rasz_pl wrote:

Review said distorted, not pausing/choppy. 200KB/s is not too much to ask from a HDD in 1993. Not to mention same thing happens if you try playing midi at the same time.

There's obviously some sort of issue or conflict, as distortion that somehow occurs only during 16-bit, 44.1kHz PCM playback isn't normal, else every review and anecdotal experience would mention it otherwise. It's certainly not something I've ever encountered. You?

Edit: Sorry, that's recording that they're mentioning with regard to the distortion, not playback.

Duke Nukem 3D stuttering with Dreamblaster X1 Duke Nukem 3D stuttering with Dreamblaster X1
+ typical SB16 hiss complaints in same thread

The MIDI "stuttering" issue is completely unrelated to anything being discussed here, and the "hiss" can be largely mitigated. Anything else?

Here's a 16bit 44.1KHz mono file to do a test with any sound card to see if just LSB can be detected from the final output as a 1KHz tone. Since the tone is a squarewave you will have every 3rd harmonic visible too on the spectrograms. If you cannot see any harmonics at all then the input sensitivity might not be sufficient on the measuring device or that LSB really did get eaten up for some reason and you don't have 16bit output happening.

Well, how about this: http://www.fte.com/WebHelp/AES/content/docume … ynamicRange.htm
The amount of bits of a DAC determine its dynamic range.
Plug the values into the formula, and you know what dynamic range you should be getting, theoretically.
So:
8-bit: 20*log10(1/((2^7)-1))) = -42 db
12-bit: 20*log10(1/((2^11)-1))) = -66 db
16-bit: 20*log10(1/((2^15)-1))) = -90 db

Now, you can put a 16-bit CODEC on a card, but if the signal-to-noise ratio is around 66 db, effectively you won't get more than 12-bit resolution out of it, because the least-significant bits get lost in the noise.
I believe early SB cards have trouble exceeding 70 db signal-to-noise, which might explain why purists might call it 12-bit instead of 16-bit audio.
I suppose the Adlib Gold is on the other end of the spectrum: physically has a 12-bit DAC, but a clean signal path, so you do get all your 12 bits out of there. Might actually sound better than a noisy 16-bit sound card in practice.

A quick google dug up this test:
https://books.google.nl/books?id=eMKimy4DFaEC … ty%20db&f=false
It includes two AWE32s and an SB16 value, none of which are anywhere near 90 db.
So I suppose the really bad SB16s would be around 70 db as well, or worse.

That's pretty helpful. As relates to SnR, I'd posted this in a different thread a few months back:

Keyboard Magazine published a set of comprehensive soundcard test results in late 1994, arriving at the following signal-to-noise ratios for the cards involved:

85.51 dB - Turtle Beach Monterey
80.09 dB - Antex Z1
78.81 dB - Roland RAP-10
78.29 dB - AWE32 (CT2760)
77.40 dB - Ensoniq Soundscape
75.82 dB - Media Vision Premium 3-D
74.48 dB - SB16 MCD (CT1750)
65.32 dB - Logitech Soundman Wave
63.16 dB - Wearnes Classic 16

These weren't even optimized tests, but rather, "out-of-the-box" installs/configurations, coupled with maximized Wave and Master mixer-volume settings. Where the SB16 MCD and AWE32 share a similar architecture, the discrepancy between the two is largely explained by the amplified output being used on the former, and the line-level output on the latter.

Even with just those results, we're still looking at "~14-bit output."

Tiido wrote:

Here's a 16bit 44.1KHz mono file to do a test with any sound card to see if just LSB can be detected from the final output as a 1KHz tone.

Thanks, Tiido. Time-permitting, I'll grab some measurements from a few of my own cards.

Cloudschatze wrote:

74.48 dB - SB16 MCD (CT1750)
Even with just those results, we're still looking at "~14-bit output."

74.48 / 6.02 = ~12

rasz_pl wrote:

74.48 / 6.02 = ~12

Care to explain? I understand what you're getting at, but don't think it applies.

