On Tue, 20 Aug 2013 15:13:10 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:
>On 8/20/2013 2:13 PM, John Larkin wrote:
>> On Tue, 20 Aug 2013 13:35:44 -0400, Phil Hobbs
>> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>>
>>> On 8/20/2013 12:14 PM, John Larkin wrote:
>>>> On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>>>>
>>>>
>>>>
>>>>> I was trying to measure both log conformance and beta linearity in one
>>>>> go, to pretty good accuracy. The usual trick of using tee networks in
>>>>> the feedback loop plus really really low offset amplifiers wouldn't work
>>>>> because chopamps all seem to have about 200 pA of input bias current.
>>>>
>>>> Chopamps usually are shooting healthy packets of charge out of both input pins.
>>>> Offset voltage and bias current can depend on what impedances the input pins
>>>> see, especially whether they see capacitance or resistance.
>>>
>>> Some seem to be a lot better than others, e.g. the OPA378 doesn't seem
>>> nearly as bad as the OPA2188.
>>>
>>> Cheers
>>>
>>> Phil Hobbs
>>
>> I did a 3-opamp 6-pole SK filter with AD8628s. That seems to be an
>> especially bad configuration... one opamp input has a cap to ground
>> and the other input connects to the output (unity-gain follower.) I
>> got something like 5 uv offset per amp. Had to add a tiny trimpot to
>> tweak the overall offset to zero. This was for some little plug-in
>> filter boardlets.
>>
>> https://dl.dropboxusercontent.com/u/53724080/Circuits/Filters/22A451A1.jpg
>>
>> https://dl.dropboxusercontent.com/u/53724080/Circuits/Filters/22A451A3.jpg
>>
>>
>> Next time, I'll breadboard anything critical that uses a chopamp; the
>> data sheets don't generally mention charge injection.
>
>Sometimes they do come clean, e.g.
><https://dl.dropboxusercontent.com/u/105025837/sed/OPA2188ChopperNoise.pdf>
>
>Cheers
>
>Phil Hobbs
Thanks. The paint can is cool.
--
John Larkin Highland Technology, Inc
jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com
Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro acquisition and simulation
Reply by Phil Hobbs●August 20, 20132013-08-20
On 8/20/2013 2:13 PM, John Larkin wrote:
> On Tue, 20 Aug 2013 13:35:44 -0400, Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>
>> On 8/20/2013 12:14 PM, John Larkin wrote:
>>> On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
>>> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>>>
>>>
>>>
>>>> I was trying to measure both log conformance and beta linearity in one
>>>> go, to pretty good accuracy. The usual trick of using tee networks in
>>>> the feedback loop plus really really low offset amplifiers wouldn't work
>>>> because chopamps all seem to have about 200 pA of input bias current.
>>>
>>> Chopamps usually are shooting healthy packets of charge out of both input pins.
>>> Offset voltage and bias current can depend on what impedances the input pins
>>> see, especially whether they see capacitance or resistance.
>>
>> Some seem to be a lot better than others, e.g. the OPA378 doesn't seem
>> nearly as bad as the OPA2188.
>>
>> Cheers
>>
>> Phil Hobbs
>
> I did a 3-opamp 6-pole SK filter with AD8628s. That seems to be an
> especially bad configuration... one opamp input has a cap to ground
> and the other input connects to the output (unity-gain follower.) I
> got something like 5 uv offset per amp. Had to add a tiny trimpot to
> tweak the overall offset to zero. This was for some little plug-in
> filter boardlets.
>
> https://dl.dropboxusercontent.com/u/53724080/Circuits/Filters/22A451A1.jpg
>
> https://dl.dropboxusercontent.com/u/53724080/Circuits/Filters/22A451A3.jpg
>
>
> Next time, I'll breadboard anything critical that uses a chopamp; the
> data sheets don't generally mention charge injection.
On Tue, 20 Aug 2013 13:35:44 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:
>On 8/20/2013 12:14 PM, John Larkin wrote:
>> On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
>> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>>
>>
>>
>>> I was trying to measure both log conformance and beta linearity in one
>>> go, to pretty good accuracy. The usual trick of using tee networks in
>>> the feedback loop plus really really low offset amplifiers wouldn't work
>>> because chopamps all seem to have about 200 pA of input bias current.
>>
>> Chopamps usually are shooting healthy packets of charge out of both input pins.
