On Wed, 21 Feb 2018 18:14:28 -0000, "Kevin Aylward"
<kevinRemovAT@kevinaylward.co.uk> wrote:
>"Jim Thompson" wrote in message
>news:p8bp8d96iul23if1s60ccp1vucht0atpgm@4ax.com...
>
>On Tue, 20 Feb 2018 21:23:38 -0000, "Kevin Aylward"
><kevinRemovAT@kevinaylward.co.uk> wrote:
>
>>>"Gerhard Hoffmann" wrote in message
>>>news:feuve1Fud5qU1@mid.individual.net...
>>
>[snip]
>>
>>>The idea of reducing 1/f noise by feedback
>>>is as old as 50 years, when some Russians discovered that a 33 Ohm
>>>emitter resistor can drop that by 20-30 dB. Was propagated by NIST then.
>>
>>>In general, local feedback does not reduce 1/f noise. The claims that say
>>>they do, are in general, false. The basic reason is that there is no
>>>feedback when the oscillator is limiting. You need to compare apples with
>>>apples. None of the papers I have seen that claimed feedback reduced noise
>>>actually did a proper A B test. Most papers are confused. I have run many
>>>tests on claimed designs where the actual feedback bit was engaged or not
>>>engaged with the circuit otherwise staying the same, and nothing changed
>>>noise wise.
>>
>>>My paper here illustrates one example where the authors, essentially,
>>>claimed it was feedback, but they were mistaken.
>>
>>>http://www.kevinaylward.co.uk/ee/phasenoise/FlickerNoiseNullification.xht
>>
>>
>
>
>>Careful, Kevin! Being right around here garners you enemies who will
>>proceed to make it their lifetime goal to bury you in vituperatives
>>;-)
>
>It's the twilight zone. I can understand that many might just accept what
>they are spoon fed from their Masters, but once obvious flaws are pointed
>out, why people defend the indefensible, is pretty stunning.
>
>In the case here, without even understanding the somewhat complex math of
>the A. Demir paper (on my site) , a basic understanding of electronics
>trivially refute HL.
>
>>Just look at the shit I have to endure just from pointing out what a
>>piece-a-crap LTspice is.
>
>>But I know we'll both hang in there... it's _such_ fun kicking idiot's
>>asses >:-}
>
>It will be a sad day for all of us, when you are no longer here to hung on
>in, Jim. Age is the final master of as all.
>
>-- Kevin Aylward
>http://www.anasoft.co.uk - SuperSpice
>http://www.kevinaylward.co.uk/ee/index.html
Yep. It keeps trying to get me, but I keeps fighting back... I'm not
done yet >:-}
I was just musing last night, whilst sipping Kim Crawford Sauvignon
Blanc and watching the figure skating competitions... I'm coming up on
the 48th year since I first started using Spice... so I see no problem
becoming expert in LTspice within a week or two ;-)
Then I can address bloviators with, "Ehhhh! Really?"
...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 |
It's what you learn, after you know it all, that counts.
Reply by Kevin Aylward●February 21, 20182018-02-21
"Jim Thompson" wrote in message
news:p8bp8d96iul23if1s60ccp1vucht0atpgm@4ax.com...
On Tue, 20 Feb 2018 21:23:38 -0000, "Kevin Aylward"
<kevinRemovAT@kevinaylward.co.uk> wrote:
>>"Gerhard Hoffmann" wrote in message
>>news:feuve1Fud5qU1@mid.individual.net...
>
[snip]
>
>>The idea of reducing 1/f noise by feedback
>>is as old as 50 years, when some Russians discovered that a 33 Ohm
>>emitter resistor can drop that by 20-30 dB. Was propagated by NIST then.
>
>>In general, local feedback does not reduce 1/f noise. The claims that say
>>they do, are in general, false. The basic reason is that there is no
>>feedback when the oscillator is limiting. You need to compare apples with
>>apples. None of the papers I have seen that claimed feedback reduced noise
>>actually did a proper A B test. Most papers are confused. I have run many
>>tests on claimed designs where the actual feedback bit was engaged or not
>>engaged with the circuit otherwise staying the same, and nothing changed
>>noise wise.