Cloudschatze wrote:

rasz_pl wrote:

74.48 / 6.02 = ~12

Care to explain? I understand what you're getting at, but don't think it applies.

https://en.wikipedia.org/wiki/Signal-to-noise … tio#Fixed_point

rasz_pl wrote:

https://en.wikipedia.org/wiki/Signal-to-noise … tio#Fixed_point

Right, but the context for that formula is an analog-to-digital conversion process.

rasz_pl wrote:

Cloudschatze wrote:

rasz_pl wrote:

74.48 / 6.02 = ~12

Care to explain? I understand what you're getting at, but don't think it applies.

https://en.wikipedia.org/wiki/Signal-to-noise … tio#Fixed_point

Given frequency response (typically) can vary by quite a bit across the entire frequency spectrum, I wonder how they arrived at 74 .48dB? (I think to keep things fair it could be averaged across the range, but bounded between say ~40 and 18,000 Hz). Here's a review of Creative's Sound Blaster Z showing a relatively flat response curve with a 108 dB of dynamic range.

Cloudschatze wrote:

rasz_pl wrote:

https://en.wikipedia.org/wiki/Signal-to-noise … tio#Fixed_point

Right, but the context for that formula is an analog-to-digital conversion process.

Formula for input to ADC and reconstructed output from DAC is the same
https://www.maximintegrated.com/en/design/too … -conversion.cfm
https://www.analog.com/media/en/technical-doc … 805090AN282.pdf

I think you have to take into account the state of PC audio at the time. I hear a lot of complaining about the effective bit-depth and noise floor of cards that were being sold at a time where nobody had the disk space to store 16-bit 44kHz PCM audio clips anyway. I was clamoring for "CD-quality" DACs at the time, but mainly so I could play my MOD files with 16-bit mixing at 44kHz. I wanted those 8.3kHz samples mixed to the highest quality possible, to bring out all the nuance of the source audio samples, which were mostly recorded via a potato (that is, an Amiga.)

Even by the time I had an AWE32, I was using up a non-trivial amount of disk space to keep a ripped track from a friend's CD -- in 22kHz stereo ADPCM. I'm sure that 70-ish dB SNR was *MORE* than adequate. The noise floor wasn't black, compared to the better-than-I-deserved stereo I was using at the time, but when you were watching cinematics in a game, and the audio faded out with what sounded like thighs rubbing together in corduroy pants, the hardware was hardly the bottleneck.

Creative certainly had their issues, but c'mon guys. It isn't going to stand up to a Metric Halo LIO-8 for sure, but jesus... it was the 90s.

rasz_pl wrote:

Formula for input to ADC and reconstructed output from DAC is the same

I remain unclear on the general applicability of that formula. Assuming that it does apply though, a statement along the lines of, "under certain conditions, the effective number of bits represented by the analog output of a SB16 can be as low as 12," is likely fair.

Even so, and to SirNickity's point, when referenced within the context of a larger comparison, the common disparagement intent is mostly lost:

85.51 dB, ENOB ~14 - Turtle Beach Monterey
80.09 dB, ENOB ~13 - Antex Z1
78.81 dB, ENOB ~13 - Roland RAP-10
78.29 dB, ENOB ~13 - AWE32 (CT2760)
77.40 dB, ENOB ~13 - Ensoniq Soundscape
75.82 dB, ENOB ~13 - Media Vision Premium 3-D
74.48 dB, ENOB ~12 - SB16 MCD (CT1750)
65.32 dB, ENOB ~11 - Logitech Soundman Wave
63.16 dB, ENOB ~10 - Wearnes Classic 16

SirNickity wrote:

I wanted those 8.3kHz samples mixed to the highest quality possible, to bring out all the nuance of the source audio samples, which were mostly recorded via a potato (that is, an Amiga.)

Allow me to straighten some facts:
1) There is no 'fixed' samplerate for instruments in Amiga mods. Sure, early samples may date from the late 80s (think ST-00), and due to a combination of lack of diskspace/internal memory and low quality samplers, these may not have been too high quality, and effectively around 8.3 kHz.
However, later MODs have much better quality samples, so this was certainly not through any inherent limitation of the Amiga or the MOD format.
2) The Amiga had no audio recording whatsoever, so whatever the samples were recorded with, the Amiga can not be blamed for the quality or lack thereof.

Aside from that, you are drawing a direct relation between 16-bit audio and PCM-encoded music. As your own MOD example already proves, there's more to 16-bit audio than just having prerecorded PCM data.
There's plenty of PC trackers that supported 16-bit audio, including in the instrument data. There's also been a thread on here a while ago about games that use 16-bit sound effects.
Aside from that, the early 90s were also the time when CD-ROM became a thing, and games could include full audio tracks and videos. Need For Speed is a fine example of a game that uses the CD-ROM to show video and play audio, which it does in 16-bit, and no it's not CD-ROM audio. They use a custom format that is streamed from CD and played through the sound card.

