HF current pulse measurements

[danmcb]

I went the Tektronix 2230/2232 route for the reason you identify, low repetition rate signals.  Except for lack of isolation, they work great for switching power supply startup problems.

Tektronix also made the 2210/2211 which is a 2225 with added digital storage capability.  I would not normally recommend them because you have a DSO now and they lack peak detection but they would work for something like this and at least in the US can be had for less than $100.

I think the CRO/DSO combination is absolutely ideal for what I do. I do mostly audio or audio related, but sometimes there is microcontroller stuff and so on involved. But when there is, it is usually at lowish frequency (this project has a PIC, I clock it at 500kHz because it just doesn't need more). The only reason I might upgrade from here is if I start working on high speed digital at some point.

[David Hess]

I think the CRO/DSO combination is absolutely ideal for what I do. I do mostly audio or audio related, but sometimes there is microcontroller stuff and so on involved. But when there is, it is usually at lowish frequency (this project has a PIC, I clock it at 500kHz because it just doesn't need more). The only reason I might upgrade from here is if I start working on high speed digital at some point.

As a low cost option for someone who can maintain an old instrument, they are acceptable but the same applies for any test instrument of a similar age.  Otherwise if you can live within their limitations, a modern cheap DSO is usually a better option unless the old instrument provides a unique capability like high bandwidth, high sensitivity, or accurate RMS noise measurement.

[T3sl4co1l]

Well, because the C is directly across the voltage source, the instantaneous current is theoretically infinite!

One would be a pretty poor theorist to ignore an entire fundamental property of the universe -- magnetism!

In fact, source impedance, wire impedance, and so on, give the limit, but it is going to be pretty high. And of course, the finer the time resolution of the measurement, the higher the current you are likely to observe.

The current peak is given, to second order accuracy (hard to beat that! ), by adding the source resistance, and wiring inductance.

In practice, the source won't be an ideal voltage source, but the lowest impedance it has will be, guess what -- another capacitor!  So to be accurate, one must have the C and ESR of that, and connect it in series with the load side ESR and C.  (If Cload >> Csource, then it could take so long to charge that the transformer, rectifier, and maybe AC mains source, have to be taken into account.  But the current will be much lower.)

ESL is the tricky one, though.  You can guesstimate stray inductance based on wiring length: the inductance of free space (that is, the inductance due to a current simply flowing through a distance) is 1.257uH/m (more correctly, mu_0 == 4*pi x 10^-7 H/m).  The inductance of most cables is around 1/2 to 1/4 of this (because the magnetic field inside a cable isn't in free space, but it's trapped between conductors that oppose!), so you can guesstimate 0.3 to 0.6 uH/m.

So, a typical 2m cable is on the order of 1uH, and paired with 47uF, the sqrt(1uH/47uF) ~= 0.14 ohms.  (So, ESR > 0.14 ohms gives good damping, and the peak current from 24V will be under 160A.)

But this has been a really instructive exercise, as it led me to the (now obvious) idea of putting an RC snubber right on the input, to limit voltage (damping oscillations) without too much of a current pulse. And the DSO is invaluable as it allows me actually find good values of R and C by observation and measurement, rather than guesstimating, which is where I'd be at without it. Great tool!

Yup.  And, you also know the value of RC required: the C needs to be more than double the non-resistor C value, and R needs to be equal to sqrt(L/C).

sqrt(L/C) shows up so often, it has a name: characteristic impedance, Zo.  Note that it does indeed have units of ohms (Google Calculator will show this works).  Which should tell you something interesting will happen when you add a resistor of similar value to the circuit!

In practice, electrolytic capacitors have enough ESR to avoid ringing except for very long cables (but then, the cable resistance itself may be significant, too).  The inrush current can be quite large, so it needs to be directed away from sensitive circuits.  Tantalum capacitors have a range of ESR available (so can be used for damping, or can be prone to more ringing), but more importantly, shouldn't be exposed to surge currents and voltage spikes, which can ignite them!  Polymer caps usually have very low ESR (like a tiny little 25V 47uF aluminum polymer having an ESR of 20mohms!), so they can cause big problems with ringing in power supplies and have to be used appropriately.

