On Wed, 12 Sep 2018 20:15:38 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 9/12/18 11:45 AM, John Larkin wrote: >> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max >> offset, EMI hardened, pA bias current. But it has a hangup mode as a >> follower: if the output goes to V+ the input back-to-back diodes keep >> both inputs above the legal common-mode range. Pity it's not RRIO. >> >> Similar amps from TI and Maxim seem to have the same issue. >> >> Here's one fix: >> >> https://www.dropbox.com/s/tdzs1ydi7quzbs7/ADA4522_hangup.JPG?raw=1 >> >> We had the depletion fets there already, to protect the amp, so all >> we'll do is add the dual zener. Above 8 volts or so, small zeners [1] >> have tiny pre-breakdown currents, so this should be fine for +-10 volt >> operating inputs. We'll test some dual zeners to be sure. >> > >Another approach would be to put a parallel RC in the feedback path.George suggested something like that... see my reply. This opamp is pretty slow and the load is known, so just a resistor should work. One resistor is a pretty tolerable kluge.> >I first ran into this sort of problem with the then-new Motorola >MC34084, which was sort of a turbo TL084. Its output swing was bigger >than its input CM range, and it had the FET op amp "phase inversion" >problem (*), so there were lots of entertaining lockup states if you >used noninverting amps as part of a larger feedback loop. ;) > >Cheers > >Phil Hobbs > >(*) It isn't really _inversion_, of course--if the positive CM limit is >exceeded, the PFET input stage turns off and the output rails low, >regardless of which input is higher. > >Cheers > >Phil Hobbs-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
ADA4522 opamp hangup
Started by ●September 12, 2018
Reply by ●September 12, 20182018-09-12
Reply by ●September 12, 20182018-09-12
On 9/12/18 8:30 PM, John Larkin wrote:> On Wed, 12 Sep 2018 20:15:38 -0400, Phil Hobbs > <pcdhSpamMeSenseless@electrooptical.net> wrote: > >> On 9/12/18 11:45 AM, John Larkin wrote: >>> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max >>> offset, EMI hardened, pA bias current. But it has a hangup mode as a >>> follower: if the output goes to V+ the input back-to-back diodes keep >>> both inputs above the legal common-mode range. Pity it's not RRIO. >>> >>> Similar amps from TI and Maxim seem to have the same issue. >>> >>> Here's one fix: >>> >>> https://www.dropbox.com/s/tdzs1ydi7quzbs7/ADA4522_hangup.JPG?raw=1 >>> >>> We had the depletion fets there already, to protect the amp, so all >>> we'll do is add the dual zener. Above 8 volts or so, small zeners [1] >>> have tiny pre-breakdown currents, so this should be fine for +-10 volt >>> operating inputs. We'll test some dual zeners to be sure. >>> >> >> Another approach would be to put a parallel RC in the feedback path. > > George suggested something like that... see my reply. This opamp is > pretty slow and the load is known, so just a resistor should work. > > One resistor is a pretty tolerable kluge. >Sure. It does degrade the noise some out near the unity-gain cross, where there isn't enough loop gain to suppress it. What do the switching spikes look like on this amp? I'm doing a product right now using OPA2188s, which are my fave HV chopamps, but am certainly open to better options! Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●September 12, 20182018-09-12
