Presently use Creative Labs EMU1212 2 channel at 192kS/s above 89% of Nyquist. There's a sharp slope of 'can't use anymore' from around there to Nyquist rate where it goes to 'zero usability'.
Problem is that I need a higher sampling rate Data Acquisition System with those bits AND that linearity.
Have a look at .
-- -TV
looks good! THANK YOU! Redmond, WA must use their products on Boeing parts, eh?
BA-DA-BING!!
Carl, do you work in this area?
This application will NOT be for listnening. Purge your mind of any contributions relating to listening, does not apply. Again, listening was brought up only in the 'aside' comments.
Again, not ALLOWED to. would like to, but not allowed to.
You obviously don't work in this area, or simply have not faced development challenges like these. You should not assume what you do not understand. as in "...serves no useful purpose." How would you know? You have no idea of the application. And have erroneously assumed the application relates to sound reproduction, in spite of the fact that very early on in this thread, I said does not relate to sound reproduction. But is ok, I understand the origin of your mindset [maybe]. I too dislike the hype [and the creation of an industry] surrounding all that bogus hifi stuff.
interesting area to look. thanks.
Again, simulations predict spec requirements. Of course, those relating specifications are tempered with reality.
For what it's worth, noise floor [which most people concentrate on] is the least damaging to the final system. Aperture jitter [sampling time uncertainty] is the final limiting specification. Not usually mentioned, because it is so easy to get decent aperture jitter performance at audio rates. But go up a bit in speed and its contribution will dominate. And, if you work in this area, there is NO way to remove it. I don't care if you use EVERY sample you ever obtained, jitter will kill your performance. Unless someone in this group knows how to remove its effect? Please, jump in. I need to remove the effect of the 70fS 'uncertainty' to the high frequency samples and so far, do not see a way to do it.
Yes, it does. Only because it is 'unfamiliar' and generally 'not accepted'. Such social pressures limit mankind's advancment, albeit control it too. Perhaps, good, perhaps, bad. You wouldn't even begin to believe other examples, because you seem to have no personal reference for them. Do NOT accept what others teach. You have been 'taught' to be less than you are capable of. As in, 'crippled from birth'. Tell a child that something is not possible [or have no expectations] and for that child it is not possible. Languages are a primary example of how a developing brain 'wraps itself' around a problem. I would bet you could not discern half the phonemes in foreign languages, just because were not exposed to them as a child [during brain development] A more public example: for years the statement that nobody can run a mile in less than 4 minutes held as fact. Thought just physically impossible. Held for a long time, too. Not your fault that society has stomped so much out of you. Except, don't let it. have some faith in your own abilities. From your descriptions to date. you obviously have a reasonable mind and more importantly discernment. You know how to separate true reality potential from unreal wishful potential.
Thank you for that link. Made interesting reading, would have liked to see some scale numbers on the Fletcher-Munson curve, though. just have to overlook those unsupported statements and some errors, but all in all pretty decent.
If you want to make a career out of hearing, keep in mind that there are MORE nerves going FROM the brain TO the ear, than coming FROM the ear TO the brain. The significance of that? ...No one probably understands hearing too well ...yet.
Not really, I just had that benefit explained to me by an engineer when I asked him why our Fourier transform mass spectrometer used a 12 bit converter while the competition used 9 and 10 bit converters, on time domain data that had 5-6 bits of dynamic range. Ah, the good old days of grad school.
BA-DA-BING!!
Carl, do you work in this area?
RobertMacy
to
waveform.
source,
shifts.
the
making
Do you really need 200,000,000 samples per second at 16 bits? That would be a very expensive ADC.
?-)
channels
I
Well here is a likely candidate part for you:
These seem to come from medical imaging applications.
?-)
channels
I
Some more from Intersil:
plenty of speed not so good on the bits
?-)
Sadly, yes. so far 14 bit at 1GS/s comes in at around $20,000.
stack eight of them? no way! just joking
Thanks for the URL! Someone [sorry can't remember now] recommended Verasonics in Redmond, WA, a supplier to the ultrasound industry, which has a ready made system. but 14 bits up to 27MHz. yet 64 to 128 channels
Someone must have put this TI part into Ultrasonic Imaging, so there should be an available system out there, too.
unfortunately, need more bits than speed, but at least at 16 bit 'resolution' and 50MHz Nyquist, may make this work yet.
Have to be careful, though. jitter is extremely distructive to the system's performance.
