Hi,
as others before have already mentioned the problem is polyphony related. With a modern SF2 soft synth this Midi file easily produces 64+ voice polyphony that is more than double what early SC devices can produce. Besides the limited polyphony SC-55/SC-55 mkII has a default voice reserve algorithm that is not ideal and perform particularly bad with this title. Namely the default voice reserve algorithm gives more preference for 1st Midi channel and neglects any channels after channel 10. But this Midi file gives strong emphasis on Channel 11-14 what is more uses a 2-voice instrument (French Horn) on Channel 13.
You can try the following GS SysEx message (after a GS Reset) at least on your SC-55 mkII:

1F0 41 10 42 12 40 00 7F 00 41 F7
2F0 41 10 42 12 40 01 10 02 02 02 02 02 02 02 02 02 02 02 02 01 01 01 01 13 F7

It distributes the voices more evenly this way :

More Info:
Re: Falcosoft Soundfont Midi Player

You can use FSMP to test the SysEx message directly:
Right click on the first aid like Reset button at the right hand side of the player. Then choose 'Select/Send Custom SysEx'. In the opening dialog select 'Use Hex String' then paste the above SysEx messages. Close the dialog and then Right click on the first aid like Reset button again and make sure that 'Custom SysEx' option is selected. Now the custom SysEx is sent each time you load and play the song but you can also send the SysEx by pressing the Reset button manually.

@Edit:
Instead of the above SysEx (that is more appropriate for generel usage) you should try this instead since your Midi file does not use channels 15 and 16 at all:

1F0 41 10 42 12 40 00 7F 00 41 F7
2F0 41 10 42 12 40 01 10 02 02 02 02 02 02 02 02 02 02 02 02 02 02 00 00 13 F7

@Falcosoft, is there a way in FSMP to detect how much polyphony a track uses?

Spikey wrote on 2022-01-30, 19:33:

@Falcosoft, is there a way in FSMP to detect how much polyphony a track uses?

Hi,
If you use an external Midi Out port then on Visualization (Channels) dialog you can see the count of active Note On messages.
But active voices/polyphony count can only be displayed when you use the internal Bassmidi SF2 soundfont synth. This is because when you use a standard Midi out port/external synth the player has no idea how many layers/samples/voices an external synth uses to create a sound at a given pitch for a given instrument. The player only knows that it sends a Note On message. It entirely depends on the external synth and it is unknown by the player whether this note uses 1 or 2 or 4 voices of the synth. It is also unknown by the player how an instrument reacts to a Note Off message (many instruments still use sustained voices even after a Note Off is sent).

With the help of Munt VSTi you can see the active partial count that is closely related to active voices (and maybe can hint more about a real gear than Bassmidi's voice count).
The above Midi file produces at some places 90+ active partials when you paly it in GM dual synth mode of Munt VSTi.

Edirol Virtual Sound Canvas has a peak polyphony meter.

The peak polyphony of everquest_theme.mid is 77.

Stretch wrote on 2022-01-30, 20:02:

Edirol Virtual Sound Canvas has a peak polyphony meter.

The peak polyphony of everquest_theme.mid is 77.

Thanks, it's very informative but somewhat misleading since maximum polyphony is NOT an objective property of a Midi file/track. Maximum polyphony (or even actual while playing a Midi file) is only meaningful in the context of a given concrete synth.
The maximum number of Note On messages (for a given time slice) is an objective property of a Midi track but polyphony is not.

BTW the maximum polyphony of everquest_theme.mid when you play it with Yamaha S-YXG50 VSTi is 71.

@Edit:
Here is a demonstration about the problems of counting polyphony (without knowing the exact implementation of a synth):
https://youtu.be/P6tqLA0V9u4

Falcosoft wrote on 2022-01-30, 20:16:

Thanks, it's very informative but somewhat misleading since maximum polyphony is NOT an objective property of a Midi file/track. […]
Show full quote

Stretch wrote on 2022-01-30, 20:02:

Edirol Virtual Sound Canvas has a peak polyphony meter.

The peak polyphony of everquest_theme.mid is 77.

Thanks, it's very informative but somewhat misleading since maximum polyphony is NOT an objective property of a Midi file/track. Maximum polyphony (or even actual while playing a Midi file) is only meaningful in the context of a given concrete synth.
The maximum number of Note On messages (for a given time slice) is an objective property of a Midi track but polyphony is not.

BTW the maximum polyphony of everquest_theme.mid when you play it with Yamaha S-YXG50 VSTi is 71.

@Edit:
Here is a demonstration about the problems of counting polyphony (without knowing the exact implementation of a synth):
https://youtu.be/P6tqLA0V9u4

Thanks, the video makes it easier to understand the problems with counting polyphony. Now, I want to install my sound blaster live and determine if I can find the peak polyphony with the same midi file.

Heres a recording on original hardware, CT3900 in AWE32 mode (620) with the SYNTHUSR.SF2 loaded. (No added effects, Bass/Treble on 0dB)

Spikey wrote on 2022-01-30, 17:30:

Interesting, thanks for educating me. Composed for an AWE32 seems unlikely in 1999, though? (see below for answer)
I grew up with an AWE32, and was constantly frustrated at how bad it sounded, especially compared to the SC-55! This is better than I ever heard it sound, although I'm not crazy about the effects/EAX processing/whatever it is.

I sold my AWE32 years ago because I didn't care for how it sounded either. I did buy an AWE64 later since it didn't take up as much space in my collection. I regret this decision as it makes it difficult to do my own recordings for testing purposes. Used market prices have gone stupid and it's just not worth it to me to buy another one for what people are asking nowadays.