On Wednesday, September 12, 2018 at 8:09:49 PM UTC-4, John Larkin wrote:> On Wed, 12 Sep 2018 16:26:08 -0700 (PDT), George Herold > <gherold@teachspin.com> wrote: > > >On Wednesday, September 12, 2018 at 3:26:42 PM UTC-4, John Larkin wrote: > >> On Wed, 12 Sep 2018 11:16:42 -0700 (PDT), George Herold > >> <gherold@teachspin.com> wrote: > >> > >> >On Wednesday, September 12, 2018 at 11:45:38 AM UTC-4, John Larkin wrote: > >> >> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max > >> >> offset, EMI hardened, pA bias current. But it has a hangup mode as a > >> >> follower: if the output goes to V+ the input back-to-back diodes keep > >> >> both inputs above the legal common-mode range. Pity it's not RRIO. > >> >Huh that is a nice opamp. (Thanks, well slew rate not so snappy) > >> >What do you mean by hangup? > >> >Does it latch? Or take a long time to recover? > >> > > >> >George H. > >> > >> If the customer input is fairly high impedance, like 20K or so, and he > >> briefly blips the voltage high, the input can hang high, even when the > >> customer backs down. > >> > >> It just occurred to us that we could have that stage have a gain of > >> +1.15 or something, which would keep the inverting input away from > >> Vcc. That just needs two resistors and has no zener leakage concerns. > >> We can cal around the gain. > >OK, so a little gain kills it. If you didn't care too much > >about the output impedance.. or you knew what it was driving, > >you could throw away the gain on the output. > >(but you know that...) > >George H. > > I hadn't thought of that but you're right, the voltage at the > inverting input has a gain of 1.000, after the divider. Unfortunately > there is a load downstream. > > Hmmm, I think it works! The downstream load is about 5K, so we just > add one 500 ohm resistor in series with the opamp output. > > https://www.dropbox.com/s/d7cpms6fcde6my2/ADA4522_fix.JPG?raw=1Huh, well that is not at all what I had in mind, but it meets all the criteria.. constraining voltages. with less parts and no matching. looks OK... even if the 5k ohm is not that firm. GH (sometimes even the blind pig finds an acorn.)> > Gain is 1.000, opamp input voltage is constrained to the safe CM > range, and the resistor tolerance and TC don't matter. > > Does look weird, but it should work. > > Good one. > > > -- > > John Larkin Highland Technology, Inc > picosecond timing precision measurement > > jlarkin att highlandtechnology dott com > http://www.highlandtechnology.com
Reply by ●September 12, 20182018-09-12
On 9/12/18 7:31 PM, George Herold wrote:> On Wednesday, September 12, 2018 at 7:26:13 PM UTC-4, George Herold wrote: >> On Wednesday, September 12, 2018 at 3:26:42 PM UTC-4, John Larkin wrote: >>> On Wed, 12 Sep 2018 11:16:42 -0700 (PDT), George Herold >>> <gherold@teachspin.com> wrote: >>> >>>> On Wednesday, September 12, 2018 at 11:45:38 AM UTC-4, John Larkin wrote: >>>>> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max >>>>> offset, EMI hardened, pA bias current. But it has a hangup mode as a >>>>> follower: if the output goes to V+ the input back-to-back diodes keep >>>>> both inputs above the legal common-mode range. Pity it's not RRIO. >>>> Huh that is a nice opamp. (Thanks, well slew rate not so snappy) >>>> What do you mean by hangup? >>>> Does it latch? Or take a long time to recover? >>>> >>>> George H. >>> >>> If the customer input is fairly high impedance, like 20K or so, and he >>> briefly blips the voltage high, the input can hang high, even when the >>> customer backs down. >>> >>> It just occurred to us that we could have that stage have a gain of >>> +1.15 or something, which would keep the inverting input away from >>> Vcc. That just needs two resistors and has no zener leakage concerns. >>> We can cal around the gain. >> OK, so a little gain kills it. If you didn't care too much >> about the output impedance.. or you knew what it was driving, >> you could throw away the gain on the output. >> (but you know that...) >> George H. > > ps > let me just add this question. I like the fact that there are more > HV (>36V) opamps... any ideas of what's driving the market? >Dunno, but it's sure a nice change from 10 years ago! Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●September 12, 20182018-09-12
