On Thu, 19 Mar 2015 14:06:31 -0400, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 03/19/2015 12:30 PM, George Herold wrote: >> On Wednesday, March 18, 2015 at 11:10:45 PM UTC-4, Winfield Hill wrote: >>> George Herold wrote... >>>> >>>> Just the final update on apex opamps... >>>> (I'm still waiting on DK, but I had to leave early today.) >>>> It turns out that I'd forgotten the first apex update, >>>> where the PA141 went to the PA241... so here a noise plot >>>> (noise density is uncalibrated.. though I have the gain settings.) >>>> >>>> https://www.dropbox.com/s/1vguffiuj4noghz/Graph1.BMP?dl=0 >>>> (The black line has a slope of 1) >>>> >>>> Turns out the opamp got progressively worse... >>>> What's weird is that the "1/f" noise >>>> is not 1/f, but f^-2/3. >>>> >>>> Well since I don't understand 1/f, (1/f)^2/3 is no weirder. >>> >>> We discuss in detail how this can happen in our lengthy >>> Chapter 8 in Art of Electronics 3rd edition, 40 tons of >>> which is shipping now from the printer. It's largely >>> due to offset shifts (i.e., popcorn-noise style) as >>> well as large thermal offset-voltage drifts. >> >> Hi Win, I assume you are talking about 1/f noise ...(and not the 2/3rds power dependence.) I've read some stuff on 1/f. (Van-der-Ziel for one.) >> I guess the classic 1/f noise is from the old carbon comp. resistors. >> Here's a plot.... well it's a bit busy, I found a tiny hint of 1/f noise in (cheap, Xicon) metal film resistors. >> https://www.dropbox.com/s/y1jv60eawkj9oxd/1_OVER_F.BMP?dl=0 >> >> George H. >>> > >IIRC most MF resistors are specified at 1E-7V/V per decade of frequency. > I measured some as a voltage divider with a 9V alkaline, battery, and >they were nowhere near that bad. > >Cheers > >Phil HobbsHigh-value (like, 100M) thickfilms seem to have excess noise. Metal films are hard to find at those values, but are much quieter. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Apex HV opamp issues.
Started by ●March 13, 2015
Reply by ●March 19, 20152015-03-19
Reply by ●March 19, 20152015-03-19
On Thursday, March 19, 2015 at 2:06:35 PM UTC-4, Phil Hobbs wrote:> On 03/19/2015 12:30 PM, George Herold wrote: > > On Wednesday, March 18, 2015 at 11:10:45 PM UTC-4, Winfield Hill wrote: > >> George Herold wrote... > >>> > >>> Just the final update on apex opamps... > >>> (I'm still waiting on DK, but I had to leave early today.) > >>> It turns out that I'd forgotten the first apex update, > >>> where the PA141 went to the PA241... so here a noise plot > >>> (noise density is uncalibrated.. though I have the gain settings.) > >>> > >>> https://www.dropbox.com/s/1vguffiuj4noghz/Graph1.BMP?dl=0 > >>> (The black line has a slope of 1) > >>> > >>> Turns out the opamp got progressively worse... > >>> What's weird is that the "1/f" noise > >>> is not 1/f, but f^-2/3. > >>> > >>> Well since I don't understand 1/f, (1/f)^2/3 is no weirder. > >> > >> We discuss in detail how this can happen in our lengthy > >> Chapter 8 in Art of Electronics 3rd edition, 40 tons of > >> which is shipping now from the printer. It's largely > >> due to offset shifts (i.e., popcorn-noise style) as > >> well as large thermal offset-voltage drifts. > > > > Hi Win, I assume you are talking about 1/f noise ...(and not the 2/3rds power dependence.) I've read some stuff on 1/f. (Van-der-Ziel for one.) > > I guess the classic 1/f noise is from the old carbon comp. resistors. > > Here's a plot.... well it's a bit busy, I found a tiny hint of 1/f noise in (cheap, Xicon) metal film resistors. > > https://www.dropbox.com/s/y1jv60eawkj9oxd/1_OVER_F.BMP?dl=0 > > > > George H. > >> > > IIRC most MF resistors are specified at 1E-7V/V per decade of frequency. > I measured some as a voltage divider with a 9V alkaline, battery, and > they were nowhere near that bad.I don't really like that noise plot for the metal films... I did it a while ago. (Maybe I'll repeat it some day.) But now you are making me think. 