Math is fun, for guessing first one correct and the rest wrong:
1/5 chance of getting it right
9/24 chance of getting the next 4 wrong (Number of derangements - OeisWiki)
But we have 5 ways of doing that (5 different positions for the one correct match).
Overall, for 5 opamps, probability of getting exactly 1 right is 45/120 (3/8). 3 or 5 matches, now, there would be something much more indicative if we redid this a number of times (all the math above is based on ONE test). But then the math changes up a whole lot.
The Matching Problem (Example 24)
(If I've completely botched the math, please, please, please correct me; I did the first time around because, well, I get things wrong. 🙂)
Last edited:
Yeah! What would you know about any of this!? What would Walt say about repurposing DSL drivers for other purposes? 😀
To reform the original statement by Robert, go have a look at the AD815, too. (I forgot the roughly equivalent DSL line drivers by other manufacturers, but they must have gotten it all right for audio, too)
*Which sure as heck isn't to say these all aren't excellent for (specific) audio purposes!
I've got a pair of THS6012 eval boards* I always meant to make into a headphone amp. The specs on that are a good reminder of how pathetically easy audio is.
* Full disclosure freebies as I new a member of the design team for that part. I believe it sold fairly well. Not seen anyone yet try and use openwrt to get audio out the ADSL port tho...
* Full disclosure freebies as I new a member of the design team for that part. I believe it sold fairly well. Not seen anyone yet try and use openwrt to get audio out the ADSL port tho...
Last edited:
Your previous post was very dismissive of B.P.
Probably he refused to deal with the esoterica neurosa common with high end audio.
Maybe it’s time to demystify the ‘deeper aspects’ of audio ?
He didn’t really care about what specifically that should be of interest to one who designs electronic amplification devices?
Form and attributes of possible worst case real audio signals?
Adequate but realistic safety factors (design for x times the expected worst case)?
A practical how to measure the Slew Rate of a signal at the output of an audio gear (proposed by P. Baxandall during vinyl era)
1. Fit 1st order high pass filter to output of your audio gear, C=1000pf, R = 1000 Ohms, so that CR = 1us. Output is Vdiff
2. Play a track with a sinusoidal test frequency, measure peak level of high pass filter output Vdiff
3. Calculate slew rate SR = Vdiff / CR
e.g Vdiff = 2.8V
Result SR = 2.8V/1uS = 2,800,000 V/s=2.8V/us
George
I've used it. It is very good and does what it says in the data sheet. It has an outrageous SR . . . and commendably low distortion but I am not sure that's the reason it sounds so good.
Last edited:
Just one comment on Mooly's test.
It makes nothing clear, because he mixed and matched opamps. At least I think that is what was in his text?
If it was going to show something clearly, it would of necessity need to be only one type of opamp per "test"/track. Assuming nothing else was going to drift, or oscillate, or otherwise be improperly implemented that is...
I think it was a test in the right spirit, but really needs to be one opamp type per signal path.
...you may now return to squabbling.
_-_-
It makes nothing clear, because he mixed and matched opamps. At least I think that is what was in his text?
If it was going to show something clearly, it would of necessity need to be only one type of opamp per "test"/track. Assuming nothing else was going to drift, or oscillate, or otherwise be improperly implemented that is...
I think it was a test in the right spirit, but really needs to be one opamp type per signal path.
...you may now return to squabbling.
_-_-
That particular one was a small air-core inductor used as a stopper.
Dan, you're right that you have a 20% chance of getting the first one correct (or one arbitrary one), but I think the calculation is easier and that the right number for getting an arbitrary choice correct is a bit higher than what you presented (more like 60%). I was sloppy in that I should have said "at least one correct when selected randomly," which simplifies things a bit as well.
Jay, thanks for trying, but that is kind of a standard name.
What title? From which manufacturer?
Do they have everyone by their sensitive parts?
Ok, I have to do my part.
Interesting books by Graeme, author are pricey.
I thought I put it in my cart at amazon-NOT.
It had 5 solid stars in review 20 - 40 reviers?
I want to say price was $200-$400.
Anyone know of an author from Serbia?
for some reason that sticks in my head.
Doesn't mean it is correct.
It was linked from a forum in DIYAudio,
also with a comment that they had you
by the you know whats,
I've looked and can't find what ever it was.
Maybe it was a link to google books.
Could it be a design book from Philips?
