SoylentNews is people
SoylentNews
Google is offering to produce free chips for you. They have to be open source, they are using 20 year old technology and you'll get 100 of them. Could someone reverse engineer a SID-chip and have Goggle start to crank those suckers out?
https://www.theregister.com/2020/07/03/open_chip_hardware/
https://fossi-foundation.org/2020/06/30/skywater-pdk
This discussion has been archived.
No new comments can be posted.
Google Will Fab Your Chips for Free
|
Log In/Create an Account
| Top
| 67 comments
| Search Discussion
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Wow! Look!! A stray meatball!! Let's interview it!
(Score: 1) by anubi on Wednesday July 08 2020, @09:36AM (9 children)
LM358... You saw the schematic. Common mode voltage range includes ground, and a couple of tenths of a volt below. Sometimes that eliminates a lot of biasing circuitry on single supply circuitry. You can still have linear operation with both inputs right around ground. But not much lower.
Like non-inverting amplifiers or comparator usage.
Like you said, some opamps latch up. I try like the Dickens not to use them. Just one errant noise pulse will ruin your project. A latchup is a snake in the grass!
I've never seen an LM358 latchup. If you have, please post how you did it so I don't do the same!
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2) by RS3 on Wednesday July 08 2020, @04:06PM (8 children)
Yes, good point re: biasing. I've always done, as is very common, a simple resistor divider (with cap) to establish a mid-supply voltage reference. Just to keep out of potential non-linear ranges. Negative feedback / stabilization should take care of that, but in audio especially, well I do a lot in audio, but the point is that negative feedback isn't perfect due to normal circuit propagation delay (causes distortion that "golden ears" people truly do hear). I mean to say, you have to have negative feedback, but don't expect it to perfectly fix dynamic problems because if you're getting into nonlinear operation ranges, it's too late to fix. Feedforward helps, but can't always be done.
I'd heard of "latchup" but had never seen it in person until maybe 10 years ago when I bought an otherwise awesome digital audio sampling interface- Motu 896HD. It uses the aforementioned INA163 which has very low noise, but can latchup. I reverse-engineered the circuit and I see the problem and even designed a fix. I think it's a design failure and things like that are discouraging- dovetails with the (heated) discussion on repairability but I won't go into that now. If I had my way they'd have to fix it. I'm an EE and if I designed something that crappy I'd want to be forced to fix it. Point is- nobody's perfect and I think the world would progress better if we learned from, and fixed mistakes.
Frankly the fact that an op amp can latch up to me is a complete design failure and no such thing should ever have been produced!! Not sure if you know what an SCR is, but it's a 4-layer silicon (or other "semiconductor") device that due to device physics latches itself ON. It's not linear. It can only be turned OFF by removing the electric current, so they're great for AC power control. A "triac" is basically 2 SCRs in parallel- 1 PNPN and 1 NPNP if you will, and they're used quite extensively for AC power control.
Again, AFAIK most op amps don't have the latchup problem- only a few and they should be avoided, or very protected.
(Score: 2, Interesting) by anubi on Thursday July 09 2020, @01:15AM (7 children)
Oh yes, I have used SCR a lot. Mostly in resonant power converters and tap switchers.
100% agree with you... A latchup in a linear amplifier is a serious design flaw and should never have left the factory.
Hence my inquiry as if you ever had a 358 latch on you. I use the 358 a lot in control systems and having one latch on me could have serious consequences, not to mention my reputation.
Also agree the LM358 is not natively a very good high fidelity audio amp. The designers did a trade-off in the output source and sink circuit, running it class B for minimal quiescent current, but with significant crossover distortion as the output switches modes from sink to source and back. This can be minimized by forcing the output stage into a either source or sink mode with a pulldown or pull-up resistor to ground or supply rail. You audiophile guys need ultralinear class A biasing.