Much of the effort from those nerves is for tuning and training the ear to identify the phonemes used in spoken language during the first two or three years of a child's life. That's why it is difficult to learn significantly different languages as an adult or teenager - you can learn the grammar and vocabulary, but you will never fully learn the phonemes that you did not hear regularly as a young child.
But it also shows that the signals from the ear to the brain are not a simple sound single - the ear has already done much of the work of identifying particular types of sound (mainly speech), sorting signals from noise, etc. And it shows that hearing is a very complicated sense.
On a sunny day (Wed, 02 Apr 2014 09:28:40 +0200) it happened David Brown wrote in :
I dunno, some languages I learned later, I speak Dutch, French, German, English, and a bit Portuguese. I give you that we started here in kindergarten with French (required) on top of Dutch, so used to 'other languages', French is a Roman language unlike Dutch that is a German one, very different. What helps immensely is that when you actually live in such a country, and speak the language every day. Watching teevee in that language also helps, provided you have also studied some grammer and learned words etc.
I look at (and I really do not know what 'current consensus is, and actually don't care) the ear as a feedback system. What happens is a bit (say completely) different from what is normally parrotted. I think it works like this: Sound activates the eardrum, and via the bones the canal filled with fluid with all them hairs (wave pattern created in it). Normally (and I think the OP too) people say those hairs are frequency sensitive, form some spectrum analyzer sort of thing, multi frequency detector. But it is not quite that way, The 'hairs' are _active_ vibration generators, and create, while listening to a spectrum, a continuous vibration compensating the vibration in the fluid canal. So it is a feedback system, well known in electronics, to keep a 'null' vibration in that canal (or there about). While listening the (neurons in) the brain reproduce the original that way, in fact 'play along'. When people are disharmonic the brain has to follow that disharmonic sound (create it in motion), and will get very tired, as it is a non-natural task for the neural net (that has its own perfect harmony frequency). So, there is more to it, and I do actually not want to get into this now, but just a feedback system. The same way we experience (or some people) buzzing and beeping sound (tintinus? these things (hairs) vibrate by their own. this has actually been measured by putting a sensitive microphone in the ear-canal. Quite a bit of amplitude.
So, whatever .. its really simple in a way, but given a few thousand years, dinosaurs, and civilization collapsing glow ball worming too of course and do not forget WW3, maybe somebody will re-discover how it works. Or maybe not. :-)
You could surely describe it in abstract terms of looking for a small signal buried in a powerful carrier wave.
Your time series is going to be pretty well cluttered with AM & VHF broadcasts and right up to the airband at the proposed sampling rate.
I have a pretty good idea of the capabilities and specs of these fast DACs. I knew where to look. We used something similar for MALDI-TOF and they are also used in transient recorders. I also know just how hard it is to get true linearity in high precision scientific ADC systems.
Even if it isn't sham spec audio it still appears to be a fantasy spec.
Something that is actually manufacturable would be a start.
Your highest frequency being around 250MHz or 4ns. The 70fs sample jitter should come out in the wash when you average the signals
40000000/70 is still a fairly large number. Weak law of large numbers should protect you when the time series data are averaged or FFTd.If you described in abstract terms the sort of signal(s) you are trying to measure it would be a lot easier to make helpful comments.
You now appear to want not just unicorn horn but to resurrect your unicorn from the severed horn and then bolt wings onto its back!
It is harder to learn a foreign language as an adult that is certainly true but it is by no means impossible. The hardest part is recognising when vowel length or tone is significant in Japanese and Chinese. I learnt Japanese in Japan as an adult but to nothing like native speaker standard. It was still better than my schoolboy French though. Total cultural immersion is quite a good incentive to learn a language.
Actually there had been attempts to try and beat the 4 minute mile barrier going back quite a way to the 1880s. Professional sports coaching made a big difference once it was allowed in the Olympics.
-- Regards, Martin Brown
Were you thinking of NI's RIO by any chance? Good and fast but the ADC elements are only 12 bits. Not sure if they have improved that spec yet though. You would need to add input protection and galvanic isolation if you are using in industrial settings though.
-- ******************************************************************** Paul E. Bennett IEng MIET..... Forth based HIDECS Consultancy............. Mob: +44 (0)7811-639972 Tel: +44 (0)1235-510979 Going Forth Safely ..... EBA. www.electric-boat-association.org.uk.. ********************************************************************
Thank you for that link. Made interesting reading, would have liked to
The key point is not the grammar and vocabulary - that can be learned later (though it is always easier if you already have experience with learning other languages). It is the phonemes, the "atoms" of spoken speech, that are important. I don't remember the exact figures, but I think there are something like 250 phonemes in common use around the world, with each language using between 70 and 150 of them. It's these phonemes - along with features such as tonal and volume variation - that your ears learn as a baby and toddler.