1E-7 V/V per decade, So with ten volts of bias and let's say a 1 Hz to 1 kHz bandwidth, that would be 3E-6 V(rms)(?) and a noise density (at 1 kHz) of ~1E-7V/rtHz... that seems huge! (even it I did loose a factor of pi in there somewhere.) George H.> > Cheers > > Phil Hobbs > > > -- > Dr Philip C D Hobbs > Principal Consultant > ElectroOptical Innovations LLC > Optics, Electro-optics, Photonics, Analog Electronics > > 160 North State Road #203 > Briarcliff Manor NY 10510 > > hobbs at electrooptical dot net > http://electrooptical.net
Reply by ●March 19, 20152015-03-19
> But now you are making me think. �1E-7 V/V per decade, � >So with ten volts of bias and let's say a 1 Hz to 1 kHz bandwidth, >that would be 3E-6 V(rms)(?) and a noise density (at 1 kHz) of >~1E-7V/rtHz... that seems huge! (even it I did loose a factor >of pi in there somewhere.)By my few measurements, the real number is at least 30 dB lower than that. Cheers Phil Hobbs
Reply by ●March 19, 20152015-03-19
John Larkin wrote...> > High-value (like, 100M) thickfilms seem to have excess noise. Metal > films are hard to find at those values, but are much quieter.Yes. The "excess noise" I've observed is usually highly voltage dependent, proportional to voltage, etc. It may be non-existent at low voltages, like under 10V, but severe above 50 to 100V to 1kV. And it can vary dramatically from part to part. E.e., made some +/-2KV amplifiers, with sub 1ppm noise levels, and 20% of the 10kV feedback resistors were very noisy and had to be replaced. But troublesome FET transistor noise may not have classic 1/f characteristics at all, e.g., "telegraph noise" and "popcorn noise" are not strictly 1/f. Let's talk after you've digested our measurements and the discussion in Chapter 8. The book is being shipped as we speak. My order is in. -- Thanks, - Win
Reply by ●March 19, 20152015-03-19
On Thursday, March 19, 2015 at 9:33:52 PM UTC-4, Winfield Hill wrote:> John Larkin wrote... > > > > High-value (like, 100M) thickfilms seem to have excess noise. Metal > > films are hard to find at those values, but are much quieter. > > Yes. The "excess noise" I've observed is usually highly voltage dependent, > proportional to voltage, etc. It may be non-existent at low voltages, like > under 10V, but severe above 50 to 100V to 1kV. And it can vary dramatically > from part to part. E.e., made some +/-2KV amplifiers, with sub 1ppm noise > levels, and 20% of the 10kV feedback resistors were very noisy and had to be > replaced. > > But troublesome FET transistor noise may not have classic 1/f characteristics at > all, e.g., "telegraph noise" and "popcorn noise" are not strictly 1/f. Let's > talk after you've digested our measurements and the discussion in Chapter 8. > The book is being shipped as we speak. My order is in. > > > -- > Thanks, > - WinExcellent! I look forward to the discussion once my piece of the 40 tons arrives. I wonder if the voltage dependence is a surface thing. (could be the outer surface or some layer underneath.) George H.
Reply by ●March 20, 20152015-03-20
On a sunny day (Thu, 19 Mar 2015 19:45:19 -0700 (PDT)) it happened George Herold <gherold@teachspin.com> wrote in <4527cdae-051b-4973-94cc-0dc648fef4a1@googlegroups.com>:>I wonder if the voltage dependence is a surface >thing. (could be the outer surface or some >layer underneath.)Yes, probably above some voltage you may get arcing over minuscule 'islands' of resisytro material.
