I don't know of any books, I read some of his papers (written with Purcell) on NMR. Very nice stuff, much better than F. Bloch's IMHO. But that is mostly likely was due to Purcell.
He (like me) may not have been a very good writer.
George H.
On a sunny day (Mon, 08 May 2017 15:25:24 -0700) it happened John Larkin wrote in :
I usually think in voltages, say a divider, 2.5V across lower resistor, 1V across top. That means 2k5 and 1k Or 5k and 2k or 500 versus 200 Ohm. etc, depending on what loads it, smallest one is most important. No software needed. I have boxes full of SMDs, I know the values by now... You often see those volt calculations in the corner of my great artistic circuit diagrams ;-) But I have only a few resistors with wires that I use for testing over and over again in a box on the table.. When one value of SMD runs out I switch to an other...
We must have a thousand resistors in stock, from 5% to 0.05%, various case sizes and tempcos. If I want some ratio, with some precision, I could spend hours with the stock listing and a calculator. It's more fun to write a program.
Some of our precision resistors are not standard E96 values, like 2.2K
0.1% for example. Just having a few oddballs in stock opens up a huge range of possible ratios that are impossible in E96.-- John Larkin Highland Technology, Inc lunatic fringe electronics
On a sunny day (Mon, 8 May 2017 21:15:52 +0100) it happened "Kevin Aylward" wrote in :
That is really the Luddite approach.
It is a truly great approach. It would be impossible to have what we have today without such an approach.
Of course it is. I already explained. Its a fact. Period. Every single modern IC you buy today has been designed in the virtual world.
I agree many mathematicians and applied mathematicians that refer to themselves as physicists, have lost the plot.
-- Kevin Aylward
I did said flatband. For oscillators in the 10MHz to 50MHz range, this is of the order of 100Khz+ offset. This is range where xtals work best.
Its always 1Hz BW
PN is always specified relative to carrier in dBc. The actual power/voltage is a separate spec.
If noise in dB is written, it means dBc.
Note we are talking about phase noise, not amplitude or additive noise. Oscillators can get very good flat band phase noise, but the majority of systems require a square wave. The limiter always dominates flatband noise.
Close in PN is highly dependant on the c1 of the xtal. That is the same design will give vastly different results for a fundamental with a series c1 of 5ff verses a 3rd overtone with 0.3ff.
A 10MHz fundamental design, in a small package, low current, might be say
100dBc at 10Hz offset. 3rd OT, -120 dBc.My pn papers are at :
-- Kevin Aylward
Oh dear..., you do realise that that was tongue in cheek comment?
There are systematic ways to design by simulation by varying components.
Actually no. Monte Carlo is useful for severeal things, but in general, its not a main driver. Most ic blocks do not require MC at all. The most important design process is worst case WC
and multivariable parameter sweeps.
As an example. Suppose its an LDO. One needs to theoretically know where to put compensation networks, and an understanding of how the poles and zeros should be placed.
Say, there are 3 comp caps and two comp resistors, and the load cap is to be
10n to 10uf at load currents from zero to 1A.One sets up parameter sweeps to vary these components in a systematic way whilst outputting at loop gain and phase margins. This is all done all over again using transient runs. This is done while also examining the results over process corners and temperature.
That is for stability. Other issues are PSRR. This requires technical knowledge of what determines PSRR. Its usually a rock and a hard place. Getting the right device sizes and operating voltages, so devices don't crush, are not too large, etc...
Typically it takes 10,000s of runs to converge to an acceptable solution. There is still an art in it in choosing the best runs though. Grads and the majority that gain 1 year of experience 20 times over, typically take 3 months for such a task, those like me and Jim T, might take a day or so, depending on spec.
The route explanation of all of this, is random. Once one really thinks about the problem, there is only one rational solution. Essentially, an infinite number of universes each with different laws of physics, we just happen to be in one where the laws allow us to exist.
-- Kevin Aylward
[snip] [snip]
My typical TOTAL design time for a relatively complex Analog/Mixed-Signal chip is only 3-4 months, including interim and final design reviews... all by my lonesome, except for my layout associate.
Some un-named "teams" take longer than that cutting and pasting from existing IP and text-book "solutions"... AND have a high rate of "re-spins" >:-} ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| STV, Queen Creek, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Thinking outside the box... producing elegant solutions.
