How is accuracy specified?

J

John Smith

Guest
Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?

Thanks for your help.

John
 
On Sat, 13 Nov 2004 18:02:09 GMT, "John Smith"
<kd5yikes@mindspring.com> wrote:

Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?

Thanks for your help.

John
Click to expand...
Depends on what the spec looks like. If you have a simple 1% accuracy,
then he is right, and you are out of spec. Can you show us the spec
you are working to?

d
Pearce Consulting
http://www.pearce.uk.com
 
On Sat, 13 Nov 2004 18:02:09 GMT, the renowned "John Smith"
<kd5yikes@mindspring.com> wrote:

Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?
Click to expand...
Yes, or use percent of span. Sometimes percent of reading is used as
*part* of an accuracy spec, but there's always another factor in
there.

Check out the specs of any number of instruments on the market.

But, I would carefully check your design. Usually if a device is to
meet accuracy specs without adjustment in production it will be MUCH
closer than the limits typically. More like +/-0.3%. Your device is
right on the edge.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
 
John Smith wrote:
Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?

Thanks for your help.

John
Click to expand...
You can always go for an accuracy specification such as maximum of 1% or
10uA- which is not uncommon and takes care of the low end.
 
John Smith wrote:

Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?
Click to expand...
I'm *almost* with you.

4-20 mA is a 16mA range so 1% accuracy would be +/- 0.16mA

If the target current is 8mA then it could be 7.84 - 8.16 mA. So you're in
spec.

The wording of the spec ( and his opinion ) may interpret things differently
though.


Graham
 
Jim Thompson wrote:

And I've had them come back a year or more later asking for me to fix
the mess they've gotten themselves into.
Click to expand...
And don't they *just* do that ! ?

Had to fix a few other contractors / employees cock-ups actually. Quite interesting
- especially when you read the design notes !


Graham
 
On Sat, 13 Nov 2004 19:03:15 +0000, John Smith wrote:

"Spehro Pefhany" <speffSNIP@interlogDOTyou.knowwhat> wrote in message
news:t2lcp0l8uaid0bc0b6d9kbbdf43vtahr1b@4ax.com...
On Sat, 13 Nov 2004 18:02:09 GMT, the renowned "John Smith"
kd5yikes@mindspring.com> wrote:

Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?

Yes, or use percent of span. Sometimes percent of reading is used as
*part* of an accuracy spec, but there's always another factor in
there.

Check out the specs of any number of instruments on the market.

But, I would carefully check your design. Usually if a device is to
meet accuracy specs without adjustment in production it will be MUCH
closer than the limits typically. More like +/-0.3%. Your device is
right on the edge.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers:
http://www.trexon.com
Embedded software/hardware/analog Info for designers:
http://www.speff.com

Hi, Spehro -

Yes, I knew it was a bit low. But, I had not counted on their interpretation
of accuracy. I guess I'm too naive for this kind of work.

Well, I'm not out much and it was fun working on it. I can still fix it
easily enough if I want to pursue a possible order of maybe 100 per year. I
just didn't want to order another set of prototype PCBs.

You're not supposed to have ordered boards anyway, until the bench lash-up
Click to expand...
works. ;-)

Good Luck!
Rich
 
"Spehro Pefhany" <speffSNIP@interlogDOTyou.knowwhat> wrote in message
news:8b2dp0935gdpp4dm84ce87k08906k517bh@4ax.com...
On Sat, 13 Nov 2004 19:32:59 GMT, the renowned
donaldun@spamfreepearce.uk.com (Don Pearce) wrote:


Not really, because 1% of FS has no meaning at half scale.

Sure it does. Output must be 12mA +/-200uA at 0.500V in.

It's not an uncommon spec.


Best regards,
Spehro Pefhany
Click to expand...
Try +/-160uA.

The span is 16mA and 1% is 160uA.

'It's not an uncommon spec.'

What's the fucking point then?

DNA
 
On Sat, 13 Nov 2004 22:26:29 GMT, Rich Grise <rich@example.net> wrote:

On Sat, 13 Nov 2004 19:03:15 +0000, John Smith wrote:

Well, I'm not out much and it was fun working on it. I can still fix it
easily enough if I want to pursue a possible order of maybe 100 per year. I
just didn't want to order another set of prototype PCBs.

You're not supposed to have ordered boards anyway, until the bench lash-up
works. ;-)
Click to expand...
---
More edicts?

Oft times (lately, that is (where've you been?)) the prototype boards
_are_ the bench lashup...


--
John Fields
 
"Fred Bloggs" <nospam@nospam.com> wrote in message
news:4196A007.9070809@nospam.com...
Genome wrote:
"Spehro Pefhany" <speffSNIP@interlogDOTyou.knowwhat> wrote in message
news:8b2dp0935gdpp4dm84ce87k08906k517bh@4ax.com...

