Project idea brainstorm - low cost oscilloscope current probe?

[max_torque]

Having finally got a nice 4 channel scope, i now am missing the ability to sample current data from test devices.  High bandwidth current probes are obviously very expensive, but generally, for what i am doing, i don't need a very high current sampling rate (as inductance tends to limit RoC of current).

So,  i'd be interested to hear what people think about making a simple ISOLATED current sensor pcb for general scope use?

Questions, and initial suggestions as follows:


1)Frequency bandwith?  perhaps 1Khz be enough or maybe 5kHz?
2) Input ranges? ideally options for 0-500mA, 0 to 1A, and 0 to 10A
3) Absolute accuracy? hmm, good question see below
4) Resolution? Hmm, good question, see below
5) Circuit Topology & Current sensing method:  Hmm, see below!!
6) Input output connections - 4mm bananna inputs, BNC output, micro usb if "smart"??


In order to make the sensing isolated and to keep costs down, i'm going to suggest using an off-the-shelf hall effect current sensor, something like the Allegro ACS712:

http://www.allegromicro.com/en/Products/Current-Sensor-ICs/Zero-To-Fifty-Amp-Integrated-Conductor-Sensor-ICs/ACS712.aspx

http://www.allegromicro.com/~/media/Files/Datasheets/ACS712-Datasheet.ashx

That sensor, or a similar one, is obviously not an laboratory grade device, but i think will do the job for a low cost?  (any other suggestions for sensors - must be isolated)

The -3db bandwidth is claimed to be 80Khz for that sensor, and the datasheet shows reasonable linearity & noise values, although i have no experience of how these devices perform in the real world (could shield sensor on pcb with a nice low mU metal shield to limit stray magnetics etc?)


The the issue becomes one of absolute accuracy and output voltage sensitivity.    Obviously different scopes have different probe calibrations for voltage inputs that represent other parameters, so being able to select a suitable output sensitivity sounds useful too.  In terms of absolute accuracy, being able to calibrate the output also sounds useful, as it would allow users to perform calibration when accuracy was important (so long term drift would be less important in the basic hardware)
That points to using some sort of embedded micro to read the sensor output, and some sort of DAC to output a "corrected" value, or to buffer the sensors output with a variable gain amplifier etc.  Now this is where other, more experienced peoples ideas might be useful!  Using a micro with USB hardware also means the device could be a nice unit for direct data export to PC based assets etc?


What do people think, and what suggestions do you have?   

[void_error]

Not sure about the accuracy of the ACS712...
Another more costly but probably more accurate way to go would be and isolation amplifier powered via one of those small DC-DC isolated converters... just an idea...

[max_torque]

It is, of course, a trade off between cost and accuracy to a large degree, ie:


No way should i be trying to make a laboratory grade probe, you can already buy those.

No way should i be trying to make an el-cheapo chinese knock off probe, you can already buy those.


But, somewhere between those two extremes lies a gap.  A gap for a low cost, but useful bench current probe.  Kinda like Dave's Ucurrent, but aimed at the 0-5A or 0-10A range.  If it were open source, and made from easily accessible, easy to assemble parts then i could see it being useful to anyone with a scope.

The question is where to pitch it.  Using the off-the-shelf hall type sensors brings a relatively low level of absolute accuracy and precision (ie 21mV noise to 185mV signal @ 1Amp for the ACS712) although i suspect that can be improved on if the bandwidth is limited.  But what it does bring is low cost, and easy assembly and development. NO WAY do i want to get into a long, tedious analogue development process!   Which is why the Hall Sensor -> ADC -> micro _> DAC (or USB out) approach is interesting, and in effect i'll do all the "bodging" in software......  ;-)

Basically, like your cheap multimeter switched into "current mode" but able to send out that info to a scope or via USB to a pc.   That sounds useful to me??

