Gee, Fred, I prefer the mindless approach. SPICE is working properly? Then hit the button. Bang. 3db down at 1GHz. Goodenough Yes/No? OK build it and let's go home. Oh - you mean we have to measure it? That's a different story. 3:00am Equipment is fired up and working Yes/No? What do you mean we have a bad connector? Intermittent. +/- 60dB. Bad news.
Do what every red-blooded engineer in his right mind would do. Tear the @#$%!~^& instrument apart, fix the intermittent, measure the product, ship it if it's good, or fix it if it's bad. Then go for a cup of coffee.
Home? Wife? Children? What are these strange terms you speak of?
And it is for this reason I am hoping and anxiously waiting for any possibility that you might consider discussing your low power 2kV MOSFET amplifier:)
I believe others have also expressed considerable interest in these things:))
This circuit features a high-voltage infinite-input-impedance buffer followed by a low-voltage low-error attenuator for the A/D signal.
Note that the critical low-voltage opamps A1 and A3 are operated with no common-mode voltage, hence their CMRR isn't important. The high- voltage opamp A2 is inside A1's feedback loop, so A2's CMRR and other specs aren't important either. A2 does have to operate with 60V total supply, but in exchange, it allows you to have a Terra-ohm Zin, common for a +/-2V on a multimeter, etc., on 20V or even 200V fs input ranges. If A2 is one of my +/-1100V amplifiers, then you can go to +/-1100V with infinite Zin. This is the basic approach used in Electrometers.
The 5V floating supplies are often made using separate AC transformer secondaries, but you can do it with current sources and active zeners, powered from the higher-voltage output-amp supply.
Ahhh- good example of a very common misperception and confusion.
In the case of using dB to characterize *amplifier* performance, there is an implied unit of normalization, and that is the amplifier performance at "midband". You are taking Log ratios of gains, and not voltages or watts. The dB is always computed as
20*Log(Gain/Gain,midband) regardless of whether the gain is a voltage or power gain. To see this, note that a power gain, Gp, is Gp=Pout/Pin=(Vout^2/Ro)/(Vin^2/Rin), and therefore Pg/Pg,mb= (Vout/Vout,mb)^2=((Vout/Vin)/(Vout,mb/Vin))^2=(Gv/Gv,mb)^2, so that
20*Log(Gp/Gp,mb)=2*20*Log(Gv/Gv,mb) or Gp,dBmb=2*Gv,dbmb. In words, the power gain loss relative to power gain at midband is 2x the voltage gain loss relative to voltage gain at midband in dB. Then a plot of amplifier voltage gain normalized to midband has its vertical dB scale multiplied by 2 to show power gain rolloff with frequency. When the Gv is down 3dB, the Gp is down 6dB, when Gv is down 6dB, the Gp is down 12dB, etc. This should show you that just because you are talking about power does not mean you mindlessly apply 10*Log(.).
You're just experiencing the Adobe upgrade technique. New PDF's made with Acrobat 5,6,7 can't be read by Reader 3 :-(
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
I learned all the basic math skills at MIT, but all the nuts and bolts from hanging around my Dad's radio and TV repair shop... also from being a technician in the Woodson/Jackson/Melcher MHD lab in MIT Building 20, designing and constructing the test gear for the PhD's.
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
If the manu's are challenged to produce a "perfect buffer" internally, what hope do you have to do it externally. LTC has the "East Drive" technology, LTC2484:
Regarding the link you provided, either the group didn't meet long enough, or perhaps they met for too long. I am not impressed with their work. Included in what the author wrote was: "A review of common decibel forms and their proper usage included dBm, dBu, dBV, dB-SPL, dB-A and others. Most people still can't use or write many of these correctly, and they have sometimes changed over the years."
He comments on people not writing the forms correctly, and then mucks-up dBv, dB SPL, and dBa.
In the "handy things to remember," they failed to remember dBv is referenced to 1 Volt *peak-to-peak*. And that's just from a quick peek.
Because it does not appear to you that the decibel has been formally redefined since the good old days, I suggest you look a lot longer before you return. Go buy a IEEE dictionary. I wish I still had mine, but I don't usually give a rip nowadays.
LMC6082 and LMC6062 both give bad CMap on my poor Win 3 Acrobat. Unreadable. Blank screens. So I have to fire up the Win 95 machine and transfer the files across. Then try to find them in Windows. (Sigh)
The good news is I made the desktop look exactly the same as Win 3.11. So at least it looks reasonable.
I did get a fleeting glimpse of 10fA input bias current, so it might be promising. Anyone happen to know the cmmr offhand while I am pushing tons of stuff out of the way to get the the Win95 machine?
Apparently someone is throwing a bunch of light switches in China. You have of course heard of the Unocal bid? A drop in the bucket. Things are going to get a lot more expensive very soon. See
"China on global hunt to quench its thirst for oil"
formatting link
Since we have little to look forward to except humungous oil prices, I am dealing with the future by replacing all the power hungry 19 inch rack mounted instruments with their updated micropower versions. A HP3456A running on clock batteries should sell like hotcakes. So don't bug me while I'm on a run:)
Room temperature. But learn to do 10fA first. Later you can worry about Sapphire insulators, Triax connectors and double-coax cables, Tribolectricity, carbon-filled coax, super-low nodal capacitance, Kelvin-connected guards, low-leakage, safety circuitry, etc., etc.
