What's a good, really hgih speed 'soundcard'

Have you ever thought what kind of aperture precision is needed to get 20 bits (1 millionth of full scale) at 200 million samples per second?

Another point to ponder is thermal noise on the bandwidth.

I think

Mark

There is no evidence to support the Golden Ear brigade and likewise your claims. However, if you wish to waste your money I won't stop you. I would wager good money that in a blind test you could not tell the difference between a good Hi-Fi and your ultrasonic version!

It's a little more than a way of thinking of the suppliers. Pesky laws of nature. They just won't go away, even if you don't believe in them.

Oh well, you only need to do about 35dB better than state of the art. Maybe you can get there with a 70dB increase in budget. Maybe...

Jeroen Belleman

as you increase the sample rate the quantization noise power stays the same but it is spread over a wider bandwidth. Having said that, is your ear sen sitive to quantization noise levels of -120dB? You talk nonsense and seems like you have been reading too many Hi-Fi mags. I expect you are a believe r in Peak power too!

Mark,

PERFECT! Thank you. well, closest so far. pricey at $7k, but it appears to be mature, easy to expand to multiple channels [unspoken future requirement] and their website was open, easy to navigate, with FULL disclosures of information. [you have no idea how much appreciated those features are.]

Will contact them

Are you associated with them, have you worked with their product(s) before, do you know anybody there???

yes, all included in the requirement.

You'd be surprised, you probably could probably tell a great deal of difference. You don't give yourself enough credit, or don't have enough confidence. Don't listen to naysayers, and others, LISTEN to yourself. The 'trick' is to NOT think about it, just do it.

You have NO idea of this application, not related to sound at all. That discussion was merely an aside.

As I pointed out in another thread. The ear has some incredible abilities. I used to tune pianos and can tell you WHICH of the two or three strings is bad. Any tuner worth his salt can do that. And yes, I have NO hifi sound system around me because I cannot stand the 'finger nails down the blackboard' presentations usually done by these systems.

Ear acuity is better than you can imagine IF ALLOWED. I have actually listened to a direct cut recording done in the 30's. Based upon that listening experience drew sketches of the person playing the guitar the chair he was sitting in, AND the room he was sitting in, AND the openings and window placements around the room. During the performance someone moved around, which actually enhanced the acoustical experience and made it possible to 'envision' the room based upon those multiple reflections. Again, doing such is not that difficult, if allowed. So my premise that the ear can discern 'quality' in a recording is based upon FACT. Not tosh. Not audio-'foolishness'. Nor Golden Ear brigade.

You want 'other' claims/facts? like you can actually hear a person verbalize a thought if you don't try to [most people in deep concentration will verbalize what they are thinking] their vocal chords moving and making sounds almost like they're talking, and thus you can 'hear' what they are thinking and 'answer' their question without them audibly asking it. VERY disconcerting, since it appears that you are a 'mind reader'. You're not. Simply relax and hear what they ask.

spread over a wider bandwidth? implying it goes down?

application is NOT sound related. previous system inherently was operating in the 1-10 ppm range, that's close to 120dB

Don't read much, as in "Why Johnny Can't Read" now in VHS! At a risk of hijacking my own thread, what are "Peak power" implications? Sounds like a justificaiton for creating excessive 'headroom' in the reproduction electronics.

No, I've never worked with them, I actually ended up going with an Agilent solution because I was after wider bandwidth. But I have heard good things from others and talked with them on the phone. Like most of the smaller companies they seem pretty easy to work with to find the right solution.

Mark

If you are interested in a smaller bandwidth than half sampling then yes. This is the principle of a sigma delta A/D convertor.

Well, it's not unusual for this kind of thing to show up in an ultrasound system; Verasonics can set you up with a 64-channel unit. It's not really going to be a quick operation to do something useful, though, and you'll HATE the sticker price.

I know nothing of the OP's requirements, but one advantage to having more bits in the ADC than the simple calculation below says is needed is to get better spur free dynamic range, so intermodulation and harmonic distortion products will still be well below the input noise level even in the pathological cases of a strong signal and a very weak signal at the same time, even after signal averaging the power spectra.

