The MT-32 --- even crappier than we thought.

[parazythum]

I have read somewhere that when people sing, they inconsciously introduce viariations because a human being doesn't naturally like "pure" sounds. If I apply this to what I could read above (with great interest)...

Second remark : many people use the MP3 format (not necessarily like it) because it is handy. And most people won't be able to tell the difference between the audio CD source and the resulting MP3 file, because either they don't listen to it in good conditions (bad speakers or headphones, bad EQ, walking in the street or on a train or a bus), or they naturally haven't good ears enough (the percentage of people with a "good" sense of hearing is very low). I can tell the difference between an MP3 and the original only if I can hear them both. I have "tested" friends on purpose about this subject in the past, and about half of them couldn't tell which song was the original source, some of them even preferred the MP3'd one

[NewRisingSUn]

First, I don't claim that it's necesarrily a good thing,

In your original post you did, but maybe that was an advocatus diaboli  statement ("provocative").

some peole might prefer the "worse" sound to a more "pure" sound, and this shouldn't be considered that strange).

A pure sound is always preferable during production because once you have a pure sound, you can choose between keeping the pure sound and adding any distortions later on. If you don't have a pure sound to begin with, you don't have that choice.

For instance, many people prefer the sound of vinyl records to CDs, or analogue synths to digital synths, because what could be called its "sound character".

That can be explained in quantifiable terms: the distortions by a vinyl record, a tube amplifier or an analogue synth are mostly even order harmonics, which the human ear likes. The distortion by digital equipment is all over the place, having a lot of odd order harmonics, which the human ear dislikes.
Again we realize that what people like and don't like is quantifiable to a large extent. Whole industries live by that.

We use lots of metaphors in all fields of our life

Unfortunately. By not calling things as they are, they make things harder, not easier to understand. Some metaphors also bring emotion to emotionless things, thereby subliminally manipulating people's feelings and therefore their judgments.
Metaphors have their place in fiction and in colloquial language, not in technical, political or scientific discourse. Originally about politics, I find this essay a very good read.

We say things such as "warm" or "cold" sound , despite sound has no temperature, or "flat" or "wide", despite sound has no shape (the sound as a perception of the senses, not as a waveform), "vivid" and "lifeless", despite sound is not a living being;

I don't use these words, and any good magazine reviewer won't (which should tell you a lot about the quality of most magazines out there). This babble about warm/cold/flat/wide/dynamic sound is mostly used by people who want to hype useless equipment, like a) $1000-a-meter cable (I'm serious), b) sound "enhancers" like the one in Creative's X-Fi sound cards, and other junk. If they told you the truth that all they're doing is a) induce a placebo effect in the buyer and b) using different dynamic compression on different frequency bands, you probably wouldn't buy them.
Methaphors cloud the mind (and yes, I know that's a metaphor, too) at best and deliberately manipulate at worst.

[Cloudschatze]

Has anyone ever actually verified that all 16 lines are connected to the LA32 chip in the 01 MT-32/CM, by tracing the lines on the PCB? I've got a suspicion that the bit shift is actually the same, but is done by purposefully misconnecting the LA32, instead of purposefully misconnecting the DAC as in the early unit.

I haven't traced out the actual PCB, but I have verified the lines using the PCB diagram in the service notes. It's the same in the CM-32L/64/500/LAPC-I/E-15 notes as well.

I know this isn't the answer you're looking for, and I'll verify against the PCB when I get home.

[shad0wfax]

Hey NewRisingSun,

I actually agree in nearly everything you say, and I'm afraid that our disagreements in fact respond to the different aspects or points were are focusing at. Of course, at least theoretically it's better to have the best possible audio signal, but it's also true that the very same fact that many times some "degradations" are consciously applied to achieve certain sound "character" (again, that damned metaphor ) shows that not always the purest sound is to be preferred. Otherwise, it would be totally pointless for instance that the Fantom includes effect algorythms such as "vinyl noise" and various types of "lo-fi". Sometimes even digital distortion and digital signal degradation are consciously applied (there are lots of examples, mostly in techno, hip hop, r'n'b and the like).

Regarding the metaphors, under ideal conditions, all of them could be eliminated by the use of technical and precise jargon. But the problem is that technical and precise concepts and vocabulary are only accessible to experts, and no one can be an expert in every field. I suspect that if you talk about digital quantization, SNR, odd and even harmonics and the like, very few people will understand, and moreover, even if someone properly understand those concepts, there's not an obvious connection between (for example) knowing what quantization is and knowing how it actually affects to the sound, prior to listening to it. On the contrary, if you say that analog sources sound "fatter" or "warmer" than digital ones, or that low-quality digital sources make the sound "grainy", most people will have a slightly better idea about how it actually sounds, even if they know nothing about the technical stuff.

