Forums

TI opamp

Started by John Larkin March 10, 2015
Am 12.03.2015 um 16:51 schrieb John Larkin:
> On Tue, 10 Mar 2015 17:22:27 +0100, Gerhard Hoffmann > <ghf@hoffmann-hochfrequenz.de> wrote: > >> Am 10.03.2015 um 15:57 schrieb John Larkin: >>> On Tue, 10 Mar 2015 08:44:40 -0500, John S <Sophi.2@invalid.org> >>> wrote: >>> >>>> On 3/9/2015 10:36 PM, John Larkin wrote: >>>>> >>>>> >>>>> Has anybody used this one? >>>>> >>>>> http://www.ti.com/product/lme49990 >>>>> >>>>> Sounds too good to be true. >>>> >>>> Note that the GBW is 110MHz on page 1 of the specs :) >>> >>> Yup. The ADI low-noise amps, still relatively noisy, are slow. The >>> LT1028 is below 1 nv/rthz but has issues. >> >> The AD797 is not significantly noisier or slower, if at all. >> I really like the ADA4898-2. >> >> Depending on how it's measured, a little bit lower ft than the LME, but >> more than twice the slew rate, no 300 KHz noise mole hill like LT1028, >> and with 2 amplifiers/SO-8 quite a good price, especially if you average >> 20 devices for _really_ low voltage noise from low impedance sources. >> >> < http://www.hoffmann-hochfrequenz.de/downloads/lono.pdf > >> >> and, related: >> >> < >> http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf >>> >> >> < >> http://www.hoffmann-hochfrequenz.de/downloads/Noise_Measurements_On_Some_Laboratory_Power_Supplies.pdf >>> >> >> regards, Gerhard >> > > All the batteries have similar curves, especially at low frequencies.
Not really. Scott Wurcer (father of the AD797) spotted immediately that some are much worse than 1/f. Having read Rubiola in the meantime: that must be simply drift. Rubiola has done a similar preamp using 3*MAT02 or sth. like that.(1)
> > Is there any chance that the presence of DC increased the preamp > noise? It wouldn't take much leakage in the input coupling cap to > upset the preamp. I assume the preamp is a bunch of paralleled jfets.
The preamp is 20 parallel ADA4898. Input capacitors are ~22 10uF WIMA foil capacitors; I could not get more of those I had made the layout for, only mechanically bigger ones. So I glued them dead bug style onto the board, connected them with wrapping wire and hid them under the merciful cover of the Hammond alu box. :-) I have made a next generation layout, but since the old one solves most of my problems, it is questionable if I ever will build it.
> > Some simple tests could discount that possibility. It looks like the > Tadiran is the lowest voltage and has the least noise below 1 Hz.
Ok, it seems to be long-time stable, but the 1/f corner is ridiculous. I bought it as a replacement for my 8970B noise figure meter, so I did not check it under load. That test would eat up some years of 8970B operation. BTW the Li battery came from Digikey in an extra box; it seems it must not fly on a plane that carries passengers also. I wonder if a noise figure meter is allowed to fly on a passenger plane. For me, NiCd rulez, and size matters. And, at 0.1 Hz, the coupling capacitor begins to isolate the preamp array from its badly needed low source resistance. So the noise current starts to play a role down there. But I'm not really sure if it pays to push the lower frequency corner that much. Most test objects have a much higher 1/f corner than the ADA4898 and accepting a higher pole and computing it away would probably result in a better, not a worse dynamic range.
> >
(1) Enrico Rubiola, Franck Lardet-Vieudrin : Low flicker-noise amplifier for 50 &#9024; sources (2) Enrico Rubiola, Claudio Francese, and Andrea De Marchi: Long-Term Behavior of Operational Amplifiers I think his web site is rubiola.org, I'm just in the middle of nowhere and have barely cell phone reception, so I cannot easily search the urls. A _lot_ of stuff is there for timing-interested people. regards, Gerhard
Am 12.03.2015 um 16:44 schrieb George Herold:


> I wonder if you say any microphonic pickup in the batteries.
