Halcyon -> RE: ECC block size of testing software (5/6/2005 5:53:42 AM)
The sampling interval is based on the chipset and the software polling the chipset:
Philips Nexperia (BenQ/Philips/NuTech): PIE rows/8ECC, PIF rows/8ECC
Mediatek (LiteOn): passive, but can be polled down to c. PIF rows/1ECC (and summed PIE rows/8ECC)
Nec (Nec): variable, rows/1-8ECC are used at least in CD-DVD Speed
ALI (AOpen/Asus some dvd-rom models): PIE/PIF rows/32ECC or rows/64ECC (depending on model)
Sanyo (Plextor/Optorite burners): PIE rows/8ECC, PIF rows/1ECC (PIE/PIF have to be scanned as separate scan runs)
Via (Toshiba/Plextor some dvd-rom models): PIE rows/8ECC, PIF c. rows/1.5ECC
Ricoh (some AOpen/Ricoh): rows/16ECC
This information comes from the author of CD-DVD speed.
As for the other questions:
Q1: "If it uses 8 then won’t it over-report PIF peaks?"
If the scanning interval is more than 1ECC block at a time, then error spikes will be underreported, because the number of errors in an 8ECC sum window are just averaged.
It is possible that a 32 PIF from 8ECC blocks (summed) comes from 6 blocks containing 3 PIF each (=18 PIF) and the rest two containing 14 PIF (32 total from all 8 blocks). Of course, the software will only report 32/8 = 4 for each, because it cannot know the distribution of those errors within the 8ECC block window.
Q2: "And at 1 it will under-report PIE?"
If the chip/drive/software reports PIE at each consecutive block (i.e. rows/1ECC), it is very easy to calculate the floating 8ECC block window from this, without any under or over-reporting.
So, only too high block sampling size is somewhat problematic, because it will hide the peaks (maxima) in each sample block.
Q3: "How do I change the settings for CD-DVD Speed Test and DVDInfoPro?
This can only be changed for drives that support it (i.e. basically Nec), through some registry hacks and I wouldn't do it, unless I knew what I was doing (really knew). It’s not worth the time, unless you really become engulfed in read error testing, imho.
Q4: "Or do these programs already test both block sizes and the displays they show are correct?"
See above listing of PIE/PIF reporting window sizes.
All in all, I wouldn't be too worried about this, because if the disc is truly bad, you will get a huge increase in PIE/PIF regardless of the sampling size.
Also, as the errors are not on the disc, but in the reading process, it is useless to measure accurate PIF/1ECC block and determine based on that whether the disc is out of spec or not. All it tells whether that particular read on that particular drive produced a read error rate within the specifications (or not).
Q5:"is this why one is supposed to ignore high individual spikes?
High spike ignoring comes from chips/drives that have glitches and occasionally over-report errors, even when they read no such errors.
For example, if PIE remains really low, but there is a huge single sample area spike in PIF, it is more than likely that the error is from the drive's reporting and that there is no problem with the disc. PIE and PIF are interrelated, PIF only happens when there is a high enough number of PIE errors (in the same block).
Q6: "The reason I ask is that I have a disk with a narrow band of around 12 PIFs – if this is testing at 8 ECC, I could reasonably expect that at the 1 ECC level I would have no PIFs above 4. If the Nero CD-DVD disk speed uses 8 ECC, then surely its quality score is meaningless?"
PIE is always graphed at 8ECC (or an average of that), PIF is always graphed at 1ECC (or an average of that).
If your PIF is already at 12 in the graph, then the maximum in all of the single blocks it AT LEAST 12 and statistically can be much higher.
To repeat: samples may be gathered at higher intervals than 1ECC (due to drives/chipset), but they are graphed at 8ECC for PIE and 1ECC for PIF (at least in CD-DVD Speed, don't know about DVDInfoPro). This means that for any drive where sample size is bigger than 1ECC, the PIF maxima will be more than the averages graphed by CD-DVD Speed.
Hence, PIF spikes are somewhat under reported, while the averages remain truthful.
Q6: "p.s. Would a test at 1 ECC take eight times longer than at 8 ECC?"
Depends on the drive, but generally no.
It's just that not all drives support reporting of read errors at 1ECC, because the drive makers didn’t see it useful to implement it.
Q7:"Is all this quality testing really necessary? I have a couple of disks that show groups of high PIFs (in the late teens) but which pass CRC and copying tests with flying colours!"
If you only make the copies for fast day-to-day use and don't need long storage life / archiving, then it's down to you to decide.
If it works for you in day-to-day use and you don’t need to store it for longer periods, never mind about the error rates
However, if you are concerned with either:
A) long time storage of your discs
B) maximum compatibility of your burns with other drives
Then read error (digital error) rate testing can be useful, but ONLY when done properly. Single scans from LiteOn drives (or any other drive for that matter) do not prove reliably the quality of the disc or the quality of the burn.