>
>>My paper here illustrates one example where the authors, essentially,
>>claimed it was feedback, but they were mistaken.
>
>>http://www.kevinaylward.co.uk/ee/phasenoise/FlickerNoiseNullification.xht
>
>
>Careful, Kevin! Being right around here garners you enemies who will
>proceed to make it their lifetime goal to bury you in vituperatives
>;-)
It's the twilight zone. I can understand that many might just accept what
they are spoon fed from their Masters, but once obvious flaws are pointed
out, why people defend the indefensible, is pretty stunning.
In the case here, without even understanding the somewhat complex math of
the A. Demir paper (on my site) , a basic understanding of electronics
trivially refute HL.
>Just look at the shit I have to endure just from pointing out what a
>piece-a-crap LTspice is.
>But I know we'll both hang in there... it's _such_ fun kicking idiot's
>asses >:-}
> But I know we'll both hang in there... it's _such_ fun kicking idiot's
> asses >:-}
>
> ...Jim Thompson
You must be pretty badly bruised up by now.
Reply by ●February 20, 20182018-02-20
On Wednesday, February 21, 2018 at 10:25:51 AM UTC+11, Jim Thompson wrote:
> On Tue, 20 Feb 2018 21:23:38 -0000, "Kevin Aylward"
> <kevinRemovAT@kevinaylward.co.uk> wrote:
>
> >>"Gerhard Hoffmann" wrote in message
> >>news:feuve1Fud5qU1@mid.individual.net...
> >
> [snip]
> >
> >>The idea of reducing 1/f noise by feedback
> >>is as old as 50 years, when some Russians discovered that a 33 Ohm
> >>emitter resistor can drop that by 20-30 dB. Was propagated by NIST then.
> >
> >In general, local feedback does not reduce 1/f noise. The claims that say
> >they do, are in general, false. The basic reason is that there is no
> >feedback when the oscillator is limiting. You need to compare apples with
> >apples. None of the papers I have seen that claimed feedback reduced noise
> >actually did a proper A B test. Most papers are confused. I have run many
> >tests on claimed designs where the actual feedback bit was engaged or not
> >engaged with the circuit otherwise staying the same, and nothing changed
> >noise wise.
> >
> >My paper here illustrates one example where the authors, essentially,
> >claimed it was feedback, but they were mistaken.
> >
> >http://www.kevinaylward.co.uk/ee/phasenoise/FlickerNoiseNullification.xht
> >
> >
> >-- Kevin Aylward
> >http://www.anasoft.co.uk - SuperSpice
> >http://www.kevinaylward.co.uk/ee/index.html
>
> Careful, Kevin! Being right around here garners you enemies who will
> proceed to make it their lifetime goal to bury you in vituperatives
> ;-)
>
> Just look at the shit I have to endure just from pointing out what a
> piece-a-crap LTspice is.
You have other faults, and most of the shit directed at you isn't prompted by your dislike of LTSpice.
> But I know we'll both hang in there... it's _such_ fun kicking idiot's
> asses >:-}
But in quite a few areas Jim presents himself as an ill-informed ass who whose backside needs kicking.
His uncritical acceptance of the nonsense that gets published on HotAir and TownHall is evidence of intellectual incompetence, and his passion for posting links to some of it on an electronics user group demonstrates regrettably poor judgement.
--
Bill Sloman, Sydney
Reply by Jim Thompson●February 20, 20182018-02-20
On Tue, 20 Feb 2018 21:23:38 -0000, "Kevin Aylward"
<kevinRemovAT@kevinaylward.co.uk> wrote:
>>"Gerhard Hoffmann" wrote in message
>>news:feuve1Fud5qU1@mid.individual.net...
>
[snip]
>
>>The idea of reducing 1/f noise by feedback
>>is as old as 50 years, when some Russians discovered that a 33 Ohm
>>emitter resistor can drop that by 20-30 dB. Was propagated by NIST then.
>
>In general, local feedback does not reduce 1/f noise. The claims that say
>they do, are in general, false. The basic reason is that there is no
>feedback when the oscillator is limiting. You need to compare apples with
>apples. None of the papers I have seen that claimed feedback reduced noise
>actually did a proper A B test. Most papers are confused. I have run many
>tests on claimed designs where the actual feedback bit was engaged or not
>engaged with the circuit otherwise staying the same, and nothing changed
>noise wise.
>
>My paper here illustrates one example where the authors, essentially,
>claimed it was feedback, but they were mistaken.
>
>http://www.kevinaylward.co.uk/ee/phasenoise/FlickerNoiseNullification.xht
>
>
>-- Kevin Aylward
>http://www.anasoft.co.uk - SuperSpice
>http://www.kevinaylward.co.uk/ee/index.html
Careful, Kevin! Being right around here garners you enemies who will
proceed to make it their lifetime goal to bury you in vituperatives
;-)
Just look at the shit I have to endure just from pointing out what a
piece-a-crap LTspice is.
But I know we'll both hang in there... it's _such_ fun kicking idiot's
asses >:-}
...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 |
It's what you learn, after you know it all, that counts.
Reply by Kevin Aylward●February 20, 20182018-02-20
>"Gerhard Hoffmann" wrote in message
>news:feuve1Fud5qU1@mid.individual.net...
>
>>> Oscillator phase noise is somewhat bizarre. It acts like FM capture
>>> effect.
>No.
Ho hum.....
>>> You can have a noise plot, find and eliminate a dominant noise noise,
>>> yet still have the "improved" topology have an essentially overlaid new
>>> phase plot. A new noise source just pops up as the dominant, and the
>>> noise don't change.
>
>> Oscillator phase noise is not bizarre at all.
>
> Yes it is, when compared to amplitude noise.
>
> Some of us have actually done 10,000s of phases simulations over the last
> 10 years...and produced product to boot...
>Yeah. Me. The simulations and the product, too. Does an oscillator
>and the complete analog part of a Dual Mixer Time Difference system
>to compare two atomic clocks, a hydrogen maser and a cesium, count?
>The flight sample had to work right after production, every change
>would have been an enormous paper work, and no solder joint to be
>heated more than 3 times. One change. Definitely. No place for
>adjusting the pulling range, and the crystals had been produced
>already.
Was that more than the one?
>
>> It is just that Spice
>> cannot handle nonlinear noise analysis. Harmonic balance analysis
>> is missing.
>
>> Spice calculates noise in the linearized circuit at the operating
>> point. But oscillators are inherently nonlinear since they need
>> a limiting mechanism or they would go super nova after some time.
>> (Or not oscillate at all.)
>
>> During each part of the oscillation cycle the OP is different,
>> gain is different, contributions are different.
>> So you need a harmonic balance simulator, as available in ADS
>> or in Microwave Office and the models for it.
>
> Granpas.. Suck eggs.....teach....don't...
>
> http://www.kevinaylward.co.uk/ee/phasenoise/phasenoise.html
>There is nothing that contradicts my words that Spice is unable
>to perform nonlinear noise analysis. There has been a group of
>people on the LTspice list who wanted to force it. That has
>turned pretty silent.
Sure, Spice can't do phase noise. What's you point?
>And with your venom against Hajimiri-Lee you are standing quite alone.
Err... No. I guess you missed the point that A. Demir mathematically
*proved* that HL was nonsense. The bit about "50 dB in error" was the nail
in the coffin. I showed some simpler spice results that proved by
simulation that HL is shit.
The reality is, if there is a non linear cap, HL fails. That means all
practical circuits.
Do you have an academic reference that refutes A. Demir to support that
your claim that he is wrong? This would be interesting indeed, as A. Demir's
method is the method used by Berkley Design Systems, bought by Mentor.
I guess you don't understand the difference between a theory that is an
approximation under known conditions, say Newton gravity compared to General
Relativity, and one that just happens to give the right answers by luck,
like Phlogiston does for fire.
A theory like a stopped watched being accurate at two time points per day,
is wrong. Period.
