Path: csiph.com!eternal-september.org!feeder.eternal-september.org!nntp.eternal-september.org!.POSTED!not-for-mail From: john larkin Newsgroups: sci.electronics.design Subject: Re: about electronics Date: Tue, 28 Apr 2026 09:22:45 -0700 Organization: A noiseless patient Spider Lines: 155 Message-ID: References: <10sg86t$7i9b$1@dont-email.me> <10shi71$j3f6$2@dont-email.me> <10sipkr$ueq9$1@dont-email.me> <10skt7b$1gnss$6@dont-email.me> <10spohh$2vate$2@dont-email.me> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Injection-Date: Tue, 28 Apr 2026 16:22:47 +0000 (UTC) Injection-Info: dont-email.me; posting-host="37feb1fd1fd119e6537d5d177ecbf437"; logging-data="3455629"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX1865bNJZoqCBz/RV7mH3DyX" User-Agent: ForteAgent/8.00.32.1272 Cancel-Lock: sha1:MpaeRBwR0mKkf6Wmu6oxjLBlr8c= Xref: csiph.com sci.electronics.design:743509 On Tue, 28 Apr 2026 17:44:47 +1000, Bill Sloman wrote: >On 28/04/2026 6:24 am, john larkin wrote: >> On Sun, 26 Apr 2026 21:34:01 +1000, Bill Sloman >> wrote: >> >>> On 26/04/2026 4:08 am, john larkin wrote: >>>> On Sun, 26 Apr 2026 02:20:43 +1000, Bill Sloman >>>> wrote: >>>> >>>>> On 26/04/2026 1:14 am, john larkin wrote: >>>>>> On Sat, 25 Apr 2026 15:07:29 +1000, Bill Sloman >>>>>> wrote: >>>>>> >>>>>>> On 25/04/2026 3:38 am, john larkin wrote: >>>>>>>> On Fri, 24 Apr 2026 10:10:53 -0700, Buzz McCool >>>>>>>> wrote: >>>>>>>> >>>>>>>>> On 4/24/2026 8:55 AM, john larkin wrote: >>>>>>>>>> Summary: very few people understand electronics. >>>>>>>>> >>>>>>>>> Edit: Very few people understand electronics as well as you. >>>>>>>>> >>>>>>>>> A few years ago you had some posts with typical interview questions you used. >>>>>>>>> I collected these and sent them to my now freshly minted EE child who will >>>>>>>>> be starting work next month. So don't think that people aren't listening. >>>>>>>> >>>>>>>> I hope your kid understands electricity. Most ce/ee grads don't, and >>>>>>>> AI may make them into grocery store clerks. >>>>>>> >>>>>>> Probably not. AI is essentially mindless plagiarism. >>>>>>> It gets stuff right about 90% of the time, rather like junior engineers, >>>>>>> but needs to be backed up by people with deeper understanding >>>>>>> >>>>>>>>> If you have more pearls of wisdom to share, please do. Years ago Phil shared >>>>>>>>> an article about how the front page of ceramic cap datasheets are a pack of >>>>>>>>> lies, which was very revealing to me. Even now in the 4th quarter of my career >>>>>>>>> I want to understand more about electronics. >>>>>>>>> >>>>>>>> Ceramic caps are awful. Buy a 22uF 10v cap and you might get 3 uF at >>>>>>>> 10v. >>>>>>> >>>>>>> Ceramic caps can be voltage dependent - but it does depend a lot on the >>>>>>> actual ceramic. Reading the data sheet carefully can be informative. >>>>>> >>>>>> It's rare for a ceramic cap data sheet to even mention the cap fallout >>>>>> with voltage, much less quantify it. >>>>>> >>>>>>> NPO ceramics are about as good as it gets. Ceramics with higher >>>>>>> dielectric constants offer more - but less predictable - capacitance. >>>>>> >>>>>> NPOs are great, but only come in small values. >>>>> >>>>> That's what I just said. >>>>> >>>>>>> https://en.wikipedia.org/wiki/Ceramic_capacitor >>>>>>> >>>>>>> does talk about this. Class 1 ceramic capacitors can be very stable. >>>>>>> Class 2 offers more capacitance but less stability, and Class 3 offers >>>>>>> loads of capacitance in a small package but it can vary a great deal. >>>>>>> John Larkin doesn't seem to read data sheets in detail, and he may have >>>>>>> bought a class 3 capacitor for a Class 2 application. >>>>>> >>>>>> Classes, and categories like Z5U, are very crude. One has to test caps >>>>>> when it matters, and then hope one can buy consistent parts. >>>>> >>>>> Crude they may be, but you didn't even mention that they existed. >>>>> >>>>>> We do buy big reels of custom-mixed caps with defined behaviors and >>>>>> tempcos. I had Capax make us 5000 pieces of 3.3 pF 0805's, with -4700 >>>>>> ppm/degC tempco. They came out close. We have a lifetime supply. >>>>> >>>>> That implies more technical expertise than you exhibited above. Did one >>>>> of your customers spell out the details, or that pesky Ph.D. that you fired? >>>> >>>> Neither. I use the NTCs to temperature compensate my instant-start LC >>>> oscillators. FR4 has a radical positive TC. I typically wind up with a >>>> parabolic frequency tempco, flat around 35c. >>>> >>>> Most things work like that. Compensate the first-order term, and >>>> what's left is mostly second-order. >>> >>> The instant start LC oscillators are a bad idea. You can get much better >>> timing out of a continuously running oscillator - you need two >>> interpolation circuits rather than just one to find out where your >>> interval started as well as where you want it to end, but the lower >>> jitter on the continuously running oscillator makes that a better choice. > > > > >> There are lots of ways to make a triggered delay with XO accuracy and >> jitter. Most involve a lot of insertion delay, and some have analog >> s/h drift that is work to correct for. > >The ones that involve less insertion delay are more interesting. > >> Many users want minimal insertion delay. > >Most of them want the timing ramp to start a few nanoseconds before they >know they need to start it. Thiotimoline would solve that problem, if it >existed. > >https://en.wikipedia.org/wiki/Thiotimoline > >It can make more sense to redesign the experiment so that you can >control when it starts, but customers don't want to do that - they do >try to buy their way out of bad system design choices. > >> The instant-start oscillator is great in that respect. > >But not in others. > >> After it starts, we observe it for a while >> and phase-lock it to a good OCXO but keep the timing based on the >> trigger. > >Which inserts a delay. About 3 nanoseconds. If the triggered oscillator is accurate and low-jitter on its own, we can observe it and phase lock it at leisure. No rush. The HP 5359A time synthesizer used a triggered delay-line oscillator and used a slow complex heterodyne system to phase-lock it to an OCXO. Similar idea. But they didn't have access to fast ADCs and DACs like we do now. My favorite triggered oscillator uses a coaxial ceramic resonator, but that has lots of real-world problems. > >>> You can't make a silk purse out of a sow's ear. >> >> That's clever. Did you make that up? > >Like the emitter-coupled monostable, it isn't original. > >> Here's our benchtop DDG. >> >> https://highlandtechnology.com/Product/P500 >> >> One nice feature is the GaN output stages that make super fast clean >> pulses over a wide voltage range. > >Pity about the instant-start timing oscillator. OK, you design a DDG and show me how. John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics