jonahn

jadeclaw is right, "Auto" mode when using optical in is fixed 0dB gain, not AGC at all. Manual setting of 23/30 (IIRC) gives the same result - this was on a NH700.

I ran some tests on the NH700 analogue output - this post might help: http://forums.minidisc.org/viewtopic.php?t=6560 The reason for the big difference in sound and the "stray frequencies" (harmonics) in your sweep test is that, with the volume maxed out at 30/30 on the HiMD, the headphone amp is clipping. I found that the onset of clipping is around a volume setting of 27/30 (even lower at sub-100Hz frequencies because of a quirk in the amp design). And yes, the output level is pretty puny! BTW I bought my NH700 in Singapore so I assume I have the same version as you (i.e. non-European model).

Yes you're right of course, but those are compensating for the physics of the media and/or the pre-emphasis applied in recording. What I meant was, in the context of "style of sound" resulting from digital audio compression, or distortion by any other means, there is nothing that an amp can do to improve the sound once it's been damaged. Yep, I was the one who found that! :grin: Actually it's compensating for losses in the too-small output coupling capacitors (which you could consider to be part of the amp itself), not deficiencies of the earbuds.

I assume you mean, compared to playing it back through a PC, correct? I really don't think this is possible. Audio codecs are designed so that the output sounds as close to the input as possible. There is no way that you can optimise a codec to sound best on minidisc, because the fact that the data is stored on a minidisc is irrelevant to the sound quality of a codec. Whichever media you're using, whether it is minidisc, hard disk, CD, flash memory etc., is just a pure digital data storage: bits out = bits in. As far as I know, there is nothing in an audio amplifier design (or a DAC, or anything else in the audio signal path) that you can optimise to deal with a particular style of sound...

Is there really any physical difference between the MD units sold in Europe and the rest of the world? Could it be that the difference in quoted power output is just due to some European law requiring a more stringent measurement method? (e.g. power output quoted at 1%THD in Europe, 10%THD for the rest of the world...)

I can explain what it's doing to the signal. The user manual mentions that optical inputs at 32 and 48kHz are Sample Rate Converted to 44.1kHz. What I've found is that the SRC is still active, even if the incoming signal is already at 44.1kHz. This is called "1:1 SRC", and it is re-sampling the signal from the original sample clock to a separate, internally-generated 44.1kHz reference (which may differ in frequency by a few Hz from the external source). It is not changing the gain nor (IMO) deliberately sabotaging the signal to prevent exact copies. As for codecs being optimised for specific devices... I don't believe that for a second. Optimised in terms of hardware complexity and power consumption, yes, but how on earth can you optimize the sound quality for optimum playback on a minidisc / ipod etc.??

I wasn't using a soundcard :wink: From what I can tell, it looks like the unit is doing sample rate conversion on the digital input, even if the sample frequency is already 44.1kHz...

Sorry to re-open an old thread... but has this been tested any further? I've done some tests of my own and I'm also finding that PCM recording from optical input is not bit accurate - can anyone confirm or deny this?

Finally got round to re-checking with a 16 ohm load connected. It does indeed flatten out the response - still not perfect but better. The response still rises initially below 100Hz, peaking around +2dB at 50Hz then drops off below that. The shared compensation circuit causes lots of crosstalk between left and right channels at these low frequencies. The amount of compensation is reduced when left and right inputs are out of phase, but not completely eliminated. Everything I said above about the non-flat respose when using the headphone socket as a line-out, and the higher possibility of output saturation at low frequencies (with or without headphones connected), still applies!

One idea... this is a long shot but were your sweeps 180 degrees out of phase between left and right channels? Because the amp's low-pass compensation circuit is shared between both channels, so I suspect it would not work with out-of-phase signals. If not... then I don't know. With the full scale (0dB) sweeps and volume turned up to 30/30, the amp probably would have been saturated across the entire band, so the frequency response wouldn't be meaningful. But it should have been visible on the lower level sweeps. BTW the amp seems to have a soft-clipping effect when it saturates, so the clipping is not necessarily obvious when looking at waveforms. I didn't realise this at first as I was finding the saturation point by looking at THD readings on the AP. KJ_Palmer, I will have another go at checking the response with a 16 ohm load connected...