Your point about metaphors and hype is quite right, but technical jargon can also be used the same way (to fool the public), and this is a well-known marketing resource.  If there's an advert of a dull product full of cryptic (for the vast majority) tech specs, it's likely than more than one will say: I don't understand a word, but this must be great!. The only solution, at the end of the day, is having a solid knowledge of the matter, so you can't be fooled either by aesthetically pleasant metaphors or by cryptic technical hype.

[Great Hierophant]

Regarding the metaphors, under ideal conditions, all of them could be eliminated by the use of technical and precise jargon. But the problem is that technical and precise concepts and vocabulary are only accessible to experts, and no one can be an expert in every field. I suspect that if you talk about digital quantization, SNR, odd and even harmonics and the like, very few people will understand, and moreover, even if someone properly understand those concepts, there's not an obvious connection between (for example) knowing what quantization is and knowing how it actually affects to the sound, prior to listening to it. On the contrary, if you say that analog sources sound "fatter" or "warmer" than digital ones, or that low-quality digital sources make the sound "grainy", most people will have a slightly better idea about how it actually sounds, even if they know nothing about the technical stuff.

Your point about metaphors and hype is quite right, but technical jargon can also be used the same way (to fool the public), and this is a well-known marketing resource.  If there's an advert of a dull product full of cryptic (for the vast majority) tech specs, it's likely than more than one will say: I don't understand a word, but this must be great!. The only solution, at the end of the day, is having a solid knowledge of the matter, so you can't be fooled either by aesthetically pleasant metaphors or by cryptic technical hype.

A technical approach to product review doesn't work for people who do not understand the technical terms or do not know what the technical specifications of a product should be compared to other products.  Also, companies have been known to manipulate technical terms to make their product seem better than another product. 

The problem is that your average consumer does not have the time to master the terminology or the product market.  The consumer wants to know whats "better".  A good salesperson should always simplify the features of a product as much as possible.  Your average consumer will vaguely be aware that the 4GB iPod is better than the 2GB iPod, but he should be able to ask the salesperson "How is it better?"  Then, having learned what the difference is, he should shop around and compare the prices of similar products. 

As NWS seems geniunely interested in assisting MT-32 emulation development, he should insist on precision and accuracy in defining the MT-32's abilities and limitations.  It is commendable that he is trying hard to correct false information as well as spend hours over technical documents to make accurate assertions.  But I believe the first goal of emulating old hardware, whether it be a sound device like the MT-32 or a video game system, is to reproduce "what was", not "what it should/could have been [in a perfect world.]"  That the MT-32 was a noisy synth is quite common given its price at the time.  The great majority of people who owned one or composed for one lived with it.  There is something compelling about analog methods, objectively poorer in quality, than lifeless digital reproductions.  You can call it "warmth", which almost truly meaningless or you can says "it produces phychologically more-pleasing harmonics", which is truly meaningless to most people.  However, due to the longevity and cheapness of digital storage, we increasingly accept DVD over film, CD over record, digital cable over broadcast signal.  But at their best, the latter examples are sublime compared to the former. 

Once we have masted the "what was", then I think interested people should move onto "what it could be."  But can you recreate in an emulated MT-32 model the analog noise by which the MT-32 actually sent it sound to a pair of speakers?  That may be unique to the components used or too-difficult and performance-intensive to accurately model. 

[MichalN]

like a) $1000-a-meter cable (I'm serious)

I wish you weren't, but I know you were   The "audiophile" class includes many members who have considerably more money than common sense. That there are people who charge one grand for a meter of cable doesn't surprise me. That there are people willing to pay so much for so little is mind-boggling.

One of my favourites is this kind of device: http://www.audioconnect.com/html/cd_mat.html  No one can quite explain how this thing is supposed to work (obviously - because it can't). But the description is worth reading because it is a perfect example of text that contains many fine-sounding words while being almost completely devoid of meaning. Orwell would be proud of them, in a twisted way.

And here's what's wrong with words like "warm", "colorful" or "detailed" when applied to audio. They are far too vague and inaccurate, and they mean different things to different people. They are not quantifiable, and are thus more or less (rather more) meaningless. It is possible for anyone to learn what "SNR" or "harmonic distortion" means, because the terms are clearly defined. The meaning of "metaphors" is not clearly or not at all defined, and that's why they should be avoided in any text that aspires to be objective, accurate and informative.

