Topic: Tascam DR-70D 4-channel audio recorder (Part 5)

[morst]

A product recall is in order.

[voltronic]

So I did some further testing.  I have a couple of SD card testing programs, although none of them write data the same way the 70D does.  The card passed them, but the write speed is much lower than what I expect: between 11 and 16 MB/s depending on the test.  That still should be plenty fast enough though.  Read speed around 33 MB/s, far lower than the stated 48 MB/s but still OK.  I haven't been able to find a program that will test using continuous writes.  All of these write data in chunks, which is a different animal, but at least it was enough to verify that I got an authentic card.

The only other device I have that takes full-size SD is my older Panasonic camera, and it worked fine for recording 720p AVCHD video after a format.

While doing all this testing on the SanDisk, I ran my old but reliable Transcend card at 24/96, 4 channels until the card was full.  No errors, checked the files and they are perfect.  As I said, I've never had a problem with this card in the 70D.

Tried to put the new SanDisk card back into the 70D, and it wouldn't let me do anything with it now: I got a "bad card" message or something to that effect, it told me I had to replace the card, and it powered down.  This was right after doing the video recording test. 

Put it back into PC, used the official SD Formatter utility to do a full/erase format.  Back to the 70D, and it formatted without incident and now seems to be working fine.  I just recorded a couple hours of 24/96, 4-channel and no errors, popped the card into the PC and verified that there are no glitches on recording either.

It's worth noting that after I had all of my problems last night I tried formatting in the 70D using both the Quick and Erase methods, and continued to get Write Timeout errors after 10 seconds.  That continued today, and the only thing that fixed it was formatting the card on my computer using the SD Formatter utility.

I'm not sure what's going on here.  Did the filesystem on my SanDisk card somehow get so corrupted that the on-board format routines couldn't fix it?  And why was it not even recognized in the 70D after formatting in my camera?  This card appears to be working fine now, but will that last?  All I can do right now is continue the same test I did before until the card is full, and see if I get any problems or this was a one-time thing.

Maybe Tonedeaf and any others who have purchased this same card would please run a continuous test also to see if this model of card really is no good with the 70D and TEAC Japan wasn't thorough with their testing, or if it's just mine.  Card model is SDSDUP-16G.

EDIT: I'm running that card in the 70D now again until the card is full to see if there are any further problems.  If I can do that 2 or 3 times with no issues, then maybe this was a fluke.  This card is going to have to prove itself worthy before I use it again to record something I really care about though.

One other thing I forgot to mention before: for the marching band show I was recording last night, the set that I cared about for our students actually came out fine.  What I do with these recordings is I replace the audio on the video track from my phone or from another camera using Vegas, and then that goes to a private share to use for the students to self-assess following each competition.  What I noticed this time was that even though this particular 10-minute recording was free of audible glitches, it drifted significantly from the video track such that I had to do a cut/crossfade to make the remainder match up.  Even without timecode sync, I never have significant drift for recordings this short - in fact it's the only time I've ever noticed a mismatch at all.  So a longer-than-normal recording may be another symptom of card write problems.

[stevetoney]

^ That's a very interesting sequence and maybe there are some clues as to what might be happening in there.  It's almost like the formatting on the card got corrupted somehow...and then the PC re-format restored the card to original or near original state.  Also, almost seems like there could be a progressive corruption...like over the course of a couple of record cycles...until the card reaches an 'unwrite-able' status. 

I'll run the test you suggested tomorrow.  A couple of questions...

By continuous test you just mean start recording to the card (I'll set it up for four channel recording) and run the card all the way until it's full?

How did you determine that there were no glitches?  Just load the files into your DAW and look for any random spiking? 

Thinking out loud here and answering my own questions...perhaps I'll set it up to record four channels of a test tone from my laptop and let it roll.  That way when I load the recorded files back into my laptop, it'll be easy to see visually if there are any dropouts or glitches on the flat/brick waveform.

[voltronic]

^ That's a very interesting sequence and maybe there are some clues as to what might be happening in there.  It's almost like the formatting on the card got corrupted somehow...and then the PC re-format restored the card to original or near original state.  Also, almost seems like there's a progressive corruption...like over the course of a couple of record cycles...until the card reaches an 'unwrite-able' status. 

That's exactly what I was thinking.  In my PC repair days, I used to see this happen all the time with failing HDDs.  What's surprising in this situation is seeing this happen with a brand new card, especially since my very old Transcend card has been perfect as I've stated many times.  The card started out performing great for several days, then all of a sudden I had glitching / skipping, then that quickly degenerated into the Write Timeout and finally not even being recognized.