Name a commercial simulator that actually uses the HL technique...
> and you typically, need something like Shooting Method when the system is
> highly non-linear.
>Bla. Your small geometry MOS transistors, resonators with a Q
>like a wet sack of sand don't go together well with state of the
>art low noise oscillators.
I have no idea as to what you are referring to here. I design commercial,
precision xtal ocxos and tcxos asics with Qs of 100,000s. They have ppb
temperature stability.
hint:
www.rakon.com
>The idea of reducing 1/f noise by feedback
>is as old as 50 years, when some Russians discovered that a 33 Ohm
>emitter resistor can drop that by 20-30 dB. Was propagated by NIST then.
In general, local feedback does not reduce 1/f noise. The claims that say
they do, are in general, false. The basic reason is that there is no
feedback when the oscillator is limiting. You need to compare apples with
apples. None of the papers I have seen that claimed feedback reduced noise
actually did a proper A B test. Most papers are confused. I have run many
tests on claimed designs where the actual feedback bit was engaged or not
engaged with the circuit otherwise staying the same, and nothing changed
noise wise.
My paper here illustrates one example where the authors, essentially,
claimed it was feedback, but they were mistaken.
http://www.kevinaylward.co.uk/ee/phasenoise/FlickerNoiseNullification.xht
-- Kevin Aylward
http://www.anasoft.co.uk - SuperSpice
http://www.kevinaylward.co.uk/ee/index.html
Reply by Gerhard Hoffmann●February 18, 20182018-02-18
Am 17.02.2018 um 16:31 schrieb Kevin Aylward:
> "Gerhard Hoffmann" wrote in message
> Am 17.02.2018 um 09:21 schrieb Kevin Aylward:
>>>> I have a noise report in SS where I order the noise of each
>>>> component and
>>>> state its % to the total.
>>
>
>> In LTspice just plot onoise(R1) / gain to get it scaled to the input
>
>
> Missed the point.
>
> I said an *ordered* *list*, for example
No. I didn't miss the point. The few parts that
are suspects are worthy to be clicked upon. And
then you see them in the right order when you
scan the plot from top to bottom. And also
who is the worst offender, at just one sight and at
each & every frequency.
Not a table for each frequency that really nobody
with a working brain wants to scan line by line.
> 32 qn1_rc 4.05183p 218.248p %
> 33 qn3_rc 3.51699p 164.435p %
>
> Total Output Noise 1.93939u
>
>
>>> Oscillator phase noise is somewhat bizarre. It acts like FM capture
>>> effect.
No.
>>> You can have a noise plot, find and eliminate a dominant
>>> noise noise, yet still have the "improved" topology have an
>>> essentially overlaid new phase plot. A new noise source just pops up
>>> as the dominant, and the noise don't change.
>
>> Oscillator phase noise is not bizarre at all.
>
> Yes it is, when compared to amplitude noise.
>
> Some of us have actually done 10,000s of phases simulations over the
> last 10 years...and produced product to boot...
Yeah. Me. The simulations and the product, too. Does an oscillator
and the complete analog part of a Dual Mixer Time Difference system
to compare two atomic clocks, a hydrogen maser and a cesium, count?
The flight sample had to work right after production, every change
would have been an enormous paper work, and no solder joint to be
heated more than 3 times. One change. Definitely. No place for
adjusting the pulling range, and the crystals had been produced
already.
The whole system does what JL's time distribution system does, only
that one of the partners goes around the earth every 2 hours.
Navigation on a mm scale.
It is completely integrated by now, waiting for SpaceX to bring it to
the space station.
>
>> It is just that Spice
>> cannot handle nonlinear noise analysis. Harmonic balance analysis
>> is missing.
>
>> Spice calculates noise in the linearized circuit at the operating
>> point. But oscillators are inherently nonlinear since they need
>> a limiting mechanism or they would go super nova after some time.
>> (Or not oscillate at all.)
>
>> During each part of the oscillation cycle the OP is different,
>> gain is different, contributions are different.
>> So you need a harmonic balance simulator, as available in ADS
>> or in Microwave Office and the models for it.
