"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...
the consequence may then be : a minimum Avol is required to achieve that maximum percent linearity error relative to full dynamic range. ( and here I assumed a few implicit meanings..., disregarding the dynamics, and only looking at the steady state or DC "final value".)
Jure Z.
So just overkill a 0.25% best-straight-line fit by 10:1 or so, and they can't complain.
John
John Larkin a écrit :
Sure thay can:
... " That's overkill! Couldn't you have done it cheaper?"
"Not and be sure of maintaining production yields".
Best regards, Spehro Pefhany
They know HOW to multiply and divide.
They just don't know WHAT to multiply and divide.
:-)
Mark
You've got big problems with that DC gain of 300 assuming reasonable full scale output range specified, seeing as how most of your stuff is LV MOS. The parrots seem to be hung up on "best" straight line fit etc when in fact the specification is maximum deviation of actual from ideal as a percent full scale range...
You _can_ draw a straight line through any three points if you're a great comic!!!
Depends if you are the customer or supplier. If you are the supplier then you want the widest barn door.
Assume Gain = K = dVout/dVin, measured over the widest possible swing of Vout, dVoutmax/dVinmax.
"0.25% linearity" is the departure from the straight line as a percentage of Voutmax... = Voutmax/400.
"DC-offset" is the value of Vout when Vin=0 for the particular device under test.
Vout = K*Vin +/- (Voutmax/400) +/- (DC-offset).
So to verify any particular DUT.
Measure and record DC-offset and K for that device.
Then measure Vout for a range of Vin = 0 to Voutmax/K.
If the DUT is in spec, then for all plotted values of Vin:-
(Vout - K*Vin -/+ DC-offset) =< +/- Voutmax/400
You can get a barn door a little wider than this. You can at least do a best fit onto a straight line so that the gain value you come up with gives the lowest distortion number for no phase shift. This means that your way of defining offset has to go as I explain below.
Phase shifts could be dealt with by adjusting the assumed delay of the input sinewave for the best fit too.
Even order distortion will be worse if you do this. You really want to do the offset as part of the fit to the straight line. This way, you will center the "ideal" on the even distortion component too.
"Linearity" is ususally not the same as absolute accuracy, but that's a matter of application and definition.
John
Naaah! You're talking to the master here. I'm getting 0.03% deviation from a best-fit straight line. Plus I'm nulling (auto-zero) the amplifier before each measurement.
...Jim Thompson
What I'm doing, though I'm sure it over states the error is...
(V(OUTP,OUTN)-(MAX(V(OUTP,OUTN))-MIN(V(OUTP,OUTN)))/6m*V_VSW)
in PSpice Probe, since I have 65 process corners to test :-(
(6mV is the input sweep, V_VSW is the name of the source)
The gives the classic error "S-curve" _except_ the end point are at zero instead of the usual even spread about zero error.
Since I have 65 sets to do I'm trying to avoid best-fit for each :-(
...Jim Thompson
Yup, gain is cheap these days. Linearity is dominated by the feedback resistors.
John
Like computers without programs, or one sitting idle without a current running process.
All potential, no work done.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required