oscillations at the output. Nor do I see any oscillations at the bias of the sensors which U1-2 regulates. This is good, although I still don't like this structure of the bias circuit, it has 2 poles which can cause instability or at the very least noise peaking which might still be inside the BW of the amplifier after it. I will try to remember to probe this with the audio measurement setup, to observe the spectrum there, might reveal something more interesting that way.

The current draw of my unit is ~11.4mA, which is higher than the 10mA you've measured, but its expected there would be variation among units, especially with the value of the sensors resistance varying so much. Still, not too impressive life time out of it.

What I found interesting is the value of the DC rails. The positive rail I've measured at 3.22V, so ok for a 3.3V LDO with slightly lower voltage than nominal (I see the schematic marks it as 3.2 nominal, in which case its almost spot on). The negative rail though is -3.8V, so D4 might be a 3.9V variant (I see the schematic you've posted is marked with 4.2V, also possible as these have a fairly soft knee and a few percent tolerance - I admit I was too lazy to check). Why do they plan for such a significant voltage for the negative rail? The bias of the sensors, and as a result the CM voltage at the inputs of the op-amps are derived from positive rail, so I'd expect them to shift the rails in favor of the positive rail if anything, not the negative rail. 3.2V+3.8V = 7V, so if we leave a bit of headroom the battery must have >7.2V for normal operation, seems like inefficient use of battery energy.
Additionally, because of the way the low battery threshold is sensed, the positive rails needs to drop quite a bit for this indicator to light up. So this only happened at 6.8V in my unit at which point the positive rails was already 0.3V lower than nominal at 2.95V. This means that performance might degrade even before the low battery indicator lights up.

Might be of value to replace D4 with a lower voltage shunt regulator (just so its sharper than a zener, but a zener is ok too), perhaps 2.5V for instance. Then move the voltage sensing for low battery to 3.5V instead of 3V. Then 3.5V+2.5V would allow operation with a 6V supply instead of 7.2V.

It'll probably take a little while before I order and get the parts for mods, but I will post if I see anything interesting before any changes are made to it.

BTW, what I found quite funny is that if the battery voltage drops slightly lower than the value designed as the limit, the LED turns off. Probably because U6 has very limited ability to drive positive output voltage, so that combined with the LED drop is too much. So at 6.8V it lights up, but at 6.5V you can barely see it 

[toli]

I replaced C4 with a 1000uF/10V high quality cap

Did you use a bipolar cap or still a polarized cap?

I see the original capacitor is polarized, which means the negative voltage range is very limited if we want leakage to remain low. However, if we go to a bipolar cap, we can live with an increased voltage over this cap,
which means we can reduce its weight in the circuit (increase value of R23). This will make it less sensitive to leakage of this capacitor. There's still the effort of finding a cap with low leakage though.
Edit: its possible to do the same trick with a polarized cap obviously, but then there a risk of the reverse voltage being large enough to cause increased current.
If we go too far we might need to increase R269 too, and then there's the question of how the leakage of the cap depends on the voltage over it when the applied voltage is so much lower than rated voltage