This is probably not an issue, I have found all major software packages are fully IEEE compliant, even Gates does not have to balance his budget to 15 digits.
It's generally less serious than having his head about to explode, dear Nobel Prize winner.
syn08; said:
Last edited:
Rohm did extensive listening tests comparing to both of the other 'heavy hitters' They didn't elaborate on how those tests were conducted.
The absolute numbers are slightly down on the others but it's still in the ball
park and they've obviously sacrificed numbers for a reason.
The DAC's architecture looks very good from a perspective of offering
plenty of analog stage options. It likely will pick up where BB left off, we will
see.
TCD
The logic here escapes me.
IME the BOM on a transparent DAC should be no more than $100 or so translating into an MSRP of around $500. I'm being very generous knowing what the components cost in the real world. As I posted someone fixed the IMD hump an ESS design with maybe $2 and a little savvy. You would do your cause well to get some test equipment and do some bench engineering.
Last edited:
Since I’m into programming these days (fighting that XMOS monster) I gave scipy a shot; starting from the same time limited sine input, the differences in Fourier skirts (non windowed) between scipy and Excel are day and night. Unless the Excel FFT is fundamentally flawed (which I doubt) there isn’t any other explanation than the limited Excel FP precision (which, BTW, is much lower than 15 digit, more like 8 digits for the standard FP, give it a try). Could be other types with 15 digit, I didn’t dig too far into this.
OK. First... I dont find those are an extraordinary claims.
WE said we heard same difference between one with and one without. Mark thought the without was better. Where is the extraordinary claim in that?
Live voice vs repo voice comparison.... Nothing extra ordinary in that either. Done all the time by recording personnel.
Sounding worse with 16/44 as freq goes up with transient signals (not continuous sine waves) is not a new first expose. Nothing extraordinary about that either with 16/44.
Nothing extraordinary about AES 24/96 sounding more accurate.
THx-RNMarsh
Last edited:
That's interesting a quick look at that link showed some problems with de-normalized numbers. If you say so, I have benchmarked Mathematica, Matlab, and Python on MAC, PC, and Sun and never found any discrepancies except Mathematica on early MAC's used 80bit floats. That might have been true on the Sun too, it's been a long time IIRC the 80bit float was more accessible on early MOTO FPU's. There are 80 bit and 128 floats on Intel processors I think but all the simulators and most software uses 64bit as max.
I never would use Excel for anything, I tried to help someone here trying to use Excel's optimizer for IIR filters (what a nightmare).
EDIT - In the beginning the head marketing guy at XMOS was an ADI acquaintance.
Last edited:
This aligns with designer's comments at Benchmark. The analog makes the greatest audible differences, not the digital was his opinion.
THx-RNMarsh
yes, it is really high. Not saying it should be 10K BUt costs for limited production is pretty high especially the chassis, meters (if any) and construction methods. Electronic parts costs isnt a big part of most costs these days.
For a pair of Damirs amps, I had to pay for the chassis layout person's time, technician to assemble and test time and many other people's time involved from one-off silk screen to pcb and chassis milling and buyer/purchasing personnel. At cost it was $2500 USD for 2. Not counting the loaded pcb parts/board from Damir. A bargain for me but not so great if you have to market, distribute and sell it at retail.
I cant even imagine the cost if it was all made in USA.
THx-RNMarsh
Last edited:
As I said someone eliminated the ESS IMD hump with about $2-3 in external parts and you know, measurements and engineering know how.
Last edited:
No prob. I overlayed the cos(2.5khz) waveform to get close to the signal which has a known sideband, so that we know what the exponential envelope produces in sideband content. What sample rate is best for you?
Jn
I agree with the observations of others with respect to your statements. It is quit obvious what I am saying. That you choose to ascribe entirely different things is your choice, and an obvious diversion.
I simply said, most designs do not control the current paths. As a result, bad things happen. I do not care if it is audible, that is for others to decide.
I choose to discuss the actual technology. If you wish to argue audibility, you will have to find somebody else.
Jn
So 5.21 is the best rate, or 2.6. (Microsecond per sample).192K would work I could possibly do 384K, I hope we're still friends.
I will post JPEG and equation Tuesday.
I am befuddled by your question..why would we not be?
Jn
Ps.. I am happy today, I got to use befuddled in a sentence.
Hi Bob,
I used to be one of these audio hobbyist's and it just like you say. My yearly spending was about half of yours, unless I had a bump year.
Then I found DIY audio, and I stopped the buying and selling of gear.
I changed to buying parts and learning how to solder and make things. I have EE friends who do the designs. I make the rest happen.
This is all a lot more rewarding.
- Status
- Not open for further replies.