You are so right. Although high gain and negative feedback is great for precision instrumentation of relatively static phenomena, it leaves a lot to be desired when time dynamics of the feedback loop come into play. I can easily see you running VHF parts. That's why the old vacuum tube amplifiers sounded so good... The amplifiers all ran class A, very linear, with little to no feedback sans cathode resistors.
The 358 does not source very well and will not make it anywhere near the supply rail unless it's output is pulled up, then for practical purposes it's more like a LM339 comparator. Then the really high fidelity units ran a class AB push-pull output stage. Not power efficient, but sure sounded good!
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 1) by anubi on Thursday July 09 2020, @02:00AM (6 children)
Forgive me, that last sentence got misplaced and was meant for the part on vacuum tubes.
I am doing this on my phone, and I am very apt to make this kind of mistake because of my impatience with the page refresh lag of this thing.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2) by RS3 on Thursday July 09 2020, @02:22AM (5 children)
It's all good, really good stuff, thanks!
So you're obviously an EE? Sounds like you've done some cool stuff. I haven't gotten to do as much EE as I'd like- fell into mostly system, some software (C, assembler, ...), general IT, systems admin.
"Feedforward" can fix the AB crossover distortion somewhat.
You know, I never thought about the VHF op amp idea. Hmmm. The only real design project I have going is a vacuum-tube preamp. It's been on hold for a couple of years due to higher priority problems.
Again, the NE5532 is a great op amp- you'll like them I promise. :)
As far as I know op amps that can latch up are a known thing. IE, you know which ones have the problem design. So I don't think a 358 can latchup.
Mucho kudos to you- I couldn't do anywhere nearly this much typing on a phone. Maybe with speech-to-text, but I hate talking to things. Well, I talk to things, but not ones that actually listen. :)
(Score: 2, Interesting) by anubi on Thursday July 09 2020, @11:42AM (4 children)
Yup... Old EE. Brought up with vacuum tubes.
Dabbled in damn near everything you would find in the field or lab.
Got to work for an oil company, alongside brilliant geologists, then at an aerospace company.
Things were quite rosy until this thing got out among the executive management crowd. Some called this training "charm school". I had another name for it.
Needless to say, I soon developed an "attitude problem". I can't make a plane fly by signing some paper with the words "responsible engineer" under the line. I have to get my hands dirty to know what I am doing. Not from a cubicle. I usually have to personally build the prototype, just so I can see what I am doing wrong. Very rarely do I nail it on the first attempt. But I iterate pretty fast to the correct design.
I would rather do without than face the ignominy of being forced to build junk. And face the agony of failure of untested designs.
So, now I am a private builder of things.
Most of the stuff I work on is micropower custom Arduino architecture stuff. Like you, assembler and C++. Hardware, software, and any interfaces to the real world. I take a lot of pride in my work, and simply can't stand it to have some management type prodding me to cheapen the design and overlook detail. You know as well as I do: The devil is in the detail. Especially if big stuff is involved.
About all I know of IT is writing simple TCPIP stacks for embedded microcontrollers. Very limited subset.
Just for minimal connectivity on the net and maybe custom protocols that make it through the routers.
IT quickly got do complicated I lost track of how to tame the monster. I just know enough to address and mail postcards. So I can recognize mine, or send some.
I'll look into those NE5532.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2, Informative) by anubi on Thursday July 09 2020, @12:35PM
Oh yes... one for you...LMC662
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 3, Interesting) by RS3 on Thursday July 09 2020, @03:47PM (2 children)
We're much alike. You write much gooder tho. IT ain't a passion, it's just, well, sometimes for whatever reason I fall into it. I'm pretty good at diagnostics and fixing stuff, and IT being foundational, it gets top priority when something goes wrong.
I've always been a hands-on engineer, although I admit there are times I'd love to have a tech build stuff for me while I do one of my too many other interests in life, then I'll come back and test and tweak.