On Wed, 12 Sep 2018 20:49:45 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 9/12/18 8:30 PM, John Larkin wrote: >> On Wed, 12 Sep 2018 20:15:38 -0400, Phil Hobbs >> <pcdhSpamMeSenseless@electrooptical.net> wrote: >> >>> On 9/12/18 11:45 AM, John Larkin wrote: >>>> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max >>>> offset, EMI hardened, pA bias current. But it has a hangup mode as a >>>> follower: if the output goes to V+ the input back-to-back diodes keep >>>> both inputs above the legal common-mode range. Pity it's not RRIO. >>>> >>>> Similar amps from TI and Maxim seem to have the same issue. >>>> >>>> Here's one fix: >>>> >>>> https://www.dropbox.com/s/tdzs1ydi7quzbs7/ADA4522_hangup.JPG?raw=1 >>>> >>>> We had the depletion fets there already, to protect the amp, so all >>>> we'll do is add the dual zener. Above 8 volts or so, small zeners [1] >>>> have tiny pre-breakdown currents, so this should be fine for +-10 volt >>>> operating inputs. We'll test some dual zeners to be sure. >>>> >>> >>> Another approach would be to put a parallel RC in the feedback path. >> >> George suggested something like that... see my reply. This opamp is >> pretty slow and the load is known, so just a resistor should work. >> >> One resistor is a pretty tolerable kluge. >> > >Sure. It does degrade the noise some out near the unity-gain cross, >where there isn't enough loop gain to suppress it. > >What do the switching spikes look like on this amp? I'm doing a product >right now using OPA2188s, which are my fave HV chopamps, but am >certainly open to better options! >I don't know; I'll ask Rob to scope it. I wonder if the EMI hardening is bilateral! The noise plot on the data sheet looks awfully good. I think it chops at 800 KHz. Spread-spectrum? -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●September 12, 20182018-09-12
On Wed, 12 Sep 2018 20:20:07 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 9/12/18 8:09 PM, John Larkin wrote: >> On Wed, 12 Sep 2018 16:26:08 -0700 (PDT), George Herold >> <gherold@teachspin.com> wrote: >> >>> On Wednesday, September 12, 2018 at 3:26:42 PM UTC-4, John Larkin wrote: >>>> On Wed, 12 Sep 2018 11:16:42 -0700 (PDT), George Herold >>>> <gherold@teachspin.com> wrote: >>>> >>>>> On Wednesday, September 12, 2018 at 11:45:38 AM UTC-4, John Larkin wrote: >>>>>> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max >>>>>> offset, EMI hardened, pA bias current. But it has a hangup mode as a >>>>>> follower: if the output goes to V+ the input back-to-back diodes keep >>>>>> both inputs above the legal common-mode range. Pity it's not RRIO. >>>>> Huh that is a nice opamp. (Thanks, well slew rate not so snappy) >>>>> What do you mean by hangup? >>>>> Does it latch? Or take a long time to recover? >>>>> >>>>> George H. >>>> >>>> If the customer input is fairly high impedance, like 20K or so, and he >>>> briefly blips the voltage high, the input can hang high, even when the >>>> customer backs down. >>>> >>>> It just occurred to us that we could have that stage have a gain of >>>> +1.15 or something, which would keep the inverting input away from >>>> Vcc. That just needs two resistors and has no zener leakage concerns. >>>> We can cal around the gain. >>> OK, so a little gain kills it. If you didn't care too much >>> about the output impedance.. or you knew what it was driving, >>> you could throw away the gain on the output. >>> (but you know that...) >>> George H. >> >> I hadn't thought of that but you're right, the voltage at the >> inverting input has a gain of 1.000, after the divider. Unfortunately >> there is a load downstream. >> >> Hmmm, I think it works! The downstream load is about 5K, so we just >> add one 500 ohm resistor in series with the opamp output. >> >> https://www.dropbox.com/s/d7cpms6fcde6my2/ADA4522_fix.JPG?raw=1 >> >> Gain is 1.000, opamp input voltage is constrained to the safe CM >> range, and the resistor tolerance and TC don't matter. >> >> Does look weird, but it should work. >> >> Good one. >> >> >Cute! Watch out for capacitive loading though. > >Cheers > >Phil HobbsThe load is just 5 pF plus strays, and the chopamp is slow. We'll verify that, but it should be OK. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●September 12, 20182018-09-12