Nothing at all more complicated than algebra-level arithmetic!

Tim

[danmcb]

Thank you Tim!

well, as is often the case, there turned out to be more to this than meets the eye. The "more" in this case being the earth loop from the PSU being hot-plugged to the UUT via the scope.

To summarise - the sense resistor being wirewound was not the issue. I still got too high current pulse readings using an 0.1R carbon resistor instead of the 0.05R wirewound. Readings up in the 20-40A range. With a 3R3 resistor in the RC snubber. Something is not right.

The issue turned out to be the earth loop.  Although it measures 0R, it is also quite inductive. When I used a short length of thick wire to short out the earth loop (tricky, holding it on the barrel of the DC jack plug while plugging), the readings became honest. Now I got readings with my sense resistor that tally with the expected 8A or so of initial charge current.

I was also curious about the lead inductance, and decided to try to measure it. What I did was,  instead of the RC snubber, connect a 470pF 1% cap. Big enough to swamp the 14pF or so of scope probe capacitance, not big enough to kill the ringing when hot plugging.

Then catch the voltage waveform with ringing, and measure the frequency. Rearranging the well known tuned cct formula, we get:

L = 1 / (C * (2 * pi * f)^2 )

With the UUT ground NOT directly connected to the PSU ground (i.e. with the earth loop inductance in place) I got about 2.6MHz, which equates to 8uH.

With the shorting link in place, the frequency goes up to about 10MHz => 540nH! OK, the ground  loop as an RL cct may be oversimplified, but it suggests that there can be quite some reactance, which can resonate and create a lot of extra current.

In fact, I think there is maybe even more to it than that - I noted that it was hard to get ringing AT ALL if I didn't leave the whole circuit to settle down for a while before hot plugging (even if I took care to discharge the cap). I don't know why that should be.

But all of this is enough to persuade me that HF measurements are very tricky with the scope earth loop.

This is why I like bench work - you don't always get all your questions answered, but you find out a lot about what does and doesn't work, and things to look out for. In this case, it is pretty useful, as the charger might be hotplugged into a unit which is, or is not, grounded already. So good to know that this can have a big effect on the current pulse magnitude. Never would have picked that from a simulation, I think.

Thanks all!

[T3sl4co1l]

Moral of the story:
There is no such thing as absolute "ground", at AC.

Everything has a nonzero impedance.

The only point where "ground" is true, is at the point it's defined.  The scope defines its input connector as ground, so that's also what you're measuring with respect to.

Any voltage drop in the probe cable, ground clip, and other outside wiring, is part of the circuit you're measuring: it ain't ground!

To avoid the ground loop, simply short it out better.  Instead of hot-plugging the barrel jack, leave it connected, with the ground connected straight in.  Switch only the "hot" wire.

0.54uH sounds perfectly plausible.  Like I said, about that per meter, give or take whatever the exact cable length is.

8uH also sounds reasonable for an about room-sized ground loop.  The inconsistency you saw may have to do with the two contacts of the barrel jack touching at different times, or voltage between the ground points (ground loop voltage, which is induced in the ground wire due to mains current flow).

Tim

[danmcb]

indeed! You always have to think about what you are measuring, and with respect to what. If I was really bothered about refining this, I'd have cut the connector off the charger, soldered a new one on with an extra short wire ... etc ... but really I just wanted to figure out what was going on. Measuring 25A via a 3R3 resistor  and 25V just ain't right! The extra voltage of course being across the "ground" connection.

It kind of sucks that most differential probes are silly money, because they are designed to also provide isolation. What is needed is a diff input that has reasonable protection against overvoltage, a sensible stepped attenuator (pretty much like a standard scope - and really good CMR. 50MHz BW would be fine for many apps, but it would get rid of the troublesome groundloop issue. Things are the way they are because they've always been done like that, but they are not really as good as they should be.