The AD737 data Sheet mentions that the two 8k resistors inside the AD737 can vary from 6.4k to 9.6k.
What I couldn't seem to find in the data sheet is how well matched they are to each other.
The 8k resistors aren't matched.
Being "not matched" requires a much larger trim-pot range (3k+) for scaling - as in production - the two could be at extreme opposites. The RMS-to-DC Conversion Application Guide 2002 (figure 12) hints that a 1k adjustment (499R fixed + 1k variable) will suffice.
It makes me sort-of wish that the Cc (pin-1) resistor be external to the package - so I can used 0.1% resistors in both places. Whether that would mean a scaling adjustment is necessary I don't know.
By the way for anyone interested. Both the application note and datasheet for the AD736 and AD737 have a few schematic errors - that can be confusing (sorry - too many to list here). On the other hand, there are some gems of information in there if you choose to work with these parts.
If one assumes that the trim metric is conversion accuracy, not absolute resistor values, then the external component values shown in figure 12 are quite sufficient.
I'd be interested in the schematic errors, if there are 'too many to list' that sounds like they could use some attention, so perhaps you could list just a few, say, in your judgement the top 5 most egregious of the lot, and we'll see about fixing them..
If you would like to select the I-to-V and V-to-I resostor values, the new AD8436 has optional pins for this purpose.
Probably better not to publish this.
In fact you may decide to remove my comment about “errors in the document” from my previous posting – or simply refer to errors in the older “RMS to DC Conversion Application Guide”.
There’s a bit of “egg-on-my-face” as the AD737 data sheet appears to have no noticeable errors. But then I have only glanced at it – where-as a few months ago I spent many hours repeatedly going over the documents trying to make sense of it. For some reason, what I recall is errors and confusion.
In reality, I have pretty much spent most of my time focused on the October 2002 Appendix D New Products Appendix to the RMS-to-DC Conversion Application Guide. It was a really helpful document but contained the errors. See way below for my comments.
I also built two versions – but my version was a dual supply type. So the RMS/Average Value Measurement circuit just didn’t behave properly and I had to do a significant redesign. The second version worked properly – with the scaling trim-pot being my only issue.
I’m not locked in to the AD737 and will take a look at the AD8436 later. It does AVERAGE as well – which is something I wanted. Thanks for directing me to it.
From the “RMS to DC Conversion Application Guide”.
Page 3 – Figure 2. The top right hand connections (7 and 8) are swapped.
On some of the drawings, the term AD736/AD737 is used generically, yet pin-3 is shown as POWERDN in some and as Cf in others. See page 5 figure 5 & 6 for an example. It can be confusing when you first look at it.
Page 7 - Figure 9 has quite a few errors - but they have been corrected in the AD737 datasheet. However, I didn’t use the data sheet as my reference - as there was no text to describe that particular circuit. All the good stuff was in the Applications Guide.
These stand out - but I also had confusion while reading these documents too. Possibly part of getting familiar with them.
Things I would have liked to see.
A comment about the two 8k resistor match – even if it is to say “they don’t match”. I still find this a bit strange, as I would have thought the two resistors might use the same die “area”, and at least be “close” to each other in value - even if that value may be somewhere between 6k4 and 9k6.
Page 8 of the guide: paragraph “The rms converter IC has two inputs:…
It mentions a high impedance input (voltage?) and a wide dynamic range input (Current?). But it never expanded on why one had a “wide dynamic range” and the other didn’t. That would be a good aspect to expand.
Data sheet page 1 product highlights item 1. Capable of computing Average Rectified, Absolute or True RMS. That is three different things. But it never really talked about Average Rectified again – unless the terminology changed. Stuff is covered on page 12 of the data sheet – but they talk about Average Absolute. I would like to have seen something that mentioned these three methods in depth.
Thanks for your help
Welcome to the human race. I've had so much egg on my face in my lifetime I have developed a fondness for eggs-benedict .
As you probably guessed by now I'm the ADI pps engineer for rms-dc. I asked about ds errors because I can do something about fixing them, and you'll notice the data sheet has been revised many many times, mostly by me. However I was'nt here when the device was developed so I don't have the kind of historical insight to accurately respond to certain design intricacies.
As you point out, resistors are indeed designed with equal geometry, however certain products benefit from trimming where the values are adjusted with a laser. However as with any resistor, the values can vary quite a lot in manufacturing, and the balance of the circuit can contribute conversion errors as well, so the trim ends up being dependent on conversion error, not resistor absoluete value or match. I'm somewhat familiar with the AD8436 manufacturing process, but not familiar with the AD736/7, so I'm extrapolating my knowledge, basically it's an educated guess but it's the best I can offer.
I'm not surprised you found errors in the apps guide. I don't receive much feedback on apps guides, and so concentrate on data sheets because they get a lot of scrutiny inside and outside the company. I used to be an ADI customer, and now work for the company so I know a little bit about what designers would like to see, so I keep an red-line mark-up in a notebook and change the data sheet every so often. We did add average rectified notes in the AD8436 (pg 16), and in the AD737 (pg 14), but not yet in the AD736. (this ds is about due for a revision so I made a note to add a notre.) Average rectified is preferred bysome custoemrs but I've not learned why as yet.
Thanks for the feedback!