Cheers,
Sync
Last edited:
After pulling the first (right guess, 1/5 probability) there are 4 left, so chances to get the next wrong is (1-1/4)=3/4. Then the next is (1-1/3)=2/3. then the next is (1-1/2)=1/2 So the chance to get the next four wrong is 6/24=1/4
So chances to get exactly one right is (1/5)*(1/4)*5 (five ways to get the desired result, as you mentioned, combinations of 5 taken by 1) which makes a round and nice chance of 25%.
Chances of getting exactly two right is (1/5)*(1/4)*(2/3)*(1/2)*10 (ways to do it, combinations of 5 taken by 2)=16.666%
Chances to get exactly three right is (1/5)*(1/4)*(1/3)*(1/2)*10 (ways to do it, combinations of 5 taken by 3)=8.333%
Chances to get exactly four (or five, since if the first four are right we already know the fifth) is (1/5)*(1/4)*(1/3)*(1/2)*5 (ways to do it, combinations of 5 taken by 4)=4.166%
The combinations above are of course the binomial coefficients.
Last edited:
Ah, yes, if you look at the link (and subsection 24, as it's a perfect representation of the question at hand!) you'll see that P(n=5, k>0) would be 63.33%
I answered the question of exactly one correct, which is 37.5%. 🙂
http://www.math.uah.edu/stat/urn/Matching.html
Last edited:
To get an opamp stable you sometimes you have to work at it - by that I mean read the data sheet carefully and then read it again carefully.
I was working on a 0dB buffer recently that oscillated at about 20 MHz when the volume was set to zero and when it was set to max - i.e. both conditions where the + input source impedance was very low. Decoupling was by the book, all PCB tracks very short etc.
Mucho head scratching so I re-read the data sheet and discovered it clearly indicated that for unity gain NI operation be sure to have a ~100 ohm resistor in the feedback path and in series with +input. Problem solved.
I was working on a 0dB buffer recently that oscillated at about 20 MHz when the volume was set to zero and when it was set to max - i.e. both conditions where the + input source impedance was very low. Decoupling was by the book, all PCB tracks very short etc.
Mucho head scratching so I re-read the data sheet and discovered it clearly indicated that for unity gain NI operation be sure to have a ~100 ohm resistor in the feedback path and in series with +input. Problem solved.
That was the error I made the first time around! 😀 You have to be careful with the combinatorics of ensuring the wrong answers are matched up correctly.
Walt did the 744/811 and 823/815 composite line stages. Lots of happy DIYer's built one of them so it's sort of old news. xDSL requires IMD performance out to ungodly frequencies but only at the 65-70dB level but the use of huge GBW and GOBS of feedback gets pretty good performance at 20k.
Is this an analog or digital mixing console?
How much actual processing is the console doing?
The power supply is probably at the other end of board
for noise reasons? Yes?
If PS there at the Mic section of the board then you go
to the Mix section of the board, or the output/send, then
what? There are other time delays as flow of energy rate?
around light speed correct?
And if we are processing all these slow audio type signals
Heck, then why not get a mini-cray super computers with
limited wire length just for that reason and re-purpose it
for audio use.
Or, what am I missing? Perhaps Doug Self or other
mixing console designers might care to chime in.
Or is the problem with the clock?
Why not use a GPS clock for timing? Seems well eenough
used to find people objects and other stuff why not use
it to Snyc the mixing console?
What am I not seeing?
What I calculated is the chance to get exactly K out of N right guesses if you check the correctness after each extraction. In this way, the probability to get at least 1 right out of 5 (without adding back) is exactly 1 (since if you know the first 4 matches you also know the fifth). If you don't look up until you get all the five extracted, then your result is correct.
Your math is matching closer Mooly's experiment, since the HUT (Human Under Test) doesn't not know up front if his m'th guess is right or wrong, so he can screw all five guesses.
Ah, okay. Yeah, I was trying to match up with Mooly's experiment. I wrote exactly what you did first, but then tested a couple cases and it wasn't meshing up (again, not checking along the way)
Last edited:
'Specific audio IC..SSM2019...for use as a microphone preamplifier in high fidelity applications'. But wait, there's more !
(Basic Linear Design, AD)
SSM was a PMI thing, anyway if you are actually interested in details I see no special attention to the usual audiophile issues. It is a straight forward descendant of the Demrow data amplifier of 1968 for instrumentation. We were talking op-amps not special purpose functions for the small audio industry. THAT is successful in the same domain now.
The SSM catalog contains several special purpose IC's for audio, I don't know of any of them that were not designed to meet a set of specifications or were compromised in any way to suit some audiophile belief.
The concept "designed for audio" as presented here is something more.
Last edited:
- Status
- Not open for further replies.