>> Offset voltage and bias current can depend on what impedances the input pins
>> see, especially whether they see capacitance or resistance.
>
>Some seem to be a lot better than others, e.g. the OPA378 doesn't seem
>nearly as bad as the OPA2188.
>
>Cheers
>
>Phil Hobbs
I did a 3-opamp 6-pole SK filter with AD8628s. That seems to be an
especially bad configuration... one opamp input has a cap to ground
and the other input connects to the output (unity-gain follower.) I
got something like 5 uv offset per amp. Had to add a tiny trimpot to
tweak the overall offset to zero. This was for some little plug-in
filter boardlets.
https://dl.dropboxusercontent.com/u/53724080/Circuits/Filters/22A451A1.jpghttps://dl.dropboxusercontent.com/u/53724080/Circuits/Filters/22A451A3.jpg
Next time, I'll breadboard anything critical that uses a chopamp; the
data sheets don't generally mention charge injection.
--
John Larkin Highland Technology, Inc
jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com
Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro acquisition and simulation
Reply by Jim Thompson●August 20, 20132013-08-20
On Tue, 20 Aug 2013 13:10:50 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:
>On 8/20/2013 11:44 AM, Jim Thompson wrote:
>> On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
>> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>>
>>> On 8/20/2013 2:22 AM, Jan Panteltje wrote:
>>>> On a sunny day (Mon, 19 Aug 2013 16:50:36 -0400) it happened Phil Hobbs
>>>> <pcdhSpamMeSenseless@electrooptical.net> wrote in
>>>> <-b2dnULzjrWBGI_PnZ2dnUVZ_hidnZ2d@supernews.com>:
>>>>
>>>>>> That depends, take an RF amp, AGC done by changing Ic, signals are
>>>>>> small, no feedback.
>>>>>
>>>>> But that's a transimpedance effect, not a current-gain effect. If beta
>>>>> is constant, changing I_C doesn't change the gain of a current-mode
>>>>> amplifier.
>>>>
>>>> I was always under the impression that beta depends on Ic???
>>>> Confirms your own statement that beta depends on a lot of things :-)
>>>
>>> Not in a predictable way, it doesn't.
>>
>> "Predictable"? Sure it does, see...
>>
>> http://www.analog-innovations.com/SED/BipolarTransistor_FromPSPCREF.pdf
>>
>> Also some curves from previously posted questions...
>>
>> http://www.analog-innovations.com/SED/BetaCurves.pdf
>>
>> http://www.analog-innovations.com/SED/Beta_Curves_SED.pdf
>>
>
>Models do, and variations among devices on the same chip are relatively
>small. However, things like defect density and variation of doping vs.
>depth vary from wafer to wafer, and those affect the shape of the beta
>vs I_C curve, do they not?
Not much anymore. You're living in the past ;-) when process
variations were much greater.
Manufacturer to manufacturer... now that's a different story.
>
>Besides, the context was Jan's claim that his BJT AGC amps work
>primarily by controlling beta via I_C, whereas I claim that it's a
>transconductance effect. A purely beta-controlled amplifier might
>easily have to use a 100:1 ratio of I_C to get a 10 dB gain range, and
>the available range would depend a lot on the selection of the
>individual device.
Jan doesn't understand that what happens in a typical RF stage, AGC'd
that way, is the input impedance pads down the gain.
>
>OTOH in a transconductance amp it's only a nice repeatable factor of
>sqrt(10), corresponding to a delta-V_BE= kT/e*ln(sqrt(10)) ~30 mV
>(single-ended).
>
>Would you build a beta-controlled AGC amp?
Not a chance, I don't rely on beta period!
>
>Cheers
>
>Phil Hobbs
...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
Reply by Phil Hobbs●August 20, 20132013-08-20
On 8/19/2013 11:18 PM, bloggs.fredbloggs.fred@gmail.com wrote:
> On Monday, August 19, 2013 10:20:26 PM UTC-4, Phil Hobbs wrote:
>> On 8/19/2013 10:09 PM, bloggs.fredbloggs.fred@gmail.com wrote:
>>
>>> On Monday, August 19, 2013 10:30:43 AM UTC-4, Phil Hobbs wrote:
>>
>>>
>>
>>>>
>>
>>>> The base current is really a loss mechanism--an ideal BJT has
>>>> zero
>>
>>>> base current, because all of the emitter current makes it to
>>>> the
>>
>>>> collector and none recombines in the base region.