It's a little more than a way of thinking of the suppliers. Pesky laws of nature. They just won't go away, even if you don't believe in them.

Oh well, you only need to do about 35dB better than state of the art. Maybe you can get there with a 70dB increase in budget. Maybe...

Jeroen Belleman

Maybe a labjack? lotsa inputs, I think one did 50-100kHz at reduced resolution.

George H. (it runs w/python.)

On 31/03/14 08:35, John Larkin wrote: [Snip]

Single-chip 16-bit USB soundcard chip? Like the TI one in my Behringer U-Control UCA-202? The chip is a PCM2902. Add as many channels as you want, plugged into USB ports on a Linux PC. Just want to take care of ground noise from the PC and arrange to deal with timing jitter.

We used 10v fsd and a 22bit conversion but at a *much* slower sampling rate synchronous with local mains (ie. a Solartron 7060 replacement). Prior to perfecting it in the late 1980's multicollector high precision isotope ratio mass spec kit could use up to 8x 7060s per instrument.

They were calibrated and linearised to almost 20 bits (6 decimal digits) during final test - the trick was weeding out any with missing codes and removing their systemic quadratic bias. ISTR the official position was that the AD1170 was good to 18 bits. We got a bit more.

I think you have to do a bit more than assert that laws of physics do not apply here. If the thing is extremely cold then maybe you might have a point, but usually there is a more elegant means to extract a signal out of noise than brute force iff you know what you are doing.

Perhaps because they fully understand the limitations of fast sampling ADC hardware. You are pretty much asking for whole unicorn horn here.

If you insist on a crazy specification you must expect to pay insanely high prices for very little (probably no) improvement in performance.

Actually there often is a *very* obvious audible difference. The ones with insanely high frequency response tend to pick up local taxis.

If you do this and keep going until you can't hear any difference you will buy insanely expensive kit that is only marginally better for orders of magnitude more money. A friend fell into this bear pit of snake oil sales and has been known to pay thousands for wires allegedly spun by virgin mermaids in an oxygen free atmosphere! I sometimes have a laugh at his catalogue of insanely priced "magic" interconnects. eg.

(even my friend isn't that cash rich and gullible)

It seems like the placebo effect in medicine the more you have paid for an irrelevant component the better it sounds to you. Amusingly some esoteric amplifiers are actually unstable with fancy speaker cables.

Perhaps you would care to enlighten us then?

At the moment you appear to be hung up on an impossible to realise ultra precise fast ADC specification that serves no useful purpose.

SAR and broadband radio astronomy use something like this approach but have a much saner analytical approach to how many bits are enough and at what oversampling. You seem to want to digitise like hell and pray.

Seems a shame. Depending on your musical tastes Quad, Naim, JBL, B&W and the like all make perfectly good kit.

The listening room tends to impart a lot of colouration to any recorded reproduction since most are smaller than a recording studio and the wall coverings and soft furnishing all affect sound quality.

If you say so. This is beginning to sound like kooky crank stuff.

I completely agree with this.

Some Software Defined Radios (SDR) might do ADC at quite high sampling rate (MHz), but the final detection bandwidth is a few kHz after decimating. The things work due to the band (and front end) noise supplying the dithering noise.

In the early days of digital audio recording using VCRs as the storage medium (e.g. PCM1) , I was puzzled, why the input stage had a zener diode isolated by a capacitor in the microphone input line. Later on, I found out that this was the dither generator.

Since the mains frequency varies on a 24 h basis, did you use the actual mains frequency or the nominal frequency for this. Using the nominal frequency would give some dithering advantage.

Linearizing something will help you to the nominal (ideal) accuracy, after that you need dithering.

Without quite limited detection bandwidth, you are going to break the laws of physics.

With audio bandwidths (20 kHz) we are still talking about 120 dB SNR (20 bits)

Going much below that and we are talking about things floating in liquid helium :-)

For a similar requirement, I had considered something like this:

It seems to fulfil some of your requirements though having a finite number of samples might of course be an issue.