Help with 1957 Allen C-1 for a New Player

[davidecasteel]

Re: Help with 1957 Allen C-1 for a New Player



al and arie v, I was most interested in your comments about the early Allen digital instruments. I had access to one of those in the late 1970s that had the original version of the card reader and was curious about how that feature worked. Allen did not provide any information about it, so, as a computer programmer with some understanding of acoustics, I figured it out. I used that knowledge to make new cards for the chapel organist addressing some of her desires for changes. In the 1980s I wrote a document that set down what I had discovered, which is not immediately available to me (loaned or misplaced); however, I have recently reconstructed it as an Excel spreadsheet (which does most of the extensive arithmetic automatically!) and accompanying Word document. There is another post in this topic about this same issue.</P>


The card reader accepts regular IBM 80-column punched cards containing encoded values of the 16 waveform sample points arie v talked about, which are stored in the instrument and used (when selected) by the tone generation circuitry instead of the standard data built into the organ.</P>


I'm not sure about the anti-aliasing discussion and the "data loss", but it seems to be possible to create waveforms using the cards that contain frequencies higher than the 1' pitches (the limit by the number of samples available)--for example, a fair approximation of a square or triangular wave can be produced if desired; those will sound terrible, of course, and won't have all the upper harmonics intact, but it would appear to be possible to do so. Are you saying that doing that will harm the instrument?</P>


David</P>

[radagast]

Re: Help with 1957 Allen C-1 for a New Player

[quote user="davidecasteel"]


I'm not sure about the anti-aliasing discussion and the "data loss", but it seems to be possible to create waveforms using the cards that contain frequencies higher than the 1' pitches (the limit by the number of samples available)--for example, a fair approximation of a square or triangular wave can be produced if desired; those will sound terrible, of course, and won't have all the upper harmonics intact, but it would appear to be possible to do so. Are you saying that doing that will harm the instrument?</P>


David</P>


[/quote]</P>


Allen used the MOS-1 computer for a couple of RMI keyboards. For those instruments, Allen developed synth wave tone cards like sawtooth, different width pulse waves and so on.</P>


Bill</P>

[radagast]

Re: Help with 1957 Allen C-1 for a New Player

[quote user="al"]


Hi, Arie</P>


Hi, Arie</P>


In the interest of accuracy (lol), my statement regarding Allen's synthesis is in fact correct: The pipe samples were analyzed and low-pass filtered to prevent aliesing due to the Nyquist limit which states that any frequency equaling or exceeding one half of the the digitizing sample rate, will result in non recoverable data loss. This was mainly due to the limited sample rate, a result of the slow processors and memory chips available at that time. As you said, one half of the cycle was stored owing to the inherent waveform symmetry and the cost of memory. I spoke to Robert Deutch at the time he developed the system before it was sold to Allen. You can see very interesting details if you obtain copies of his patents, subsequently assigned to Allen. I had interest as I was developing a low distortion frequency synthesizer utilizing sampled waveforms. As far as I know, he did develop a frequency domain synthesis technique which was sold to Yamaha, I think. Of course the Hammond B3 has had this right along.</P>


Allen reminds me of what used to be Hewlett-Packard, manufacturer of the finest electronic test equipment which was designed to be solid, rugged and serviceable and well-engineered. You don't find that kind of quality anymore. Allen is a trueexception.Sound, well I still have 7 ranks of pipe to go with my Allen. Plus I have Dove on the way. Still it has a String Celeste that is very nice and a Tuba (Wurly sample) that is just gorgeous but you still hear the difference when the air machine speaks forth! </P>


Al</P>
<P mce_keep="true">[/quote]</P>
<P mce_keep="true">Al,</P>
<P mce_keep="true"> I'm going to have to go with Arie on this one. Every account I have read about what Allen did on the MOS systems states that Allen recorded pipe samples. However, then they did Fourier analysis to determine how loud each harmonic was. Then there was a program run on an IBM computer that allowed Allen techs to stipulate the loundnesses of each harmonic (up to 24 I think), which the program would then calculate a single cycle waveform and also determine what each point would be on the 16 points of the half-cycle wave. The result was 16 points that corresponded with the positive half of the waveform. The MOS computer would simply read out the postive values, then the negative values for the same 16 points, over and over creating a single cycle waveform. If all these accounts are accurate, then technically speaking, MOS organs used resynthesis techniques by a very crude form of additive synthesis.</P>
<P mce_keep="true">This form of additive synthesis was the same as very early computer synthesis systems like the Alpha Syntauri synth for the Apple II computer. You could specify the amplitudes of various harmonics and the computer would crunch the result into a single cycle waveform, like reverse Fourier analysis. You could not specify the harmonics to have any phase of pitch changes because the actual synth wave was a wavetable. The harmonics were all locked in phase and relative pitch to the fundamental.</P>
<P mce_keep="true">For what Allen did on the MOS organs to be called sample playback, they would have had to sample a pipe, then extract a single cycle wave and then encode it into their format. No account I have read from Allen or anyone else says they did that.</P>
<P mce_keep="true">Bill</P>

[Guest]

Re: Help with 1957 Allen C-1 for a New Player



Hi, All,</P>


Interesting......The Nyquist limit is a fundamental tenet of sampling theory. It means that if you don't properly filter an analog waveform before digitizing, the resulting digital datawill have errors that can't be removed. It just won't sound good. Sure, the waveforms Allen used were highly filtered and tweaked. They had to do this because of the lower data rates available at that time. The Allen is NOT a frequency domain synthesizer. That came a little later for others when the speed of the mathematical calculations involved in the Inverse Fourier Transforms increased sufficiently to permit real time operation.The math is far more complex than simply summing sample points. </P>


Let's have some fun.I gotta go and practice Widor's Tocatta. Is there any speed record for this piece? So easy for you guys and so damn hard for me. I'll never get through it but got to try! Ok, now, what's the most difficult organ piece written and how long did it take you to play it?</P>


Al </P>

[arie v]

Re: Help with 1957 Allen C-1 for a New Player



Bill, Al, friends,</p>

I don't know how much filtering the Allen computer did of the waveform that it read out. My ear tells me that the typical buzz these organs produced is because of the stair-stepping between sample points. That is why each sound had what I call hardware induced artifacts, and made the Allen computer sound unique.</p>

It always amazes me when I come across an early Allen digital, that they actually sound as good as they did, considering the limited amount of data that made the tone, and the primitive nature of the digital computer. And I guess enough people thought that they sounded very good, that they bought them and made Allen's owners rich.</p>

Arie V</p>

[radagast]

Re: Help with 1957 Allen C-1 for a New Player

[quote user="al"]


Hi, All,</P>


Interesting......The Nyquist limit is a fundamental tenet of sampling theory. It means that if you don't properly filter an analog waveform before digitizing, the resulting digital datawill have errors that can't be removed. It just won't sound good. Sure, the waveforms Allen used were highly filtered and tweaked. They had to do this because of the lower data rates available at that time. The Allen is NOT a frequency domain synthesizer. That came a little later for others when the speed of the mathematical calculations involved in the Inverse Fourier Transforms increased sufficiently to permit real time operation.The math is far more complex than simply summing sample points. </P>


Al </P>


[/quote]</P>


Al, I don't think you understand what I am saying. I did not say thatAllen MOS computersdo ANYTHING in real time OTHER than read a ROM wavetable. The question is how the wavetable was created. It was NOT created by snipping a sample. I'm sorry it's just the facts. Ask Allen. Read Jerome's book that you can buy on their website.</P>


Like the Alpha Syntauri and also the Synclavier, the hardware and software allowed users to specify amplitudes of harmonics. Then the computer calculates the ROM wave that corresponds to the amplitudes that were specified. The result IS NOT real time additive synthesis but additive synthesis techniques are used to create the waveform. Like I said, it is the reverse of Fourier analysis. Instead of analyzing a waveform and breaking it down into harmonic components, the harmonic components are used to create a static waveform.</P>


With real time additive synthesis like you are describing with Musicom based organs, for instance, the hardware has oscillators that can be controlled dynamically with changes to each harmonic's amplitude and pitch. The system I described for Allen's MOS organs was not dynamic. The process only allowed static non-changing amplitude values foreach harmonic, which would then be used to calculate a static waveform.</P>


Allen had a reprint of an article by a well known pipe organ builder who worked with Allen on their analog and later their digital organs. He described what I am telling you. His name was Lawrence Phelps.</P>
<P mce_keep="true">Allen used this fact in their marketing campaigns for the MOS organs. They boasted that analog oscillators and filters could only approximate what a pipe stop sounded like, whereas with their system they could specify exactly what the harmonic structure would be. Their tone card library was filled with examples of taking one stop and slightly varying harmonics. Several tone cards might be different variations on one stop. They didn't sample a different sound for each stop, they just changed the harmonics. But the computer still read out a static waveform from the card.</P>
<P mce_keep="true">Another more recent example was a program called SoftSynth that was for Apple Macintosh computers. In that program, samples were broken down into their harmonic components. Softsynth would graph the waveform and allow the user to change harmonic structures including both amplitude and pitch changes on each harmonic. To hear the result, the computer would recaculate the sample into a new wavetable. The changes to the pitch and amplitude envelopes for each harmonic was static, meaning that once you specified them and cruched it back into a sample, it was fixed. Musicom on the other hand is real time. Like the Yamaha DX-7 for instance. The DX-7 had 6 digital sine waves, but they were generated in real time, so dynamic changes could occur with velocity or pressure or other controllers. That is what made the DX-7 so expressive.</P>
<P mce_keep="true">SoftSynth created waves on the Mac were not expressive. When ported back to a sampler, the only dynamic changes that could be done, were with filters, volume and envelopes.</P>
<P mce_keep="true">So SoftSynth took the Allen idea andexpanded on it. It was not real time additive synthesis, but it still used additive synth techniques to come up with wavetables. Samples by the way are also wavetables, just longer ones. But not all wavetablescome from samples.</P>
<P mce_keep="true">What I am saying here was echoed by a former engineer who worked for Allen, whoisa memberof the Allen Users Group on Yahoo Groups.</P>
<P mce_keep="true">In the ADC era, they started experimenting with getting more sample like results in the later systems, but when MDS came out, Allen was using true long word samples. Ofcourse now the MDS samples don't look very "long", but "long" describes the fact that transient attack data as well as some sustainsound is sampled.</P>
<P mce_keep="true"></P>
<P mce_keep="true">Cheers,</P>


Bill</P>

[radagast]

Re: Help with 1957 Allen C-1 for a New Player

[quote user="arie v"]


Bill, Al, friends,</P>


I don't know how much filtering the Allen computer did of the waveform that it read out. My ear tells me that the typical buzz these organs produced is because of the stair-stepping between sample points. That is why each sound had what I call hardware induced artifacts, and made the Allen computer sound unique.</P>


It always amazes me when I come across an early Allen digital, that they actually sound as good as they did, considering the limited amount of data that made the tone, and the primitive nature of the digital computer. And I guess enough people thought that they sounded very good, that they bought them and made Allen's owners rich.</P>


Arie V</P>


[/quote]</P>


It doesn't surprise me that Allen MOS organs were a great success. I have many of Allen's demo records that they put out of the early MOS organs. The bigger organs sounded good. In big chuches, those negative details were not as evident. It was a matter of how they sounded compared to analog organs. Three computer and higher models had both waveforms that were much more close to pipe tone than any analog system, and an ensemble that approximated the ensemble fullness of a pipe organ. One or two computer organs didn't have that fullness. The market at that time was: Pipe - very high cost, analog - cheaper, Allen digital - about the same price as analog, but much closer to the pipe sound than analog. The success makes perfect sense.</P>

[arie v]

Re: Help with 1957 Allen C-1 for a New Player



Bill,</p>

I don't think the early standard multiple computer organs had computers or waveforms that were any different than single computer organs. In fact dual computer organs had 2 computer boards in them with exactly the same data in them, just that they were slightly detuned. Usually the 3rd computer handled Mixtures and some other functions. It was only by the late 70s that their custom organs, I think they were called SDDS organs, had tonal improvements in them.</p>