By continuous test you just mean start recording to the card (I'll set it up for four channel recording) and run the card all the way until it's full?

Yes, and I'd go with 24/96 to max out the data rate.  FWIW, what I was doing last night was manually stopping and starting new tracks for each band, about 10 minutes each track.  It shouldn't make a difference if you keep it running though, and then it can be unattended, making new files after ever 2GB.  If anything, I would imagine the potential for errors would be greater in a larger file than a smaller one anyway.

How did you determine that there were no glitches?  Just load the files into your DAW and look for any random spiking? 

Pretty much, and then I spot-check listened at many places throughout the files.  I'm actually now recording 4 channels but only with 1 mic plugged in.  Here's my reasoning - when experiencing the problem before, it happened on all 4 channels (only 2 mics, but was using DUAL REC to record a duplicate at -12dB).  Knowing that, for this test, I can load all of the tracks into my DAW and quickly look for any place where the 3 "silent" channels are not flatlined electronic noise of the recorder.  But there is still one channel that recorded audio so if I do find a problem on one of the "silent" tracks, I can see if there is something audible going on with background noise in the room as a reference.  It's the same amount of stress on the SD card since there are still 4 channels recording.  Before when I was testing the Transcend card I did actually connect 4 mics, but I realized it's not necessary since that proves there's no correlation with the number of phantom powered channels or something wacky like that.

Thinking out loud here and answering my own questions...perhaps I'll set it up to record four channels of a test tone from my laptop and let it roll.  That way when I load the recorded files back into my laptop, it'll be easy to see visually if there are any dropouts or glitches on the flat/brick waveform.

That's probably an even better procedure than what I'm doing.  Thanks for trying this out.

[stevetoney]

One other thing I forgot to mention before: for the marching band show I was recording last night, the set that I cared about for our students actually came out fine.  What I do with these recordings is I replace the audio on the video track from my phone or from another camera using Vegas, and then that goes to a private share to use for the students to self-assess following each competition.  What I noticed this time was that even though this particular 10-minute recording was free of audible glitches, it drifted significantly from the video track such that I had to do a cut/crossfade to make the remainder match up.  Even without timecode sync, I never have significant drift for recordings this short - in fact it's the only time I've ever noticed a mismatch at all.  So a longer-than-normal recording may be another symptom of card write problems.

Great sleuthing work!  This is also very interesting. 

The above reminds me that that one possible answer that's been suggested is that the DR70D may have some kind of an issue with how it's writing data to the card.

You mentioned you experienced significant drift.  I wonder if it would be worth doing a test of the internal clocking on the DR70D against a calibrated clock to see how much drift there is over...say 24 hours.  I'm not sure what it would mean if the DR70D clock is way off, but if somehow there's a mismatch between the data write speed due to an mis-calibrated write and how the card is formatted...perhaps like fitting a square peg in a round hole?!?  I don't really even know the technicalities of what I'm getting at here honestly, but I do find it VERY interesting that you saw drift over a relatively short ten minute time frame when you hadn't seen it before.  (But then the question becomes...why some cards and not others?!?  LOL. )

EDIT TO ADD:  I'll try something out tomorrow.  What I'll do is start my test tomorrow and note the EXACT clock time i start, down to the second...or perhaps I'll set a stopwatch on my phone.  Then I'll do the same when I manually stop the recording just before the card fills up.  Then I'll load my files into the DAW and see how much elapsed RECORDED time the DAW says is on the recording.  I think it'll take four hours to fill the card at 24/96 and four channels, so I'll check to see how many seconds, if any, of drift there is between actual time and the time the recording shows as.

[voltronic]

One other thing I forgot to mention before: for the marching band show I was recording last night, the set that I cared about for our students actually came out fine.  What I do with these recordings is I replace the audio on the video track from my phone or from another camera using Vegas, and then that goes to a private share to use for the students to self-assess following each competition.  What I noticed this time was that even though this particular 10-minute recording was free of audible glitches, it drifted significantly from the video track such that I had to do a cut/crossfade to make the remainder match up.  Even without timecode sync, I never have significant drift for recordings this short - in fact it's the only time I've ever noticed a mismatch at all.  So a longer-than-normal recording may be another symptom of card write problems.

Great sleuthing work!  This is also very interesting. 

The above reminds me that that one possible answer that's been suggested is that the DR70D may have some kind of an issue with how it's writing data to the card.