[Tom]

Well, I totally disagree with the discussion against using such adjectives as "warm" (full), "bright" (lively), etc. in describing sound.  I believe that among the music community, terms such as these serve a purpose and are perceived in the same manner.  But perhaps it's a generational thing, and the technical descriptions of sound have replaced the old vernacular.  Though I remember musicians sharing terms such as these as far back as the 1960's when I got my first electric guitar and first started getting into the music creating thing.

[MichalN]

Well, I totally disagree with the discussion against using such adjectives as "warm" (full), "bright" (lively), etc. in describing sound.

I should point out that I differentiate between sound/audio and music. Sound is a measurable and quantifiable phenomenon. Behaviour of sound waves in air and other environments is well understood, electrical transmission of sound is well understood, digital processing of sound is well understood. Human perception of sound may be slightly less well understood, but I believe that current research into psychoacoustics is very advanced (see mp3 and other audio compression schemes).

Music is a whole different ball game. Even though on the physical level music is just sound, the way humans perceive music varies greatly between individuals. It depends on cultural background, age, current mood, and a myriad of other factors which are extremely difficult to describe, understand and predict. It's so subjective that pretending objectivity would be a waste of time.

I believe that among the music community, terms such as these serve a purpose and are perceived in the same manner.  But perhaps it's a generational thing, and the technical descriptions of sound have replaced the old vernacular.  Though I remember musicians sharing terms such as these as far back as the 1960's when I got my first electric guitar and first started getting into the music creating thing.

If you can precisely describe what the terms mean and make sure that everyone who uses them understands them in the same way you do, then everything is as it should be. If not, misunderstandings are bound to happen. BTW if you have some handy reference to what these terms mean, I'd love to be educated.

Maybe there is just a difference between people who have a technical/scientific background and those who do not. I have studied some physics and I am a computer programmer by trade. Computers are famous for doing exactly what they're told, not what the operator meant. For me, precise communication has great value because misunderstandings are costly. Anything that can reduce the likelihood of miscommunication is a good thing, and conversely, vague and imprecise words and phrases need to be avoided as much as possible.

[Cloudschatze]

Has anyone ever actually verified that all 16 lines are connected to the LA32 chip in the 01 MT-32/CM, by tracing the lines on the PCB? I've got a suspicion that the bit shift is actually the same, but is done by purposefully misconnecting the LA32, instead of purposefully misconnecting the DAC as in the early unit.

I haven't traced out the actual PCB, but I have verified the lines using the PCB diagram in the service notes. It's the same in the CM-32L/64/500/LAPC-I/E-15 notes as well.

I know this isn't the answer you're looking for, and I'll verify against the PCB when I get home.

Verified.

[NewRisingSUn]

[Lord Nightmare]

Wait... Does this mean the rev 00 and early rev 01 mt-32s are actually BETTER quality(only one bit missing) than the late rev 01s, the cm-32l, cm64, etc?

Lord Nightmare

[NewRisingSUn]

No. First, because the rev 01 reverb still has the full 16 bit resolution vs. 15 bits in rev 00. Second, because in rev 00 the DAC bit 0 is grounded and thus not changing with the sign (whereas the two unused LA32 bits in rev 01 are changing with the sign), but constantly zero, meaning you'll get an unwanted change of one LSB every time the waveform changes sign, yielding one-bit noise (6 dB).

In other words, the one "dry signal" bit the rev 00 has over the rev 01 is wholly eaten up by the one-bit noise from the grounded DAC bit 0 plus the one bit less for the reverb, making for an overall lesser-quality sound.
(On the other hand, if you modified the rev 00 unit making the original sign bit (bit 15) go to both DAC bit 15 and the previously-grounded bit 0, this should improve the quality somewhat. Of course, if you're at this, you might as well undo the bit shift completely and connect all bits the way they should have been.)

Furthermore, the main volume setting changes the digital amplitude in rev 00, whereas it changes the analogue amplitude in rev 01. This is good for preventing the digital signal from overflowing, but even further reduces the available resolution for volume settings below 100.

Lastly, I have noticed that some rev 00 units suffer from crosstalk in the demultiplexer circuit --- I can specifically hear the selector signals SH1-SH3 in the analogue output (albeit very quietly), which should never happen.

