program inconsistencies

In the past couple weeks or so, a number of threads discussed some basic topics such as certain aspects of transform results, dither, sampling rates, and such. I have done a number of experiments suggested to me by some of these. I only recall writing about one of those, in the debate about 16 bit transforms: dithering vs not dithering. I started to open new threads about a number of the other experiments, but things became too confused for me to know what to write. That is the basic inquiry here: Has anyone else observed inconsistencies such as those I hope to put into words?

The latest efforts, earlier today, were again about dither. The initial experiment was:
generate a tone @32 bit/44.1kHz (CE's 440A preset @ -40dB, 5 seconds)
transform/amplification/fade to zero, linear fade
make several copies
convert one to 16 bit without dithering
convert one to 16 bit with dithering, no noise shaping
convert one to 16 bit with dithering and noise shaping

The three 16 bit results were very different, different both from each other and from the original 32 bit version. This was much more extreme than I expected.

I repeated with the Bell preset. In this case, differences between the treatments were indistinguishable to me by listening and only just discernible by viewing in Spectral View.

I repeated the first set. Results matched the Bell preset tests.

So maybe I screwed up the first time and did something I didn't realize I was doing. However, I've run a variety of different sets of things now, repeating the above and trying new ideas. Results are inconsistent from time to time. Especially when I close the program, then restart, I often get something different from an earlier trial.

Converting to 16 bit sometimes produces lots of harmonics, clearly visible in Spectral View and Frequency Analysis. Sometimes the amount of noise added, after converting with dithering but no noise shaping, differs considerably, and very noticeably, from one time to another.

I've had problems where something got corrupted and I had to reinstall the program, but on those occasions CE2K was essentially unuseable. Now everything seems to work, only various transforms don't seem to work quite the same way from one use to another.

These weirdness seems only to occur when working with generated tones; I've noticed no problems with the LP cleanup that I do almost every day. I will have to reinstall and see if things get better, but how to be sure things are better? I hoped to get some sort of insight first. It doesn't seem to be an overall program problem. I don't find any viruses. The idea that someone wrote malware that would be so sneaky and complex, to target this one application, seems ludicrous. Is my experience totally foreign to everyone?

These weirdness seems only to occur when working with generated tones; I've noticed no problems with the LP cleanup that I do almost every day.

Doctor, Doctor, it hurts when I stick my fingers in the stove...

I can't tell you why you get inconsistent results, but surely the best way not to is to stop doing large numbers of completely unneccessary experiments...

There is one thing you should be aware of, though. And that is that unfortunately, every version of CEP/AA has a tone generator that is less than perfect. Because no A-D conversion is involved, it will cheerfully generate alias tones if you are not careful with it. The bell preset certainly does at the start and end of the sweep, simply by virtue of starting and stopping abruptly. Before you do anything at all with the output from the tone generator, you should run it through a low-pass filter, unless you are absolutely sure that you are only producing unmodulated in-band sine waves. And to be absolutely sure, fade whatever you produce in and out over a few milliseconds, or at least make sure that it starts and ends at zero, not a positive or negative value.

I'd say that if you are getting acceptable results from your processing, be grateful; you are looking a gift horse in the mouth.

The incorrect assumption there is "completely unneccessary (sic)." Not so. That's my life you're trying to discard.

Perhaps more importantly, if I can't get consistent results doing one thing, how do I know the results are even likely to be consistent next time I do something more "conventional?"

Perhaps this has something to do with running out of 16 bit GDI resources, or maybe some of it has to do with such, but I don't know how to verify it. The evidence there (from Cacheman) is inconsistent too.

Eventually I got tried of trying to make any sense of things and turned off the computer. A little later I realized I have the program installed on my recording computer, so I tried some experiments there. Things made sense and worked as expected.

Later I came back to this computer and ran a series of tests. Those also behaved. When I started the earlier round, the machine had been on for hours, I had used half a dozen or more different applications, and I had been on and off the web a number of times. The whole picture isn't exactly comforting; I guess I just have to be vigilant.

The incorrect assumption there is "completely unneccessary (sic)." Not so. That's my life you're trying to discard.

