Meter with two consumer-side connections

[JohnW2]

That interwebby thing must be full of info on how they work?

Most of what I've found so far simply make assertions similar, often almost identical, to what endecotp said, and which I don't yet understand. For example, as usual, one of the first hits was the Wikipedia, which says "A Bitstream or 1-bit DAC is a consumer electronics marketing term describing an oversampling digital-to-analog converter (DAC) with an actual 1-bit DAC (that is, a simple "on/off" switch) in a delta-sigma loop operating at multiples of the sampling frequency. The combination is equivalent to a DAC with a larger number of bits (usually 16-20)".

In the context of this thread, consider a meter designed to be able to meter a max peak current of about 70A (i.e. max of about 50A RMS). I would have thought that one would then set the 0/1 changeover point of the DAC at about mid-point - i.e. about 35A. To my simple mind, that means that when the peak current consumption is <35A (RMS <~25A), the changeover point will never be reached, no matter how fast one samples, so the output of the DAC will be a constant '0', regardless of what actual current is flowing.

All I can think of at the moment is that perhaps the "delta-sigma loop" actually alters the DAC's 0/1 changeover threshold each time it iterates through its looping. By so doing, I suppose it could effectively emulate an N-bit ADC - but I would then call that an "emulated N-bit ADC", not a "1-bit" one In effect, one would be time-multiplexing a 1-bit ADC is order to make it functionality of a N-bit one - but whether that would or would not result in a net reduction in the number of logic elements, I wouldn't like to guess/say!

Kind Regards, John

 

[ban-all-sheds]

http://www.beis.de/Elektronik/DeltaSigma/DeltaSigma.html

 

[endecotp]

Hi John,

I don't want to get too deep into this because it's more than a decade since I last properly understood it and I'll end up getting myself confused. But basically....

A 1-bit DAC is easy enough to imagine; it's just a switch - but that's not very useful. If you then put an analogue integrator (i.e. a capacitor) on the output you have something that can produce a useful analogue waveform.

To make an ADC, you can use a DAC of that sort plus a comparator and a feedback loop.

Pictures: http://www.beis.de/Elektronik/DeltaSigma/DeltaSigma.html (I haven't read all the words, but the waveforms look right.)

 

[JohnW2]

A 1-bit DAC is easy enough to imagine; it's just a switch - but that's not very useful. If you then put an analogue integrator (i.e. a capacitor) on the output you have something that can produce a useful analogue waveform. .... To make an ADC, you can use a DAC of that sort plus a comparator and a feedback loop. Pictures: http://www.beis.de/Elektronik/DeltaSigma/DeltaSigma.html (I haven't read all the words, but the waveforms look right.)

Thanks. Yes, I'm currently trying to get my head around it - BAS posted that same link an hour or two ago. I'll report back if/when I get anywhere (or don't!)

Kind Regards, John

 

[SimonH2]

ADC is not the same as DAC !
Yes, a one-bit DAC is easy enough - very closely related to Class D amplifiers.
SigmaDelta ADC is a different beast, and while they have their uses, they also have some significant drawbacks. In fact, if you look closely at the waveforms given in that article, you'll see one of them in plain view - if your oversampling rate is high enough, and your comparator window small enough, then you can get significant errors - note how there's a significant portion of the sine wave (leading up to, and around the crest) where the digital output doesn't change, because the sampling is too "coarse". It's a pity he didn't include the equivalent recovered waveform by feeding the bitstream into a DAC where you'd see this as a steady ramp before the waveform starts stepping down again.
Something else the article doesn't mention is what to do with the bit stream. The arrangement show would be OK if all you want to do is transmit the stream somewhere and feed it into a low pass filter to recover an analogue output. If you try using it to create a numeric output (by feeding it into an up/down counter) then you'll find it tends to suffer from problems caused by offsets in the analogue stage which mean that the bitstream doesn't have exactly 50% 1s, and so the counter will drift or down up until it saturates - and it'll bump along like this, sort of (partially) resetting itself by distorting each crest as it fails to count bits that would take it past the limits of the counter.

