What you have is only part of the picture. You need to get the total attenuation with your 140uF and several ceramics bypassing the emitter to ground.
You might be able to play some tricks by increasing the input signal to the max, reducing the bandwidth as much as possible, and averaging. But it is doubtful the AM502 will let you go down far enough to see the resulting noise at the emitter.
Then you need to look at the result with the base filter connected to the collector to get the overall performance.
Your earlier statement
"The Tek AM502 and a digital scope, signal averaging, has a nice signal- recovery capability. This would easily resolve a 1 nV signal"
is probably not attainable.
That is doable. Multiple low-noise devices in parallel, maybe 4. Get the signal up high enough, then sum the outputs and continue from there.
Looks like the source impedance will have to be 15 ohms or less, so I'm not sure what you would measure with it.
I'll comment on these in a later post. John, what's the 33-ohm resistor for?
That's not correct. One can easily manipulate the BJT equations to show this is the case. In fact, in Early's original 1952 IEEE paper, "Effects of space-charge layer widening in junction transistors," he starts by evaluating small changes in collector current vs collector voltage, or a change in conductance if you wish, but immediately he proceeds to analyzing his new base-voltage feedback factor, mu_ec, to get a base-voltage change mu_ec * Vc. Here mu is independent of current, and describes the primary effect. mu has a value of 10^-4 in his measured transistors. mu is supposed to be independent of collector voltage, although Early observed it increasing at very low Vce, and he saw a small increase in mu with increasing collector current. (There's also a small change in rb' base-resistance effect that he spends a page evaluating, but dismisses.)
In his 1952 paper Early didn't use the term Early effect, of course, and as it happens he also didn't use the concept of Early voltage, which came later and is used by spice to model bipolar transistors. These are related V_A = V_T / mu_ec, where V_T = kt/q = 25mV, etc. That would mean his 1952 transistors had V_A = 250V, which must be pretty good for old low-voltage germanium parts.
Further commenting, you can measure the Early effect through change in current vs Vce, with constant base current (change in beta = Hfe), or as the change in collector current with constant base voltage, which is more Ebers-Moll friendly. They extrapolate back to nearly the same value for V_A on the Vce plot. The theory Early details is based on the base voltage rather than the current gain, however, and I suppose the Early-effect spice modeling must be the same.
That's correct, but as stated your other assertion is not. It's not very useful to simply model it as a resistance, because the the value r_o changes with collector current, whereas V_A does not. You can calculate r_o = VA/Ic, which works for any current.
To keep the NPN from RF oscillating. It may not be needed if the emitter were hard bypassed to ground. Being burned many times, I always add a base resistor to emitter followers, sort of like garlic and vampires.
In olden days, "Transistor Manuals" had a lot of transistor theory, and datasheets had a lot more data. Lately, the datasheets are skimpy and stuff like Early effect has to be measured experimentally. Spice is nice but I like to verify.
As you load the output the voltage drops. like the internal resistance of a power supply. I had 1k ohm feeding the base and this gave about 10 ohms of impedance (1 kohm divided by the current gain of the transistor.)
Amplify the piss out of it, Send it though a known filter, and then measure the RMS voltage with an analog multiplier. It's our latest physics 'toy'.
formatting link
If you click on the "Annotated Photographs" link there are some pics and words.
With a cap on the ADJ pin it isn't too bad but the BCX70K is certainly the component of choice here, nice low 1/f knee. I have used them in cap multipliers as well with laserdiodes. Then, the client increased the power requirement about 8-fold and we had to say good-bye to the BCX70K :-(
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
On a sunny day (Thu, 03 Jun 2010 06:53:27 -0700) it happened John Larkin wrote in :
Well, your minimum measured noise is about .1 mV I think LM317 with a decent cap on the output can match that. But I have not tried.
But from your other posting I see you need nano Volts, so curious what you measure there, the Vbe drop is hugely temp dependent too in your diagram. So, temp changes, if these happen fast, will wipe out your results.
It's the +14 power rail for the front end of a photodiode and the following jfet+opamp gain stages. It's not very sensitive to very low-frequency noise on the power rail, but I do want DC regulation and very little noise above, say, 1 KHz. My ultimate noise level at the jfet input should be around 1 nv/rthz, so the power rail noise shouldn't be much worse than, say, 10 nv/rthz. My ultimate power supply is 15 volts from a switchmode wall-wart, figure a couple of hundred millivolts of crud maybe. So I need maybe 1e7:1 noise rejection from the 15 volts to this rail.