Self-condemning...

Is my experience totally foreign to everyone?

Not at all     ...I have days when I muck things up and can't see why, too!

Perhaps this has something to do with running out of 16 bit GDI resources, or maybe some of it has to do with such, but I don't know how to verify it. The evidence there (from Cacheman) is inconsistent too.

In this day and age these considerations should not be occupying your time.

I had a weird experience the other week - I was called on to debug something in a machine running Windows 95 .   I had (almost) forgotten how bad things used to be before MS decided that all versions of Windows should be based on a real operating system core.

Paul

My recording computer runs on Win95. It has no problems or deficiencies. No other OS could improve on its performance. Of course there are many things Win95 doesn't do well, but that computer isn't employed for them. When one has some time but no money, guess which considerations rise to the top?

What audio oriented tasks does SteveG think are more ‘necessary' for me to do? This is of course a rhetorical question, since nothing except value judgments are involved, but I don't see his point in making them.

Not that I begrudge him, or anyone else, their opinions.

What audio oriented tasks does SteveG think are more ‘necessary' for me to do? This is of course a rhetorical question, since nothing except value judgments are involved, but I don't see his point in making them.

Doesn't sound like a rhetorical question at all, and the answer certainly isn't - but it does contain a metaphor.

The purpose of audio processing in your case is surely to translate from an obsolete medium, and sometimes improve the audible results of, the material at your disposal. Ultimately, you listen to the results, and these are either acceptable or not, and you take action accordingly. What you don't do is repeatedly test the system and find it wanting. It's like taking the car shopping - you don't rev it up for 2 hours in the garage first to find out where it runs out of revs, do you? You certainly don't learn anything about driving the car that way, and it doesn't get the shopping done.

So, the direct answer to your question is that I think it's neccessary for you to process material, and unneccessary to keep testing the system.

I seem to recall that several years ago, when you started corresponding about this, that a few of us gave you the same advice, and you seem to be slipping back to your old ways. Stop testing the system and writing about it incessantly, and process some records!

And don't complain about me saying that, because you asked...

My recording computer runs on Win95. It has no problems or deficiencies.

Perhaps this has something to do with running out of 16 bit GDI resources, or maybe some of it has to do with such, but I don't know how to verify it. The evidence there (from Cacheman) is inconsistent too.

Okaaay...

Paul

My recording computer which runs on Win95 never runs out of resources. That's covered under "no problems or deficiencies." This computer, or some applications running on it, under Win98, occasionally has difficulties. Upgrading now is not an option.

Occasionally there have been problems, like those with the Mia, and I test to attempt to solve them. Most of my testing, however, is done to see how things work. I start with the basic assumption that the system functions properly.

This latest endeavor, looking at the effects of dither vs no dither, was for my edification, not to see if CE could dither properly or consistently. When the computer, the OS, or the application, don't work, enlightenment is limited. Often enough it is all so confusing that its hard to figure out where the problem is: my understanding of the(theoretical) process, my expectations, my interpretation of what I find, a faulty program ... . I ask questions. All that is probably painfully obvious.

What apparently isn't obvious is that transferring LPs to CD, with everything that goes into it, is also done for my entertainment. It isn't any more important to me, and it certainly does not matter to any other person on this planet. The only "necessity" is doing something I feel like doing, which is hardly the same thing all the time.

The best hybrid 32-bit/16-bit OS (I mean not a true 32-bit OS) is Windows 98 SE, I don't see any reason why anyone should use 95 anymore.  

If one has a Win95 computer that does its job without fault, why would he change it (other than from boredom)?

Here is the difference if you would like to know.
But if your happy with 95 then you are happy with 95.

http://en.wikipedia.org/wiki/Windows_98_SE

[Jim]

The best hybrid 32-bit/16-bit OS (I mean not a true 32-bit OS) is Windows 98 SE, I don't see any reason why anyone should use 95 anymore.  

I was still using Win98 SE for audio apps until earlier this year, and it did just fine -- well, as long as I rebooted it 3-4 times a day. Seriously, though, it is much better than Win95 (but I find that XP Pro only needs to be rebooted about once a day).