Reply by ●March 20, 20152015-03-20
On Thu, 19 Mar 2015 19:45:19 -0700 (PDT), George Herold <gherold@teachspin.com> wrote:>On Thursday, March 19, 2015 at 9:33:52 PM UTC-4, Winfield Hill wrote: >> John Larkin wrote... >> > >> > High-value (like, 100M) thickfilms seem to have excess noise. Metal >> > films are hard to find at those values, but are much quieter. >> >> Yes. The "excess noise" I've observed is usually highly voltage dependent, >> proportional to voltage, etc. It may be non-existent at low voltages, like >> under 10V, but severe above 50 to 100V to 1kV. And it can vary dramatically >> from part to part. E.e., made some +/-2KV amplifiers, with sub 1ppm noise >> levels, and 20% of the 10kV feedback resistors were very noisy and had to be >> replaced. >> >> But troublesome FET transistor noise may not have classic 1/f characteristics at >> all, e.g., "telegraph noise" and "popcorn noise" are not strictly 1/f. Let's >> talk after you've digested our measurements and the discussion in Chapter 8. >> The book is being shipped as we speak. My order is in. >> >> >> -- >> Thanks, >> - Win > >Excellent! I look forward to the discussion once >my piece of the 40 tons arrives. >I wonder if the voltage dependence is a surface >thing. (could be the outer surface or some >layer underneath.) > >George H.All resistors have the same Johnson noise at zero volts. Some, not metal films, have shot noise with current flowing, which is simple electron arrival statistics. I conjecture that some nonmetallic resistors, like cermets, may have shot noise when biased. They also have excess, 1/f ish, noise, maybe from their polycrystaline confused nature. Tiny temperature variations must play heck with a grainy cermet-like structure. I've done a little testing on high-ohm resistors, but just enough to get me through a project. I needed 100 Mohms and wound up using two Dale custom-ordered 50M axials in series in mid-air, hopping over some surface-mount parts. Cermets were awful. This is fairly difficult stuff to measure, but would be a good student project or something. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●March 20, 20152015-03-20
On Friday, March 20, 2015 at 2:28:25 PM UTC-4, John Larkin wrote:> On Thu, 19 Mar 2015 19:45:19 -0700 (PDT), George Herold > <gherold@teachspin.com> wrote: > > >On Thursday, March 19, 2015 at 9:33:52 PM UTC-4, Winfield Hill wrote: > >> John Larkin wrote... > >> > > >> > High-value (like, 100M) thickfilms seem to have excess noise. Metal > >> > films are hard to find at those values, but are much quieter. > >> > >> Yes. The "excess noise" I've observed is usually highly voltage dependent, > >> proportional to voltage, etc. It may be non-existent at low voltages, like > >> under 10V, but severe above 50 to 100V to 1kV. And it can vary dramatically > >> from part to part. E.e., made some +/-2KV amplifiers, with sub 1ppm noise > >> levels, and 20% of the 10kV feedback resistors were very noisy and had to be > >> replaced. > >> > >> But troublesome FET transistor noise may not have classic 1/f characteristics at > >> all, e.g., "telegraph noise" and "popcorn noise" are not strictly 1/f. Let's > >> talk after you've digested our measurements and the discussion in Chapter 8. > >> The book is being shipped as we speak. My order is in. > >> > >> > >> -- > >> Thanks, > >> - Win > > > >Excellent! I look forward to the discussion once > >my piece of the 40 tons arrives. > >I wonder if the voltage dependence is a surface > >thing. (could be the outer surface or some > >layer underneath.) > > > >George H. > > All resistors have the same Johnson noise at zero volts. Some, not > metal films, have shot noise with current flowing, which is simple > electron arrival statistics. > > I conjecture that some nonmetallic resistors, like cermets, may have > shot noise when biased. They also have excess, 1/f ish, noise, maybe > from their polycrystaline confused nature. Tiny temperature variations > must play heck with a grainy cermet-like structure.I've never seen excess shot noise in resistors.. (excess noise that is white in frequency.) but I haven't looked hard either. According to Landauer a resistor should reduce the shot noise as 1/N, where N is the number of scattering events as a "single electron pulse" travels the length of the resistor. I'm not sure how to think about "scattering events" and 10 ns pulses.