AC...DC Waveform

[Alan Douglas]

Ah yes, you would have to commutate the magnetic field to get a unidirectional current. I suppose you could predict this by the right-hand rule, but it's more instructive to guess and get it wrong.

[Nick]

What would be the DC offset of a 5 volt pulse sitting on a common rail of zero volts.

Rather than DC offset think of it as the equivalent DC value. There is no such value for a non-periodic event (think about it) but generally the solution here is to calculate the RMS value which is simple enough with an opamp and a few oyher components or a dedicated chip (e.g. AD8436) - its the integration over time of the area under the curve with respect to some reference point, typically 0V. In practice this is generally done over a few cycles of the input and continuously recalculated using a sliding window i.e. it's approximated over a set time constant.

The simplist way to think about the RMS value of a waveform is that it is the DC equivalent of the waveform such that when applied to a pure resistance it produces the same amount of heat as the waveform under test. Indeed, this is exactly how early true-RMS meters worked... they measured how hot a resistor (or equivalent) became when the arbitrary waveform was applied and compared that with values measured when heated by a known DC value (there are a number of ways of doing this)...

Most modern cheap DVM/DMMs which purport to be "RMS" or "true-RMS" meters (both should be the same really) do no such thing as

1. they are AC coupled so ignore any DC component
   
2. they don't integrate the input properly anyway (they assume a sine wave)
   

"proper" true-RMS meters like those from Tektronix, Agilent or Fluke etc. optionally include any DC component and thus will correctly compute the real RMS value... The DC component is removed to limit the input dynamic range of the converter, the RMS value of the varying part of the waveform calculated, and then the DC part is added back in...

[Mark Hennessy]

Measuring AC is surprisingly difficult. Measuring "true RMS" is even harder

For example, just compare the DC and AC specification of a good DVM - let's take the Fluke 87V: 0.05% vs 0.7%

The USP of the heater method is that it works over a very wide bandwidth. Very important to know the BW of your test gear (a fairly pedestrian 20kHz for the 87V - the 187 does 100kHz).

As far as I'm aware, you can't build an RMS rectifier with just a single op-amp; the best you could come up with would be "peak or mean sensing, RMS calibrated", meaning that the output is simply scaled to give the right RMS answer from whatever detection method you employ. As such, it becomes inaccurate with anything that isn't a sine wave. To do genuine RMS detection, you need an IC containing very complex analogue circuity - usually employing analogue multipliers and other such gems to do the root-mean-square "calculation".

They are like miniature analogue computers! For years, the AD536 was commonly used in precision voltmeters - indeed, I note that it is still in production. They were even used in audio - some compressor/expanders used them to measure the amplitude of the signal, and it was claimed that RMS detection gave better results subjectively than peak or mean sensing. Maybe, maybe not, but using one of these ICs was a lot easier than building an op-amp based rectifier, if you didn't mind the much higher BOM cost. But I digress... Have a look at the datasheet - it's fascinating stuff:

www.analog.com/static/imported-files/data_sheets/AD536A.pdf

Back to multimeters, they do still seem to do the RMS conversion in the analogue domain - perhaps surprisingly in this day and age. So if you see "True RMS" flagged on the front panel, you will find a dedicated chip in there doing just this - you can verify this by looking for a specification for "crest factor". In my experience of Fluke and similar models, they do AC-couple the converter. The higher-end models - such as the 187, 189, 287, 289 - do have AC+DC modes, where they display both quantities separately, or add them together for you. Very nice, occasionally very useful. I also have that facility on an old Tektronix DMM914, and the Fluke 45 (a basic bench meter).

So in general, it seems that most DMMs AC-couple the rectifier - true-RMS or not - and I don't think that's a massive problem. Should you really need to do the measurement on a machine that doesn't have the AC+DC mode, there's nothing stopping you taking both a DC and an AC measurement, and manually summing them.

It's worth saying that a multimeter connected to Lawrence's pulse would give readings on both AC and DC - the accuracy and utility would depend on both the signal and the instrument, but it proves beyond all reasonable doubt that we have a mixture of AC and DC signals here

[Alan Douglas]

I have a couple of older meters (a Ballantine and an HP) calibrated for true RMS, but as a sanity check I've been known to connect a hot-wire voltmeter which by definition reads true RMS at any crest factor you like. It does of course load the circuit down.

[pwdrive]

In all of this I still can't see where the flow of current is reversed on the waveform.

Lawrence.

[Kalee20]

In all of this I still can't see where the flow of current is reversed on the waveform.

That's down to semantics then. You can call a waveform whatever you like. One which changes polarity symmetrically about zero, most people would call AC. One which stays always positive, but rises and falls cyclically without ever going negative, I suppose isn't 'alternating', but it is convenient to considerit as having a steady average value, with wiggles which add and subtract from this average. And it's this wigglyness which gets called the AC component of the waveform.

Sometimes it does not pay to take abbreviations too literally. Describing the mains as 240V AC is horrid if you expand AC into alternating current. But if you regard AC as behaving like an adjective, confusion disappears.

[Alan Douglas]

Sometimes it does not pay to take abbreviations too literally.

There's some sage advice. Just because someone called it that, 130 years ago, doesn't mean you can't think of it in some other way.

[Nick]

The AD8436 is one of the more modern single-chip solutions (http://www.analog.com/static/imported-fi...AD8436.pdf) and is accurate to 0.25% of the reading - not bad really for something with a -3dB point at 1MHz costing between USD 3.5 and USD 5 in one-off quantities...

I, too, have a DMM916 and I love it - built like a tank. Proper meter.

[Mark Hennessy]

In all of this I still can't see where the flow of current is reversed on the waveform.

I really don't know what I can add. All I can do is reiterate:

Any signal consists of an AC component and a DC offset. That's it. If this wasn't true, electronics would be impossible!

As I've already said, into a purely resistive load, the current doesn't reverse. But your signal still has a not insignificant 5V peak-to-peak AC component.

In real-world circuits, the current may well reverse. But usually, the current isn't important anyway; most often it's the voltage that equipment input stages look for.

Perhaps AC is a term that should be reserved for talking about power engineers, and perhaps electronic engineers should say "signal" instead. However, we use both synonymously, and it's not a problem because it's usually clear from the context.

[pwdrive]

In the context of the original post I will contact the bloke who asked me the question and tell him it's pulsed DC, I will include the link to the text book I found and he can take it from there.

Lawrence.