On 9/12/18 9:55 PM, John Larkin wrote:> On Wed, 12 Sep 2018 20:49:45 -0400, Phil Hobbs > <pcdhSpamMeSenseless@electrooptical.net> wrote: > >> On 9/12/18 8:30 PM, John Larkin wrote: >>> On Wed, 12 Sep 2018 20:15:38 -0400, Phil Hobbs >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: >>> >>>> On 9/12/18 11:45 AM, John Larkin wrote: >>>>> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max >>>>> offset, EMI hardened, pA bias current. But it has a hangup mode as a >>>>> follower: if the output goes to V+ the input back-to-back diodes keep >>>>> both inputs above the legal common-mode range. Pity it's not RRIO. >>>>> >>>>> Similar amps from TI and Maxim seem to have the same issue. >>>>> >>>>> Here's one fix: >>>>> >>>>> https://www.dropbox.com/s/tdzs1ydi7quzbs7/ADA4522_hangup.JPG?raw=1 >>>>> >>>>> We had the depletion fets there already, to protect the amp, so all >>>>> we'll do is add the dual zener. Above 8 volts or so, small zeners [1] >>>>> have tiny pre-breakdown currents, so this should be fine for +-10 volt >>>>> operating inputs. We'll test some dual zeners to be sure. >>>>> >>>> >>>> Another approach would be to put a parallel RC in the feedback path. >>> >>> George suggested something like that... see my reply. This opamp is >>> pretty slow and the load is known, so just a resistor should work. >>> >>> One resistor is a pretty tolerable kluge. >>> >> >> Sure. It does degrade the noise some out near the unity-gain cross, >> where there isn't enough loop gain to suppress it. >> >> What do the switching spikes look like on this amp? I'm doing a product >> right now using OPA2188s, which are my fave HV chopamps, but am >> certainly open to better options! >> > > I don't know; I'll ask Rob to scope it. I wonder if the EMI hardening > is bilateral! > > The noise plot on the data sheet looks awfully good. I think it chops > at 800 KHz. Spread-spectrum?Maybe. The fact that they plot its performance at Av=100 is super suspicious--it's not as though the marketeers have been manipulating noise plots since forever, of course. :( The original LT1028 datasheet's voltage noise plot cut off at 10 kHz, because of the gross >20dB noise peak near 300 kHz. Chopamp datasheets have continued this nonsense, with occasional outbreaks of candor as with the OPA2188--which is a big reason that it's my fave. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●September 13, 20182018-09-13
On Wednesday, September 12, 2018 at 10:01:01 PM UTC-4, Phil Hobbs wrote:> On 9/12/18 9:55 PM, John Larkin wrote: > > On Wed, 12 Sep 2018 20:49:45 -0400, Phil Hobbs > > <pcdhSpamMeSenseless@electrooptical.net> wrote: > > > >> On 9/12/18 8:30 PM, John Larkin wrote: > >>> On Wed, 12 Sep 2018 20:15:38 -0400, Phil Hobbs > >>> <pcdhSpamMeSenseless@electrooptical.net> wrote: > >>> > >>>> On 9/12/18 11:45 AM, John Larkin wrote: > >>>>> ADA4522 is an amazing amp. 55 volt supplies, low noise, 5 uV max > >>>>> offset, EMI hardened, pA bias current. But it has a hangup mode as a > >>>>> follower: if the output goes to V+ the input back-to-back diodes keep > >>>>> both inputs above the legal common-mode range. Pity it's not RRIO. > >>>>> > >>>>> Similar amps from TI and Maxim seem to have the same issue. > >>>>> > >>>>> Here's one fix: > >>>>> > >>>>> https://www.dropbox.com/s/tdzs1ydi7quzbs7/ADA4522_hangup.JPG?raw=1 > >>>>> > >>>>> We had the depletion fets there already, to protect the amp, so all > >>>>> we'll do is add the dual zener. Above 8 volts or so, small