Q8: "I have a suspicion that the only meaningful failed test is a series of tests on the same disk at 16x, 8x, 4x, 2x and 1x that shows a high PIF peak in the same spot in all test."
That is a reasonable working theory, but unfortunately somewhat incorrect.
It would only hold true for discs, where the errors are truly encoded in bit stream on the disc (the actual bits stored on the disc are erroneous).
However, such discs only appear in specially manufactured test discs and not in ordinary burning.
The other read errors (PIE/PIF/POE/POF) are the results of focus errors, tracking errors, eccentricity, reflectivity/birefringece problems, ftracking/slicer level adjustments and the jitter that results in the analog RF signal from those errors/problems. This jitter is then translated to bit-level errors in the DA stage of the drive.
To repeat: PIE/PIF/POE/POF errors are NOT on the disc. They are the results of the reading process, when the drive is inable to properly track/read/decode the analog RF signal stored on a DVD disc.
A normal read error related PIF (not chipset glitch) will almost invariably increase when the read speed is increased. This is due to increased difficulty in controlling RF jitter when the linear velocity increases.
Sure, it can be argumented that there is a problem on that part of the disc, where the spike occurs, regardless of the read speed. However, the spike level will be mostly dependent on read speed (excluding scratch/smudge/spotty polymer layer issues for now).
Q9: "Obviously, high PIE / PIF at higher speeds will slow down transfer rates, but if a small problem area of the disk can be read in its entirety for at least one of the speeds, then the disk is readable, and any slow down will be small."
This is another unfortunate and somewhat erroneous assumption, often propagated in various forums.
I have several drive/disc combos that produce excellent speed graphs (no slow-downs). However the same discs in the same drives report very high PIE/PIF rates.
Clearly the transfer speed graphs and PIE/PIF rates do not ALWAYS correlate highly.
This means that transfer rate graphs on their own are unreliable measures of readability quality of a disc/drive combo.
It is unfortunate that many people have not really tested this themselves and continue to propagate the idea that a transfer speed graph is a reliable measure of disc quality in general.
Sure, transfer rate graphs can weed out the obviously problematic discs. However, there can be serious compatibility/quality issues long before transfer speed graphs show you that there’s anything to worry about.
So, it can be a useful tool in measuring, but it’s not reliable on it’s own.
Q: "What I want to know is this – if I use PIE / PIF testing to find a media type that generally has low errors, can I rely on a combination of data verification, CRC testing and careful, scratch free, disk storage to safeguard my data?"
On the average: yes.
However there are dependencies.
First of all, to truly/reliably find a compatible media, you would need to:
Scan the burns in VARIOUS different drives (not just one!) and ensure proper low error rates in all of them.
This means that the burn is compatible in a large number of drives (i.e. it is good as far as the burn quality is concerned).
The second problem is that for longer term archiving, it's not just the initial read error rates that matter.
You need to consider the quality of the disc ingredients (particularly dye) and the quality of construction (sputtering/bonding in particular).
Good quality discs pass these requirements (dye/sputtering/bonding is almost impossible to test reliably with home grown methods, so you need to trust 3rd party independent scientific data like that from National Institute of Science and Technology).
The third problem is that, even if you find a good disc (low error rates in various of burners and very good proven disc quality overall), then you cannot be 100% sure that all discs from all batches are as good as the ones you've measured, if you only do CRC/transfer speed tests with them.
Both CRC and transfer speed test will only show you when you have really bad disc, that is when the burn quality is already at such level that the drive reading it almost cannot read the data back 1:1 bit accurately. A disc like that is already an inch away from being thrown to the garbage can.
Transfer speed rates are of course good for discs that have low read error rates with most drives. But you never know for sure if it's a good or bad disc, just by looking at CRC/transfer speed tests.
But because testing is cumbersome and sometimes difficult, as a practise it is perhaps best to find a compatible disc that:
- burns well with your burners (no coasters, low read error rates in all drives you test)
- is compatible with all the drives you want to use the discs in (no slow-downs, no read problems)
- comes from a high quality manufacturer (TY, Maxell, Ricoh, MCM, etc.)
- isn't an obvious fake (cheap imitation costing 1/3-1/6 of the real branded discs)
Then keep on using those as long as they work. However, be also aware the dye formulations and power calibration instructions (and firmware) do change with discs/drives.
This means that a disc that used to work ok, may become less compatible in the future.
If the change is big enough, you'll probably notice it as an increase in coasters or read compatibility issues.
"The only disks I can’t read reliably are Datawrite Grey (prodisk dye) on my LG combo drive and these fail CRC check. I am a person who easily becomes a little obsessed by this kind of thing and I would really like to be able to just forget about PIE/PIF quality testing every single burn I do!"
I feel your pain.
It's not easy I can tell you, but if you want to get easy out of it, then just buy the most compatible burner recommended by cdrinfo/cdrlabs/cdfreaks (they must be in unison in their findings) and the most compatible disc (don't be cheap, quality costs) and stick to those as long as they work for you.