The c-multiplier doesn't regulate of course, so its lf rejection primarily depends on the base RCRC thing. At intermediate frequencies, the Early feedthrough dominates, -40 to -60 dB depending on Vce. Once the output lowpass kicks in, at around 1 KHz maybe, we have maybe 2 ohms of Re driving the output caps and their ESR.
This is not a simple circuit! There are maybe four pole-zero things in the graph of dB rejection vs frequency.
I'm actually using a different circuit for the +14, and Vbe multipliers on other supplies that don't need tight regulation. Some of this is new to me so I'm still learning; there are noise sneak paths everywhere you look.
On a sunny day (Thu, 03 Jun 2010 11:12:39 -0700) it happened John Larkin wrote in :
Bought a very nice Chinese switch mode, 12 V 3A, 15 Euro. This looks good enough to supply with professional equipment: ftp://panteltje.com/pub/chinese_12V_3A_switcher_top_img_1985.jpg ftp://panteltje.com/pub/chinese_12V_3A_switcher_bottom_img_1990.jpg
Note the Euro mains connector, nice! this one is for LED strips, 60 RGB LEDs per meter, testing now. And the other miracle is that the DC connector fitted my other devices...
Actually, to reduce the number or 'wallwarts' I made a multi way cable for the DC side... Should patent that, it frees up mains sockets :-)
I dunno yet if this one is short circuit or overload proof, will find out one day.
Darlingtons are noisy because the driver stage runs at really low emitter current, and hence has a lot of shot noise. You get rid of that with a moderately big bypass cap on the output, which rolls it off at frequencies you care about. Going from 4 nV to 1 nV is the easy part.
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
Battery operation is a crutch. That's what good layout and cap multipliers are for! Twenty years ago, everybody thought you had to have linear regulators for sensitive analogue stuff. Not any more.
Nonsense. A two-phase synchronous detector is exactly equivalent to a narrow bandpass filter followed by a frequency mixer. The envelope is preserved.
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
I'm not so sure. Looks like you took input vs output measurements. It may be more informative in terms of characterizing the filter circuit performance to take unfiltered output versus filtered output measurements, at the load. function gen====>load function gen====>circuit===>load
I'm using a 15-volt wart that we stock, and going to a 24-volt one would push up the dissipation on other regulators. So I want a super-low-noise, low-power LDO.
This is what I did:
ftp://jjlarkin.lmi.net/P14_reg.gif
which is maybe not too bad. The opamp has PSRR rejection better than the Early feedthrough of an NPN, and the 15 ohm resistor makes a lower corner with the output caps than would 2 ohms of Re. And it regulates and current limits.
A 2-stage c-mult might be better in other situations, or maybe an LDO followed by one c-mult.
This data looks very suspect to me. You observe the emitter follower failing its job when it still has Vce = 0.75 volts, which makes no sense to me, unless perhaps your Vac voltage is really much higher than you thought it was.
I ran a spice model of the above circuit. Intusoft's BCX70 library has the BCX70's Early voltage VAF = 120 volts, which seemed a bit conservative to me, but OK. We calculate mu = VT/VA = 2.1 x 10^-4. For 200mV on the collector, that's 42uV of Early effect at the base, and arguably the same at the emitter. Nonetheless, the spice engine reports 53 to 56uV, which I'm happy to accept. But it's much less than you report. The spice model also has the BJT continuing to function well until Vce is under 200mV, which it should do.
I'll say this, narrow base-width high-beta transistors have a much worse VA than ordinary parts. I see you reported beta = 670, maybe that's an issue here. My spice part was a BCX70G, with BF = 234. NXP's BCX70 datasheet says the 70G parts have beta from 120 to 220, and K parts from 380 to 630. My spice library has a BCX70J with BF = 400 and VAF = 80 volts. Maybe that's what's going on here. Super-beta parts need not apply.
My eyes would burn, trying to work with that colour set ;) Especially the black background, the stuff would be floating, drifting in space for me after a while.
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.