Another arrangement is to feed the digital output into an up/down counter clocked by the clock. Then the output value ramps up/down according to whether the signal is higher/lower than the feedback loop. Remove the analogue integrator from the diagram as the counter is now a digital integrator. The DAC in the feedback loop now becomes a multi-bit DAC of whatever number of bits you need in the system. Any offset in the analogue stage simply creates an offset in the digital representation.

One key detail about this type of DAC/ADC ...
They require incredibly precise and stable timing. While multi-bit converters need incredibly precise analogue components (such as a ladder of very closely matched resistors) - any difference between switch-on and switch-off times in any of the logic, or any variation in clock timing, directly feeds into the converted signal. Eg, if a gate takes slightly longer to switch on than to switch off, this will cause the output to drift as a few "on" bits will get flipped to "off". This shifts some of the accuracy requirements from the analogue to digital domain.

On the upside, no sample/hold stage is needed since conversion is continuous - at any time, the value in the counter is the representation of the analogue input up to that point in time. So instead of your software triggering a conversion, waiting for the conversion to take place, then reading the result - it just reads the counter.

But as discussed previously in the software vs dedicated chip subthread - there are many tradeoffs/design considerations. There most certainly is no such thing as "the best" converter type - only the best compromise for any given situation.

 

[oldsparkyuk]

Poor old RickDastardly, he only asked a question and look where it has gone, bet lots of you don't even know what the question was about now.

 

[stillp]

What question?

 

[ban-all-sheds]

Deleted

 

[John D v2.0]

Diynot is a forum where we can all come together to learn something, enjoy our time, and help other people. Oh and if you like, have a pedantic discussion about who's insulted who as if our life depended on it...

 

[John D v2.0]

You make it sound so bad! But when it's random people on the internet I fail to get excited.

 

[ban-all-sheds]

I'm not going to drag out an argument with you. ***********

so let's leave it there.


----------------
You're right, you're not. With anyone. Unless you want the ban to be permanent
Mod
----------------

 

[RickDastardly]

Poor old RickDastardly, he only asked a question and look where it has gone, bet lots of you don't even know what the question was about now.

Tell me about it Even I've forgotten what I asked...

Bits of the thread have been interesting though... I was thinking about making myself a little kWh meter to check things out around here and see where I can save a few quid. The cheap ones in the shops always seem to get bad reviews for accuracy ('cos they're cheap I guess).

I've decided it's not worth pulling the iron out of retirement though, even though I have more than enough bits to make a few with up in the loft; both fully-MCU driven and various alternatives. Could even roll an FPGA solution if I got truly bored. With and without an embedded CPU core

Ten minutes with a Fluke true RMS multimeter is quite enlightening though. Some of those wall-warts powering various bits of kit around here are horrendously inefficient!

 

[endecotp]

Ten minutes with a Fluke true RMS multimeter is quite enlightening though. Some of those wall-warts powering various bits of kit around here are horrendously inefficient!

Beware, measuring current - even "true RMS" current - is going to give significantly wrong answers for loads with much capacitive or inductive component. A transformer (*) with little or no load on the secondary presents a near-perfect inductive load to the supply. You'll measure significant current, but it is almost 90 degrees out of phase with the voltage, so no power is consumed.

The characteristics of a switch-mode supply are more complicated.

(*) Yes, an actual transformer with coils of wire!

 

[SimonH2]

Beware, measuring current - even "true RMS" current - is going to give significantly wrong answers for loads with much capacitive or inductive component

Indeed, which is why those clip on "meters" can give such "interesting" results.

The characteristics of a switch-mode supply are more complicated.

Can often be capacitive due to filtering components if the manufacturer didn't omit them from the production units
Some of the plug in units can be quite good. Accuracy might not be too hot, but at least they can show you real-power, VA, and power factor which can be quite revealing. Cheap wall-warts often take a few VA and have a carp power factor when unloaded - which'll be why there's that urban myth that they always take their rated power even when not charging/powering something

 

[ban-all-sheds]

deleted