It was to isolate the mirror bypass from affecting the OpAmp loop compensation... fixed with emitter-driven mirrors (as I previously opined) and now reposted with that method included...
But the cheap way, but still not too shabby, is with simple mirrors...
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 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.
You never answer questions in an honest way; always coy, always affected. You must be seriously annoying in real life.
I know the bad things those 5K resistors will do. 2N3904 min beta at 1 mA is 70, less at lower current, and there is no max specified. In your circuit there's a double-whammy effect of beta on collector current. You are counting on Vbe matching, too, on a board with obvious point heat sources. Unless you have some very clever way out, it's you who don't know stuff. I suppose Master Circuit Designers assume that all transistors are exactly matched, because Spice says so.
-- John Larkin, President 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
ve the
, 0.8
e to
htly
will
the
to
tage.
actorily. It is more compact than a diff amp or current source with inducto= r, but there's no cost savings with DK's pricing. It is standard current so= urce level shift except it is voltage controlled by the 3.3V so the shift t= racks with power supply drift.
ckage.
stors.
rust
tie on
of
c-e
puts Pd deviations in the .1 mw or less range for 0.03oK or 0.06mV Vbe or = less, 0.03mV deviation, which makes current ratio errors +/-1% range.
ve problems anyway.
drive on that pHEMT then a tried and true expedient would be a wide-band R= F autotransformer to boost it up to 1.6Vpp.
eveloped a line for single ended to differential conversion for high speed = A/D converter inputs about 15 years ago, weren't real pricey AIRC.
nce the first PNP emitter is approximately GND, it is only the 3.3V to GND = potential that determines the current.=20
cheaper to make that way. Shouldn't the pairs be monolithic at least? That= 's all you really need.
This part looks good
But that's beside the point because you have so much drift in output common= mode.
[snip]
And you... You are always obfuscating. If you paid attention to the thread development you'd know I opined that emitter driven would solve the issue with OpAmp compensation, which update was posted prior to your beta comment.
And, as Bloggs pointed out, you have you head up your butt about heat.
Read my most recent posts before commenting further. ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 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.
the
satisfactorily. It is more compact than a diff amp or current source with inductor, but there's no cost savings with DK's pricing. It is standard current source level shift except it is voltage controlled by the 3.3V so the shift tracks with power supply drift.
package.
transistors.
on
puts Pd deviations in the .1 mw or less range for 0.03oK or 0.06mV Vbe or less,
0.03mV deviation, which makes current ratio errors +/-1% range.problems anyway.
drive on that pHEMT then a tried and true expedient would be a wide-band RF autotransformer to boost it up to 1.6Vpp.
developed a line for single ended to differential conversion for high speed A/D converter inputs about 15 years ago, weren't real pricey AIRC.
the first PNP emitter is approximately GND, it is only the 3.3V to GND potential that determines the current.
cheaper to make that way. Shouldn't the pairs be monolithic at least? That's all you really need.
mode.
VCM drift doesn't matter, _if_ you do it this way....
Not the multiple input VCM's (top three plots), but the output (bottom plot) stays put. ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 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.
I keep seeing the base-driven version.
Cute and too shabby. It has Vbe and beta and temperature errors all over the place. Not to mention needing *four* transistors. And the resistor value sensitivities apply to the full DC offset.
-- John Larkin, President 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
Refresh the page? Pages 3 & 4.
Instead of running your mouth, how about running the math? IF you can do it ;-) ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 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.
the
satisfactorily. It is more compact than a diff amp or current source with inductor, but there's no cost savings with DK's pricing. It is standard current source level shift except it is voltage controlled by the 3.3V so the shift tracks with power supply drift.
package.
transistors.
on
puts Pd deviations in the .1 mw or less range for 0.03oK or 0.06mV Vbe or less,
0.03mV deviation, which makes current ratio errors +/-1% range.problems anyway.
drive on that pHEMT then a tried and true expedient would be a wide-band RF autotransformer to boost it up to 1.6Vpp.
developed a line for single ended to differential conversion for high speed A/D converter inputs about 15 years ago, weren't real pricey AIRC.
the first PNP emitter is approximately GND, it is only the 3.3V to GND potential that determines the current.
cheaper to make that way. Shouldn't the pairs be monolithic at least? That's all you really need.
mode.
Several of the circuits posted here take out the Q:Qbar common-mode voltage.
And monolithic transistor arrays are expensive and usually sole-sourced. Simple 2-transistor mirrors have errors, too.
-- John Larkin, President 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
I keep seeing the old circuit. After clearing my cache.
Heat happens on high-speed boards. Everything is very close together, and I have one chip-scale package nearby that dissipates a watt, and several others dissipating fractions of a watt. There will be thermal gradients all over this board. Why have to deal with that?
-- John Larkin, President 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
How about something like an LM319 running from +3.3 and -5, 2k from ground to gate, 10k from gate to the 319's output, and a big cap from the ECL output to the gate? ISTM that all you need is for the cap to be big enough to avoid pattern-dependent bias shifts.