You'll love this stupidity story: about 1.5 years ago I was trying to troubleshoot a stupid circuit design that had gone into oscillation. Some otherwise smart engineer (or maybe non-engineer) had put a power resistor in series with an LM7815 to lower the 7815's dissipation. 2 MHz or so. And frankly it might have been okay but they didn't filter the output so the 2 MHz was getting into everything and spoiling the soup. It's a bit too long of a backstory, but strong rules prevent me from making any significant changes (huge red-tape). I did sneak some bypass caps in that quieted the singer but I haven't done a formal analysis.
Here's the stupidity part: in experimenting with bypass cap values, I had some small drop-type tantalum caps, maybe 4.7 uF. I held one between my thumb and index finger and poked it onto the live circuit. Well, I must have put it on backwards for an instant and it went CRACK!! and I had a tiny blackened hole in each finger and a good deal of pain for a day or so. I didn't know they were so intolerant or violent! The sacrifices I make fixing bad designs. Harumph!
I hear you re: big companies, cubicles, etc. I don't mind the office/cubicle environment for sw development as long as it's fairly quiet / minimal interruptions.
5523 is a workhorse chip. Thanks for the tip on the Lockheed Martin Corporation 662. :-}
I might be doing some Raspberry Pi programming soon...
(Score: 1) by anubi on Friday July 10 2020, @12:51AM (1 child)
Raspberry Pi, in my estimation, is the future of small run custom embedded. I need to get into that myself, but believe it or not, I find myself swamped by the capabilities of these toy Arduino hobby microcontrollers. Granted I had to make a few changes to the I/O schemes to give me virtually unlimited I/O space via multiplexing schemes.
I hope one day to release my designs with the hope that if they are good enough, they will become a standard design that will continue to be useful long after I am gone. But for now, I am making damn sure I have all the inconsistencies out of it.
I am making them to help out on the farm and truck, as information gatherers, and converting sensor data into common file formats...like .csv files... Without involving huge way overcomplicated systems that require all sorts of terms and conditions.
Lockheed Martin 660. I had to go look that up. I had a dual CMOS opamp in mind. I had used some before and they worked as expected on a laser current stabilizer.
Yup... Sure sounds like we have done a lot of fun stuff. Problem is, I got used to it. The longer I did it, the more things I learned NOT to do, and finally reached the level I could make nearly anything. You sound like you've hit that too, where you just look at a schematic and see the gotchas in it.
You do not receive experience and wisdom until just AFTER you really needed it.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2) by RS3 on Friday July 10 2020, @01:43AM
Sorry- very nerdy punny "humor"- LMC = Lockheed Martin Corporation. My dad worked there for many years.
I've seen some schematics and thought "wow, these people are brilliant!" But some, yeah, can't believe that ever worked, let alone was produced and shipped.
Truck and farm, that's my language. I'm not a farmer, but my mom's father grew up in the mid-west on huge farms so I've always thought it was in my blood. I like the idea of helping them gather and compile data. This could be a really good thing. I'd love to help- not sure if I could or how...
The aforementioned Pi project is a brand new possible gig. I have to go meet other company principal on Monday, but it should go well. Right now they're using some kind of 3rd-party interface gizmo that looks way overly complicated and expensive, to get a few lines of binary data. It's not necessarily 5V stuff so the guy was worried it would zap the Pi's inputs. I told him I could design a simple interface circuit to "clean up" the signals and protect the Pi. He kind of hesitated. I didn't push for it at all. Not sure if there's a "shield" or "hat" for the Pi to do that, but it's pretty trivial using a perfboard "shield".
I worked for a short time at a tiny company that makes A/D stuff- little single-board converter / mux. Some of them have essentially infinite input impedance. Mostly sells to military and science research labs. The input muxes were off-the-shelf analog switches, like 4066, and the op amps mostly TL084 family. Or maybe it was a MOS input one but off-the-shelf stuff for sure. Had to be CLEANED very very well- scrubbed very carefully, rinsed in very filtered water, and dried under warm lamps.
And they would do DC- they had an internal calibrate cycle to keep them DC "honest".