Meter with two consumer-side connections

[JohnW2]

John I linked to a datasheet in an earlier post. It has two sigma-delta ADCs for voltage and current, and then fixed-function DSP that multiplies and integrates - plus some other clever stuff like a high-pass filter that eliminates any DC offset in the current sensing. The result of that is a digital signal of some sort, e.g. a pulse per Joule or something. That's not a microprocessor.

Thanks. I think there may simply be a terminological issue here. I may be wrong, but I would have regarded "a DSP that multiplies and integrates" as a "microprocessor"! Indeed, when I typed DSP into Google, the first hit I got (Wikipedia) said:

A digital signal processor (DSP) is a specialized microprocessor (or a SIP block), with its architecture optimized for the operational needs of digital signal processing.

It surprises me that Analog Devices have gone to the trouble of producing a whole range of chips to do it in this way if it's also possible (and presumably cheaper) to do it with a "bog standard" microcontroller and its regular ADCs, but what do I know?

Good point - and, as you say, there presumably has to be a reason. That Wikipedia article went on to say that dedicated DSPs often have much better power efficiency than do general-purpose microprocessors - so I wonder if they have perhaps been persuaded by electricity suppliers to minimise the power consumption? Although, on the face of it, the amount of power involved would appear to be trivial, I suppose it could be more of a factor when multiplied by millions of meters!

Kind Regards, John

 

[JohnW2]

A very quick price comparison at one off quantities gives: AD71056 $3.11
ATmega48 $1.45 ... Of course with the AD71056 you still need a microcontroller for the ir comms and display driving plus reading storage, so I guess it's a no brainer (as they say).

Fair enough, but couldn't is a bit of a vicious circle (or 'chicken and egg')? ...i.e. might not the much lower price of the latter simply reflect the fact that, for whatever reason, the large-volume buyers have decided to go down that route?

Kind Regards, John

 

[DetlefSchmitz]

Fair enough, but couldn't is a bit of a vicious circle (or 'chicken and egg')? ...i.e. might not the much lower price of the latter simply reflect the fact that, for whatever reason, the large-volume buyers have decided to go down that route?

Kind Regards, John

I guess this isn't your field. The Atmega48 has been a very high volume device for a long time. It is capable of producing the correct functionality. The AD device could never reach the equivalent volume, as the only market is metering. Volume is the key to pricing.

You have to compare the saving produced by the designer not being prepared to implement the required algorithm against the cost of a low volume extra chip. You can't blame AD for trying, they make some great chips and I use them in high value compliance designs. But in this case they got it wrong.

 

[endecotp]

DSP just means digital signal processing.

You can do DSP using fixed-function logic, or using a general-purpose processor, or using a processor that is programmable but optimised for this sort of task; the latter is what that wikipedia article is referring to.

The most obvious applications for fixed-function DSP are those where the performance requirements are high and hence difficult to achieve in a software, even on a "DSP processor". Examples include encoding/decoding (including error detection and correction) for high-speed RF, i.e. mobile phones, wifi, digital TV etc.

Probably the most familiar applications of DSP processors are for video, i.e. for decoding an MPEG data stream to pixels in a digital TV. Here the fact that they are programmable is useful as they can be reprogrammed to handle new data formats.

A decice is not a "microprocessor" unless it has a stored program.

 

[JohnW2]

DSP just means digital signal processing. ... You can do DSP using fixed-function logic, or using a general-purpose processor, or using a processor that is programmable but optimised for this sort of task ....

Indeed but, probably wrongly, I would normally call all of those "microprocessors" (which, after all, literally means something very small that 'processes' )

A decice is not a "microprocessor" unless it has a stored program.

That is what I suggested you were probably saying, and it illustrates the difference between our respective usage of the terminology. To me (again, perhaps wrongly), a microprocessor can either have, or not have, a stored program, but can also have 'hard-wired' instructions (which, as far is the chip is concerned, is surely not materially different from having a {permanently} 'stored program'?).

I don't think we are disagreeing about any of the concepts - but I may be using incorrect terminology!

Kind Regards, John

 

[JohnW2]

I guess this isn't your field. The Atmega48 has been a very high volume device for a long time. It is capable of producing the correct functionality. The AD device could never reach the equivalent volume, as the only market is metering. Volume is the key to pricing.

It's certainly not "my field", but isn't that exactly what I said?? - that the reason for the Atmega48 was much cheaper was probably because "high volume buyers" had chosen to purchase it, not because it was necessarily any more/less suitable than the dedicated metering product (for the metering application). As you say (and as I was implying) volume is the key to pricing.