>>>> (Recombination is
>>
>>>> where the base current comes from.)
>>
>>>
>>
>>> That's not quite true. Base current is also due to minority
>>> carrier
>>
>>> diffusion as well as recombination.
>>
>>
>>
>> True. I think that's a small effect in practical devices, though,
>> since
>>
>> it wouldn't vary much from device to device whereas beta is all
>> over the
>>
>> map.
>>
>>
>>
>>>
>>
>>>>
>>
>>>> The fundamental control mechanism of a BJT is the base-emitter
>>
>>>> voltage, which provides pretty tight voltage feedback in an
>>
>>>> emitter follower.
>>
>>>
>>
>>> Maybe, but you don't drive transistor ports with ideal voltage
>>
>>> sources.
>>
>>
>>
>> I'm not sure I'm getting your point here. The OP was thinking of
>> the
>>
>> BJT solely as a current-gain device. That idea doesn't in itself
>>
>> constrain the base-emitter voltage at all, whereas thinking about
>> it in
>>
>> terms of transconductance does.
>>
>
> I don't know how gm helps you in this case. For purposes of
> determining the transistor loading on the zener, most people agree it
> will be (beta +1) x Rload. If this estimate is all over the map
> because of beta, there's nothing you can do to improve on it by using
> gm, the loading will be still be all over the map. Manufacturers help
> by providing beta variation with Ic and min and max values, things
> that help bound the variation.
The usual way to solve it is by the Victorian technique: "Add mass until
nothing breaks." ;)
That is, crank up the zener current or use a reference with internal
feedback.
My claim isn't that knowing beta is useless, far from it. We all use it
all the time for exactly this sort of calculation. But it isn't enough
on its own, because all you have is beta, you can calculate how far the
zener will sag, but not how far the emitter will sag. Putting a 10k
resistor in series with the base inside the transistor package won't
change beta, but it sure will trash the load regulation.
If you know that the transconductance g_m = eI_C/kT, and that the base
current is small, you can do a lot better than if all you know is beta.
Which is why these sorts of regulators use BJTs and not FETs. ;)
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 USA
+1 845 480 2058
hobbs at electrooptical dot net
http://electrooptical.net
Reply by Phil Hobbs●August 20, 20132013-08-20
On 8/20/2013 12:14 PM, John Larkin wrote:
> On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>
>
>
>> I was trying to measure both log conformance and beta linearity in one
>> go, to pretty good accuracy. The usual trick of using tee networks in
>> the feedback loop plus really really low offset amplifiers wouldn't work
>> because chopamps all seem to have about 200 pA of input bias current.
>
> Chopamps usually are shooting healthy packets of charge out of both input pins.
> Offset voltage and bias current can depend on what impedances the input pins
> see, especially whether they see capacitance or resistance.
Some seem to be a lot better than others, e.g. the OPA378 doesn't seem
nearly as bad as the OPA2188.
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 USA
+1 845 480 2058
hobbs at electrooptical dot net
http://electrooptical.net
Reply by Phil Hobbs●August 20, 20132013-08-20
On 8/20/2013 12:10 PM, Jan Panteltje wrote:
> On a sunny day (Tue, 20 Aug 2013 11:21:41 -0400) it happened Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net> wrote in
> <kv01i7$udr$1@dont-email.me>:
>
>>>> A base resistor functions a lot like emitter degeneration, except it
>>>> isn't as predictable because beta varies orders of magnitude more than
>>>> g_M for a given collector current.
>>>
>>> Well, it is early morning, I will save the word salad for dinner.
>>
>> What's so hard to understand? A resistor R in series with the base
>> provides the same negative feedback as R/beta in series with the
>> emitter, except for the effects of shunt capacitance and beta
>> nonlinearity. (*)
>
> It is not hard to understand, my error, I was not 100% awake yet...
>
>
> As to beta versus Ic, I will have to look up some transistor specs,
> RF transistor specs...
> I see for example the BFR92 has a flat curent gain versus IC,
> while the BFR96 falls of below 40 mA, making it more suitable for AGC.
> Both are 5GHz wideband small signal NPN...
> But I would probably use a dual gate MOSFET for AGC...
>
>
>
>
>>> If normally somebody would show me such a thing for a design review I would suggest looking
>>> if there was an other way..