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 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Check out this stuff:
-- Regards, Joerg http://www.analogconsultants.com/
Min beta 70. That's a 1.4% loss on Q1 and another 1.4% on Q4 from just base current. Add in, say, 2% diffs on the 5K resistors and 2% for the
1Ks. Close to 7% worst-case. Apply that to 2 volts of offset to be corrected. That's blown my error budget, without even considering Vbe matching or temperature gradients.One would also have to consider the effects of hanging Q4's collector on the critical high-speed node.
-- John Larkin, President 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
satisfactorily. It is more compact than a diff amp or current source with inductor, but there's no cost savings with DK's pricing. It is standard current source level shift except it is voltage controlled by the 3.3V so the shift tracks with power supply drift.
puts Pd deviations in the .1 mw or less range for 0.03oK or 0.06mV Vbe or less,
0.03mV deviation, which makes current ratio errors +/-1% range.problems anyway.
drive on that pHEMT then a tried and true expedient would be a wide-band RF autotransformer to boost it up to 1.6Vpp.
developed a line for single ended to differential conversion for high speed A/D converter inputs about 15 years ago, weren't real pricey AIRC.
the first PNP emitter is approximately GND, it is only the 3.3V to GND potential that determines the current.
cheaper to make that way. Shouldn't the pairs be monolithic at least? That's all you really need.
mode.
Unfortunately that's not the only source of error.
Since you have a signal that indicates quiescent state (+0.1V?) why not just regulate all that out? The control loop would have to go open, capacitively stored, while a pulse happens. Then temperature drift, tolerances and whatnot just don't matter anymore.
-- Regards, Joerg http://www.analogconsultants.com/
Q4 should be a hotrod RF transistor, low capacitances. A 3904 is not so cool there.
-- Regards, Joerg http://www.analogconsultants.com/
CAUTION! CAUTION! CAUTION! HANDWAVING HAZARD AHEAD....
The resistors you use are that bad? (Buying Chinese shit to improve your bottom line ?:-) What 2 volts of offset? You don't understand how it works. Like I say, run the math, and lose the snarky mouth.
You can't do math... we ALL realized that when you posted your "filter" ;-)
What? You don't trust the simulation? That's transistor models there, NOT behavioral.
Sulk all afternoon, in the corner please, so as to minimize annoying civil people... Me? I'm going to a birthday party ;-) ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 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.
Gawd, Joerg! Why? The coupling cap knocks the socks off of it. ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 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.
That's similar to the PNP current-steering thing, where the fast path is a cap from Qbar to the gate, and the PNP pair is the slow DC-cleanup path. Both have the issue that the DC path has to accurately average the logic swings. The flipflop output could pulse for a few ns, or a few us, up to 200 MHz maybe, and could have almost any duty cycle. A 319 couldn't follow that, and a PNP pair probably couldn't very well either. The error budget is picoseconds.
About the only way to observe and accurately process the logic outputs is to passively lowpass filter first. My circuit works on the lowpass filtered differential signal, and others here work on the lowpass filtered common-mode voltage.
-- John Larkin, President 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
We use 0.05% Susumus when we have to. I'd rather not do that here.
The 2 volts that you are trying to drop between the PECL levels and the phemt gate. That has to be accurate, because any error becomes direct error in the gate voltages.
You don't understand
You didn't sim the PECL.
-- John Larkin, President 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
the
satisfactorily. It is more compact than a diff amp or current source with inductor, but there's no cost savings with DK's pricing. It is standard current source level shift except it is voltage controlled by the 3.3V so the shift tracks with power supply drift.
puts Pd deviations in the .1 mw or less range for 0.03oK or 0.06mV Vbe or less,
0.03mV deviation, which makes current ratio errors +/-1% range.problems anyway.
drive on that pHEMT then a tried and true expedient would be a wide-band RF autotransformer to boost it up to 1.6Vpp.
developed a line for single ended to differential conversion for high speed A/D converter inputs about 15 years ago, weren't real pricey AIRC.
the first PNP emitter is approximately GND, it is only the 3.3V to GND potential that determines the current.
cheaper to make that way. Shouldn't the pairs be monolithic at least? That's all you really need.
mode.
The rep-rate could be 200 MHz, and the duty cycle could be high. There's no good time to sneak in there.
A couple of schottkies could do a DC-clamp-restore thing on the gate, sort of like you suggest but fast. That might almost work.
-- John Larkin, President 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
What if the ECL driver at Q only wears size 8 socks? And it'll yank up the emitter resistor as well as its base. Can largely be avoided, for a few pennies more. I like the BFS17, around 0.5pF Ccb and 0.25pF Cce. That is around an order of magnitude lower than the MMBT3904. Best of all, it's still only around 5c a pop. So about a 3.5c "cost increase".
-- Regards, Joerg http://www.analogconsultants.com/
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.