[LAU] "discovery" on db50xg clones (NEC XR385); example/demos w/ analog modification

I've discovered something interesting about the db50xg clones from ebay (NEC XR-385 http://cgi.ebay.com/XR385-YAMAHA-DB60XG-DB50XG-MIDI-Synth-Daughter-Board-/2… ). You know, the rock-solid $10-$20.00 XG synths that work great with http://qtractor.sourceforge.net/ (due to ability to reuse people's cakewalk instrument defs for XG), also graphical parameter editing via http://qxgedit.sourceforge.net/ . This discovery is the result of a simple analog modification of the output of the synth-card and the input of its host soundcard, a $1.00 Dynex dx-sc51. The results can be heard by listening to the OGG encodings of some example songs, see links below.... I heard somewhere that the NEC clones of these synthcards were used in Karaoke boxes, especially for their built-in DSP-based vocal-removal, and ability to mix in beat-sync'd (to the current midi track) vocal effects and reverb ( http://www.studio4all.de/htmle/main96.html ) .... Probably for such purposes, I guess they didn't have the sound on the synth turned up to "11" but left some dynamic range for the drunken vocalist; Or it was connected to an analog amplifier and not a soundcard's A/D section that clips over 1vrms. As I mentioned before ( http://linuxaudio.org/mailarchive/lau/2010/6/28/170833 ), my analog modification of the NEC XR-385 removed the output opamps and tap directly into http://electro-music.com/forum/phpbb-files/ak4510_codec_683.pdf -- producing zero-2.80V(full scale) into 10Kohm from its "on chip output buffer with single-ended output" and "on chip post filter". That output is fed directly via large electrolytic coupling capacitor to a vt1616 codec input whose full-scale input is reached with 1 volt * 2 * sqr(2) = 2.82842712 volts peak to peak -- same as the DAC's full-scale. Prior to my analog modification, MIDI tracks had to be turned down via "master volume" ; they now play correctly at full volume. I assume other's complaints (read on the internets) about these cards clipping&distortion can be attributed to this property of the NEC XR385. But the reality is that the clipping may have been happening because of the extra gain (2x?) provided by the buffer opamp driving the waveblaster 26 pin connector analog out. It certainly wasn't observed on any full-scale waveforms I sent through it, visualized in 'yass' ( http://www.kokkinizita.net/linuxaudio/downloads/yass-0.0.2.tar.bz2 ). The bottom line is the extra output gain nor its buffering is needed, given the AK4510 DAC's builtin buffer. The original circtuitry probably served its purpose for karaoke mixing, but for its reincarnation as a synth card, it's best to give the synth, not the drunken karaoke vocalist, the full, limited dynamic range of the 16 bit DAC. Below are some example renderings of public MIDI using no volume adjustment at the MIDI level. http://nielsmayer.com/npm/Intelligent-Summer-on-analog-modded-dynex+6db.ogg http://nielsmayer.com/npm/Is_god_-_Free_demo_Copyright_Yamaha_-on-analog-mo… http://nielsmayer.com/npm/chpn_op53-on-analog-modded-dynex+15db.ogg ... http://nielsmayer.com/npm/Skyline_Herbie_Hancock_on_analog-modded-dynex+13.… Note I didn't re-do the Herbie Hancock track because it needed the volume increase at the midi level (via Kmid's volume control [ http://kmid2.sourceforge.net/ ] ). However, now, turning up the analog level on "classical" tracks, such as the Chopin above, is necessary. The downside, for quieter tracks, is that with turning up the "capture gain" amplification on the soundcard (via vt1723 and vt1616 codec) you get more noise: At +3db capture gain jnoisemeter ( http://www.kokkinizita.net/linuxaudio/downloads/jnoisemeter-0.1.0.tar.bz2 ) gives: w/ DC filter, flat response, RMS L = -83.4 R= -83.4 At +4.5db capture gain jnoisemeter gives: w/ DC filter, flat response, RMS L = -82.0 R= -81.9 At +6db capture gain jnoisemeter gives: w/ DC filter, flat response, RMS L = -80.5 R= -80.4 But that's not bad given that the synth-card's DAC has an -84db min. S/N ratio (-89dB nominal)... and the whole bricolage I constructed sits unshielded in a micro-tower case. Given those noise figures, there's no point in taking my "next step before going all digital" approach which was to make a balanced output for hookup to external DAC ( http://www.dancetech.com/article.cfm?threadid=59&lang=0 ). Without even anything connected, I measured -80db into the prospective circuitry I was considering using: an external "active transformer" (aka differential op-amps in a box) driving 20ft of Canare balanced microphone cable, into a switcher/router (passive, balanced) and into an external a/d converter. I'm better A/Ding at the source and not having to worry about converting any ground noise/loops and other hum.... other than the massive DC bias present on the signal.... As best example of my analog modification approach, I believe the above sound examples best match the full 16 bit 44.1K dynamic range of this synth-card. I hear a significant difference between the latest "on-analog-modded-dynex" versus the original non-modified ones recorded on a Terratec DMX6Fire 24/96. For comparison, here are the old ones, and original MIDI sources: http://nielsmayer.com/npm/Intelligent_Summer.ogg (orig http://www.megatrade.ru/Midi/Isummer.zip -> Intelligent Summer.mid) http://nielsmayer.com/npm/Is_god_-_Free_demo_Copyright_Yamaha_-_XG.ogg (orig http://www.blueman.name/Jazz.php -> Is god - Free demo Copyright Yamaha - XG.mid NOTE: demo of the guitar sound capabilities extolled in http://en.wikipedia.org/wiki/Yamaha_XG ) http://nielsmayer.com/npm/Skyline_Herbie_Hancock_TG300B.ogg (orig from http://www.blueman.name/Jazz.php -> "Skyline\ -\ Herbie\ Hancock\ -\ TG300B.mid" ) http://nielsmayer.com/npm/chpn_op53_coggie.ogg (orig http://www.piano-midi.de/midis/chopin/chpn_op53.mid ) -- Niels. http://nielsmayer.com