dMZX Forums: Many SAM files from Bernard the Bard play at wrong frequency -> MegaZeux Bugs -> Tracker

STR:
Watch the title screen of Bernard the Bard. There are a bunch of sound effects played at a lower frequency than they used to be in much older versions of MegaZeux, such as the crowd cheering effects. Other effects, such as the shooting sounds, are played at the correct frequency.

I've compared several of the samples that sound off (2.51 in DOSBox vs 2.91b). The problem isn't their frequency as far as I can tell.

This is SAM 1000 "KRABOOM.SAM" (the one that sounded the most different to me). The top is 2.51, and the bottom is 2.91b. DOS is inverted and plays the SAM on one channel, but they're at essentially the same frequency. If anything, 2.91b looks slightly faster.



Zoomed in, 2.51 and 2.91b. 2.91b might be losing some detail.



The .SAM file imported into Audacity, zoomed to the same spot.



There's some clipping, but it has several recognizable features that are in 2.51 and aren't in 2.91b. This is probably why 2.91b sounds lower. Maybe resampling modes?

Revvy, on 22 January 2018 - 02:07 PM, said:

Record video of both dos MZX and 2.91b on that title screen. I just listened to both and I can't tell the difference, so that would help get an idea of what you're experiencing.

Here's an MP3 showing the difference. 2.51 plays first, followed by 2.91b.

Updating status to: Cannot Reproduce

Any difference here appears to be caused by a different resampling mode being the default. Old MZX samples were low bitrate and probably are affected a lot more by resampling mode than samples used today. I can't reproduce any other condition that is altering the pitch of the effect, sorry.

Moving to: Archived MegaZeux Bugs

Moving to: MegaZeux Bugs

Updating status to: Confirmed
Updating severity to: 3 - Medium

After taking a deep dive into why sam 0 doesn't work properly for .sam files I think I figured out what's going wrong here: the old SAM to WAV conversion code was accidentally generating stereo WAV files and the overall frequency of all sound effects was halved to compensate for this (easily fixable) bug. When asie replaced the SAM to WAV converter with a proper SAM loader they started getting treated as signed 16 bit mono samples instead due to wav_read_data not having an implementation for SAMPLE_S8. This ended up sounding roughly the same as the "stereo" WAVs so it wasn't really noticed by any of us.

I still don't know why this wasn't noticeable to me before when comparing the Bernard the Bard samples, but it is definitely obvious with AD_LAZER.SAM comparing what played before with what plays with my (in progress) fix.

Updating status to: Awaiting Feedback
Updating severity to: 1 - Low

The sam 0 bug was fixed in GIT ec47ada7. Revvy clarified on Discord that he specifically remembers the samples playing at about double the speed/frequency as they currently do. I can't really reproduce this particular thing but it's specific enough that it might be worth checking e.g. some of the lesser used BWSB audio drivers for oddities. Either way this is kind of subjective and not particularly urgent.