C-multiplier again
in [Design]
|
Prev: Futuristic weapon question: Anti-matter-gun, would it have a signature ?
Next: EEVblog Live Event
From: Mike on 28 May 2010 16:29 George Herold <gherold(a)teachspin.com> wrote: [...] > Thanks for all the advice Mike. I'm not even sure that is the same > Alum-electro that I'm using. I didn't 'spice' this circuit till a few > days ago. About a year or so ago, I proto-ed it on copper clad and > played with values till the noise was low enough. (A bit of a pain > since to test it I had to close it up inside a metal box.) Then it > oscillated at certain bias currents, and I f'ed around till that went > away. (I'm mostly just a circuit 'hack' though I hope I'm starting to > understand things.) > > George H. (Sorry for the Walt confusion on this thread, 'my' Walt > knows nothing about circuits, but makes some 'killer' mustard.) Hi George, I kinda jumped the gun. But I love mustard, especially on Polish sausages. So either Walt is fine:) I just posted a test attachment using Fairchild's model for the MPSA14. I assume this is converted to Base64 during transmission, then back to plain ascii when received. My newsreader automatically saves attachments to a download directory. I just checked, and sure enough the files were there and worked fine. If you have a minute, see if your newsreader will recover these files properly and see if they will run in LTspice. I'll give google groups time to archive the files, then see if they show up there. If so, and everyone can read these files in the multitude of newsreaders out there, maybe this solves the problem. And maybe it will also save zip files. That would be handy. Thanks, Mike
From: Phil Hobbs on 28 May 2010 21:46 John Larkin wrote: > On Thu, 27 May 2010 22:04:34 -0700 (PDT), whit3rd <whit3rd(a)gmail.com> > wrote: > >>>> I have an interesting idea. How about a blue LED as the reference. >>> I just used two diodes in series to make a low-noise -1.5 volt >>> shunt-type supply. I could have used an LED, which would be cool - >>> they light up! - but I didn't want any stray light inside our box. >> GAAA! If you want low noise from AN OPTICALLY OPEN DIODE >> you need to shield from light. Incandescent will cause 120 Hz >> input, fluorescent 60 Hz and 120 Hz, electronic ballasts and >> CCFL can go from kilohertz to megahertz. >> > > > I wonder what the numbers are like here. Suppose one used a > front-panel type green LED as a power indicator and voltage reference. > The dynamic impedance of the LED will be ohms. I'd guess that any > fluorescent light induced current would be nanoamps. So we'd have > nanovolts of optically-induced noise. A cap across the LED would kill > the high frequency stuff, like from electronic ballasts. > > Of course, the current source would have to be a lot better than your > average 5-volt-supply-resistor thing. > > John > If you hang an emitter follower on the LED, the tempcos cancel out to leading order, so you can make a pretty reasonable voltage reference--much quieter than a bandgap, though less accurate. It's pretty unlikely for a LED to detect more than a microamp in any indoor setting, but it could be as much as 10 uA in sunlight. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
From: dagmargoodboat on 29 May 2010 02:08 On May 28, 11:04 am, Mike <s...(a)me.not> wrote: > dagmargoodb...(a)yahoo.com wrote: > > Yeah, but Walt *was* the manufacturer, designing in his own part, > > knowing what's inside. > > > I see that AD's 25mA limit for the AD797 input diodes was a > > recommendation, not an absolute instantaneous limit. Not knowing > > what's inside, withstanding a worst-case spike at 7x the recommended > > d.c. limit seems entirely plausible. Or maybe Walt just didn't want > > to give up any more noise performance. > > > Maybe he was just designing to withstand start up and shut down > > transients. > > > If the circuit doesn't protect against hard dead shorts, then it > > wasn't supposed to. > > > If you are happy with unknown fault response and 1988 noise performance, > then by all means stick with that approach. Shoot, I wasn't saying that--I'm not using the thing. I thought the cap bootstrap was pretty cute, and that the circuit might be a useful data point for John, that's all. 1.25nV/rtHz at the output is pretty decent, vs. 15nV/rtHz (typ) for the LM8261 alone. As far as exceeding the manufacturer's spec, it's quite possible that Walt, the manufacturer, was telling us from personal knowledge that this was okay. Or not. Anyway, if you dislike the 1rst stage you'll hate the 2nd--that's got 10 ohms and 1,500uF instead of 49.9 ohms and 10uF. Walt used ordinary wimpy signal diodes to protect that stage, so it looks like he recognized a pitfall, and thought two wimpy diodes sufficient. By his attention to it in the circuit, it looks like Walt thought about input protection considerations, and thought these measures appropriate. Were they? I don't know, I'm not Walt. ISTM R3 offers some protection against whatever danger Walt saw and wanted to protect. If you'd rather ditch it and go naked as you wrote up thread, by all means rock on -- you can get under 1nV/rtHz that way. -- Cheers, James Arthur