> Back in grad school we used 9V transistor batteries to bias a bolometer, > a few uA of current. (Low temperature FIR sensor.) There was this > extra noise sometimes that we eventually traced back to microphonic > pickup in the batteries... (If you talked loudly next to the battery > bias box you saw the noise on the scan.) I did not see any microphonic pickup. But then, the preamp was in a alu cast box, the measurement object was also in such a box, Semi Rigid coax between the two boxes, everything on a staple of books and all together in a massive aluminium cargo box with BNC feedthroughs only. One of the books was Conelly-Motchenbacher on low noise design, to ask the gods for a good outcome. ANY connection of the preamp and the test object box in excess of the semi rigid resulted in a loop that filled the screen with spurious. Not _some_ spurious - Nothing BUT spurious. The baseline here is pV noise densities. That isolation also helps against acoustics, and me, I'm quite quiet, so sorry, no microphonics seen. :-) regards, Gerhard
On Thursday, March 12, 2015 at 6:02:40 PM UTC-4, Phil Hobbs wrote:
> Diode lasers have a strong tendency to lock to any stray reflection--almost like a phase-conjugate mirror with gain. You can make unintentional ECDLs really easily unless you have a whole lot of attenuation or a couple of stages of isolation. >
Grin.. yeah the beauty and curse of a diode laser. Believe me I know all about this. (A bit ago I spent ~1 week measuring the 25C free running wavelength of 100's of DL's into a wave meter Jr. Whose back reflections is bad too. The back reflection "stair case" has lots of steps in ~1GHz, so ~100MHz. for a ~1m path length. Watching a spectrum seems like a good indicator. Are back reflections a parametric amplification? So that below some threshold it's "mostly harmless"? This is much smaller... it smells like a ground loop, shielding thing, which is harder, cause whatever I did ~10 years ago for grounding and shielding was f'ing around till it worked... George H.
> Cheers > > Phil Hobbs
On Thu, 12 Mar 2015 23:46:14 +0100, Gerhard Hoffmann
<ghf@hoffmann-hochfrequenz.de> wrote:

>Am 12.03.2015 um 16:51 schrieb John Larkin: >> On Tue, 10 Mar 2015 17:22:27 +0100, Gerhard Hoffmann >> <ghf@hoffmann-hochfrequenz.de> wrote: >> >>> Am 10.03.2015 um 15:57 schrieb John Larkin: >>>> On Tue, 10 Mar 2015 08:44:40 -0500, John S <Sophi.2@invalid.org> >>>> wrote: >>>> >>>>> On 3/9/2015 10:36 PM, John Larkin wrote: >>>>>> >>>>>> >>>>>> Has anybody used this one? >>>>>> >>>>>> http://www.ti.com/product/lme49990 >>>>>> >>>>>> Sounds too good to be true. >>>>> >>>>> Note that the GBW is 110MHz on page 1 of the specs :) >>>> >>>> Yup. The ADI low-noise amps, still relatively noisy, are slow. The >>>> LT1028 is below 1 nv/rthz but has issues. >>> >>> The AD797 is not significantly noisier or slower, if at all. >>> I really like the ADA4898-2. >>> >>> Depending on how it's measured, a little bit lower ft than the LME, but >>> more than twice the slew rate, no 300 KHz noise mole hill like LT1028, >>> and with 2 amplifiers/SO-8 quite a good price, especially if you average >>> 20 devices for _really_ low voltage noise from low impedance sources. >>> >>> < http://www.hoffmann-hochfrequenz.de/downloads/lono.pdf > >>> >>> and, related: >>> >>> < >>> http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf >>>> >>> >>> < >>> http://www.hoffmann-hochfrequenz.de/downloads/Noise_Measurements_On_Some_Laboratory_Power_Supplies.pdf >>>> >>> >>> regards, Gerhard >>> >> >> All the batteries have similar curves, especially at low frequencies. > >Not really. Scott Wurcer (father of the AD797) spotted immediately >that some are much worse than 1/f. Having read Rubiola in the >meantime: that must be simply drift. Rubiola has done a similar >preamp using 3*MAT02 or sth. like that.(1) > >> >> Is there any chance that the presence of DC increased the preamp >> noise? It wouldn't take much leakage in the input coupling cap to >> upset the preamp. I assume the preamp is a bunch of paralleled jfets. > >The preamp is 20 parallel ADA4898. Input capacitors are ~22 10uF >WIMA foil capacitors; I could not get more of those I had made the >layout for, only mechanically bigger ones. So I glued them dead bug >style onto the board, connected them with wrapping wire and hid them >under the merciful cover of the Hammond alu box. :-) > >I have made a next generation layout, but since the old one solves most >of my problems, it is questionable if I ever will build it. >> >> Some simple tests could discount that possibility. It looks like the >> Tadiran is the lowest voltage and has the least noise below 1 Hz. > >Ok, it seems to be long-time stable, but the 1/f corner is >ridiculous. I bought it as a replacement for my 8970B noise >figure meter, so I did not check it under load. That test would >eat up some years of 8970B operation. BTW the Li battery came >from Digikey in an extra box; it seems it must not fly on a plane >that carries passengers also. I wonder if a noise figure meter is >allowed to fly on a passenger plane. > >For me, NiCd rulez, and size matters. > >And, at 0.1 Hz, the coupling capacitor begins to isolate the >preamp array from its badly needed low source resistance. So >the noise current starts to play a role down there. > >But I'm not really sure if it pays to push the lower frequency >corner that much. Most test objects have a much higher 1/f corner >than the ADA4898 and accepting a higher pole and computing it away >would probably result in a better, not a worse dynamic range.