You mentioned you experienced significant drift.  I wonder if it would be worth doing a test of the internal clocking on the DR70D against a calibrated clock to see how much drift there is over...say 24 hours.  I'm not sure what it would mean if the DR70D clock is way off, but if somehow there's a mismatch between the data write speed due to an mis-calibrated write and how the card is formatted...perhaps like fitting a square peg in a round hole?!?  I don't really even know the technicalities of what I'm getting at here honestly, but I do find it VERY interesting that you saw drift over a relatively short ten minute time frame when you hadn't seen it before.  (But then the question becomes...why some cards and not others?!?  LOL. )

Well I've only ever seen this drift on this card, and only this one time.  I recorded a competition last week using this recorder / card combo, and there was zero discernible drift between the video and my separate audio track, and the recording was good.  And all of my previous recordings with the 70D using my other card have had no drift compared against the videos from prior weeks. 

Given that it's only happened with this card, I really don't think the clocking of the 70D is is suspect here - I'm suggesting that previously unseen clocking errors may be another symptom to look out for.  Tracks I recorded last night before this one where I found the drift did have occasional digital noise and skips, and they may very well have had clocking errors as well - I just didn't have video tracks to go along with those to try and sync them with. 

The problem with testing the 70D clocking is there is no timecode, so I'm not sure you can even test this properly.  It's not like your F8, like the test that was done where they put one in a freezer overnight with a 744 and checked the clock differences after.

I've been recording to the suspect SanDisk card for almost now and it's still going along fine (I think).  We'll see what the files look like tomorrow.

[voltronic]

EDIT TO ADD:  I'll try something out tomorrow.  What I'll do is start my test tomorrow and note the EXACT clock time i start, down to the second...or perhaps I'll set a stopwatch on my phone.  Then I'll do the same when I manually stop the recording just before the card fills up.  Then I'll load my files into the DAW and see how much elapsed RECORDED time the DAW says is on the recording.  I think it'll take four hours to fill the card at 24/96 and four channels, so I'll check to see how many seconds, if any, of drift there is between actual time and the time the recording shows as.

That's a good plan.  In reality, you'll only probably get a little less than 4 hours, since you don't really have 16GB to work with.  The formatted capacity for my card shows up as 14.8 GB in Windows.  The Sound Devices calculator says that's 3:47 for 15 GB, so figure maybe 3:40.

[stevetoney]

Right.  I understood all those points. 

I'll try to explain better if i can what my thoughts are about the timing.  At minimum, it seems telling to me that you saw the time drift coincident with when you were experiencing the other issue on this card, yet you said that at the same time this portion of recorded audio didn't have any actual diginoise or skipping.  So as you point out, all of these issues seem to be symptoms of whatever is 'the problem'.  To me what's new is that perhaps 'the problem' can be better diagnosed knowing that there seems to be some data write drift or mis-match.  To my way of thinking, this might not be unlike what happened in cassette days when the deck used to record from wasnt calibrated right...to 1.8 inch per second or whatever... And then you play the cassette back on a different deck and it sounds either faster or slower. 

All recorders have some type of a clock... Not timecode... But a clock that keeps time and date information.  If somehow that clock gets corrupted and there's a mismatch between how much data is supposed to be written to the card (48000 bytes per second or whatever) versus how much is actually written, then you might see drift if there's a difference between the rate of data write vs the rate of data read or playback.  Or if the write rate is too high, then perhaps you'd get data bottlenecks with actual periodic losses of data, with the possible result being skipping or diginoise.

Again i don't know how formats happen but perhaps this all goes back to bad card formatting... If card sectors aren't properly segmented... Like if somehow the 70d associates data write time To a formatting segment size... or something like that

Just some thoughts without any real technical basis.

[voltronic]

Right.  I understood all those points. 

I'll try to explain better if i can what my thoughts are about the timing.  At minimum, it seems telling to me that you saw the time drift coincident with when you were experiencing the other issue on this card, yet you said that at the same time this portion of recorded audio didn't have any actual diginoise or skipping.  So as you point out, all of these issues seem to be symptoms of whatever is 'the problem'.  To me what's new is that perhaps 'the problem' can be better diagnosed knowing that there seems to be some data write drift or mis-match.  To my way of thinking, this might not be unlike what happened in cassette days when the deck used to record from wasnt calibrated right...to 1.8 inch per second or whatever... And then you play the cassette back on a different deck and it sounds either faster or slower. 