> > I've done a little testing on high-ohm resistors, but just enough to > get me through a project. I needed 100 Mohms and wound up using two > Dale custom-ordered 50M axials in series in mid-air, hopping over some > surface-mount parts. Cermets were awful. This is fairly difficult > stuff to measure, but would be a good student project or something.I wired up my LTC6090 HV opamps today. Fired it up on the test bench. no smoke, but all I saw noise-wise, was first the room lights, and then 60Hz, with cross-over spikies from the linear supply.... I wanted to go watch the UB bulls play in the "big dance".* So I'll have, to put it in the box tomorrow...(Monday) Measuring noise can be hard. I sometime wonder how I ever made things work in the past.... George H. *UB lost, we trailed the entire game, tied it up with ~3:00 to go, and then lost. Still it was a great emotional ride for me. +1, I hope Bobby Hurley stays and we go back. (intellectually sports makes no sense at all, for me, it's all about the emotion.)> > > -- > > John Larkin Highland Technology, Inc > picosecond timing precision measurement > > jlarkin att highlandtechnology dott com > http://www.highlandtechnology.com
Reply by ●March 20, 20152015-03-20
On Fri, 20 Mar 2015 16:04:17 -0700 (PDT), George Herold <gherold@teachspin.com> wrote:>On Friday, March 20, 2015 at 2:28:25 PM UTC-4, John Larkin wrote: >> On Thu, 19 Mar 2015 19:45:19 -0700 (PDT), George Herold >> <gherold@teachspin.com> wrote: >> >> >On Thursday, March 19, 2015 at 9:33:52 PM UTC-4, Winfield Hill wrote: >> >> John Larkin wrote... >> >> > >> >> > High-value (like, 100M) thickfilms seem to have excess noise. Metal >> >> > films are hard to find at those values, but are much quieter. >> >> >> >> Yes. The "excess noise" I've observed is usually highly voltage dependent, >> >> proportional to voltage, etc. It may be non-existent at low voltages, like >> >> under 10V, but severe above 50 to 100V to 1kV. And it can vary dramatically >> >> from part to part. E.e., made some +/-2KV amplifiers, with sub 1ppm noise >> >> levels, and 20% of the 10kV feedback resistors were very noisy and had to be >> >> replaced. >> >> >> >> But troublesome FET transistor noise may not have classic 1/f characteristics at >> >> all, e.g., "telegraph noise" and "popcorn noise" are not strictly 1/f. Let's >> >> talk after you've digested our measurements and the discussion in Chapter 8. >> >> The book is being shipped as we speak. My order is in. >> >> >> >> >> >> -- >> >> Thanks, >> >> - Win >> > >> >Excellent! I look forward to the discussion once >> >my piece of the 40 tons arrives. >> >I wonder if the voltage dependence is a surface >> >thing. (could be the outer surface or some >> >layer underneath.) >> > >> >George H. >> >> All resistors have the same Johnson noise at zero volts. Some, not >> metal films, have shot noise with current flowing, which is simple >> electron arrival statistics. >> >> I conjecture that some nonmetallic resistors, like cermets, may have >> shot noise when biased. They also have excess, 1/f ish, noise, maybe >> from their polycrystaline confused nature. Tiny temperature variations >> must play heck with a grainy cermet-like structure. > >I've never seen excess shot noise in resistors.. >(excess noise that is white in frequency.) >but I haven't looked hard either. >According to Landauer a resistor should reduce the shot noise >as 1/N, where N is the number of scattering events >as a "single electron pulse" travels the length of the resistor. > >I'm not sure how to think about "scattering events" >and 10 ns pulses.10 ns is still pretty slow.>> >> I've done a little testing on high-ohm resistors, but just enough to >> get me through a project. I needed 100 Mohms and wound up using two >> Dale custom-ordered 50M axials in series in mid-air, hopping over some >> surface-mount parts. Cermets were awful. This is fairly difficult >> stuff to measure, but would be a good student project or something. > >I wired up my LTC6090 HV opamps today. Fired it up on the test bench. >no smoke, but all I saw noise-wise, was first the room lights, and then >60Hz, with cross-over spikies from the linear supply.... I wanted to >go watch the UB bulls play in the "big dance".* So I'll have, >to put it in the box tomorrow...(Monday) Measuring noise can be hard. >I sometime wonder how I ever made things work in the past.... > >George H. > > *UB lost, we trailed the entire game, tied it up with ~3:00 to go, >and then lost. Still it was a great emotional ride for me. >+1, I hope Bobby Hurley stays and we go back. >(intellectually sports makes no sense at all, >for me, it's all about the emotion.) >Was that basketball? I never understood that game. It seems to be a bunch of guys with endocrine problems, running back and forth all night. Looks chaotic to me. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Reply by ●March 20, 20152015-03-20