zeners [1] > >>>>> have tiny pre-breakdown currents, so this should be fine for +-10 volt > >>>>> operating inputs. We'll test some dual zeners to be sure. > >>>>> > >>>> > >>>> Another approach would be to put a parallel RC in the feedback path. > >>> > >>> George suggested something like that... see my reply. This opamp is > >>> pretty slow and the load is known, so just a resistor should work. > >>> > >>> One resistor is a pretty tolerable kluge. > >>> > >> > >> Sure. It does degrade the noise some out near the unity-gain cross, > >> where there isn't enough loop gain to suppress it. > >> > >> What do the switching spikes look like on this amp? I'm doing a product > >> right now using OPA2188s, which are my fave HV chopamps, but am > >> certainly open to better options! > >> > > > > I don't know; I'll ask Rob to scope it. I wonder if the EMI hardening > > is bilateral! > > > > The noise plot on the data sheet looks awfully good. I think it chops > > at 800 KHz. Spread-spectrum? > > > Maybe. The fact that they plot its performance at Av=100 is super > suspicious--it's not as though the marketeers have been manipulating > noise plots since forever, of course. :(I didn't realize this is a chopper amp. That's what zero-drift means? I saw the noise plot and wondered why it was only for a gain of 100. Maybe birdies up at higher freq.? George H.> > The original LT1028 datasheet's voltage noise plot cut off at 10 kHz, > because of the gross >20dB noise peak near 300 kHz. Chopamp datasheets > have continued this nonsense, with occasional outbreaks of candor as > with the OPA2188--which is a big reason that it's my fave. > > Cheers > > Phil Hobbs > > > -- > Dr Philip C D Hobbs > Principal Consultant > ElectroOptical Innovations LLC / Hobbs ElectroOptics > Optics, Electro-optics, Photonics, Analog Electronics > Briarcliff Manor NY 10510 > > http://electrooptical.net > http://hobbs-eo.com
Reply by ●September 13, 20182018-09-13
>I didn't realize this is a chopper amp. > That's what zero-drift means?Yup. Chopper, autozero, or other such schemes. They've been getting better at it--I remember the old ICL7650. It was pretty amazing in 1980, but you could practically draw sparks from the switching spikes. ;)>I saw the noise plot and wondered why it was only for a gain of 100. >Maybe birdies up at higher freq.?That's my suspicion as well--either isolated switching harmonics or maybe an elevated noise floor if they're doing the spread spectrum thing. What you usually have to watch out for is current spikes coming out the inputs--the bias current is far from constant, so chopamps are bad news when the source is high-impedance. That's why even truthful noise voltage plots don't necessarily tell you the whole story at all, at all. Cheers Phil Hobbs
Reply by ●September 13, 20182018-09-13
On Thu, 13 Sep 2018 05:59:55 -0700 (PDT), pcdhobbs@gmail.com wrote:>>I didn't realize this is a chopper amp. >�That's what zero-drift means? > >Yup. Chopper, autozero, or other such schemes. They've been getting better at it--I remember the old ICL7650. It was pretty amazing in 1980, but you could practically draw sparks from the switching spikes. ;)I think that was the amp that took seconds to recover from railing.> >>I saw the noise plot and wondered why it was only for a gain of 100. � >>Maybe birdies up at higher freq.? > >That's my suspicion as well--either isolated switching harmonics or maybe an elevated noise floor if they're doing the spread spectrum thing. > >What you usually have to watch out for is current spikes coming out the inputs--the bias current is far from constant, so chopamps are bad news when the source is high-impedance.Input offsets tend to shift as a function of the capacitance that the two inputs see. I'm optimistic that the ADA4522 is better. -- John Larkin Highland Technology, Inc lunatic fringe electronics