[Richard Head]

If you are prepared to tolerate the inconvenience of a shunt then you can simply build a high slew rate DC amplifier (using op-amps) and connect it to your scope. This can be done up to a few hundred amps easily. The shunt can be easily cal'd using a 6 1/2 digit multimeter. The main drawback is no isolation.
If you want an isolated probe for AC only then you can use a simple current transformer, again connected to your scope via a burden resistor. CT's can have a very wide bandwidth if correctly designed.
If you want AC and DC with isolation and the convenience of a split core then it's probably easiest to buy a used LEM PR30 probe or similar. They are a little pricey but very convenient. They are good to 100khz. There are several options for probes that have a 50-100khz bandwidth. Yokogawa also make them as do others.
The next step up is to the Tektronix P6302 and P6303 current probes with an AM503 or AM5030 probe amplifier. This stuff is pricey but excellent (50Mhz bandwidth).
Dick

[tomizett]

If you're keen on using digital signal path (ADC/uC/DAC) then you could consider putting your isolation barrier on the digital side  - this way you avoid any problems having to build a wide-bandwidth linear isolation barrier.
If you want to isolate in the analogue domain, you could possibly combine techniques such as optical and transformer coupling. I had a bit of a play with such a circuit a while back; it never quite worked as well as I wanted, but certainly showed promise.

[cosmicray]

In order to make the sensing isolated and to keep costs down, i'm going to suggest using an off-the-shelf hall effect current sensor, something like the Allegro ACS712

Pololu has a line of current sensor micro-cards based on Allegro chips. I don't see the ACS712 being used, but they do implement the ACS709, ACS711, ACS714 and ACS715. Prices are reasonable for what you're getting.

http://www.pololu.com/category/118/current-sensors

[Seekonk]

There are always those who want bigger, better, more accurate.  Give them an extra bit and they will want another.  99% of the time you just need to know generally what is happening and when in a circuit. I remember back when you were thrilled just to get a 5% resistor.

[tszaboo]

If you are working with <5-10V common mode voltages, you can just put an instrumentation amplifier on our shunt, and set the desired gain. Use batteries to power it, I've used the USB port on the scope for power, with a small isolated DC-DC (like traco 0515D). There are also isolated shunt amplifiers from TI, AD, which are simple as a rock.

[Alex Eisenhut]

You mean this

http://www.testecvw.com/

?

There was another non-clip current probe a few years ago. Anyone?

[SeanB]

Simplest current sniffer is a ferrite toroid ( material depends on frequency, dust iron up to about 1MHz, then you can use low loss ferrites higher, or for low frequency permalloy) with a few ( 10 works fine) turns of thin wire ( PTFE is best but enamelled copper wire for lower isolation with some extra sleeve on the sensed wire) and a 50R resistor soldered across it, and then thin coax to the scope with 50R termination at the scope if needed.

Put wire to sense through the toroid and connect to scope. Smallest i made was one with a ferrite bead, it fitted the lead of a TO220 transistor that I wanted to see the current through the device to see why it ran so hot.

[max_torque]

Whatever i make it's got to be isolated, as doing that prevents a whole range of accidental foobars, even for experienced operators!  (like me    )

It's also got to be cheap, because otherwise, why not just spend a little more on a proper probe as mentioned up^^^ there a few times.

I mean, lets not kid ourselves, no one is going to be using some cheap, open source hardware they found on t'net to calibrate the Large Hadron Collider (or even a Small Hadron Colider for that matter).

It's got to be practical and useful, something to keep handy, that you can just plug into your scope and away you go.


The digital path (sensor->ADC->DAC) is interesting imo, especially at this low cost point, and i suspect it might be possible to whip up something with say a useable 25Khz bandwidth or perhaps even a bit more.  Using a cheap but low resolution micro with a reasonably fast ADC (say 10b) with a couple of gain ranges, also opens up the possibility to oversample for more resolution when you don't need the speed, and here, some noise is actually good thing (as long as it's "nice" noise!)


Basic costs would be:


1) power supply (Lipo or 9V battery & direct USB power)
2) Micro (small 8bit-er) not expensive
3) Current sensor - again using a hall sensor makes that cheap
4) DAC, probably the most expensive bit, say 12Bit
5) A nice-ish voltage reference, few opamps and maybe some digi pots to get some range changing capability
6) Connectors (2 x 4mm banana, 1 x BNC, 1 x USB)
7) Enclosure - lots to chose from, smaller the better

Could be an interesting little project, with a mix of analogue design and DSP stuff too!

[Kalvin]

Melexis, for example  MLX91206, MLX91208 or MLX91205, might also be possible solutions.

datasheets:
http://www.melexis.com/Assets/MLX91206-Datasheet-5942.aspx
http://www.melexis.com/Assets/MLX91208-Datasheet-6285.aspx
http://www.melexis.com/Assets/Datasheet-MLX91205-5505.aspx

I haven't been playing with these devices, but googling revealed that some people have been using these for isolated, low current (<1A)  measurements.