All recorders have some type of a clock... Not timecode... But a clock that keeps time and date information.  If somehow that clock gets corrupted and there's a mismatch between how much data is supposed to be written to the card (48000 bytes per second or whatever) versus how much is actually written, then you might see drift if there's a difference between the rate of data write vs the rate of data read or playback.  Or if the write rate is too high, then perhaps you'd get data bottlenecks with actual periodic losses of data, with the possible result being skipping or diginoise.

Again i don't know how formats happen but perhaps this all goes back to bad card formatting... If card sectors aren't properly segmented... Like if somehow the 70d associates data write time To a formatting segment size... or something like that

Just some thoughts without any real technical basis.

I'm with you on all of that, and I agree with your cassette analogy.  When I was saying "clock" I was actually referring to the oscillator on a chip that all of these things have to set their sample rate, etc. which sounds like what you're talking about also.  What you describe as the possible data rate mismatch is exactly what I think is happening.  The question is - what is causing it?  A formatting / filesystem snafu, or something else?

Back to testing stuff: ran the card last night until full as I mentioned, and no problems.  I loaded each of the files into RX as it is a much more powerful tool for visual analysis, and took a couple of screenshots you can see below.  I had one Naiant X-Q plugged into Ch 1, and the other 4 channels were recording but nothing plugged in, and phantom turned off.  You'll see some periodic noise in the spectrum around 30kHz for Ch 1, but it is inaudible.  The other little noises are actual environmental sounds.  No glitches or dropouts anywhere in any of the files.

I'll format and run the same thing again tonight.

[stevetoney]

I'll format and run the same thing again tonight.

Your successful restoration of the Sandisk card on your PC tends to make me want to do my card formats on my PC and not the DR70D.  If there's a possibility that the format process in the unit is glitchy, perhaps conducting multiple formats in the unit would be a way to simulate use over time since many/most of us are in the habit of quick formatting before use.  Obviously, one of my main goals is to figure out/isolate where the issue lies and, if possible, make it repeatable.  This would eliminate uncertainty about the source of the issue and restore my confidence in using the unit.

[Life In Rewind]

Regarding the drift with no apparent errors...

Sounds like just another version of "fail" - just not to the point where the PCM decode fails completely (and makes noise/static). Possibly just another manifestation of the same problem.

Also - I know you want to fill the card for the tests...but I think you should start and stop a few times - and power cycle the deck...things we all do when using the deck...

The one thing we almost never do - is fill the card...and it certainly doesn't sound like there is any correlation in reports so far. I'd get the card -close- to full - like a normal use pattern.

Voltronic - are you certain all your errors are happening real time? Myself and perhaps one other member have had previously good files get corrupted once a card starts failing... with the same problems the subsequent files have.

I mention that only because I'm not sure we know if these errors are related to real-time processes - or something wrong with shut down...boot - or other action - that corrupts the files.

[Life In Rewind]

I'll format and run the same thing again tonight.

Your successful restoration of the Sandisk card on your PC tends to make me want to do my card formats on my PC and not the DR70D.  If there's a possibility that the format process in the unit is glitchy, perhaps conducting multiple formats in the unit would be a way to simulate use over time since many/most of us are in the habit of quick formatting before use.  Obviously, one of my main goals is to figure out/isolate where the issue lies and, if possible, make it repeatable.  This would eliminate uncertainty about the source of the issue and restore my confidence in using the unit.

That's been my feeling as well. We tapers probably format more than most users. Probably the first thing we all did - grab a good card from the pile, hit FORMAT

The introduction of the ERASE FORMAT function in later firmwares gives a clue they know something is deficient.

[Life In Rewind]

It's worth noting that after I had all of my problems last night I tried formatting in the 70D using both the Quick and Erase methods, and continued to get Write Timeout errors after 10 seconds.  That continued today, and the only thing that fixed it was formatting the card on my computer using the SD Formatter utility.

This is powerful evidence that the TASCAM Format functions are not up to snuff...or the SD Card controller could be buggy. Not sure the latter is fixable.

[Life In Rewind]

One other tidbit.

I tried to use my "bad" card to transfer a few photo files from one Windows PC to another.

The files copied to the card - but when they were transferred to the other PC there were quite a few that were corrupted (only half the photo showed up).

[stevetoney]

One other tidbit.

I tried to use my "bad" card to transfer a few photo files from one Windows PC to another.

The files copied to the card - but when they were transferred to the other PC there were quite a few that were corrupted (only half the photo showed up).

Did you reformat the card in your pc prior to doing this?