Interesting TIA design
in [Design]
|
From: Joerg on 24 Oct 2009 20:13 John Larkin wrote: > On Sat, 24 Oct 2009 09:48:32 -0700, Joerg <invalid(a)invalid.invalid> > wrote: > [...] >> Always nice to enter a new field. I just entered that of injectors and >> high pressures. It does command some respect, if something flies off in >> a 4500 psi situation things can get ugly. > > Pinhole leaks, like from bad welds or defective hoses, can make > invisible fluid jets that cut like a sword, or inject hydraulic fluid > like a hypodermic. Bad stuff. > > Guys in ships, looking for steam leaks, would probe around carefully > with a broomstick. A leak would slice the end off. > That's why I never understand how people can operate a wood splitter without eye protection. They don't realize what's going on inside those hoses. -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
From: Phil Hobbs on 24 Oct 2009 20:56 Joerg wrote: > Phil Hobbs wrote: >> Joerg wrote: >>> Phil Hobbs wrote: > > [...] > >>>> Besides the TIA, the two most challenging parts were: >>>> (1) Achieving a 1% linearity spec over the full range, including the >>>> gain error from a 1x-128x programmable gain amp. Doing a built-up >>>> PGA with good accuracy at 1 MHz is surprisingly hard with jellybean >>>> parts, I discover--the on-resistance specs of muxes are the pits. >>>> (2N7002s to the rescue.) >>>> >>> >>> If you do a low-volume (meaning non-consumer) design check out canned >>> PGAs. One of my favorites is the AD603. They raised the price a >>> couple of bucks over the years but it is a great chip. >> >> Nice part. It won't make the 1% accuracy spec, though. >> > > Well, you need to linearize it a bit or servo it with an out-of-band > pilot tone. But for a mass product it ain't the ticket, demasiados Dolares. This is a lab mockup for something that might become a mass-market product. I don't know exactly what it's for. I was trying not to be too curious, but it might be for something like a Wii--no data transmission, but a fairly tight linearity spec. I built it with one kind of op amp etc. to make life easier for the folks prototyping it, because they won't want to fly me across the Pacific to help. The linearity error of the AD503 is a nice Chebyshevish thing with about 10 ripples, so linearizing it would really need that pilot tone. Detecting the pilot tone would need the back end of another AM radio, which would be a bit more complicated than I'd really like...that linearity spec again. Linear, wide range amplitude detectors are surprisingly hard to make when you really want 1% accuracy right down into the mud. (You might well have better tricks for that than I do, of course.) I'd probably try to avoid that problem by attenuating the pilot tone at the input of the amp, using something predictable like a BJT diff pair or one of your 4007/op amp gizmos, and servoing the gain to bring the pilot tone to a constant amplitude at the output. That way the amplitude detector would run at almost the same level at all times. The spec required binary weighted gains of 1 to 128, so I used a 3-> analog mux with the common terminal grounded, and just dorked the resistor values to take account of the on-resistance. It was sort of cute, using one mux to control the gain of two amplifier stages. The first stage had to run at lower impedance to reduce the noise, so the Ron of the mux was too high. I used the first 3 mux outputs digitally to drive three 2N7002s with their gates pulled high via 100k resistors. A couple of Schottky diodes turned the 1-of-N mux output into a 3-bit thermometer code, so that output A0 turned off all three FETs, A1 turned off #2 and #3, and A2 just turned off #3. If the gain setting was 2**3 through 2**7, all three FETs were turned on, so that the first stage ran at full gain and outputs A3-A7 changed the gain of the second stage the usual way. >>> Another option is to get a chip with at least two uncommitted >>> N-channcels on there, such as the CD4007 (cheap). Unbuffered logic >>> can work as well, even cheaper. Now servo them by running one of them >>> in a current source with opamp feedback. So your DAC commands a >>> voltage and this sets the exact Rdson because it fudges the gate >>> until that condition is met. The other FET that is part of your RF >>> divider will follow, as long as this is monolithic and not a >>> multi-chip