Levels of abstraction

Well differential nonlinearity is a fundamental ADC spec, if you use an ADC this is something you must have seen on its datasheet.

Assuming the DNL is close to 100% (one LSB, this is a successive approximation ADC, they are not suitable for MCA purposes because of this) which you have scaled down from 10 to 7 bit you must have something like 10% DNL visible in your spectrum. Far from great, the spectrum will look hairy, ADC-s for spectroscopy do DNL about 1% - which is why they are not successive approximation type (Wilkinson, Gatti in the old days, with today's DSP techniques the DNL of the ADC (which can again easily be close to 100%) is spread among samples, dithered by noise etc. and the DNL is de facto not measurable.

DNL of MCU ADCs is almost always horrible--2 LSB or thereabouts. Makes histogramming rather entertaining, if not very informative. ;)

Cheers

Phil Hobbs

Come out for Photonics West and be prepared to reconsider.

On a sunny day (Mon, 1 Nov 2021 21:48:41 +0200) it happened Dimiter_Popoff snipped-for-privacy@tgi-sci.com wrote in <slpgap$fqn$ snipped-for-privacy@dont-email.me:

Better look up the 18F14K22 dataheet The secret of the system I use is not in the 10 bit ADC possible error but in the sample and hold circuit from the PMT. There is a reason in the schematic drawing on my site where the secret sauce is covered by a big 'Steadler Mars plastic' eraser:

ADC is pretty good, used it for audio too. So no need for paranoia.

And an other thing, in case you have limited screen resolution you may do math to 30 digits behind the dot (comma for Europeans) but you will still have to rightshift many times, practical example in 'scope_pic'

I do the Fourier transform with high precision, but can only display vertical 64 and horizontal 128 dots so simply did a couple of right shifts driving the display, is purvekt Lets Go Brandon! Zorry

Had no idea the Russian version - which is, like I said, Dmitriy in Cycillic - is written up so in English. That Shostakovich chap was also a "Dmitriy", apparently known as Dimitri, I just did not know.

We have a Meyer lemon tree just our side of the fence. When they moved in, we told them they could pick all the lemons they could reach.

Meyers grow well in our cool damp climate and make tons of lemons all year. We are the neighborhood lemon source.

So C (the Bulgarian lady) made lemon cupcakes (very pretty ones) and gave us a few. Act of war.

Last shot was our giving them some Guinness Cake. Top that, you foreigners!

A cute trick is to add a lot of dither with a DAC and subtract it out of the ADC data. That gives excellent DNL without smearing resolution.

Yes, this is the Gatti method (aka "sliding scale" IIRC). Have never designed one of these, the DSP era came for us by the end of the 90-s. My first attempt on a Wilkinson ADC (done during one night in Cologne back in 1990 or 1991) was with synchronous counters, probably 74xx161 or something. Guess what, the DNL was no good of course; one could see in the spectrum how many bits had toggled for each channel :-). I then resorted (routing directly on the board) to asynchronous counters, 74xx393 or whatever. This worked well, DNL was within 1%. I have seen much weirder approaches. A guy had designed a Wilkinson ADC with synchronous counters (the mistake I had made at my first attempt), 74xx193 IIRC. Now in order to wrestle the DNL problem he must have hit he was somehow producing some random number of the same width and preloaded the counters with it... Then he must have added/subtracted etc. , I only briefly looked at the schematics some 30 years ago, but I got impressed by the hoop jumping he had subjected himself to :).

That Guinness cake looks good, but wars are known to be hard to end... Let us see to what they will resort. Lucy used to make various cakes and nice things, but we had not enemy in sight back then (the neighbours who bring things are new, she was already ill when they came). I used to say "our cake - the best one"...

Have they targeted you with "banitsa"? Probably not, this is too much an everyday thing here but can be good. (

, last banitsa Lucy made).

That is letter of the law. Let us see how it works in practice. There was patent (in name of Feynmann) on plane with nuclear propulsion, issued around end of WWII. Cleary and that time nobody, however skilled, could not build such plane. So, basically idea was obvious and main/only thing were technological problems yet idea was patented. Also photcopy (Xerox) patent: people close to the top Xerox management admited that when Xerox bought the patent it was inpractical, they had to do serious research before they came with process that really worked. But the original patent gave them monopoly over the field...

Well, I prefer to avoid political discussion. But there is one obvoius solution: let Chinese get rich. When Chinese workers are paid as well a Americans they will no longer be able to compete on labour cost alone, so one big motivation to send manufacturing to China will be gone. OTOH, if you send manufacturing to China because only they are technology or organization to (economicaly) make the product, then there is question who have main input to the product? Maybe instead of trying to conrol whole operation more sustainable way would be to sell design to Chinese (and possibly act as distributor on some markets)?

In different spirit: there are laws beside patent law that are intended to protect you from unfair competition.

I'm about due for a PWest run, but we're expecting to be super busy round then, so I'll have to see. It would be great seeing you folks, for sure.

I'm all for custard and stuff--my issue with bread pudding is mainly the texture.

Cheers

Phil Hobbs

Mine is fruity and crispy/chunky on the top and custard on the bottom. Not a monolithic mess.

You seem to misunderstand the value of patents.

If you have an original idea and patent it, you can justify investing in its development. If it doesn't work, the patent won't get in anybody else's way. There are also diligence requirements--a patent can be invalidated if the inventor doesn't work pretty hard to reduce it to practice within a reasonable time. That's intended to discourage people from filing nuisance patents out ahead of likely technological directions.

The article isn't political, it's a blog post about how to structure contracts, patents, and trademarks in China to avoid problems.

It's mostly gone already.

The linked article and the rest of the Harris Bricken blog has a lot of stuff about how to do that sort of thing. It's far from simple.

A patent is a property right, like the deed to your house. Of course you could rent an apartment or sleep under a bridge, but those are far from equivalent.

Cheers

Phil Hobbs