So, haven't posted for quite some while now. Several years in fact.
Anyway, I wrote a paper on the Sigma-Delta linear model. As it is at variance to every paper I have ever found on the web in these 3 universes, it would be interesting to get some comments on it.
Hey Kevin, welcome back. Perhaps Genome will turn up now.
Saved for later, I look forward to reading it, hopefully I will comprehend enough to be useful.
I spent a bit of hobby time playing with a voltage-ratio to duty-ratio converter circuit. As I considered adding more things to it I realised one day that it was just turning into a sigma-delta!
Hey Kevin, welcome back. Perhaps Genome will turn up now.
The simplified summary is that the comparator measures time, not voltage in a SD converter. Apparently all the mathematicians that write the conventional theory, don't know that a limiter can't have feedback applied to make its output different from its limit value. Its been a case of the "Emperors New Clothes". All papers repeat the same erroneous argument to arrive at what is known to be the answer they want.
I added a somewhat novel feedback distortion paper to clarify this
formatting link
Yes. Fascinating how independent ideas merge. I had a good read of the original patent on Fractional-N PLL by Wells in 1985. He basically invented what is now known as a Sigma-Delta process for reducing the noise. He states that you use a Pascal's triangle to get the numbers for changing divide codes, but the patent does not say how he got there. SD theory gets there by the binomial expansion of (1-Z)^n
I don't understand how people make d-s ADCs so accurate. We buy parts that are linear, and bipolar ratiometric on the unipolar reference, to a few PPM.
A delta-sigma dac has many more edges than PWM - up to Fclk/2 compared to Fclk/N
- and the number of edges varies with the input. If an adc were to use the common "block-diagram" +/- feedback voltage summing into the integrator, the edges would have to be femtosecond precise.
I'm guessing that they actually use some very fancy charge dispensing architecture for both the input and feedback paths into the integrator. Still, charge injection errors must be huge compared to the ADC accuracy. Anybody know how they actually work inside?
--
John Larkin 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
Hi Kevin, Welcome back! Looks like an interesting read. ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 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 imagine it's that the integrated charge injection is independent of duty cycle. The summing point can move up and down a bit, but it can't go very far, so the sum of all the nonlinear capacitance contributions is bounded by +-1 count's worth.
Just the thermal drifts in the nonlinear capacitance would be important at 24 bits!
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 USA
+1 845 480 2058
hobbs at electrooptical dot net
http://electrooptical.net
Yeah. And a charge dispenser will have to "empty out" to PPM emptiness every single shot. The data patterns can vary from alternate + and - shots, all the way to very long runs in one direction. Pretty impressive, to spend a few bucks on a chip that's better than a systems voltmeter that used to cost as much as a house.
Hey, not to change the subject (who, me?) but you can buy ST or FC multimode glass fiber patch cords up to 1000 um core diameter! That's like putting light through a fire hose.
--
John Larkin Highland Technology, Inc
jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com
Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro acquisition and simulation
Yes, but they're not very flexible, which makes them sort of vulnerable. Fibre bundles are good news where you need mechanical flexibility.
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 USA
+1 845 480 2058
hobbs at electrooptical dot net
http://electrooptical.net
In a certain sense, they might not be. They can have spurious tones at specific input voltages that are not included in the claimed DC accuracy.
Yes, for PWM at fixed frequency. Measuring time to 1/100,000 within a 1MHz clock is a tad difficult. If time is measured to one clock period. Count
100,000 clocks and your resolution/accuracy is 0.001%. The capacitor/integrator is a bit like the dual slope, things cancel out.
There is a basic misconception on higher order converters and oversampling. There is a standard claim that you can have a lower oversampling ratio ( 1 bit types) than the raw N.Fs number. This is based on the random noise bit analysis of the SD converter, idle tones are not included. In fact the paper:
?Understanding Sigma-Delta Modulation: The Solved and Unresolved Issues? Joshua D. Reiss, J. Audio Eng. Soc Vol 56 no. , 2008 January/February
makes statements that the presence of idle tones is not understood at all. In particular, that "...there is no theoretical explanation for a F=Vdc.Fs "experimental" relationship for idle tones." In fact, it's trivially obvious, and I was gob smacked on discovering that there was any misunderstanding in the industry on this issue.
Have a read of the standard tutorial papers, e.g. analog devices, maxim, crystal etc. There seems to be a universal non appreciation of the fundamental tau/T=Vi/Vref nature of the SD delta converter. For rational ratios, the loop will force a constant tau/T, at some constant frequency, but noise or pseudo sampling can spread this frequency. Ratios near 1 and 0 only have one solution for tau/T = Vi/Vref. For a non rational ratio, i.e. values in-between quantum values, T might vary all over place so spurious tones may well be hidden.
This can all be checked in SuperSpice :-)
I did discover that the spice accuracy has to be set quite tight, like retol=10u, other wise the system is chaotic. In practise, it seems that circuit noise and this chaotic behaviour masks a lot of potential idle tones, but it doesn't really defeat the principle of no free lunch. For example, it seems the dominant audio SD ADC use multibit to achive a lower OSR.
Oh... in my absence I also wrote some phase noise papers, equally controversial :-)
--
So Jim dies and goes to Heaven, and when Saint Peter lets him through
the Pearly Gates, Jim notices that it's uncomfortably hot inside.
"Wow!", Jim exclaims, "it's hot in here. Can't God air-condition the
place?"
"What, for the three of us?" replies Peter.
Naaaah! 73 is not old. If I hang around as long as my father, I've got 17 years left to go. ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 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.
Methinks "hell" is probably cooler than Arizona ;-)
(121°F a few weeks ago) ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 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.
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.