Post by thread:[EE] More On Proper Unused Op-AMP Termination

> While on the Pericom website, I stumbled across an > excellent > EMI Appnote (#11) that states categorically: > > "If there are OpAmps in a design, terminate unused opamps > in dual > or quad packs by grounding the positive input to ground, > then connecting > the negative input to the output." That is an excellent method. Amplifier becomes a follower and has a defined input. For a dual rail system around ground that's about as good as you'd get. Only issue MIGHT be if amp was single supply & non rail-rail and had some issues, probably minor with out of range output. Odds are it would seldom be a real problem. I'd suspect that slightly better may then be using a signal-less input at enough above ground to get the amp in operating range. This could be a resistive divider across supply (reconstituted pseudo ground) or any local voltage that was essentially invariant. Russell

Bob Axtell <engineer <at> cotse.net> writes: > "If there are OpAmps in a design, terminate unused opamps in dual > or quad packs by grounding the positive input to ground, then connecting > the negative input to the output." "Always assuming that the Opamps have input-to-ground specification" (unlike certain popular FET input opamps ... The rule of thumb is that there is no rule of thumb. Connecting the unused + inputs to a bias source in the CMMR range and the outputs to the - inputs would be a valid way to nail it I think. Excepting for noise and other factors like RF rectification. Peter P.

On Jan 29, 2008 3:46 PM, Bob Axtell <spam_OUTengineerTakeThisOuTcotse.net> wrote: > While on the Pericom website, I stumbled across an excellent > EMI Appnote (#11) that states categorically: > > "If there are OpAmps in a design, terminate unused opamps in dual > or quad packs by grounding the positive input to ground, then connecting > the negative input to the output." > Yup. Leaving them uncommitted causes problems with the other amps on the die. I've seen oscillation, and noise on adjacent stages.

> Bob Axtell <engineer <at> cotse.net> writes: >> "If there are OpAmps in a design, terminate unused opamps >> in dual >> or quad packs by grounding the positive input to ground, >> then connecting >> the negative input to the output." > "Always assuming that the Opamps have input-to-ground > specification" (unlike > certain popular FET input opamps ... The writer of that advice MAY have been living in a dual rail only world where ground is always well away from the rails. Russell.

>>> "If there are OpAmps in a design, terminate unused opamps >>> in dual >>> or quad packs by grounding the positive input to ground, >>> then connecting >>> the negative input to the output." > >> "Always assuming that the Opamps have input-to-ground >> specification" (unlike >> certain popular FET input opamps ... > >The writer of that advice MAY have been living in a dual >rail only world where ground is always well away from the >rails. However to be fair, on a single rail system there is generally a bias point that gets regarded as "analogue ground", and that is effectively what is suggested for the connection, without it being implicitly said.

> "If there are OpAmps in a design, terminate unused opamps > in dual > or quad packs by grounding the positive input to ground, > then connecting > the negative input to the output." >>The writer of that advice MAY have been living in a dual >>rail only world where ground is always well away from the >>rails. > However to be fair, on a single rail system there is > generally a bias point > that gets regarded as "analogue ground", and that is > effectively what is > suggested for the connection, without it being implicitly > said. I disagree. Or, if that was the case, they should undergo a brief course of self criticism :-). ie they clearly knew what they were trying to achieve, but they wrote a specific instruction which would certainly be taken literally by many people. If it is not what they meant they should have said differently - as eg you and I and others have suggested. Excess simplicity can be fatal. Russell

> -----Original Message----- > From: .....piclist-bouncesKILLspam@spam@mit.edu > [piclist-bouncesKILLspammit.edu] On Behalf Of Bob Axtell > Sent: Tuesday, January 29, 2008 12:47 PM > > While on the Pericom website, I stumbled across an excellent > EMI Appnote (#11) that states categorically: > > "If there are OpAmps in a design, terminate unused opamps in > dual or quad packs by grounding the positive input to ground, > then connecting the negative input to the output." > Check these references, both written primarily for single-supply op amps: www.maxim-ic.com/appnotes.cfm/an_pk/1957 http://focus.ti.com/lit/an/slod006b/slod006b.pdf see section 17.7.3, page 375 in the PDF. GM

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Tue, Feb 12, 2008 at 06:32:48PM -0800, GlennM wrote: {Quote hidden}> > While on the Pericom website, I stumbled across an excellent > > EMI Appnote (#11) that states categorically: > > > > "If there are OpAmps in a design, terminate unused opamps in > > dual or quad packs by grounding the positive input to ground, > > then connecting the negative input to the output." > > > > Check these references, both written primarily for single-supply op amps: > > www.maxim-ic.com/appnotes.cfm/an_pk/1957 > http://focus.ti.com/lit/an/slod006b/slod006b.pdf see section 17.7.3, page > 375 in the PDF. Looks to me like if you have a design that must be low power, the best bet is to avoid having the uncommited op amp in the first place. I assume the resistor string shouldn't be too high value, two 1M resistors at 3V would still be 1.5uA leakage, more power draw that the rest of the circuit might be. - -- http://petertodd.org -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) iD8DBQFHsoNi3bMhDbI9xWQRAj0HAJ9QZNLRyOtGNgCf4BG87g21qrbPawCfbPTk ds+5NVpsENT1/bWyuAp5v/E= =yKyb -----END PGP SIGNATURE-----