If I wanted to record this MIDI file on my AWE64, does it support these sound fonts? Does the GM synthesizer sound any different than the AWE32?

Spikey wrote on 2022-01-30, 12:32:

The two I would limit your scope to would be the SC-55 and SC-55ST, which again should be the same other than minor quality differences/polyphony. Those should reproduce a 95-99% identical result. Right now there's too many synths and too many variables.

I think you mean the SC-55mkII and the SC-55ST?

Spikey wrote on 2022-01-30, 12:32:

The conclusion is more likely, after all, to be "some minor mistake by Kahenraz" than "half his synths are broken or randomly acting strange".

I spent a lot of time perfecting a consistent routine for these recordings and I welcome someone to do better and challenge my findings. The more data the better. But I will look into this further as your arguments and suggestions are reasonable.

I am having a lot of trouble getting a consistent recording. There are subtle differences no matter how I reset the device; by a hard reset or a GS/GM reset.

Here is the reset .mid file I am using. See the following posts for a collection of various recordings. I'm splitting the next information drop into several posts because there is a limit to how many attachments can exist for any given post.

These three recordings were made after a full power-on reset of the Roland Sound Canvas SC-55ST. This was done by removing the power cable, pressing the power switch twice, then reconnecting the power. There is no battery inside the SC-55ST so this is not only a full reset but will set the device to whatever defaults it is configured with at startup.

These three recordings were made sequentially. Between each recording the gs-reset.mid was played. No other tracks were played on this device after powering it on.

Here are three excerpts from the track that was from the full power-on reset. There are subtle but clear differences between each one. They are difficult to spot when listening to the entire track by itself unless you are already familiar with how certain parts should sound. I have isolated a very small portion to emphasize the difference. It may take several listens to spot but they're there.

I don't see why this track would play differently hardware resets. No other midi files were played between each sequential reset and subsequent recording.

These last recordings are the same part but from the recordings which had gs-reset.mid played between. No power-on reset was made between these recording.

These are subtly different between themselves and the previous recordings.

Kahenraz wrote on 2022-01-31, 07:30:

I am having a lot of trouble getting a consistent recording. There are subtle differences no matter how I reset the device; by a hard reset or a GS/GM reset.

Here is the reset .mid file I am using. See the following posts for a collection of various recordings. I'm splitting the next information drop into several posts because there is a limit to how many attachments can exist for any given post.

No matter what OS you use timers are not infinitely precise so when a player sends Midi messages to ports real time there can be some milliseconds jitter. Together with the voice reserve algorithm and this particular Midi file this can result in some subtle differences.
Have you tried the optimized voice reserve SysEx what I suggested before? (try it 1st with your SC-55mkII).

Falcosoft wrote on 2022-01-31, 07:44:

No matter what OS you use timers are not infinitely precise so when a player sends Midi messages to ports real time there can be some milliseconds jitter. Together with the voice reserve algorithm and this particular Midi file this can result in some subtle differences.
Have you tried the optimized voice reserve SysEx what I suggested before? (try it 1st with your SC-55mkII).

Not yet. Why do you suggest trying it first with the SC-55mkII? Can you tell which part of the track might be the best place to listen for a change?

These recordings are a baseline, as suggested by Spikey, to rule out anything inherited from other MIDI files between recordings. This seems to confirm that, at least for this one track, there are issues playing this file on an SC-55ST. Whether the same problems exist on the other synthesizers is unknown. It's possible that the other recordings were just lucky enough to sound better at that location for that particular recording.

I actually have two SC-55STs; a black and a white one. I added "z197763w" to the filenames of the most recent recordings which includes the serial number and color (white) to help me compare them later just in case the issue is found to be with this specific synthesizer.

Is there any randomization involved in the default voice reserve algorithm? does it always apply the exact same choices, when the same MIDI is played repeatedly?

Shreddoc wrote on 2022-01-31, 08:10:

Is there any randomization involved in the default voice reserve algorithm? does it always apply the exact same choices, when the same MIDI is played repeatedly?

I don't know if there is a way to verify this other than subjectively. It would appear through the recordings I just made that there is some kind of randomness which affects the performance to some degree, depending on the demand.

Shreddoc wrote on 2022-01-31, 08:10:

Is there any randomization involved in the default voice reserve algorithm? does it always apply the exact same choices, when the same MIDI is played repeatedly?

The algorithm itself stays the same, no randomness is involved. But when the algorithm works it can result in different outcome depending on the order and timing of incoming Midi messages. The order is usually guaranteed by most Midi players but the exact timing is not.

@Kahenraz:
I have made 2 versions of the original Midi file for easier testing. Both contain GS reset at the beginning in the Midi file itself so you do not have to bother with separate initialization step.
The 1st one only contains the GS Reset (everquest_theme_GS.mid).
The second one also contains the modified voice reserve SysEx (everquest_theme_GS_mod.mid).

BTW, What Midi player do you use for testing? Mine is using QPC timer so it is as precise under Windows as it can be. Please, try it if it makes any difference:
https://falcosoft.hu/midiplayer_61_test.zip

Falcosoft wrote on 2022-01-31, 08:22:

BTW, What Midi player do you use for testing? Mine is using QPC timer so it is as precise under Windows as it can be. Please, try it if it makes any difference:
https://falcosoft.hu/midiplayer_61_test.zip

I used Windows Media Player as included in Windows 10 for all previously attached recordings.