>>>
>>> Moisture, high humidity, you'd need a sealed box too.
>>
>> Oh, I'm with you there. I certainly wouldn't do it for anything but a
>> lab one-off, because it takes too much babysitting.
>>
>> I was trying to measure both log conformance and beta linearity in one
>> go, to pretty good accuracy. The usual trick of using tee networks in
>> the feedback loop plus really really low offset amplifiers wouldn't work
>> because chopamps all seem to have about 200 pA of input bias current. I
>> could have used a charge dispensing loop, but it's a lot easier to store
>> samples of a voltage. The right approach would have been to use more
>> complicated MUXing and a bunch of online calibration, but I'm not as
>> fast at MCU stuff as you are, and since the job was for a university
>> group with limited funds, I really wanted to keep the hours down. It
>> works fine as long as it's me driving it.
>>
>> Probably most of us have a bunch of those one-offs lying around that
>> we've built over the years, and they come in really handy sometimes.
>> (Of course I also have a Keithley 405 Micro-Microammeter that I got for
>> $5 on eBay--it has a 100 fA FS range, but takes a good two hours' warmup
>> to get down that low. It's really fun to use, though--electrometer
>> tubes are actually pretty amazing.)
>
> Nice, good price too.
>
>
>
>> I bought a Keithley 602 for very cheap, mostly for the meter, range
>> switch, and box. All solid-state, several years newer, 100 times less
>> sensitive. I'll put new guts in it one of these days when I'm motivated.
>
>
> I have ordered a Sony super HAL CCD 'starlight' camera module.
> what it can do:
> http://www.sony.net/Products/SC-HP/effiowld/ 0.001 lux...
> ebay:
> http://www.ebay.com/itm/150705214606?ru=http%3A%2F%2Fwww.ebay.com%2Fsch%2Fi.html%3F_from%3DR40%26_sacat%3D0%26_nkw%3D150705214606%26_rdc%3D1
>
> I had almost pinged you for help trying to make sense of the specs...
> It is supposed to give a clear picture with just starlight illumination (I mean at 30 fps).
> On board is a DSP too that reduces contrast so you can get contrast from areas while in
> in other areas there is strong light,
> Will go in my 'drone' alias 160 km/h SDcard trafficing ..
> This is to bypass NSA (Hello) of course.
> It is faster than internet...
I feel your pain. CCD datasheets are uniformly horrible.
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 USA
+1 845 480 2058
hobbs at electrooptical dot net
http://electrooptical.net
Reply by Phil Hobbs●August 20, 20132013-08-20
On 8/20/2013 11:44 AM, Jim Thompson wrote:
> On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>
>> On 8/20/2013 2:22 AM, Jan Panteltje wrote:
>>> On a sunny day (Mon, 19 Aug 2013 16:50:36 -0400) it happened Phil Hobbs
>>> <pcdhSpamMeSenseless@electrooptical.net> wrote in
>>> <-b2dnULzjrWBGI_PnZ2dnUVZ_hidnZ2d@supernews.com>:
>>>
>>>>> That depends, take an RF amp, AGC done by changing Ic, signals are
>>>>> small, no feedback.
>>>>
>>>> But that's a transimpedance effect, not a current-gain effect. If beta
>>>> is constant, changing I_C doesn't change the gain of a current-mode
>>>> amplifier.
>>>
>>> I was always under the impression that beta depends on Ic???
>>> Confirms your own statement that beta depends on a lot of things :-)
>>
>> Not in a predictable way, it doesn't.
>
> "Predictable"? Sure it does, see...
>
> http://www.analog-innovations.com/SED/BipolarTransistor_FromPSPCREF.pdf
>
> Also some curves from previously posted questions...
>
> http://www.analog-innovations.com/SED/BetaCurves.pdf
>
> http://www.analog-innovations.com/SED/Beta_Curves_SED.pdf
>
Models do, and variations among devices on the same chip are relatively
small. However, things like defect density and variation of doping vs.
depth vary from wafer to wafer, and those affect the shape of the beta
vs I_C curve, do they not?
Besides, the context was Jan's claim that his BJT AGC amps work
primarily by controlling beta via I_C, whereas I claim that it's a
transconductance effect. A purely beta-controlled amplifier might
easily have to use a 100:1 ratio of I_C to get a 10 dB gain range, and
the available range would depend a lot on the selection of the
individual device.