You are mistaken. You are confusing the possible *first* use of the dB with what it actually means after the fact.
The dB is a totally general term. It simply signifies that a log of a number was taken. Thats it. End of story.
The db is not even a real unit, that is, it is not a dimension. The term under the log must be a unitless. You cant take a function of something with dimensions. e.g.
I=Io.exp(qV/KT)
note, qV/KT is dimensionless.
db's are applied in many fields, the power use is no more than a historical incidental.
You are trying to claim that, for example, if it is stated that an amplifier has 25dB gain more than another, that such a statment is wrong. You will be pretty much alone on that idea.
Kevin Aylward snipped-for-privacy@anasoft.co.uk
formatting link
SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design.
Nothing wrong with "Peak envelope power", if we're talking about SSB transmitters. Rather fundamental, in fact.
--
"Electricity is of two kinds, positive and negative. The difference
is, I presume, that one comes a little more expensive, but is more
durable; the other is a cheaper thing, but the moths get into it."
(Stephen Leacock)
I send you 300dB of Thanks:) I currently have a blinding mold headache, so I can barely read your post. But it has what I am looking for. This is really, really important to me. I need to divorce myself from commercial equipment that becomes obsolete and cannot be fixed when it fails.
Your +/- 1100V with infinite Zin is pure magic. Do you intend on posting details for A2?
BTW, I don't know if I should tell you this or not. I am working on a very modest constant current supply. 330V with variable current from 100uA to 6mA. Trivial, except I want to control the current by means of a voltage referenced to ground.
So I have this simple TLO74 op amp driving a NPN which drives a PNP current mirror which drives the load which supplies current to the load resistor.
The current mirror is current limited so when the feedback is broken, it only supplies ~8mA so nothing gets electrocuted or welded. Still no problem.
The interesting part is I followed your posts on MOSFET models in SPICE, and read how they fail at subthreshold values.
I happen to have a bunch of power MOSFETS, and I had tried to use them in the SPICE model. They always failed. After your threads, I figured it was because of the subthreshold problem.
But you encouraged me to give it another try. And guess what? I found the simple method of measuring the open loop gain desribed by John Larkin and Mike Englebert works great. All you do is insert a voltage source in series with the op amp feedback pin, and measure the gain on both sides of the voltage source. I think I got that right but will check when the headache dissipates.
And do you know what? The reason my circuit failed with the MOSFET was because I screwed up the open loop gain. I had included 1nf bypass caps on the input to the op amp to eliminate rfi from a local taxicab transmitter. Their antenna is about 10 feet from my workbench, and everything folds when they key their mike.
My rfi solution put terrible wrinkles in the feedback. This was impossible to see looking at the closed loop response, but it stood out like a sore thumb when you plot the open loop response. Then it was easy to fix.
So my grateful thanks to you for opening up the subject, and to John and Mike for describing a simple way to make the measurement.
TO be fair, so do those BSEEs going into industry. My boss liked the UIUC types because we had practical experience. Theory is goodness but many, even from MIT, did try screwing fuses up their ass. ...then scolded others for being so stupid as not to turn them left. That's not the way to get a produuct out the door.
Mike, in normal circumstances my tolerance knows almost no bounds ! I am
*phenomenally* tolerant. Ppl who know me well comment on it ! When it comes to plain lies however I have *zero* tolerance.
Rubbish ! Gain means a ratio > 1. Attenuation means a ratio < 1. These are
*explicit terms*. Note that I have also mentioned ratios and ratiometric measurements in my discussion of this topic and intentionally avoided any reference to unity.
Oh please ...... I accept you're being apologetic for Tony but is it necessary to stoop that low ?
Indeed, he helped contribute to my call for assistance re the 3525. Thanks, Tony.
Being well considered and an able person does not absolve him from criticism when he makes totally fatuous comments though. Indeed, as a respected member it's incumbent upon him to be accurate in the advice he gives.
Ungenerous ? Maybe.
Was he talking crap ? Yes !
For those who haven't got it yet - pls consider the VIR 'triangle' that defines V=IR, I=V/R and R=V/I.
To talk about power dB ratios being *pure* and voltage dB ratios somehow
*unpure* is bonkers. It smells like some perverse electronics fundamentalism or whatever.
If P=V.I , then by inspection it's equally I^2.R or V^2/R. None can exist in isolation.
I'm staggered that anyone can even fail to realise this and see that you simply adjust the relevant equation according to take account of any square power terms.
Note that the dB ( indeed the Bel actually ) is just a mathematical convenience that's great for dealing with large ratios in particular.
And as for whoever said someone wrote a book 71 yrs ago that said dBs only applied to power..... It was wrong then and it's still wrong ! I'll write a book refuting it if you like - nerr nerr !
Mary Mother of Jesus ! Leaving the Twilight Zone now......
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.