"It is not in doing what you like, but in liking what you do that
is the secret of happiness." -James Barrie
On a sunny day (Thu, 11 May 2017 19:05:06 +0100) it happened "Kevin Aylward" wrote in :
Na, I could follow you for most parts up to here. But the infinite universes thingy does not jive with me. Apart from the fact that if one says 'Universe' everybody has his / her own ideas... I do not think the laws of physics are that different.. If you define 'universe' as that what we currently can see (detect), yes then there can be many big bangs, _could_ have happened, just like there are many stars etc. But in each of those similar matter and laws as there are in each star we observe. Some basic constants may be very different in those 'bangs', or maybe even have been very different in our own bang (think speed of light for example). There are several theories, like for example tired light, to explain the redshift, but maybe the medium changed, that what the EM waves propagate in. maybe u0, there are plenty of unsolved problems, dark force, dark matter, what not. but that takes us to physics. Back to the MOSFETs :-)
[snip] [snip]
It is interesting that most hirers simply don't understand that there are those that are 10+ times as productive as others.
I would say I average out at 100 blocks, 50 transistors per block, per year. I would say most do about 4-5, if that, and those still have problems and re-spins.
Some blocks might take me an hour or so.
Yeah... Have a look at my linked in profile :-)
-- Kevin Aylward
Such huge offsets are of interest mainly with half-duplex repeaters. Commercial half-duplex repeater stations usually work with several MHz duplex intervals. However amateur radio repeaters usually work with only 100 kHz @29 MHz and 600 kHz @51 and 145 MHz bands.
When a common Rx/Tx antenna is used, of course the receiver needs a notch (several cavity resonators) tuned to the Tx frequency, but in practice, you also need one or more cavity resonators in the Tx chain to notch out the phase noise from the Tx falling on to the Rx passband. The aim should be that the Tx phase noise leaking into the receiver should be less than the receiver thermal noise and the band noise (which is huge at 29 and 51 MHz).
For this kind of applications, your figures are impressive, making it possible to drop one or perhaps all cavity Rx notch filters from the design.
However, for general receiver design, the local oscillator (LO) phase noise has an other nasty feature. If there is a very strong (say 1 MW) transmitter station close to the desired weak station, the phase noise from the LO and the strong unwanted signal will form _reverse_ mixing products falling into the pass band of the receiver. Thus strong signals at the nearby channel can be quite harmful. for LF/MF/HF signals the nearby signal can be 5/9/10 kHz, for aviation 8.33/25 kHz and for general FM service 12.5 or 25 kHz away.
For this reason, the phase noise at 5-25 kHz offset is of great interest.
So did I assume, but you never know, which tricks specification writers use these days :-)
For practical purposes, the interesting offset is when the oscillator phase noise is drowned by the thermal noise and for LF/MF/HF/VHF receivers, when it drops below the band noise.
For this reason, you really need significant power levels in order to stay away from the thermal noise floor. For low power designs, the maximum dBc figures are simply limited by the thermal noise.
I assume that by flatband noise you are referring to the thermal noise floor ?
Oh, they understand that. They have no idea how to deal with it though.
You would be incorrect.
My arguments are here:
-- Kevin Aylward
On a sunny day (Fri, 12 May 2017 13:52:51 +0100) it happened "Kevin Aylward" wrote in :
Introduction, time Time is simply the observation that objects have changed their position relatively to other objects. agreed. Something From Nothing
I think you are twisting reality here a bit (or a lot). What we call 'law of physics' is simply our observations reduced to a mathematical model. Indeed it requires, 'us' and the 'world around us' to exist.
A little sidestep 'gedanken experiment' something from nothing is strongly related to QM, randomness, probabilities (electron is here or there but we know not... cannot measure both ... blah blah) I do not like QM, it is like epicycles. There are other theories, I am with De Broglie's 'Pilot Wave Theory',
Fundamental Axioms Of Physics
If you assume the space is empty, then nothing will randomly manifest. Einstein did away with a medium for EM (light) and put whatever there was under the carpet making it look clean, same for his mathematical model for gravity. OF COURSE there is something that allows all those things. 'Virtual particles popping in and out of existence' is often quoted but simply not good enough.
Big And Micro Bangs
Well first I must admit that I am not even sure there was a big bang, that theory has caused many observational problems, and still does so to this day. The previous Pope told Hawkins 'it would be good if there was a bang, as that would explain everything'. I have no idea what he meant by 'everything', and he stepped down anyways... :-)
Beware of singularities and other mathematical traps.
