I guess these would be the "pedestrian" designs which merely amplify the input signal? How boring!morinix said:
I while back I responded to one fellow's article about opamps. I mentioned that opamps are best used in series pairs to minimize distortion. Some time later he described his latest invention where he was now using two opamps and getting 10 dB better results.
As this is the second coincidence where he has followed my adventures and made "New" advances, I will only mention my current version of best use of opamps requires six units to produce one improved version. I get almost 10 dB of noise improvement and about 15 dB less distortion.
So pedestrian designs can be boring. But the elephant in the room is that for audio purposes some kinds of distortion are perceived as an improvement and others not so.
I'll take a badly distorted Picasso for my wall rather than a precise snapshot.
Especially the ones copied directly from data sheets and application notes? I have seen some degreed engineers who mostly stick to that, at least as much as possible. It can work I guess, but I wouldn't consider it all that creative or insightful.
Actually, come to think of it, someone with a BS degree in engineering usually comes out of school with a tool bag of mathematical tools, and very little practical experience. Depending on where they go from there, find a mentor, or learn on their own, they don't always end up as well as we might hope.
I agree that someone who only knows the electronics taught at university is not ready to design any serious audio gear. Having 'demonstrated' in second-year undergraduate labs I was shocked at how little they knew about quite basic circuits. That is not a criticism of good engineering, but a comment on modern education. Some of the anti-EE folk on here seem to have similar levels of understanding to a modern EE graduate.
@Mark: If you look at how few DACs out there actually meet data sheet performance then possibly following the reference design might be a good thing for more audio designers to do! (note also that Benchmark are one of the ones who do).
Oh goodness, this again. What on earth are you expecting from them? This IS their intro to EE, right then and there. First year is almost exclusively general science and math classes. No wonder I continued to battle almost every professor I ever TA'ed under -- there is such a massive gulf between reality and expectation.
Anyhow, we're back to our old scuffles. But, if we're going to go with "engineer tropes", many of us are mechanistic thinkers. An explanation bereft of at least a decently-developed hypothesis why a thing works is going to be held in an incredulous light. Which is why knowing *why* you're moving away from DS designs is kind of important. 🙂
On complex circuits such as sota adcs or bluetooth transcievers you follow the reference design and layout or you are in terra incognito. Its too costly to improvise there. Opamp implementation is less risky. I dont see how 6 opamps is better than one unless you are working outside the performance limits of them. Parallelingfor lower noise brings costs in input bias and cap. Valid in some cases.
Sent from my LG-H811 using Tapatalk
Sent from my LG-H811 using Tapatalk
Ed was trolling someone who seemed to be borrowing his ideas and passing them off as their own!
I suppose I should add that the people who produce the chips and the reference designs are also generally EE/physics grads. But of course we all knew that 🙂
I suppose I should add that the people who produce the chips and the reference designs are also generally EE/physics grads. But of course we all knew that 🙂
Then why do British EEs know so little science and maths? Maybe they slept through their first year classes. By the second year most of them know less about electronics than I did when I was 14.DPH said:
Dunno what the course is like at my Alma Mater now cw the 80s, but there was plenty of science and maths. At the end of the day a kid who is into electronics and has support will know a lot before they hit uni. And don't forget people have got PhDs by 14 (and serious issues). Vive la difference!
No argument here against following reference designs when that is the best choice. Just an observation that rather than designing with discrete components as a starting point, the starting point is at a higher, more modular level. Going in the other direction, I guess some people might have occasion to design their own components.
As design has often become more modular, not for everything, but in many cases, sometimes less highly skilled workers can produce working devices.
Certainly we see a trend like that in software, where virtually anyone can now program, but very few can troubleshoot in assembly or machine code, even if they really need to.
As design has often become more modular, not for everything, but in many cases, sometimes less highly skilled workers can produce working devices.
Certainly we see a trend like that in software, where virtually anyone can now program, but very few can troubleshoot in assembly or machine code, even if they really need to.
Most of the software types I encounter think HTML PHP and the contemporary frameworks for web pages is software. Assembly would be greek to most. Few having ever heard of it. Is there an assembly compiler for ARM?
I suspect designing the insides of a microprocessor or decode for a CISC processor is something only a handful of engineers do today (like in the 1970's).