>> Check these references, both written primarily for >> single-supply op amps: >> >> www.maxim-ic.com/appnotes.cfm/an_pk/1957 >> http://focus.ti.com/lit/an/slod006b/slod006b.pdf see >> section 17.7.3, page >> 375 in the PDF. > > Looks to me like if you have a design that must be low > power, the best > bet is to avoid having the uncommited op amp in the first > place. Agre, but > I assume the resistor string shouldn't be too high value, > two 1M resistors > at 3V would still be 1.5uA leakage, more power draw that > the rest of the > circuit might be. Probably not if op-amp are involved. Unless you are dealing with ultra-special devices the supply current for opamps will usually be higher or much higher than that. An exception which proves the rule is MAX406BCSA at 1.2uA max supply current. You pay for this (totally expectedly) with a 4 kHz gain-bandwidth product and 5 mV/uS slew rate (!!!!!!!!!!!!!!!!!!!!!!) Overview http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1468 Data sheet http://pdfserv.maxim-ic.com/en/ds/MAX406-MAX419.pdf Russell

>> Looks to me like if you have a design that must be low power, the >> best bet is to avoid having the uncommited op amp in the first place. I thought the general conclusion was to configure a voltage follower whose input was well within the range allowed for that particular op- amp. For most modern op-amps, doesn't that mean you can parallel the input with any of the USED amps in your package ? Their inputs are already supposed to stay within allowed limits. There may be circuits that are particularly sensitive to loading at certain points, but there are almost certainly other points that are pretty robust, and the average modern opamp input is not much of a load... BillW

William "Chops" Westfield wrote: >>> Looks to me like if you have a design that must be low power, the >>> best bet is to avoid having the uncommited op amp in the first place. > > I thought the general conclusion was to configure a voltage follower > whose input was well within the range allowed for that particular op- > amp. For most modern op-amps, doesn't that mean you can parallel the > input with any of the USED amps in your package ? Yes, or possibly not exactly parallel an op-amp input but another analog level point in the circuit that's guaranteed to stay within limits. If you have an extreme low-power design with op-amps, I think it's quite likely that you have a point somewhere that's not too sensitive WRT the extra load, that's within the limits and that doesn't change too fast. Gerhard

>>>> Looks to me like if you have a design that must be low >>>> power, the >>>> best bet is to avoid having the uncommited op amp in >>>> the first place. >> I thought the general conclusion was to configure a >> voltage follower >> whose input was well within the range allowed for that >> particular op- >> amp. Yes - but not having it at all is even better power wise :-). {Quote hidden}>> For most modern op-amps, doesn't that mean you can >> parallel the >> input with any of the USED amps in your package ? > Yes, or possibly not exactly parallel an op-amp input but > another analog > level point in the circuit that's guaranteed to stay > within limits. If you > have an extreme low-power design with op-amps, I think > it's quite likely > that you have a point somewhere that's not too sensitive > WRT the extra > load, that's within the limits and that doesn't change too > fast. Agree. ie if you parallel another amplifier with both inputs you both add bias currents and whatever power is dissipated by having the amplifier change level with time. If you drive no load with the second output then incremental power may be low but non zero. However, if you have another amplifier with one input which is varying less than the other then configuring your 'spare' as a follower and then connecting non inverting input to the low variation point will do an OK job - you are still potentially adding bias current. So, as Gerhard says, IF you can find a point that varies less than anything connected to an opamp input that may be better. It needs to be of low enough impedance that the opamp input doesn't overly affect it. However, the main point here was correctly terminating the amplifier inputs and it bears restating the advice in reasonably complete form: TERMINATION OF UNUSED OP AMPS: Connect unused opamps as a follower and then connect the non-inverting input to a point that is within the common mode range of the amplifier (preferably well within). - Where there is a choice - Select a connection point with minimum AC variation. - Select a point where the amplifier will have minimum impact on the original functionality. - Consider creating a simple 2 resistor divider where the above conditions cannot be well met. - Not having any spare sections may be the best choice if possible. Russell McMahon

Apptech wrote: > TERMINATION OF UNUSED OP AMPS: > > Connect unused opamps as a follower and then connect the > non-inverting input to a point that is within the common > mode range of the amplifier (preferably well within). > > - Where there is a choice > > - Select a connection point with minimum AC variation. > > - Select a point where the amplifier will have minimum > impact on the original functionality. One such point is probably the /output/ of the existing op-amp, rather than paralleling the input. In general, the input range is wider than the output range, and it's usually a low impedance point. > - Consider creating a simple 2 resistor divider where > the above conditions cannot be well met. > > - Not having any spare sections may be the best choice > if possible. Gerhard