>
> Granpas.. Suck eggs.....teach....don't...
>
> http://www.kevinaylward.co.uk/ee/phasenoise/phasenoise.html
There is nothing that contradicts my words that Spice is unable
to perform nonlinear noise analysis. There has been a group of
people on the LTspice list who wanted to force it. That has
turned pretty silent.
And with your venom against Hajimiri-Lee you are standing quite alone.
There are people who built on it with further developments.
Rohde comes to mind.
> and you typically, need something like Shooting Method when the system
> is highly non-linear.
Bla. Your small geometry MOS transistors, resonators with a Q
like a wet sack of sand don't go together well with state of the
art low noise oscillators. The idea of reducing 1/f noise by feedback
is as old as 50 years, when some Russians discovered that a 33 Ohm
emitter resistor can drop that by 20-30 dB. Was propagated by NIST then.
Now that Tina has made a belly landing at TI, isn't that THE opportunity
for SS to fill the void? I mean, you seem to have connections to TI?
Cheers, Gerhard
(Please bear with me, I have to think in a foreign language.)
Reply by Steve Wilson●February 18, 20182018-02-18
Gerhard Hoffmann <gerhard@hoffmann-hochfrequenz.de> wrote:
> Am 17.02.2018 um 16:56 schrieb Phil Hobbs:
>> Good oscillators generally have ALC, though, so the self-limiting
>> behaviour isn't such an issue with simulations (at least outside the ALC
>> bandwidth).
> No, not at all generally. It is easy to spoil a lot and the advantage
> is small in reality.
> But I do have a design that works well.
> cheers, Gerhard
Any chance to post the schematic?
I can show you how to do instant start in LTspice so you don't have to wait
for eons for the oscillator to start up and stabilize. It is independent of
the oscillator Q, and allows you to inspect the oscillator waveforms in the
first few cycles instead of after milliseconds of startup. It saves you
much time and allows a much finer analysis of the oscillator waveforms than
you can achieve after milliseconds of startup and settling. It is very
simple.
It works in LTspice IV and XVII, but you may have to change the INI file to
get it to run properly. There is no damage to the file and you can swap
back to the original whenever you want.
If you can't release the schematic, post a different one that you like, and
I will show you how to apply the technique so you can use it on any high Q
oscillator.
Reply by mixed nuts●February 18, 20182018-02-18
On 2/18/2018 1:15 PM, Phil Hobbs wrote:
> On 02/18/2018 12:10 AM, mixed nuts wrote:
>> On 2/11/2018 5:33 PM, Phil Hobbs wrote:
>
>>> I'd be very interested to hear about yours and others' fave
>>> chopamps--there's a depressing amount of specsmanship going on in
>>> their datasheets.
>>
>> ADA4528: noise peak at 200 kHz but otherwise well behaved. 2.2-5.5V
>> supply.
>>
>> I'm also using an LT6655-2.5 as a (DAC) reference - seems to have the
>> similar no 1/f with a noise peak characteristic. 6-8 uF on the output
>> flattens the noise nicely.
>
> Looks like an interesting reference. How low does that "no 1/f"
> behaviour go?
Don't know for certain. I'm not tooled to measure it directly. It's
all part of a control loop to stabilize a magnetic field to ~ 1 ppm (NMR
magnetometer). Temperature is the primary driver but our oven wanders
around about 0.1C which is also resolution of the system's temperature
measurements. We can see doors opening and A/C cycling when the system
is in the oven. The measured temps correlate with the correction
currents within a ppm or two over hours so there's no evidence that the
reference is worse than that.
--
Grizzly H.
Reply by Gerhard Hoffmann●February 18, 20182018-02-18
Am 17.02.2018 um 16:56 schrieb Phil Hobbs:
>
> Good oscillators generally have ALC, though, so the self-limiting
> behaviour isn't such an issue with simulations (at least outside the ALC
> bandwidth).
No, not at all generally. It is easy to spoil a lot and the advantage
is small in reality.
But I do have a design that works well.
cheers, Gerhard