On Sat, 13 Nov 2004 19:32:59 GMT, the renowned
donaldun@spamfreepearce.uk.com (Don Pearce) wrote:


Not really, because 1% of FS has no meaning at half scale.

Sure it does. Output must be 12mA +/-200uA at 0.500V in.

It's not an uncommon spec.


Best regards,
Spehro Pefhany


Try +/-160uA.

The span is 16mA and 1% is 160uA.

'It's not an uncommon spec.'

What's the fucking point then?

DNA



Well even 1% stinks- almost anything does 0.1% these days. Who contracts
for a custom 1%? Nobody- the product is not real.
Click to expand...
I think the product might be real.

I couldn't word a way of shafting the originality of a 4mA-20mA control loop
that has been adapted as an industry standard and is still adopted today.

Legacy, it's a word that needs redefining.

DNA
 
On Sun, 14 Nov 2004 00:32:05 GMT, the renowned "Genome"
<dna@nothere.net> wrote:


I think the product might be real.

I couldn't word a way of shafting the originality of a 4mA-20mA control loop
that has been adapted as an industry standard and is still adopted today.

Legacy, it's a word that needs redefining.

DNA
Click to expand...
You a legacy man or a boobacy man?


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
 
John Smith wrote:

"Pooh Bear" <rabbitsfriendsandrelations@hotmail.com> wrote in message
news:419678A9.50A700C9@hotmail.com...
Don Pearce wrote:

On Sat, 13 Nov 2004 19:54:08 +0000, Pooh Bear
rabbitsfriendsandrelations@hotmail.com> wrote:



John Smith wrote:

Sorry for such a basic question, but I know I'll get the right answer
in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC
to
4-20 mA DC. The person testing the circuit is telling me that it does
not
meet the 1% accuracy that is specified. At 0.25 V input, he is
measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA
(full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near
zero
volts input?

Am I right?

I'm *almost* with you.

4-20 mA is a 16mA range so 1% accuracy would be +/- 0.16mA

If the target current is 8mA then it could be 7.84 - 8.16 mA. So you're
in
spec.

The wording of the spec ( and his opinion ) may interpret things
differently
though.


Graham

Don't follow the logic. Suppose you bought a 100ft tape measure with a
1% accuracy, and it measured ten feet as nine feet, you would take it
back.

The point is that the 'zero points' for the input and output and not both
referred to ground - 0V - or even 0mA.

The spec should be clear as to the required output for 0V in. Clearly it
can't be
4mA +/- 0mA. This 'offset zero' makes the spec doubly tricky.

Except there is no spec !

Let the best man win.

Is there any trouble ( question for John Smith ) improving the linearity /
offset
?

No. I know how to improve it (technically). If I didn't, I could look in
AoE. The purpose of the question was to learn whether my knowledge of
accuracy (measurements) was screwed up or not. I now know (from the posts in
this group) that there are several different definitions of accuracy, and
the definition of interest depends on agreements and circumstances. A
redesign of the circuit and board will allow it to meet their specs.

Thanks, Don.
Click to expand...
I'm Graham actually but no sweat !

Good to see you have a fall back position. I hope this isn't a freebie for the
custard.

( custard = CUSTomer bastARD ) see Computer Shopper ( under development )

Take them for what they're worth. Sounds like they're looking for freebie
designs so screw them as far as they'll squeak.

Consider if they can find an alternative. If not - double / triple the price.

They look like they want to screw you - so you should screw them back !


Graham
 
"Spehro Pefhany" <speffSNIP@interlogDOTyou.knowwhat> wrote in message
news:14edp09h1eqs66g4ba0fev96eck3gcve05@4ax.com...
On Sun, 14 Nov 2004 00:32:05 GMT, the renowned "Genome"
dna@nothere.net> wrote:


I think the product might be real.

I couldn't word a way of shafting the originality of a 4mA-20mA control
loop
that has been adapted as an industry standard and is still adopted today.

Legacy, it's a word that needs redefining.

DNA

You a legacy man or a boobacy man?


Best regards,
Spehro Pefhany
Click to expand...
I'm an AcyAcy man.

DNA
 
Rich Grise wrote:

On Sat, 13 Nov 2004 19:54:08 +0000, Pooh Bear wrote:


John Smith wrote:

Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?

I'm *almost* with you.

4-20 mA is a 16mA range so 1% accuracy would be +/- 0.16mA

If the target current is 8mA then it could be 7.84 - 8.16 mA. So you're in
spec.

Well, it would have been at 1% of 20 mA, but 7.80 < 7.84. ;-)

Cheers!
Rich
Click to expand...
Ummm - so we agree that written specs are better ?


Graham
 
On Sat, 13 Nov 2004 20:45:29 -0500, Spehro Pefhany wrote:

On Sun, 14 Nov 2004 00:32:05 GMT, the renowned "Genome"
dna@nothere.net> wrote:


I think the product might be real.