part. But even the Rohm multi-chips are pretty good >>> because the devices usually come off the same wafer. No guarantees >>> with multi-chip, of course. >> >> That is a truly gnarly idea. One for the tricks file. >> > > You can also use this trick with RF dual N-channels or fast arrays for > super-fine delay tuning. First time I did that it was met with some > disgust in the design review but worked like a champ. It made a TMS320 > almost unemployed and obsoleted some expensive and terribly noisy > digital delay lines. Long story short they could have never made it work > reliably with those delay lines. > >>> >>> >>>> (2) Getting rid of the DC and low frequency photocurrent without >>>> using resistors. I made a BJT current sink degenerated with 4 >>>> diodes in series with the emitter. Using N diodes reduces the >>>> transconductance by a factor of N+1, and increases the equivalent >>>> shot noise voltage by sqrt(N+1), so the resulting current sink runs >>>> 10*log(N+1) dB lower than full shot noise--7 dB for 4 diodes. So >>>> far, so vanilla. The problem was that the capacitance of the diodes >>>> dominated at low photocurrents, which essentially doubled the 1 MHz >>>> noise at 40 nA. I used BFT25A C-B junctions for the diodes, but >>>> that was the best I could get. I'd love to figure out how to make >>>> fast, wide range, sub-Poissonian current sources below 100 nA. >>>> >>> >>> No chance to chop DC and LF the old-fashioned way, with inductors and >>> capacitors? >> >> Unfortunately not at these impedance levels. I'd need a 10 H inductor. > > > Jim would build you a gyrator on silicon :-) A sub-Poissonian gyrator would be a thing of beauty. I'm not holding my breath. >>>> So I learned something about a corner of the design space that I'd >>>> never worked in before, which I thought was pretty neat. >>>> >>> >>> Always nice to enter a new field. I just entered that of injectors >>> and high pressures. It does command some respect, if something flies >>> off in a 4500 psi situation things can get ugly. >> >> My son has a 4500 PSI carbon fibre air tank for his paintball gun. I >> explained the importance of keeping it in one piece, but I'm not sure >> I made an impression. Cutting a broomstick would do it. >> > > Be careful. A woman from our village was killed by a paintball tank that > flew off. Hit her in the head, she slumped into the arms of her husband > and was dead within seconds. Brr. 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: Joerg on 24 Oct 2009 21:12 Phil Hobbs wrote: > Joerg wrote: >> Phil Hobbs wrote: >>> Joerg wrote: >>>> Phil Hobbs wrote: >> >> [...] >> >>>>> Besides the TIA, the two most challenging parts were: >>>>> (1) Achieving a 1% linearity spec over the full range, including >>>>> the gain error from a 1x-128x programmable gain amp. Doing a >>>>> built-up PGA with good accuracy at 1 MHz is surprisingly hard with >>>>> jellybean parts, I discover--the on-resistance specs of muxes are >>>>> the pits. (2N7002s to the rescue.) >>>>> >>>> >>>> If you do a low-volume (meaning non-consumer) design check out >>>> canned PGAs. One of my favorites is the AD603. They raised the price >>>> a couple of bucks over the years but it is a great chip. >>> >>> Nice part. It won't make the 1% accuracy spec, though. >>> >> >> Well, you need to linearize it a bit or servo it with an out-of-band >> pilot tone. But for a mass product it ain't the ticket, demasiados >> Dolares. > > This is a lab mockup for something that might become a mass-market > product. I don't know exactly what it's for. I was trying not to be > too curious, but it might be for something like a Wii--no data > transmission, but a fairly tight linearity spec. I built it with one > kind of op amp etc. to make life easier for the folks prototyping it, > because they won't want to fly me across the Pacific to help. > If it's in Korea your wife might ban you from the bedroom after the trip. When I came back I reeked of garlic so badly that my wife complained even though she really likes garlic. If it's Japan she might complain as well because guys like you and me would never be able to resist buying tons of parts in Akihabara, filling up the basement. > The linearity error of the AD503 is a nice Chebyshevish thing with about > 10 ripples, so linearizing it would really need that pilot tone. > Detecting the pilot tone