[Cloudschatze]

ROLAND LA32
+------------------------------------------------------------------------------------+
| PART NUMBER |  IC NAME                | SOURCE                                     |
+-------------+-------------------------+--------------------------------------------+
|             |                         |                                            |
|  15229851   |  MB87136                |   D-50 SERVICE NOTES 1ST EDITION  MAY 1987 |
|  15229851   |  MB87136A               |  MT-32 SERVICE NOTES 1ST EDITION  AUG 1987 |
|             |                         |                                            |
|  15229896   |  MB87136APF             |  D-110 SERVICE NOTES 1ST EDITION  JUN 1988 |
|  15229896   |  MB87136A               |  MT-32 SERVICE NOTES 2ND EDITION  NO  DATE |
|  15229896   |  MB87136A               |   E-20 SERVICE NOTES 2ND EDITION  DEC 1988 |
|  15229896   |  MB87136A               |  MT-32 SERVICE NOTES 3RD EDITION  OCT 1988 |
|  15229896   |  MB87136A               | CM-32L SERVICE NOTES 1ST EDITION  JUN 1989 |
|  15229896   |  MB87136A               |  CM-64 SERVICE NOTES 1ST EDITION  AUG 1989 |
|  15229896   |  MB87136A               | LAPC-I SERVICE NOTES 1ST EDITION  OCT 1989 |
|  15229896   |  MB87136APF-G-LBND-001  | CM-500 SERVICE NOTES 1ST EDITION  DEC 1991 |
|             |                         |                                            |
|  15229899   |  MB87136APF-G-LBND-001  |   D-10 SERVICE NOTES 1ST EDITION  APR 1988 |
|             |                         |                                            |
+------------------------------------------------------------------------------------+

Feel free to wildly speculate.

[NewRisingSUn]

Code:

|  15229896   |  MB87136APF-G-LBND-001  | CM-500 SERVICE NOTES 1ST EDITION  DEC 1991 |
|             |                         |                                            |
|  15229899   |  MB87136APF-G-LBND-001  |   D-10 SERVICE NOTES 1ST EDITION  APR 1988 |

Shouldn't the first part number be the second? Also, does the D-10 have the Parkison vibrato?

[Cloudschatze]

Code:

|  15229896   |  MB87136APF-G-LBND-001  | CM-500 SERVICE NOTES 1ST EDITION  DEC 1991 |
|             |                         |                                            |
|  15229899   |  MB87136APF-G-LBND-001  |   D-10 SERVICE NOTES 1ST EDITION  APR 1988 |

Shouldn't the first part number be the second?

If you were to order a replacement IC, you would do so by specifying the eight-digit part number.

Also, does the D-10 have the Parkison vibrato?

I don't believe so...
...now I'm going to have to check.

[Cloudschatze]

Also, does the D-10 have the Parkison vibrato?

I don't believe so...
...now I'm going to have to check.

No, no granny vibrato.

I don't think you came to the same conclusion that I did, in that the name is inconsequential, and that the same LA32 chip can be found in both professional and consumer products. By this reasoning, and since every LA32 reference I have states 15-bit data output, any variance ought to be the result of something else.

[Lord Nightmare]

Btw by the part number, I'm starting to suspect the LA32 is actually an MCU and perhaps not a real ASIC IC at all!
Why?
MB8xxxx parts are made by Fujitsu, and they're DSP/MCU chips.
The MB86232 and MB86233 are the 'TGP' DSP chips (also known as F-DSP4 or f-DSP5 iirc) used in the Sega Model 1 3D hardware from 1994ish.
This makes me think the LA32 may be an FDSP-2 or FDSP-3. The only problem is the dates don't line up. The f-dsp2 and i THINK F-dsp3 WAS made in 1987 iirc, so it IS early enough, I believe.
One of them, I forget which version, is called the MB8764
Being an MCU would also nicely explain the la32 part number change: The different la32 versions have a different internal ROM.

Another BIG question of mine is what sort of MCU the Reverb chip is. Based on the RWI documentation, its supposedly a simple dsp with 16 versions each of 4 different programs which run depending on what cycle a sound plays at on the 32.767mhz clock. This chip being a fujitsu MCU is also possible. It would be nice if someone had additional information on what architecture, etc it is.

If anyone has a dead mt-32 btw with fried guts, I'm interested in decapsulating the LA-32, the reverb chip, and possibly the gate array.

Lord Nightmare AKA Jonathan Gevaryahu

[Cloudschatze]

Btw by the part number, I'm starting to suspect the LA32 is actually an MCU and perhaps not a real ASIC IC at all!
Why?
MB8xxxx parts are made by Fujitsu, and they're DSP/MCU chips.

I think you're over-speculating.

[NewRisingSUn]

Still, it's worth investigating. While uncorroberated, it's plausible enough, and if it is indeed an MCU with integrated ROM, dumping and disassembling said ROM should make emulation trivial indeed.

[MichalN]

This could be a dumb question but... has anyone tried to track down the engineer(s) who designed the MT-32? I only have experience with software engineers, but they tend to be helpful and willing to answer questions about products they worked on in the past, even though after 20 years memory can be a problem