Let's take a simple example. Some guy long ago found, by observation, I = U / R. Not knowing about electrons he could easily assume that _infinite_ current would flow in a zero resistance object, he would also state that in a vacuum nothing could flow. Things changed a bit when teh first vacuum diode was demonstrated... when it was found that current was granular, and the equation breaks down at one electron at the time. MANY mathematicians like to go for the incomplete formulas, and they LOVE the infinities that come with it. Make press, people say: 'Oh it must be right, the math proves it.' And an other wormhole is created. Epicycles.
I will comment on the rest below later, after reading.
On a sunny day (Fri, 12 May 2017 13:52:51 +0100) it happened "Kevin Aylward" wrote in :
From a POV of whatsit the random movements of signals in the brain, in the human being, sometimes referred to as 'humming bean' try to make a model of the reality it observes.
Consciousness in that sense is just one part of it measuring(..) the other few neurons activity, and it has some output that emits: theory, opinion, anything.
It is easy to write a computer program that is conscious, the simplest thing I always mention is a sun-screen with photocell, it opens in the dark and closes in light, a simple control loop, just like our own eye pupil.
Add 10 lines or less code to measure the control signal and say 'it is dark it is light', and there you are, it is conscious.
A doctor will check the eye reflex to see if you are conscious. I do not see any mystery there at all.
But it makes great stuff for those who know nothing about tronix, like whatshisname 'I think so I am', How profound..
So to the 'opinions' 'theories' 'world view' etc etc many possibilities exist. In a (computer simulation) neural net instability can also exist, and the endless motion can be if a larger net, as a _hopefully_ converging activity to a balance, a weights and connections configuration that we then can call our 'current understanding' That is due to change as the net (we) are exposed to continuous inputs. From that POV our theories are worth just exactly that much. We need the models we create in our brain to exist in this world. From catching a ball to driving a car or designing some tools, or using those (evolution had it that way). Purely from that point we need no math. Math is designed by us as a side product that can help us, but just like we invented knives it can hurt us too, lead us astray. Always and always we should verify our theories with observation.
String theory does not allow us to do that, so until it can it is just a useless pattern, say noise, in the neural fireworks. No better than saying Elvis is alive and anything can be explained by that, just assign enough power to him.
:-)
Think I had nuf of this for 2daaai ;-)
High speed, ADCs, care about the whole whole flatband range to minimise jitter.
There is a standard telecom spec of 12kHz to 20 Mhz integration for jitter...
Yes. The whole electronics industry is full of lies er.. creative specmanship.
Power is the main concern. Many would like 1 ma for a whole oscillator system. A decent limiter needs several ma on its own.
If there are bipolars in the design, it may well be that shot noise is a main contributor.
Its the the bit where the noise stops rolling off with frequency and becomes flat with frequency.
-- Kevin Aylward
I was responding to Kevin. I agree with most of what he says, but I don't think you can claim that analogue asic's are more (or less) challenging that digital ones.
(Although for complex devices they are no longer purely analogue or digital, and it looks like the relative analogue silicon area as a percentage of the die of a majority of complex devices is continuing to increase.)
I can see problems ahead since the current analogue design tools are beginning to struggle. Analogue artist is not exactly a quantum leap from Harris's SLICE of 30 odd years ago.
CoolAudio is making THAT2181 quad VCAs again for a buck in quantity, get 'em while they're hot
Would you care to prove your assertion by examination ?>:-}
In today's world, digital _chip_ design is virtually automated, once the logical functionality is established.
Unless, of course, you are talking about designing the gates at the device level (then it's analog ;-) I did that once for Motorola/ON-Semi so they could move their entire 74HCxxx line from in-house processing to Chartered Semi's line.
The tool is _not_ the designer. I sketch out most of my designs on paper before entering them into a simulator. ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| STV, Queen Creek, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Thinking outside the box... producing elegant solutions.
"It is not in doing what you like, but in liking what you do that
is the secret of happiness." -James Barrie
Oh...?
A modern digital asic design is ALL done in, effectively, in software. People just write VHDL or Verilog Code. The tools take the code and generate the layout with a button press.
Software is piss easy :-)
In Cadence mixed mode, ones sets up a view to the code and runs the code in with the hard part, i.e. the analog
-- Kevin Aylward
Spoken like a real ignoramus. That's why there are hundreds of people on the design teams.
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.