Detecting the peak value
in [Design]
|
Prev: Changing the color of objects/primitives only ? (flatshading...) (Re-using verteces, is user supplied data considered "state" aswell ?)
Next: Ronald Reagan socialism commercial
From: John Larkin on 1 Aug 2010 22:19 On Mon, 02 Aug 2010 11:53:39 +1000, Grant <omg(a)grrr.id.au> wrote: >On Sun, 01 Aug 2010 14:49:08 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon(a)On-My-Web-Site.com> wrote: > >>On Sun, 1 Aug 2010 18:51:09 GMT, hal(a)nospam.com wrote: >> >>> >>><snip> >>> >>>> >Yes, it seemed the best way to use the high precision components that I >>>> >have. I was going to charge through one of the precision Rs and measure >>>> >to >>>> >get the charging current. As long as the current source is linear over a >>>> >small range (a couple volts starting at .5 or 1V - NOT in-circuit, of >>>> >course), with an accurate clock, and using Kelvin sensing for both the >>>> >reference and DUT, I felt that I should be able to achieve very good >>>> >accuracy. I realize that this is a one-dimensional measurement of the >>>> >capacitance, but it suits for what I am trying to accomplish. The fun >>>> >part, >>>> >for me, is trying to see how accurate I can get that one measurement - >>>> >ditto >>>> >for ESR. One issue I'm still working on is a current source with the >>>> >required linearity. There are quite a few designs that are accurate at a >>>> >set current, but I don't have the expertise or equipment to determine if >>>> >they are accurate over a sweeping voltage range. Suggestions, anyone? >>>> > >>>> > >>>> >>>> Pch mosfet with a resistor in the source. An opamp compares the >>>> voltage drop across the resistor to some voltage reference, like an >>>> LM4040 maybe, and drives the fet gate. That will be constant to PPMs >>>> over a huge load voltage range, and will have good short-term >>>> stability. It might be a bit noisy from the reference noise, but you >>>> can use a better reference, or average most of that out. >>>> >>>> Just use something like the top half of this, with the reference up >>>> there. Or use the whole thing, drive it from a dac, and have a >>>> programmable current source. >>>> >>>> ftp://jjlarkin.lmi.net/Isrc.JPG >>>> >>>> Chopper opamps might be nice here. >>>> >>>> John >>> >>>That's more or less the circuit I was thinking about - I had been playing >>>with it in SPICE and it _seemed_ to have the required linearity, but I >>>wasn't sure. I'll probably have 3 ranges (50ma, 200ua, 1 ua - each a >>>dedicated source). I was also thinking about choppers, although still >>>looking at specs - especially for the 1ua range. All voltages and currents >>>are drived from a buffered and filtered internal reference (MAX6341), I'll >>>have +- 5v available, +-12 (or +12) if needed for the op-amps. Any special >>>considerations using the applicable portion of the above circuit for these >>>current ranges? Anything special to consider in the MOSFET? >>> >>>Thanks >>> >>>Hal >> >>I don't think such an active current mirror will hold up to your 0.1% >>accuracy... aforementioned channel-length modulation. >> >> ...Jim Thompson > >And how does one overcome that? > >I can't see a DC effect (not saying there isn't one) because steady >state it looks like the source sense resistor is seeing the drain >current, disregarding gate leakage. > >I can see there might be a dV/dt related issue due to drain-gate >capacitance, is this what you're talking about? > >All I can see is that there might be a voltage compliance change >due to the effect you mention. > >You published similar circuit, CurrentSourceConceptual.pdf four >years ago, but I didn't check if there's any related discussion. > >Grant. The best I've done is 0.1% linearity in a 13 nanosecond linear ramp. Trick for fast ones: add a hi-Z ferrite bead in the drain. Works wonders. John