So 16-bit audio was a legitimate step-up from 8-bit. The problem with the argumentation here is that some sound cards did 16-bit audio significantly better than others. Which is why I don't buy into the '70 db SNR is enough'-argument. Might be enough for some, but once you hear a great 16-bit card, you can't un-hear it.
Besides, it's not just the SNR... The overall audio on SB cards on that era is just... unbalanced, because of their poor eq'ing of the audio.
I have various GUS cards, and also an AWE32, but playing the same tracker songs on the cards makes a world of difference. The GUS sounds full and punchy, where the AWE32 is tinny and 'scooped'.
I also have various Amigas, so I know what MODs *should* sound like. Amigas have quite decent sound circuitry with a reasonably flat frequency response, and the GUS has the same basic characteristics, but just offers 16-bit samples if you want them (the Amiga as you may know, has 4 actual DACs so performs its 'mixing' in the analog domain, making its effective sound more than 8-bit resolution. Theoretically you get about 14-bit resolution out of the 4 channels combined, taking the 65 volume levels per channel into account).
The AWE32 just doesn't compare. It has too much bass, virtually no midrange, and the high frequency range is just too 'fizzy'.

And the Amiga is 80s technology. 90s tech should sound better... Some of it did, but some 90s PC sound cards were still quite horrible.
Here's some nice recent Amiga music (in 64k): https://youtu.be/tshGRUDvsUs
Oh, and mustn't forget this one: https://youtu.be/WkHl3VoP-W8

SirNickity wrote:

(snip)
Creative certainly had their issues, but c'mon guys. It isn't going to stand up to a Metric Halo LIO-8 for sure, but jesus... it was the 90s.

Couldn't agree more. The early sound card industry was literally breaking new ground, and was very affordable for what it offered.

As proof by all the 80s and 90s CD "remastered" editions being released 30 years later (and still on CD), even the state of the art commercial audio engineering at the time was insufficient to capture the full dynamic range offered by CDs with a noise-floor at or below -96db.

Appreciations out of the way, comparing the cards at the time with hard measurements is what it's all about and I'm looking forward to the results!

krcroft wrote:

Couldn't agree more. The early sound card industry was literally breaking new ground, and was very affordable for what it offered.

I couldn't disagree more.
PC hardware was way overpriced, and underperforming (as I said before, I paid 600 guilders for my Sound Blaster Pro 2, same as I paid for my brand-new Amiga 600 at around the same time, and the Amiga 600 had better sound quality).
It also wasn't breaking new ground, other platforms had better and more mature sound solutions long before sound cards on PC took off.

krcroft wrote:

As proof by all the 80s and 90s CD "remastered" editions being released 30 years later (and still on CD), even the state of the art commercial audio engineering at the time was insufficient to capture the full dynamic range offered by CDs with a noise-floor at or below -96db.

I think it's exactly the opposite:
Loudness wars mean that old masters of CDs sound 'bad' (as in way softer) compared to contemporary music.
So CDs are remastered to 'spice them up' a bit (read: compress the hell out of them). They may sound very 'full', but they will use less of the dynamic range than the original masters. They just concentrate most of the energy in the top of the dynamic range.

Scali wrote:

I couldn't disagree more. PC hardware was way overpriced, and underperforming (as I said before, I paid 600 guilders for my Soun […]
Show full quote

krcroft wrote:

Couldn't agree more. The early sound card industry was literally breaking new ground, and was very affordable for what it offered.

I couldn't disagree more.
PC hardware was way overpriced, and underperforming (as I said before, I paid 600 guilders for my Sound Blaster Pro 2, same as I paid for my brand-new Amiga 600 at around the same time, and the Amiga 600 had better sound quality).
It also wasn't breaking new ground, other platforms had better and more mature sound solutions long before sound cards on PC took off.

The same way Starbucks didn't break new ground with their urban franchised "fast-coffee" business model in the 70s and 80s, because early middle eastern shops predate them by 500 years - and their coffee is still vastly superior? I see both points.

To be more specific about breaking new ground, I'm talking add-on cards that significantly changed the intended use. Business Machines: back when the early PCs targeted the driest of small and medium sized business purposes and little more. A small accounting shop could now justify the expense to store client's records digitally like Schwab did on their mini computers. Yes, consumers still saw value and paid excessive "business prices" to bring these things into their homes. In parallel, the Amiga was targeted from day-1 as an in-home family computer, with those deliberate furnishings already thought through - great graphics and sound, which helped the Amiga branch into profession video/audio houses.