If you arrange the batteries antiparallel, for 0V nom output, the first stage could have a decent gain, 20x at least, and be DC coupled.
>> >> >(1) Enrico Rubiola, Franck Lardet-Vieudrin : Low flicker-noise amplifier >for 50 ? sources > >(2) Enrico Rubiola, Claudio Francese, and Andrea De Marchi: Long-Term >Behavior of Operational Amplifiers > >I think his web site is rubiola.org, I'm just in the middle of nowhere >and have barely cell phone reception, so I cannot easily search the >urls. A _lot_ of stuff is there for timing-interested people. > >regards, Gerhard
Every wire connected to a battery is a thermal antenna + thermocouple. You could get 10s of uV/K from a copper wire soldered to a nickel-over-steel end cap. Millikelvin temp fluctuations could generate some of the noise. We've seen the same thing in crystal oscillators, in the 1 Hz jitter ballpark. You can really improve a cheap XO's jitter by covering it with foam. -- John Larkin Highland Technology, Inc picosecond timing laser drivers and controllers jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
On 3/11/2015 9:06 AM, John Larkin wrote:
> On Wed, 11 Mar 2015 07:22:15 -0500, John S <Sophi.2@invalid.org> > wrote: > >> On 3/10/2015 9:57 AM, John Larkin wrote: >>> On Tue, 10 Mar 2015 08:44:40 -0500, John S <Sophi.2@invalid.org> >>> wrote: >>> >>>> On 3/9/2015 10:36 PM, John Larkin wrote: >>>>> >>>>> >>>>> Has anybody used this one? >>>>> >>>>> http://www.ti.com/product/lme49990 >>>>> >>>>> Sounds too good to be true. >>>> >>>> Note that the GBW is 110MHz on page 1 of the specs :) >>> >>> Yup. The ADI low-noise amps, still relatively noisy, are slow. The >>> LT1028 is below 1 nv/rthz but has issues. >>> >>>> >>>> Nice op-amp! >>> >>> Sure looks it. >> >> I need help understanding this: How can an op-amp with an output shown >> on fig.27 of the data sheet be zero at about 5MHz when it has a GBW of >> 110MHz? > > That's the slew rate limit, which is low for this amp. That saves > power but limits it for some appplications.
I still don't understand. :( How can fig.27 show the output is zero at 10MHz if the GBW is 110MHz? Should there not be *some* output at 110MHz? Thank for your help.
> How can fig.27 show the output is zero at > 10MHz if the GBW is 110MHz? > Should there not be *some* output at 110MHz?
There will be, but not much. With a slew rate of 22 V/us, the maximum undistorted sine wave is 22E6/(2 pi * 110E6) = 30 mV peak. It's intended for audio, where that slew rate is much better than good enough. Cheers Phil Hobbs
On 3/13/2015 12:49 PM, Phil Hobbs wrote:
>> How can fig.27 show the output is zero at 10MHz if the GBW is >> 110MHz? Should there not be *some* output at 110MHz? > > There will be, but not much. With a slew rate of 22 V/us, the maximum > undistorted sine wave is 22E6/(2 pi * 110E6) = 30 mV peak. > > It's intended for audio, where that slew rate is much better than > good enough. > > Cheers > > Phil Hobbs >
Thanks, Phil. I think I understand now.