OTOH in a transconductance amp it's only a nice repeatable factor of
sqrt(10), corresponding to a delta-V_BE= kT/e*ln(sqrt(10)) ~30 mV
(single-ended).
Would you build a beta-controlled AGC amp?
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 USA
+1 845 480 2058
hobbs at electrooptical dot net
http://electrooptical.net
Reply by John Larkin●August 20, 20132013-08-20
On Tue, 20 Aug 2013 11:21:41 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:
>I was trying to measure both log conformance and beta linearity in one
>go, to pretty good accuracy. The usual trick of using tee networks in
>the feedback loop plus really really low offset amplifiers wouldn't work
>because chopamps all seem to have about 200 pA of input bias current.
Chopamps usually are shooting healthy packets of charge out of both input pins.
Offset voltage and bias current can depend on what impedances the input pins
see, especially whether they see capacitance or resistance.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators
Reply by Jan Panteltje●August 20, 20132013-08-20
On a sunny day (Tue, 20 Aug 2013 11:21:41 -0400) it happened Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote in
<kv01i7$udr$1@dont-email.me>:
>>> A base resistor functions a lot like emitter degeneration, except it
>>> isn't as predictable because beta varies orders of magnitude more than
>>> g_M for a given collector current.
>>
>> Well, it is early morning, I will save the word salad for dinner.
>
>What's so hard to understand? A resistor R in series with the base
>provides the same negative feedback as R/beta in series with the
>emitter, except for the effects of shunt capacitance and beta
>nonlinearity. (*)
It is not hard to understand, my error, I was not 100% awake yet...
As to beta versus Ic, I will have to look up some transistor specs,
RF transistor specs...
I see for example the BFR92 has a flat curent gain versus IC,
while the BFR96 falls of below 40 mA, making it more suitable for AGC.
Both are 5GHz wideband small signal NPN...
But I would probably use a dual gate MOSFET for AGC...
>> If normally somebody would show me such a thing for a design review I would suggest looking
>> if there was an other way..
>>
>> Moisture, high humidity, you'd need a sealed box too.
>
>Oh, I'm with you there. I certainly wouldn't do it for anything but a
>lab one-off, because it takes too much babysitting.
>
>I was trying to measure both log conformance and beta linearity in one
>go, to pretty good accuracy. The usual trick of using tee networks in
>the feedback loop plus really really low offset amplifiers wouldn't work
>because chopamps all seem to have about 200 pA of input bias current. I
>could have used a charge dispensing loop, but it's a lot easier to store
>samples of a voltage. The right approach would have been to use more
>complicated MUXing and a bunch of online calibration, but I'm not as
>fast at MCU stuff as you are, and since the job was for a university
>group with limited funds, I really wanted to keep the hours down. It
>works fine as long as it's me driving it.
>
>Probably most of us have a bunch of those one-offs lying around that
>we've built over the years, and they come in really handy sometimes.
>(Of course I also have a Keithley 405 Micro-Microammeter that I got for
>$5 on eBay--it has a 100 fA FS range, but takes a good two hours' warmup
>to get down that low. It's really fun to use, though--electrometer
>tubes are actually pretty amazing.)
Nice, good price too.
>I bought a Keithley 602 for very cheap, mostly for the meter, range
>switch, and box. All solid-state, several years newer, 100 times less
>sensitive. I'll put new guts in it one of these days when I'm motivated.
>> But when I had to design a big thing to drive RGB into long cables I decided to use very expensive opamps
>> with huge bandwidth that did drive symmetrically... just to be on the safe side.
>> No LTspice in those days, IBM PC was a new thing...
>
>IIRC Horowitz & Hill suggest this method, with another resistor in the
>collector to provide current limiting. I don't think I've ever actually
>used it, but series-termination is dead useful. One good thing is that
>it won't oscillate with an unterminated or short-circuited cable,
>because the transistor won't be in normal bias.
Good, yes, usually I do this:
http://panteltje.com/pub/mvp2.png
this gives 2 x gain, to compensate for 50% loss due to cable termination.
>(*) Yes, it's technically R/(1+beta), but up at frequencies where the 1
>matters, beta has a serious phase shift, so it's rarely needed.
>Designing accurate current mirrors is one of those rare places.
I have never done chip design, I did read some papers, but I think I am not ready for that ;-)