Scali wrote:

I think it's exactly the opposite: Loudness wars mean that old masters of CDs sound 'bad' (as in way softer) compared to contemp […]
Show full quote

krcroft wrote:

As proof by all the 80s and 90s CD "remastered" editions being released 30 years later (and still on CD), even the state of the art commercial audio engineering at the time was insufficient to capture the full dynamic range offered by CDs with a noise-floor at or below -96db.

I think it's exactly the opposite:
Loudness wars mean that old masters of CDs sound 'bad' (as in way softer) compared to contemporary music.
So CDs are remastered to 'spice them up' a bit (read: compress the hell out of them). They may sound very 'full', but they will use less of the dynamic range than the original masters. They just concentrate most of the energy in the top of the dynamic range.

Absolutely, for those audiences and genres that unfortunately respond with increased sales to winners of the loudness-wars.

I'm talking about the field involving extracting higher fidelity sampling of the original recording medium, modeling or measuring the recording pathway's noise as a point distribution function and backing it out of the data using deconvolution, and signal processing algorithms vastly superior than the Dolby NR switch I used to toggle on my Dad's amplifier in the 80s.

krcroft wrote:

The same way Starbucks didn't break new ground with their urban franchised "fast-coffee" business model in the 70s and 80s, because early middle eastern shops predate them by 500 years - and their coffee is still vastly superior? I see both points.

That is not an analogy to what I was talking about.
PC sound cards generally used off-the-shelf components, which had also been used in various other computers already. Hardly groundbreaking in any way, more like catching up.
And as I said, selling them at higher prices to boot.

krcroft wrote:

To be more specific, I'm talking add-on cards that significantly change the mode of use wasn't originally intended: "business machines", back when the early PCs targeted the driest of business purposes and little more.

I'm pretty sure that this was already changed by IBM themselves in the PCjr, and brought to the masses by Tandy, long before soundcards took off.
The PC was pretty well-established as a gaming platform by the time the first soundcards became successful, in the late 80s to early 90s.

krcroft wrote:

People saw their value and paid "business prices" to bring these things into their homes. The Amiga was targeted from day-1 as an in-home family computer, with those deliberate furnishings already thought through - great graphics and sound.

That's an interesting way of looking at things, seeing as the PC is the exception here, not the Amiga. Most other computers, including ones for 'professional use', such as the Apple Macintosh, at least had some kind of onboard sound chip.

krcroft wrote:

I'm talking about the field involving extracting higher fidelity sampling of the original recording medium, modeling or measuring the the noise as a point distribution function and backing it out of the data using deconvolution (imagery/film side of the house), and signal processing algorithms vastly superior than the Dolby NR switch I used to toggle on my Dad's amplifier in the 80s.

Is that even a thing in the audio world? I mean, my dad had one of the first CD players back in the early 80s. I have some of the very first CDs ever made, in their original mastering. Noise was never an issue. Even AAD-mastered CDs generally sounded very clean, because the high-quality multitrack tape machines that they used, could easily reach 80+ db SNR.
I've never even heard anyone mention noise in relation to remastered CDs. All remasterings I've ever heard were mainly 'louder'/'punchier' than their originals.
That is probably also what led to the 'vinyl sounds better'-myth: because vinyl has some very strict limits on dynamic range, you *have* to use a significant amount of compression during mastering, to keep the needle from jumping out of the groove.
As a result, vinyl tends to have a 'loud'/'fat' sound.
Early CDs were probably mastered in a way to maximize the dynamic range that was now on offer (they sound 'softer', so I ripped a few to disk back in the day, and studied the peaks, and found that the peaks are not the problem, it's the average (or root-mean-square if you like) that's lower). I feel the same about the Commodore Amiga. It offered 4096 colours, and many early games, cracktros and demos cycled them around wildly to show everything off. It didn't necessarily lead to something that was aesthetically pleasing.
So, over time, CDs became 'louder' as well, less dynamic range, but more aurally pleasing to most. Makes sense in a way... things have to be really silent to actually make use of a 90 db dynamic range. In most cases, environmental noises would drown out soft parts anyway. It's must more practical to pump up those soft parts, so you can hear them more easily, even in somewhat noisy environments.