Curious, FLAC has built in CRC error checking. I believe some of the other formats also have built in error checking. https://hydrogenaud.io/index.php/topic,38795.0.html and http://www.blisshq.com/music-library-management-blog/2015/03/31/test-flacs-corruption/
And if you want to see if the lossless file was really lossless Lossless Audio Checker
Now that I reflect, just to make clear: I wasn't alleging that the bootstrapped or neutralized input capacitance would reduce the noise, just reduce the distortion at the input with a highish Z source, given that any residual capacitance will be voltage-dependent, indeed as well nonlinearly voltage-dependent, and thus be a distortion mechanism.
But induced gate noise may well set a limit to the benefits of device paralleling.
You are taking things out of context again to suite your agenda.
I said you would not beat the cost-performance of a 5534 not that it was the lowest noise solution.
Shop around and the lowest noise version is about 50c IIRC, 1% SMD COG's in volume are pennies as are 1% thin film 1206 resistors. You get great PSRR, overload margin etc.
If you want +10 dB better noise performance then use JFETs (in parallel) and accept increased cost and complexity.
Last edited:
That would be the preeminent preamp from JR?
BTW 3 bucks for an NE5534 in 1985 dollars? We've come a long way down from that!
Jan
Last edited:
I'm imagining people who wish to use opamps like OPA1611 (datasheet) that cost $5 each but spec 1.1 nV/sqrtHz, better than the 5534. Or people who want to try their hand at taming the famously tempermental (but famously low noise) AD797 at $10.50 each. Remember to double the prices if you like to compare against the 5532 dual opamp.
I'm also imagining people who cascade a high current, possibly class A, high bandwidth output stage after their low noise IC gain stage, example below. This costs more than the $1.00 NE5534 but DIYers may want to try it anyway.
_
Attachments
Last edited:
True, my secret agenda was set over the weekend by the World Government at the Bildenberg group meeting in Dresden.
Of course some go out of their way to increase cost. LCR phono stages being a good one .Why use a 10c capacitor when you can chuck a $50 inductor at it. The Tango EQ transformers of yore seem to sell for thousands now. Common sense rarely prevails when vinyl is involved. But its only a hobby 🙂
Maybe my bad luck, one bit that defines a decode table would trash a lot in some compressed formats. You list a couple of uncompressed formats which would only trash one sample or pixel.
As to the original intent of the thread do you have any reason to suspect normal opening, closing, copying of files has an error problem at any noticeable level? Or how about cache which is exercised continuously during compute intensive operations?
But then you won't see them oscillating at 18GHz. 😉 IME these devices are pretty bad for noise at low frequencies some have 1/f corners of 10MHz.
EDIT - the noise has improved a lot but the betas still seem to be low.
Last edited:
Mr. Wurcer, at least in the mpeg compression that I am more familiar with, the individual bits can have any value within the spatial or temporal compression frames. Altering one or more bit could worst case compromise an entire frame, which may or may not be subjectively noticeable, but certainly would not crash the decompressor. There is nothing like a checksum in the mpeg standard.
Removing a random number of bits from the compressed file could screw the entire remaining file and may cause the decompressor to abort. The later is unlikely to happen due to corrupted mpeg frame structure, but because the possible corruption of meta data in the file (typical would be the optional mpeg time stamps).
When I was working on spectrometer electronics in the late 70s I was very ambitious to use the fastest parts I could find, and I managed to do so. I wish I'd taken a few photos of some of the circuit boards, particularly the places where I was close-packing TO-72 case devices, as they would look very quaint today compared to SMD. But things worked. However in time I realized that the added lossy components needed to tame parasitics reduced the bandwidth to the point I might well have been better off using slower parts to begin with.
One of the revelations was just how low the collector impedances could be for fast devices. I have a friend who defaults to treating bipolars as having zero conductance at the collector, which in some situations might be a good initial approximation but of course fails miserably in others.
No problem, Mark, my point was more to contrast the massive scope creep known as diy anything. You know the kind where a small weekend project to fix the grout around the sink ends up resulting in a new kitchen 6 months and 4 contractors later. 😀
HUMOR!
Attached is a plot of a string of twenty 2 ohm resistors fed via a 1,000 ohm unit. The frequency response is measured at each tap. The parts were on a solderless breadboard.
The second plot is a single 2 ohm resistor just clipped to the 1,000 ohm unit.
Attached is a plot of a string of twenty 2 ohm resistors fed via a 1,000 ohm unit. The frequency response is measured at each tap. The parts were on a solderless breadboard.
The second plot is a single 2 ohm resistor just clipped to the 1,000 ohm unit.
Attachments
Unfortunately, adding a discrete output stage is usually cumbersome, compared to the use of the IC by itself. Fast IC buffers are more compact, but are really expensive for what they do. Major designers often load down an IC with a load that is below its rating, for practical (for them) reasons like setting the gain or keeping the noise as low as possible. Here is where a good IC buffer could be useful, but they are still expensive and cannot easily be universally applied with cost sensitive (but hi quality) products like OPPO, in every case. This is a problem.
Maybe not bad luck, maybe just well chosen bits in my case. 🙂 I generally go 1/3, 1/2 or 2/3 of the way down the file, find some value and change it. It never passes a a checksum after that, but it doesn't crash or make a noticeable difference.
The reason I've done this was simply to check that I was getting bit perfect copies by whatever method I used to transmit them. I wanted to change just one bit by hand to see if the checksum really worked. It does. And I've had no problem getting bit perfect copies. That is the main point.
I'm game to do some more bit hacking to see if I can crash a compressed file, like FLAC. I will report my findings here.
By the time I finish a unity-gain buffer, usually I've spent six transistors, two complementary e-follower or source-follower output devices, two complementary input devices, and two current sources, usually LED-biased. It's usually something fast enough to not spoil the gain-bandwidth of the preceding stages. But it is cumbersome, and not so much expensive in terms of parts but demanding of PCB area.
The stage can be biased as richly as one desires.
Anyone remember the Burr-Brown modified TO-3 package buffer, the 3553? I see someone wants 200 US for one.
The stage can be biased as richly as one desires.
Anyone remember the Burr-Brown modified TO-3 package buffer, the 3553? I see someone wants 200 US for one.
- Status
- Not open for further replies.