would need the back end of another AM radio, > which would be a bit more complicated than I'd really like...that > linearity spec again. Linear, wide range amplitude detectors are > surprisingly hard to make when you really want 1% accuracy right down > into the mud. (You might well have better tricks for that than I do, of > course.) > The radio could be a cheap little consumer chip. I've used RSSI outputs for that kind of stuff. But you'd have to scour the market to see what's going to stick around for enough years. > I'd probably try to avoid that problem by attenuating the pilot tone at > the input of the amp, using something predictable like a BJT diff pair > or one of your 4007/op amp gizmos, and servoing the gain to bring the > pilot tone to a constant amplitude at the output. That way the > amplitude detector would run at almost the same level at all times. > > The spec required binary weighted gains of 1 to 128, so I used a 3-> > analog mux with the common terminal grounded, and just dorked the > resistor values to take account of the on-resistance. It was sort of > cute, using one mux to control the gain of two amplifier stages. > > The first stage had to run at lower impedance to reduce the noise, so > the Ron of the mux was too high. I used the first 3 mux outputs > digitally to drive three 2N7002s with their gates pulled high via 100k > resistors. A couple of Schottky diodes turned the 1-of-N mux output > into a 3-bit thermometer code, so that output A0 turned off all three > FETs, A1 turned off #2 and #3, and A2 just turned off #3. If the gain > setting was 2**3 through 2**7, all three FETs were turned on, so that > the first stage ran at full gain and outputs A3-A7 changed the gain of > the second stage the usual way. > There is yet another esoteric method of variable RF attenuation: Double-balanced mixers or DBM. Feed DC into the IF port, use LO as input and RF as output. But it could be too expensive because DBMs are all well over a buck. I believe Mini-Circuits manufactures special ones for attenuator purposes now, no idea what they cost though. [...] -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
From: Phil Hobbs on 24 Oct 2009 21:48 Joerg wrote: > Phil Hobbs wrote: >> Joerg wrote: >>> Phil Hobbs wrote: >>>> Joerg wrote: >>>>> Phil Hobbs wrote: >>> >>> [...] >>> >>>>>> Besides the TIA, the two most challenging parts were: >>>>>> (1) Achieving a 1% linearity spec over the full range, including >>>>>> the gain error from a 1x-128x programmable gain amp. Doing a >>>>>> built-up PGA with good accuracy at 1 MHz is surprisingly hard with >>>>>> jellybean parts, I discover--the on-resistance specs of muxes are >>>>>> the pits. (2N7002s to the rescue.) >>>>>> >>>>> >>>>> If you do a low-volume (meaning non-consumer) design check out >>>>> canned PGAs. One of my favorites is the AD603. They raised the >>>>> price a couple of bucks over the years but it is a great chip. >>>> >>>> Nice part. It won't make the 1% accuracy spec, though. >>>> >>> >>> Well, you need to linearize it a bit or servo it with an out-of-band >>> pilot tone. But for a mass product it ain't the ticket, demasiados >>> Dolares. >> >> This is a lab mockup for something that might become a mass-market >> product. I don't know exactly what it's for. I was trying not to be >> too curious, but it might be for something like a Wii--no data >> transmission, but a fairly tight linearity spec. I built it with one >> kind of op amp etc. to make life easier for the folks prototyping it, >> because they won't want to fly me across the Pacific to help. >> > > If it's in Korea your wife might ban you from the bedroom after the > trip. When I came back I reeked of garlic so badly that my wife > complained even though she really likes garlic. If it's Japan she might > complain as well because guys like you and me would never be able to > resist buying tons of parts in Akihabara, filling up the basement. > > >> The linearity error of the AD503 is a nice Chebyshevish thing with >> about 10 ripples, so linearizing it would really need that pilot tone. >> Detecting the pilot tone would need the back end of another AM radio, >> which would be a bit more complicated than I'd really like...that >> linearity spec again. Linear, wide range amplitude detectors are >> surprisingly hard to make when you really want 1% accuracy right down >> into