I suspect designing the insides of a microprocessor or decode for a CISC processor is something only a handful of engineers do today (like in the 1970's).
Turns out there is an assembler for ARM ARM Assembly Language Programming And its a good primer on the insides of the process.
Back to opamps.
Back to opamps.
Here's a manual for it: ARM Assembly Language Programming
Also, as far as I know, computer science undergraduates still have to take a required course in assembly, and learn about things like register architectures.
EDIT: Oops! Accidental cross post.
Anyone can program BADLY. The state of the average website is proof of that.
And to counter your argument I will use a single Acronym COBOL. So since 1958 anyone can program, just back then they didn't have access to systems to do it on.
(as an undergrad I was told FORTRAN was all I ever needed. How wrong they were)
And to counter your argument I will use a single Acronym COBOL. So since 1958 anyone can program, just back then they didn't have access to systems to do it on.
(as an undergrad I was told FORTRAN was all I ever needed. How wrong they were)
I have the 2015 box set of Beatles mono pressings, is it cheating (sinning, wrong, bad manners, amateurish) if I listen to them on my stereo with 2 speakers? They still sound nice ....
Last edited:
There's many thousands of people working in semiconductor companies today, who design microprocessor circuits (and lay them out as geometric polygons), one transistor at a time. Some circuits are automatically synthesized via silicon compilers, others are hand polished and carefully optimized and squeezed to the Nth degree. That was my job; I did that for a living, here in Silicon Valley. It was a remunerative occupation.
Over time the number of hand polished transistors per chip has risen dramatically, meaning the number of engineers per microprocessor chip who do nothing but hand-polish transistors has also risen dramatically. But the number of automatically generated (not hand polished) transistors has risen even more dramatically. So the ratio of (total engineers / total transistors) has fallen, as indeed it must.
The entire engineering staff who designed a 686 microprocessor, fit into one medium sized building. Now they don't even fit into one site in one country.
Over time the number of hand polished transistors per chip has risen dramatically, meaning the number of engineers per microprocessor chip who do nothing but hand-polish transistors has also risen dramatically. But the number of automatically generated (not hand polished) transistors has risen even more dramatically. So the ratio of (total engineers / total transistors) has fallen, as indeed it must.
The entire engineering staff who designed a 686 microprocessor, fit into one medium sized building. Now they don't even fit into one site in one country.
Last edited:
maybe a transistor to engineer ratio would be instructive. 1.4 billion transistors for 1,000 engineers? that would be 1.4 million transistors. You would need to polish pretty quickly to get them all shiny.
When I was last working in that trade we used rubylith. It took quite a while to cut all of those holes.
When I was last working in that trade we used rubylith. It took quite a while to cut all of those holes.
Thanks George, starting with 352k should be a better starting point than the 192k file you downsampled in the first experiment, downloading now.
In my first assay, I mainly listened to the record clicks in the first 10 seconds and noted the attack/decay/tails behaviour of such impulsive sounds.
These new recordings of course don't contain these kinds of short impulse sounds in 'isolation', I will see what I find.
One confounder is that I need to go into Sound>Playback Devices>Advanced>Default Format each track.
I am running Foobar as the player application, anybody know how to avoid this manual Default Format switching when comparing different sample rates ?.
Dan.
I had a lot of trouble getting the 96kHz file to play back at the correct sample rate in Windows. Windows doesn't say anything, but by default it was downsampling to 16/44.1 on the fly in realtime, which did not sound good. I could see what it was doing from the Lynx2 mixer control panel, which shows the sample rate and bit depth being received from Windows.
The fix was to go to Windows control panel, sound, select the playback device, select properties, advanced tab, then set shared mode sample rate to 24/96. After than the hi-res file plays correctly, and sounds very nice.
Maybe Windows will now try to upsample 16/44 and everything else to 24/96, I don't know, haven't tried it yet. Normally I can control playback using Reaper and setting the project sample rate, but I decided to try other applications in Windows that other users may be trying to use.
Programs playing back the hi-res file incorrectly included Windows Media Player, VLC, and Audacity.
EDIT: Confirmed Windows now upsamples everything on the fly to 24/96. Maybe this has to do with programs accessing Windows sound drivers in shared mode, rather than exclusive mode. Not sure. Don't know how most users can confirm actual sample rate received at sound card either.
Last edited:
- Status
- Not open for further replies.