I couldn't word a way of shafting the originality of a 4mA-20mA control loop
that has been adapted as an industry standard and is still adopted today.

Legacy, it's a word that needs redefining.

DNA

You a legacy man or a boobacy man?

If guys get titillated, do girls get cockillated?
Click to expand...

;^j
 
On Sun, 14 Nov 2004 02:12:34 +0000, Pooh Bear wrote:

Rich Grise wrote:

On Sat, 13 Nov 2004 19:54:08 +0000, Pooh Bear wrote:


John Smith wrote:

Sorry for such a basic question, but I know I'll get the right answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V DC to
4-20 mA DC. The person testing the circuit is telling me that it does not
meet the 1% accuracy that is specified. At 0.25 V input, he is measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA (full
scale). This means that I could be as low as 7.8 mA and still meet the
specs. Otherwise, how will I ever meet the specification at very near zero
volts input?

Am I right?

I'm *almost* with you.

4-20 mA is a 16mA range so 1% accuracy would be +/- 0.16mA

If the target current is 8mA then it could be 7.84 - 8.16 mA. So you're in
spec.

Well, it would have been at 1% of 20 mA, but 7.80 < 7.84. ;-)

Cheers!
Rich

Ummm - so we agree that written specs are better ?
Click to expand...
Yeah - I think that was everybody's first take on it - the rest of the
time we've been comparing different ways of spec'ing accuracy. :)

Thanks,
Rich
 
On Sat, 13 Nov 2004 18:15:49 -0600, "John Smith"
<kd5yikes@mindspring.com> wrote:


Is there any trouble ( question for John Smith ) improving the linearity /
offset
?


No. I know how to improve it (technically). If I didn't, I could look in
AoE. The purpose of the question was to learn whether my knowledge of
accuracy (measurements) was screwed up or not. I now know (from the posts in
this group) that there are several different definitions of accuracy, and
the definition of interest depends on agreements and circumstances. A
redesign of the circuit and board will allow it to meet their specs.

Is there no method of trimming manually? For 100 pieces, it might pay.
Click to expand...

RL
 
"John Smith" <kd5yikes@mindspring.com> wrote in message
news:B_rld.25549$KJ6.14787@newsread1.news.pas.earthlink.net...
Sorry for such a basic question, but I know I'll get the right
answer in
this group.

I designed and supplied a prototype for testing that converts 0-1V
DC to
4-20 mA DC. The person testing the circuit is telling me that it
does not
meet the 1% accuracy that is specified. At 0.25 V input, he is
measuring
7.84 mA when it should be 8 mA. He says this is -2% error.

I told him that the usual definition of 1% is with respect to 20 mA
(full
scale). This means that I could be as low as 7.8 mA and still meet
the
specs. Otherwise, how will I ever meet the specification at very
near zero
volts input?

Am I right?
Click to expand...
1% accuracy means that the answer you get is within 1% of the voltage
at the input. But there has to be a floor, so the accuracy would read
something like +-(1% of the reading + 1mv) e.g. If you feed in 0.25v,
the reading at the output will be from 0.2535 to 0.2465.

Norm Strong
 
John Fields wrote:
On Sat, 13 Nov 2004 22:26:29 GMT, Rich Grise <rich@example.net> wrote:


On Sat, 13 Nov 2004 19:03:15 +0000, John Smith wrote:


Well, I'm not out much and it was fun working on it. I can still fix it
easily enough if I want to pursue a possible order of maybe 100 per year. I
just didn't want to order another set of prototype PCBs.


You're not supposed to have ordered boards anyway, until the bench lash-up
works. ;-)


---
More edicts?

Oft times (lately, that is (where've you been?)) the prototype boards
_are_ the bench lashup...
Click to expand...
with smt you often dont have much choice!

Cheers
Terry
 
On Sun, 14 Nov 2004 17:55:01 GMT, the renowned legg
<legg@nospam.magma.ca> wrote:

On Sat, 13 Nov 2004 18:15:49 -0600, "John Smith"
kd5yikes@mindspring.com> wrote:



Is there any trouble ( question for John Smith ) improving the linearity /
offset
?


No. I know how to improve it (technically). If I didn't, I could look in
AoE. The purpose of the question was to learn whether my knowledge of
accuracy (measurements) was screwed up or not. I now know (from the posts in
this group) that there are several different definitions of accuracy, and
the definition of interest depends on agreements and circumstances. A
redesign of the circuit and board will allow it to meet their specs.

Is there no method of trimming manually? For 100 pieces, it might pay.

RL
Click to expand...
If you trim zero and span you'd have to try pretty hard (oscillation?
discrete parts?) to make it worse than so-called "0.1% accuracy" under
"reference conditions", even with carbon film resistors, an LM741 and
a zener reference.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top