On Friday, March 20, 2015 at 7:48:22 PM UTC-4, John Larkin wrote:> On Fri, 20 Mar 2015 16:04:17 -0700 (PDT), George Herold > <gherold@teachspin.com> wrote: > > >On Friday, March 20, 2015 at 2:28:25 PM UTC-4, John Larkin wrote: > >> On Thu, 19 Mar 2015 19:45:19 -0700 (PDT), George Herold > >> <gherold@teachspin.com> wrote: > >> > >> >On Thursday, March 19, 2015 at 9:33:52 PM UTC-4, Winfield Hill wrote: > >> >> John Larkin wrote... > >> >> > > >> >> > High-value (like, 100M) thickfilms seem to have excess noise. Metal > >> >> > films are hard to find at those values, but are much quieter. > >> >> > >> >> Yes. The "excess noise" I've observed is usually highly voltage dependent, > >> >> proportional to voltage, etc. It may be non-existent at low voltages, like > >> >> under 10V, but severe above 50 to 100V to 1kV. And it can vary dramatically > >> >> from part to part. E.e., made some +/-2KV amplifiers, with sub 1ppm noise > >> >> levels, and 20% of the 10kV feedback resistors were very noisy and had to be > >> >> replaced. > >> >> > >> >> But troublesome FET transistor noise may not have classic 1/f characteristics at > >> >> all, e.g., "telegraph noise" and "popcorn noise" are not strictly 1/f. Let's > >> >> talk after you've digested our measurements and the discussion in Chapter 8. > >> >> The book is being shipped as we speak. My order is in. > >> >> > >> >> > >> >> -- > >> >> Thanks, > >> >> - Win > >> > > >> >Excellent! I look forward to the discussion once > >> >my piece of the 40 tons arrives. > >> >I wonder if the voltage dependence is a surface > >> >thing. (could be the outer surface or some > >> >layer underneath.) > >> > > >> >George H. > >> > >> All resistors have the same Johnson noise at zero volts. Some, not > >> metal films, have shot noise with current flowing, which is simple > >> electron arrival statistics. > >> > >> I conjecture that some nonmetallic resistors, like cermets, may have > >> shot noise when biased. They also have excess, 1/f ish, noise, maybe > >> from their polycrystaline confused nature. Tiny temperature variations > >> must play heck with a grainy cermet-like structure. > > > >I've never seen excess shot noise in resistors.. > >(excess noise that is white in frequency.) > >but I haven't looked hard either. > >According to Landauer a resistor should reduce the shot noise > >as 1/N, where N is the number of scattering events > >as a "single electron pulse" travels the length of the resistor. > > > >I'm not sure how to think about "scattering events" > >and 10 ns pulses. > > 10 ns is still pretty slow.Well just a number out of the air... well typical PMT pulse width, that's more than one electron of course.> > > >> > >> I've done a little testing on high-ohm resistors, but just enough to > >> get me through a project. I needed 100 Mohms and wound up using two > >> Dale custom-ordered 50M axials in series in mid-air, hopping over some > >> surface-mount parts. Cermets were awful. This is fairly difficult > >> stuff to measure, but would be a good student project or something. > > > >I wired up my LTC6090 HV opamps today. Fired it up on the test bench. > >no smoke, but all I saw noise-wise, was first the room lights, and then > >60Hz, with cross-over spikies from the linear supply.... I wanted to > >go watch the UB bulls play in the "big dance".* So I'll have, > >to put it in the box tomorrow...(Monday) Measuring noise can be hard. > >I sometime wonder how I ever made things work in the past.... > > > >George H. > > > > *UB lost, we trailed the entire game, tied it up with ~3:00 to go, > >and then lost. Still it was a great emotional ride for me. > >+1, I hope Bobby Hurley stays and we go back. > >(intellectually sports makes no sense at all, > >for me, it's all about the emotion.) > > > > Was that basketball? I never understood that game. It seems to be a > bunch of guys with endocrine problems, running back and forth all > night. Looks chaotic to me.Yeah, I played it a lot when younger, which helps.. Team sports are fun. I wish I still had good knees, I hate running, but running after some ball is different. (I'm trying to win!) Basketball, soccer, tennis, volleyball... I still play ping-pong, a pretty small ball, and not much running. George H.> > > > > -- > > John Larkin Highland Technology, Inc > picosecond timing precision measurement > > jlarkin att highlandtechnology dott com > http://www.highlandtechnology.com