From: Grant on 1 Aug 2010 23:08 On Sun, 01 Aug 2010 19:19:20 -0700, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >On Mon, 02 Aug 2010 11:53:39 +1000, Grant <omg(a)grrr.id.au> wrote: > >>On Sun, 01 Aug 2010 14:49:08 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon(a)On-My-Web-Site.com> wrote: >> >>>On Sun, 1 Aug 2010 18:51:09 GMT, hal(a)nospam.com wrote: >>> >>>> >>>><snip> >>>> >>>>> >Yes, it seemed the best way to use the high precision components that I >>>>> >have. I was going to charge through one of the precision Rs and measure >>>>> >to >>>>> >get the charging current. As long as the current source is linear over a >>>>> >small range (a couple volts starting at .5 or 1V - NOT in-circuit, of >>>>> >course), with an accurate clock, and using Kelvin sensing for both the >>>>> >reference and DUT, I felt that I should be able to achieve very good >>>>> >accuracy. I realize that this is a one-dimensional measurement of the >>>>> >capacitance, but it suits for what I am trying to accomplish. The fun >>>>> >part, >>>>> >for me, is trying to see how accurate I can get that one measurement - >>>>> >ditto >>>>> >for ESR. One issue I'm still working on is a current source with the >>>>> >required linearity. There are quite a few designs that are accurate at a >>>>> >set current, but I don't have the expertise or equipment to determine if >>>>> >they are accurate over a sweeping voltage range. Suggestions, anyone? >>>>> > >>>>> > >>>>> >>>>> Pch mosfet with a resistor in the source. An opamp compares the >>>>> voltage drop across the resistor to some voltage reference, like an >>>>> LM4040 maybe, and drives the fet gate. That will be constant to PPMs >>>>> over a huge load voltage range, and will have good short-term >>>>> stability. It might be a bit noisy from the reference noise, but you >>>>> can use a better reference, or average most of that out. >>>>> >>>>> Just use something like the top half of this, with the reference up >>>>> there. Or use the whole thing, drive it from a dac, and have a >>>>> programmable current source. >>>>> >>>>> ftp://jjlarkin.lmi.net/Isrc.JPG >>>>> >>>>> Chopper opamps might be nice here. >>>>> >>>>> John >>>> >>>>That's more or less the circuit I was thinking about - I had been playing >>>>with it in SPICE and it _seemed_ to have the required linearity, but I >>>>wasn't sure. I'll probably have 3 ranges (50ma, 200ua, 1 ua - each a >>>>dedicated source). I was also thinking about choppers, although still >>>>looking at specs - especially for the 1ua range. All voltages and currents >>>>are drived from a buffered and filtered internal reference (MAX6341), I'll >>>>have +- 5v available, +-12 (or +12) if needed for the op-amps. Any special >>>>considerations using the applicable portion of the above circuit for these >>>>current ranges? Anything special to consider in the MOSFET? >>>> >>>>Thanks >>>> >>>>Hal >>> >>>I don't think such an active current mirror will hold up to your 0.1% >>>accuracy... aforementioned channel-length modulation. >>> >>> ...Jim Thompson >> >>And how does one overcome that? >> >>I can't see a DC effect (not saying there isn't one) because steady >>state it looks like the source sense resistor is seeing the drain >>current, disregarding gate leakage. >> >>I can see there might be a dV/dt related issue due to drain-gate >>capacitance, is this what you're talking about? >> >>All I can see is that there might be a voltage compliance change >>due to the effect you mention. >> >>You published similar circuit, CurrentSourceConceptual.pdf four >>years ago, but I didn't check if there's any related discussion. >> >>Grant. > >The best I've done is 0.1% linearity in a 13 nanosecond linear ramp. >Trick for fast ones: add a hi-Z ferrite bead in the drain. Works >wonders. I don't need to go that fast :) But there's a spot where I need a very slow ramp. Think around mV per minute or slower type ramps. Probably a dithery DAC (delta-sigma), since drift over hours may kill a current source into cap multiplier for such a slow ramp rate. Though years ago I did work with an amazing circuit that implemented a three hour timer by making 1uF look like 10000uF, add 68M resistor, and feed a 555! Was in a very noisy environment where CMOS counters simply failed. Used a super gain preamp in front of opamp for the cap multiplier, metal can opamps. Grant. > >John