[tautech]

I yearned for any sort of current probe many years ago, a mate had a Tek P6022 (100 MHz) and hell was I envious.
Had a rat through some old SMPS PCB's in my junk pile and there was a nice little potted CT ~ 10 mm thick and 20 H &20 W.
Soldered some nice coax on the pins, BNC on the other end and presto, so I thought. 
The resultant waveforms were OK but how do you attempt to calibrate it?
Easy for AC mains 50 or 60 Hz, but who'd need a current probe for that. 

So thinking this through.....it's from a SMPS PCB probably running ~30 KHz but triggering a threshold for limiting so waveform reproduction mightn't be too flash.
So we set up a small test ~20 KHz and connected both probes in turn.
Added a shunt resistor across the CT pins and got a calibrated(rough) output.(10mA/mV)

Plenty good enough for what I needed a the time and I still have it if you need a photo.

But I spent a good amount of time later scouring fleabay for a Tek AC probe.
I got 2 over time, both P6021's,(60 MHz) 1 with the switchable termination the other with the Tek 134 amplifier. Each cost ~US$120 IIRC.

It's got to be practical and useful, something to keep handy, that you can just plug into your scope and away you go.

Been there, done that, either get a cheapy wun hung low(shit specs) or bite the bullet and get something you can trust and be proud of.
The Teks I got have at least doubled in value, you can't often say that about gear can you? 

Remember, you often use devices well ouside the DC region as this is where you need to ensure they are operating within their specs and you'd want accurate measurement, wouldn't you?

[nctnico]

Such a device already exists (I need to push my products  ):
www.eevblog.com/forum/buysellwanted/fs-new-cp2100-low-cost-current-probes-(dual-channel)/

[T3sl4co1l]

Well, if you can find data on the CT, you can calibrate it quite closely because the turns ratio is known.  Ideally, only the burden resistor (and any matching to a transmission line, if desired) need be calibrated.

If you want 100MHz bandwidth, you're probably better off with something like a Pearson than trying to muddle on with cheap off-the-shelf parts.  The average Triad CST-306 has a first resonant frequency around 300kHz, the intensity of which depends on the position of the sensed current.  The higher resonant modes are not as easily nulled by geometry, nor can they be damped in practical terms by using a large burden resistor (the resonant impedance is very high, in the kohm range).

The fundamental problem with a toroidal current transformer is: the helical toroidal resonator modes.  These are modes of the helical waveguide system, but with the helix' axis wrapped around in a circle, so that the helix becomes a torus, and therefore rather than having resonant modes corresponding to open ends, it has periodic boundary conditions.  This causes the first standing wave to be a full wave, rather than a quarter or half wave.  As with the [straight] helical resonator, higher modes are anharmonic (it is a dispersive transmission line).  Although the wavelength increase is handy, the fact that current transformers are generally ferrite-loaded means the resonant frequency is inconveniently low (however, modes N >= 5 or so tend to be smeared out or dampened away, so only the first few modes are a prime concern.)

You can attempt to dampen the resonator modes with:
- Series resistance (the burden resistor or otherwise?) or parallel loss (lossy core -- Metglas?), to dampen the resonances.  Note that, parallel damping competes with series damping (burden), and therefore impacts calibration (gain and phase shift) towards and beyond the first resonant mode.  In general, series burden is not sufficient, because you'll be using a burden much less than 50 ohms, while the resonant impedance is in the kohms or more.
- Restrict the modes by adding shielding, which has the effect of shorting nodes and antinodes together.  the ideal complete shield shape is a solid toroid shell, cut in half like a bagel, but only along the inside or outside edge, with the windings along the inside or outside surface, and the core further within.  This acts to turn it into a lower impedance transmission line with higher frequency resonant modes.
- Use fewer turns, so the first resonant mode is above the frequency of interest.  Unfortunately, you get low ratios this way (perhaps 1:20 to 1:50), so you may need to chain several CTs to get a practical ratio (over 1:100) with reasonable bandwidth.  The same rules apply to each CT, but the subsequent CTs can be smaller (fewer total amp-turns to deal with) so they can generally be made with much higher resonant frequencies for a given number of turns.

Tim