From: Mike on 29 May 2010 03:47 dagmargoodboat(a)yahoo.com wrote: > ISTM R3 offers some protection against whatever danger Walt saw > and wanted to protect. If you'd rather ditch it and go naked as > you wrote up thread, by all means rock on - you can get under > 1nV/rtHz that way. > Cheers, > James Arthur My proposal was to add back-to-back schottky diodes across the input pins, then reduce the series resistance R3 to the electrolytics. It turns out that low noise op amps set an absolute maximum voltage across the input pins to keep the the internal diodes off. There may or may not be a corresponding current limit specified. The following is a short list of the maximum input voltage of some low noise op amps. Note "nV" stands for "nV/sqrt(Hz)," and some op amps have a minimum stable gain. AD797 : 0.90nV +/-0.7V 25 mA AD8099 : 0.95nV +/-1.8V +/-10mA G >=2 ADA4898 : 0.90nV +/-1.5V LMH6624 : 0.92nV +/-1.2V LT1128 : 0.85nV +/-1.8V +/-25mA LT6200 : 0.95nV +/-0.7V +/-40mA OPA687 : 0.95nV +/-1.2V G >= 12 OPA847 : 0.85nV +/-1.2V G >= 12 The +/-0.7V of the AD797 and LT6200 are the toughest requirements to meet. An IRF 11DQ05 was one of the first schottkys that looked suitable: http://www.datasheetcatalog.org/datasheet/irf/11dq05.pdf If you look at Fig. 1, "Maximum Forward Voltage Drop Characteristics" on page 3, you can see the forward voltage is 0.7V with a diode current of 2A at 25C. Since the instantaneous voltage across the input pins is 10V with a hard short, we can set the series resistor to R = E / I = 10 / 2 = 5 Ohms The highest forward voltage shown on the graph is 1.1V, corresponding to a forward current of 10A. This means the rest of the op amps could use a series resistance on 1 Ohm and still be below spec. John Larkin raised the issue of diode leakage and added capacitance adding a pole at the input which could make the amplifier unstable. Here is a portion of my reply: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ There are large electrolytics at both inputs. These would swamp any extra capacitance from the diodes. There are 10k and 1k resistors at the inputs. These resistors, plus the electrolytics, would swamp any noise from diode leakage. Since I may need to use a similar approach, I modeled this circuit in LTspice and uploaded the archive 3CBC2208.ZIP to abse. The Transient Analysis is shown in 3CBBA0E7.ASC. I modeled the AD797 as a single pole op amp with Avol=15Meg, GBW=110Meg, Slew=110Meg. I added RX=7.5K across the input pins to match the AD797 datasheet, and also added CX=1nF across the input pins to account for any stray capacitance from the schottky diodes. I evaluated the response with CX and RX both included and excluded in the feedback from the pertubation source V2. As shown in the Transient Analysis, there is considerable ringing at the output of the op amp after a step change at the input. This appears to be caused by driving the large electrolytic cap C5 at the output. Reducing the value of C5 causes the ringing to increase. Increasing the value causes the slew rate limiting to increase. Changing the value of CX from 1pF to 10nF has little or no effect on the response. Changing the value of the series input resistor, R3, from 2 ohms to 49.9 ohms has little or no effect on the response. Thus any extra capacitance from adding schottky protection diodes at the input will not change the loop characteristics. The Open Loop Gain is shown in 3CBC2036.ASC. The response changes considerably depending on how CX and RX are located at the input. However, in either case, the value of CX either has no effect on the open loop response, or improves it. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Next, to confirm the expected diode turnon, Jack Smith of Clifton Laboratories posted a page titled "Diode Turn-on/off Time and Relay Snubbing" at: http://www.cliftonlaboratories.com/diode_turn-on_time.htm He used a HP 8012B set to +/-3.9V. The 50 ohm series resistance gave a measured output current of about 65mA. The 11DQ05 is shown at: http://www.cliftonlaboratories.com/diode_turn-on_time.htm#11DQ05 As expected, the diode turns on and off quickly, and shows a forward drop of 0.4V at this current. Finally, some datasheets may recommend adding a small resistor in series with the input pin, presumably to prevent oscillation due to the high bandwidth of the input transistors. The extra resistor increases the overall noise, which defeats the purpose of the amplifier. It may be possible to substitute a one or two turn ferrite bead to kill any high frequency oscillation, and retain the low noise performance below 1MHz. Thanks, Mike
From: MooseFET on 29 May 2010 11:25
On May 28, 7:14 am, John Larkin <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: > On Thu, 27 May 2010 22:04:34 -0700 (PDT), whit3rd <whit...(a)gmail.com> > wrote: > > > > >> >I have an interesting idea. How about a blue LED as the reference. > > >> I just used two diodes in series to make a low-noise -1.5 volt > >> shunt-type supply. I could have used an LED, which would be cool - > >> they light up! - but I didn't want any stray light inside our box. > > >GAAA! If you want low noise from AN OPTICALLY OPEN DIODE > >you need to shield from light. Incandescent will cause 120 Hz > >input, fluorescent 60 Hz and 120 Hz, electronic ballasts and > >CCFL can go from kilohertz to megahertz. > > I wonder what the numbers are like here. Suppose one used a > front-panel type green LED as a power indicator and voltage reference. > The dynamic impedance of the LED will be ohms. I'd guess that any > fluorescent light induced current would be nanoamps. So we'd have > nanovolts of optically-induced noise. A cap across the LED would kill > the high frequency stuff, like from electronic ballasts. > > Of course, the current source would have to be a lot better than your > average 5-volt-supply-resistor thing. "Can you say bootstrap. Sure you can." If you add a digital pot to the design so that you can trim it, the accuracy can be made darn good. You need to subtract the tempco away but a silicon diodes drop is a good match to a non-superbright LEDs tempco. A current saving idea: Vcc ! --------!+\ ! ! >-------+----------Ref out ! ---!-/ ! ! ! ! ! ---------+ [R] ! ! ! ! [LED] ! ! ! ! ! GND ! --------------+---[R]---GND The operating current of many op-amps is nearly constant with voltage. I don't know if it is noisy or not. The op-amp is only given just enough gain to make the output above the Vee pin enough. Temperature compensation could perhaps be done by putting a thermistor into the divider chain. |