the mud. (You might well have better tricks for that than I do, >> of course.) >> > > The radio could be a cheap little consumer chip. I've used RSSI outputs > for that kind of stuff. But you'd have to scour the market to see what's > going to stick around for enough years. AD has come out with some really nice successive detection log amps lately--65 dB range with better than +-1 dB log conformity. Back in the day I'd use something like a MC3362 FM IF strip, but those are long gone. About the only jellybean left that I know of is the SA605/615, and those have gone up in price by a lot. Those also have two IF stages, and you have to get the interstage attenuation just right for the RSSI curve to come out straight, with no cliffs or plateaus. > > >> I'd probably try to avoid that problem by attenuating the pilot tone >> at the input of the amp, using something predictable like a BJT diff >> pair or one of your 4007/op amp gizmos, and servoing the gain to bring >> the pilot tone to a constant amplitude at the output. That way the >> amplitude detector would run at almost the same level at all times. >> >> The spec required binary weighted gains of 1 to 128, so I used a 3-> >> analog mux with the common terminal grounded, and just dorked the >> resistor values to take account of the on-resistance. It was sort of >> cute, using one mux to control the gain of two amplifier stages. >> >> The first stage had to run at lower impedance to reduce the noise, so >> the Ron of the mux was too high. I used the first 3 mux outputs >> digitally to drive three 2N7002s with their gates pulled high via 100k >> resistors. A couple of Schottky diodes turned the 1-of-N mux output >> into a 3-bit thermometer code, so that output A0 turned off all three >> FETs, A1 turned off #2 and #3, and A2 just turned off #3. If the >> gain setting was 2**3 through 2**7, all three FETs were turned on, so >> that the first stage ran at full gain and outputs A3-A7 changed the >> gain of the second stage the usual way. >> > > There is yet another esoteric method of variable RF attenuation: > Double-balanced mixers or DBM. Feed DC into the IF port, use LO as input > and RF as output. But it could be too expensive because DBMs are all > well over a buck. I believe Mini-Circuits manufactures special ones for > attenuator purposes now, no idea what they cost though. Several dollars in onesies, dunno in thousands. I love Mini-Circuits stuff--been using them since 1981 or so. Their high-output phase detector parts (RPD-1 in them days) are a good 10 dB better than ordinary DBMs for PLL use. They really saved my tuchis in my first engineering job, and I have a whole Vidmar drawer full of MCL stuff in my basement. Good medicine. Problem with DBMs as attenuators is linearity. You can linearize a diff pair with a couple of auxiliary diode-connected transistors (a la LM13700 input structure). Of course a pilot tone doesn't have much to intermodulate with until it gets into the signal path, so that would work fine here too. (I was thinking more of something like a BCV61C dual BJT.) We need a white board and a keg to do this really right, but it's fun even at a distance. 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: Joerg on 25 Oct 2009 11:35
Phil Hobbs wrote: > Joerg wrote: >> Phil Hobbs wrote: [...] >> >>> The linearity error of the AD503 is a nice Chebyshevish thing with >>> about 10 ripples, so linearizing it would really need that pilot >>> tone. Detecting the pilot tone would need the back end of another AM >>> radio, which would be a bit more complicated than I'd really >>> like...that linearity spec again. Linear, wide range amplitude >>> detectors are surprisingly hard to make when you really want 1% >>> accuracy right down into the mud. (You might well have better tricks >>> for that than I do, of course.) >>> >> >> The radio could be a cheap little consumer chip. I've used RSSI >> outputs for that kind of stuff. But you'd have to scour the market to >> see what's going to stick around for enough years. > > AD has come out with some really nice successive detection log amps > lately--65 dB range with better than +-1 dB log conformity. Back in the > day I'd use something like a MC3362 FM IF strip, but those are long > gone. About the only jellybean left that I know of is the SA605/615, > and those have gone up in price by a lot. Those