I've found out what's going on... The low-frequency boost is deliberate... sort of. The headphone amp is a TA2131, is this chip has a "low-pass compensation" feature which allows use of smaller (cheaper) output coupling capacitors. Using the smaller capacitors in combination with the headphone impedance causes a drop in low frequency response, so the amp compensates for this by boosting these frequencies. So with 16 ohm headphones connected, the frequency response should be flat (I haven't checked). But with the output connected to a high impedance load, you will see the compensated signal. BTW I assume I have the non-European version as I bought it outside of Europe... is there any visible way of checking?

Very strange! What was the lowest frequency you used in your sweeps? I'm very surprised you found no clipping even with the 0dBFS sweep... I've found that playing music with the volume turned up to 30/30, the clipping is very audible. Unless you were using the (3V) AC adapter? That might increase the output swing of the headphone amp? I was using the supplied (1.2V) NiMH battery. The method I used was optical out from an Audio Precision box into the NH700, and headphone output back into the AP analogue analyser, with the unit set to paused-record mode. The AP was set to digitally generate sine waves at 0dBFS, and the record level was set to Auto. The results I got were all as expected apart from this low-frequency gain. I double-checked that the EQ was switched off and I've never used the AVLS feature so I'm pretty sure that was switched off too. Is there anything else I could have been doing wrong?

I have access to some audio test equipment at work, so I hooked up my MZ-NH700 walkman to it to try a few things. Noticed a few interesting things, but the strangest was this: the headphone output has a frequency response that is far from flat! It is pretty flat from 100Hz to 20kHz (dropping slightly at the top end), but below 100Hz it rises sharply and reaches about +6dB at 20Hz (reference 0dB at 1kHz). This means a couple of things - firstly, if you use the headphone socket as a line-out, you will not get a signal that is true to the recording. Secondly, it means that the maximum safe output level before clipping begins is severely reduced. With a 1kHz sine wave at digital full scale, clipping begins on the headphone output at a volume control setting of about 27. At 20Hz, clipping starts with a volume setting of only 20. This applies when listening with headphones as well, so you will potentially get clipping in music with low-frequency content with the volume control set at anything over 20 - even though there is plenty of headroom left through the rest of the audio band! So unless I've made a mistake with my test, then this seems like a pretty stupid design feature to me! if they were going to put in a bass-boost feature they could have at least made it switchable... (The above test was made with the EQ switched off.)

Sounds feasible... Is was wondering if this sample difference, and/or the imprecise positioning, could also be related to the frame size used by ATRAC. HIMDRenderer allows you to specify a block size and overlap in seconds, but assuming the ATRAC frame size is 4096 samples (which I think I remember reading somewhere), then an integer number of seconds doesn't translate to a whole number of frames. Would it be worth experimenting with block sizes that are a multiple of 4096 samples instead of whole seconds? Also - am I right in thinking that the sample difference doesn't happen when converting PCM sources? If so, then the "Max allowed sample difference" could be safely set to zero - which the program currently doesn't allow. This might improve the reliability of the block lining-up process. Apologies if I'm talking complete rubbish though! :wink:

IIRC, there's an option in SS to "Delete tracks from MD after upload" or something to that effect, and I think it was enabled by default. If this option is un-checked, could it make the transfers safer, so that even if it fails you can still resort to analogue or Total Recorder capture methods? Or has anyone still had their tracks trashed even with this option switched off?

You mentioned earlier in this thread that when you step back and restart playback, the samples it gives you are not identical to the ones you got the first time around, so you need to search for near (but not exact) matches in the overlap region to stitch the blocks together. (if I've understood correctly!) So effectively, each time you request a block of audio from the Direct Show software, it gives you different results?? Any idea why this is happening?

Hi all, New here... just bought a MZ-NH700, very neat little gadget. Quick question: if I format a blank disc in standard MD mode to do some tests with it, will I still be able to re-format it to Hi-MD or will it be permanently stuck as an MD-format disc? There is a "Format" menu option, but from what I understand in the manual this is only available with a Hi-MD format disc inserted... Can anyone confirm this? Thanks in advance!

Marc, Another big "Thankyou" for the time and effort you've put into this project! An absolutely essential tool for Hi-MD owners, and I'm frankly appalled that the software supplied with the units doesn't have this function built in already. I've only tried it briefly, it did work but produced some errors in the output wav file. I think I know what might be causing them, but first a question: Why is conversion from OMA to WAV such an imprecise process, requiring all this overlap business / searching for near-matching samples etc.? Any other audio format conversion will always give exactly repeatable, unique results with no risk of repeats or clicks, so what is different about OMA files? Thanks, Jonah