As a Dutch speaker, you will have grown up with almost all the phonemes that are needed in English, German and French - despite the different backgrounds, Germanic and Roman languages use a very similar set of phonemes. But you probably find that some of the French vowels are hard for you to distinguish and pronounce. And if you listen to foreigners speaking Dutch, there are some sounds that will cause them trouble - such as the "ch" that sounds like you've got something stuck in your throat. (Scottish natives will have a better chance, having grown up with "loch" and "Bruichladdich".
Phonemes are the reason why Chinese natives have such trouble with "l" and "r" sounds in English - they don't have them as separate phonemes in Chinese, and their ears cannot distinguish the sounds. They can /say/ the two different sounds - they just don't know which one they are saying!
On a sunny day (Wed, 02 Apr 2014 11:22:24 +0100) it happened Paul E Bennett wrote in :
No, it is called Zinc? or Sink ;-) John posted a picture of it recently.
The fun thing about ethernet is that you get galvanic isolation for free.
let's see 70fS is 95dB down, in a way. That's 25 dB above where most of what I do occurs.
Simulations of present system fairly well match that system's performance, but being in audio ranges the aperture jitter was still below other noise sources and its effect, or potential effect, was completely ignored.
Simulations for operating at higher frequency show that jitter now becomes not only noticeable, but cannot be removed! You can average to your heart's content but that noise does NOT drop. Johnson-type noise, yes. but not jitter. For example, on first pass, your noise spectrum looks flat. Average some and the noise floor lowers, still looks flat, but a small ramp near Nyquist starts appearing. Average more samples and the noise floor again lowers, but that small ramp does not change, only becomes more evident, since it's being left there as the noise floor is continuing to drop out from 'under' that ramp with the increased sample averaging. Appears as a 'linear' ramp, not log function.
This is not simulation using 'calculations' based upon someone's formulas [learned to not trust that stuff] but rather simulations [performed on a system with numerical analysis limitations of around 250 to 300 dB down, and therefore ignored.] attempting to replicate the EXACT system by using DAC-like signal with its digital quantizations, added white noise with known flat spectral density using proper levels of gaussian random sequences to represent that 'johnson noise', and now that 'reacquired' signal is digitized using a system with its own quantization, johnson-type noise, and gaussian aperture jitter. Yes, averaging gets rid of that pesky white noise, but does absolutely nothing to the effects from aperture jitter.
So it appears that the unicorn stops with a high frequency limitation that can NEVER be overcome. Must be content with using only half the spectrum, the lower half of the spectrum. And, yes, I have a definitive function for describing that 'ramp' in terms of bits, sampling rate, Nyquist rate, noise density, and aperture jitter. Those five terms create a 'ramp' in the noise floor that you can then determine its intersection with the flat noise. Remember, *if* you average; as expected the more samples averaged the more the 'flat' portion moves down, leaving the 'ramp' and thus, the 'useable' spectrum keeps decreasing.
As a comparison, the present system operates near 89% of Nyquist at a sampling rate of 192kS/s, higher than I thought possible but glad it does. The 'new' system will only be useable to around 50, 60, maybe 70% of Nyquist [set by features of operation] with sampling rate of 100MS/s, limiting, but not catastrophic. That leaves 25MHz of clean operation for a
100MS/s system. Less than desired, but more than before. I have NOT yet tested the DSP and processing algorithms to see what happens with the presence of 'ramp' noise dominating over flat noise. arrrggg! more work!
THANKS FOR THAT ONE!
The acquisition system is pretty much grounded to the PC, sticks out there a bit, so can be in contact with people, and could hit AC mains! as in blow the breaker. take out a lot of expensive electronics. You must have experience with the robust EMC environment of the Industrial Market. [What caught me off guard with Industrial Products was the requirement for being almost water proof, - not gentle rain, nor submersion, but from a spray hose!] Probably would have eventually discovered that electronic sensitivity on my own, but it's better to make the system initially capable of Medical grade patient leakage and be transient hardened early. As in, 'fixing' ALWAYS costs more than 'designing'. and, definitely fixing is unplanned.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.