also have two IF > stages, and you have to get the interstage attenuation just right for > the RSSI curve to come out straight, with no cliffs or plateaus. Yeah, they priced those right out of the market. Beats me why. There are always chips like the TDA7210 but who knows when the last order bell ringeth. Another option that won't go away any time soon is WWVB receivers chips. Many have a peak output which can be used to extract the pilot tone level. It's really meant to set the AGC time constant but there ain't no law against other uses. This one is IMHO too expensive and just meant as an example because it has a little more meat in the datasheet: http://www.c-max-time.com/downloads/getFile.php?id=533 Possibly the CR output of this one works: http://www.gsg-asia.com/ic_data_sheet/ak2125.pdf The Asian ones often come with a very skimpy datasheet so essentially you'd have to buy a few and try it out. Since your client is in Asia they should be able to obtain them for you. While they were made to work on 60kHz or 77.5kHz that doesn't mean they won't work at a few hundred kHz. It's just that they are sold for "atomic clocks". Meaning Walmart pricing :-) Also, you don't need crystal filtering for your purpose if the area around the pilot is quiet enough. Later when such a scheme works you can typicaly buy the bare bones bond versions that reside under a tar blob. >> >> >>> I'd probably try to avoid that problem by attenuating the pilot tone >>> at the input of the amp, using something predictable like a BJT diff >>> pair or one of your 4007/op amp gizmos, and servoing the gain to >>> bring the pilot tone to a constant amplitude at the output. That way >>> the amplitude detector would run at almost the same level at all times. >>> >>> The spec required binary weighted gains of 1 to 128, so I used a 3-> >>> analog mux with the common terminal grounded, and just dorked the >>> resistor values to take account of the on-resistance. It was sort of >>> cute, using one mux to control the gain of two amplifier stages. >>> >>> The first stage had to run at lower impedance to reduce the noise, so >>> the Ron of the mux was too high. I used the first 3 mux outputs >>> digitally to drive three 2N7002s with their gates pulled high via >>> 100k resistors. A couple of Schottky diodes turned the 1-of-N mux >>> output into a 3-bit thermometer code, so that output A0 turned off >>> all three FETs, A1 turned off #2 and #3, and A2 just turned off #3. >>> If the gain setting was 2**3 through 2**7, all three FETs were turned >>> on, so that the first stage ran at full gain and outputs A3-A7 >>> changed the gain of the second stage the usual way. >>> >> >> There is yet another esoteric method of variable RF attenuation: >> Double-balanced mixers or DBM. Feed DC into the IF port, use LO as >> input and RF as output. But it could be too expensive because DBMs are >> all well over a buck. I believe Mini-Circuits manufactures special >> ones for attenuator purposes now, no idea what they cost though. > > Several dollars in onesies, dunno in thousands. I love Mini-Circuits > stuff--been using them since 1981 or so. Their high-output phase > detector parts (RPD-1 in them days) are a good 10 dB better than > ordinary DBMs for PLL use. They really saved my tuchis in my first > engineering job, and I have a whole Vidmar drawer full of MCL stuff in > my basement. Good medicine. > Same here :-) > Problem with DBMs as attenuators is linearity. You can linearize a diff > pair with a couple of auxiliary diode-connected transistors (a la > LM13700 input structure). Of course a pilot tone doesn't have much to > intermodulate with until it gets into the signal path, so that would > work fine here too. (I was thinking more of something like a BCV61C > dual BJT.) > You'd have to set the pilot where there is the least in intermodulation. That is where my old millimeter-grid pad usually comes out. > We need a white board and a keg to do this really right, but it's fun > even at a distance. > Maybe a keg of Yuengling or some of the other good stuff you guys have over on the east coast. I really, really miss my white board. Even just for myself. No space for it in this office and in the garage I'd only have maybe 2-3 months out of a year. All other times you'd either get heat stroke or freeze. -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM. |