From: JosephKK on 2 Aug 2010 02:11 On Sat, 31 Jul 2010 09:47:09 +1000, Grant <omg(a)grrr.id.au> wrote: >On Fri, 30 Jul 2010 16:38:57 GMT, nico(a)puntnl.niks (Nico Coesel) wrote: > >>Grant <omg(a)grrr.id.au> wrote: >> >>>On Thu, 29 Jul 2010 12:01:20 -0700 (PDT), john1987 <conphiloso(a)hotmail.com> wrote: >>> >>>> >>>> >>>>Hi, >>>>Thanks for your response. I need analog or digital solution to >>>>implement it. I have doe search on peak detector on the internet and >>>>found many circuits. But I need something that can do what I mentioned >>>>in the diagram. Ideally with out those capacitor and resistor. If you >>>>can direct me in correct way than I wil be thankful. >>> >>>Also see my post in reply to Jim T. >>> >>>You can do it in software, but not easily for 100kHz. That's >>>need DSP techniques, I think? >> >>Nope. Just do random time sampling using a simple microcontroller. As >>long as the S&H can handle 100kHz it's fine. Keep in mind that today's >>microcontrollers are 32 bit, run at >50MHz and have hardware >>multipliers and dividers. > >Erm, in my mind, a microcontroller is smaller, like a PIC chip, >I really do have difficulty realising the ARMs and other 32bit >controllers are about as cheap and have so much greater power :) > >But then I time-travel in a sense from early '90s, returning to >electronics after a long break. And, picking up where I left off, >in a sense, by using the PIC chips. That'll do for this year, >for me. > >Grant. And the top end of PICs are in that range also.
From: Grant on 2 Aug 2010 02:16 On Sun, 01 Aug 2010 23:11:11 -0700, "JosephKK"<quiettechblue(a)yahoo.com> wrote: >On Sat, 31 Jul 2010 09:47:09 +1000, Grant <omg(a)grrr.id.au> wrote: > >>On Fri, 30 Jul 2010 16:38:57 GMT, nico(a)puntnl.niks (Nico Coesel) wrote: >> >>>Grant <omg(a)grrr.id.au> wrote: >>> >>>>On Thu, 29 Jul 2010 12:01:20 -0700 (PDT), john1987 <conphiloso(a)hotmail.com> wrote: >>>> >>>>> >>>>> >>>>>Hi, >>>>>Thanks for your response. I need analog or digital solution to >>>>>implement it. I have doe search on peak detector on the internet and >>>>>found many circuits. But I need something that can do what I mentioned >>>>>in the diagram. Ideally with out those capacitor and resistor. If you >>>>>can direct me in correct way than I wil be thankful. >>>> >>>>Also see my post in reply to Jim T. >>>> >>>>You can do it in software, but not easily for 100kHz. That's >>>>need DSP techniques, I think? >>> >>>Nope. Just do random time sampling using a simple microcontroller. As >>>long as the S&H can handle 100kHz it's fine. Keep in mind that today's >>>microcontrollers are 32 bit, run at >50MHz and have hardware >>>multipliers and dividers. >> >>Erm, in my mind, a microcontroller is smaller, like a PIC chip, >>I really do have difficulty realising the ARMs and other 32bit >>controllers are about as cheap and have so much greater power :) >> >>But then I time-travel in a sense from early '90s, returning to >>electronics after a long break. And, picking up where I left off, >>in a sense, by using the PIC chips. That'll do for this year, >>for me. >> >>Grant. > >And the top end of PICs are in that range also. Yeah, it's scary. I'll stay with the 12F & 16F for now, thanks. Grant.
From: JosephKK on 2 Aug 2010 02:35
On Fri, 30 Jul 2010 07:28:22 -0700 (PDT), john1987 <conphiloso(a)hotmail.com> wrote: >Hi, > >The accuracy requirements are as follows > >1. The frequency of the sine wave is 100 kHz. +/- 3us >2 The amplitude is 2 volts peak to peak, it should be accurately >measured at +/- 1.9 volts. >Its not a home work. :) > > >Thanks > >John > +/- 3 uS out of 5 uS (for a half cycle) is pretty sloppy. A constant delay filter of 100 nS and a comparator will yeild much better results. |\ signal --------+------------------|+\ | | \ | | \ | | >--------- out | | / + (| 100 nS)------|- / | / |/ Bias etc., ommited. Somewhat subject to high frequency noise and amplitude wobble. Typically withing a few degrees. |