Kind Regards, John

 

[endecotp]

I may be using incorrect terminology!

Yes.

 

[DetlefSchmitz]

The thing behind this is that a vendor with a speciality will often try to package this in a device on the basis that the designers don't have the technical capability to do the job with already available tools. AD assumed that the designer wouldn't be able or willing to implement the power calculation algorithm.

There is an interesting example I recently investigated was the MSF time signal decoder chip. Endless vendors have produced dedicated chips, yet virtually none is currently in production. The point being that a dedicated chip is not needed as the functionality only needs an LF amplifier plus a microcontroller. Modern education seems not to train engineers for the task.

 

[JohnW2]

The thing behind this is that a vendor with a speciality will often try to package this in a device on the basis that the designers don't have the technical capability to do the job with already available tools. AD assumed that the designer wouldn't be able or willing to implement the power calculation algorithm.

If that's the explanation, I would have thought they must be fairly naive - there is a lot of expertise out there, some ready and waiting to be 'bought in', even if they don't have the requires skills in-house.

Kind Regards, John

 

[endecotp]

There have been a couple of threads here where people have discovered that their electronic meters have malfunctioned, and in the worst possible way i.e. by recording power that they haven't actually used. I'd be interested to know if the meters used in those cases were implemented using chips like the Analog Devices ones, or by some engineer at the meter company doing it himself.

 

[JohnW2]

There have been a couple of threads here where people have discovered that their electronic meters have malfunctioned, and in the worst possible way i.e. by recording power that they haven't actually used. I'd be interested to know if the meters used in those cases were implemented using chips like the Analog Devices ones, or by some engineer at the meter company doing it himself.

Interesting question. However, isn't it probable that these (presumably very rare and sporadic) malfunctions are consequences of a hardware fault (most likely a chip malfunction), rather than a systematic error in how the system had been implemented or programmed?

Kind Regards, John

 

[JohnW2]

Yes.

Fair enough. I'm very willing to be educated, so perhaps you could start by telling me how you would define a "microprocessor".

For what it's worth, if I had been asked to have a stab at a definition, I would probably have said something along the lines of .... "a complex integrated circuit ('chip') which can accept digital input(s) and provide digital output(s) and which can undertake digital processing/ manipulations in accordance with a set of provided instructions". Would you regard that as 'way off'?

Kind Regards, John

 

[endecotp]

in accordance with a set of provided instructions

The point is that a fixed-function digital IC doesn't have a "set of instructions", just as a PCB full of discrete components doesn't have a "set of instructions".

 

[JohnW2]

The point is that a fixed-function digital IC doesn't have a "set of instructions", just as a PCB full of discrete components doesn't have a "set of instructions".

Ah, I think I'm coming to understand my misunderstading

It all really comes down to the nature of the innocent-looking little bit of the AD71056 called "multiplier" in its block diagram. You are implying (I suppose actually saying!) that this multiplication of the digitised voltage and current signals is undertaken entirely by 'hard-wired' logic on the chip. If that is the case, such that the functionality of the chip is totally unchangeable, then I agree that even I would not call it a microprocessor!

I was thinking, probably almost subconsciously, that implementing (presumably fairly high precision) multiplication of two binary numbers by 'hard-wired logic' would be 'too much like hard work' (I cannot forget building a very rudimentary 'calculator' out of discrete logic elements in my mis-spent youth!), and that they would therefore probably have included some very simple standard programmable processor 'module' on the chip, along with a set of hard-wired (or burnt-in, or whatever) instructions to tell the processor how to do the multiplication - maybe not dissimilar to what one might find in a calculator chip. If that had been the case (which I now realise it probably isn't!), then I would probably have called the whole chip a 'microprocessor', even though the 'program' was not user-alterable.

Apologies for any confusion I may have introduced!

Kind Regards, John

 

[SimonH2]

There are advantages both ways - but I suspect that the dedicated energy monitor chip is more accurate and needs less (and less complex) software. In terms of development, the meter manufacturer can leave the precision stuff to the chip, and have the processor just read it regularly and tot up the consumption (and drive the display etc).
Doing it in software needs "quite a bit of care", in particular timing is critical as you need to keep the read-calculate-accululate loop running at a constant rate regardless of what else the processor is doing.

On and off I've been looking at making a multi-channel energy monitor - it's of interest both at work and home. Never had time at work, and time or money for home One board I've had my eye on is this one with 16 channels of dedicated hardware