I think there are 3 reasons why Allens sold as well as they did in the 70s and 80s. First they did sound different, technically had a more "accurate" tone than non digital organs. Second, it was promoted as the latest in high tech, the way of the future, maintenance free etc. and thirdly, they were very well built, which at least some others were not. So, for a lot of churches anyway, Allen became the "safe" choice to buy.
</p>

Anyways, even from organs of the 70s, I preferred organs from AOB, Shaw, and larger Rodgers to at least dual computer Allens. The Allens to me had a stainless steel kind of sound.</p>

In fact, listening to early Allens and then a newer one say a Rennaisance, makes me wonder why Allen hasn't progressed a whole lot more than they have. The latest Allens to me don't sound near as pipe-like as a number of other vendors of digi-organs.</p>

Arie V</p>

[Guest]

Re: Help with 1957 Allen C-1 for a New Player



Hi, Bill</P>


Don't be sorry, I think the issue was one of semantics. I called the MOS organ a sample player, perhaps incorrectly, but it really isacrude player ofwaveformdataor whatever isprogrammed into the roms. It does not store any harmonic data, just strictly amplitude. The harmonic manipulation was to yield a symmetrical waveform as thatthen reduced the memory requirements by 2 since they storeone-half of the waveform. They could get away withthis as the phase of harmonics can be changed without altering the tone. (so they say) I did not mean sample=wav file.Obviously you can't just take a Wurlitzer Tuba Wav file and punch a card or burn an eprom and have it sound like a Wurly Tuba. I wasn't concernned with the processing required to make a waveform suitable or whether you could even make a sound like a Tuba. So many factors like sample size and resolution (byte) size would not make this possible. You would get a sound but it surely would not sound like a Tuba. I mention Tuba as my newer toasterhas a Tuba sample that is gorgeous and has attack and decay. But the real pipe sound still is unmistakeable.</P>


Part of my interest in the MOS organ of late was it's availability, cost, and the ease with which I could add Midi and then have real pipe. The keyboard multiplexing scheme lended itself to a Midi adapter that didn't require 61 wires per keyboard. I like simplicity plus a beautifule console. When I added reverb, and got it tweaked up, some sounds were pretty good and some were impossible. I had a Wurlitzer maybe 4550 or whatever with the Omni synthesizer that sounded a lot better for a while until it self destructed from poor design.</P>


Cheery Regards, </P>


Al [:)]</P>
<P mce_keep="true">PS</P>
<P mce_keep="true">For info I will quote the Allen Computer Service Manual Theory Section:</P>
<P mce_keep="true">"The voice information which is stored in the computer memory consists of 16 seven bit words per voice. A word is a string of "ones" and "zeros" such as 0001001 which represents an amplitude at some point on the VOICE envelope. In order to obtain this information, the desired voice must be recoeded and analyzed for its harmonic content so that an envelope can be constructed as shown in figure 2. It must be noted that the harmonics which make up this voice have been arranged phase wise so as to produce a waveform whose first half cycle and second half cycle are mirror images of each other except for the sign. Arranging the waveform this way means that only one-half of the cycle information has to be stored in the computer memory. The first half cycle is then reconstructed as 16 amplitude samples corresponding to 16 equally spaced sample points on the half wave form, figure 3. .......The (reconstruction) procedure begins with the transfer of voice information from the specification memory (rom) ...to the registration memory (ram) ....The function of this memory is to allow two or more voices to be combined.....The object of the read out procedure is to successively read out the 16 sample points ....at such a rate as to eventually produce an audio tone (after DAC) related in frequency to the key depressed." Note there are multiple ram locations corresponding to Flute, Main, and pedal.,,,,,</P>
<P mce_keep="true">Cheery Regards Again!,</P>
<P mce_keep="true">Al</P>
<P mce_keep="true"></P>
<P mce_keep="true"></P>


</P>

[davidecasteel]

Re: Help with 1957 Allen C-1 for a New Player



I wish I had had access to that theory document in the late 1970s--it would have made my analysis of how the process worked a lot easier! It appears that I did come to the correct conclusions about it, however, and I was successful in producing a number of interesting and satisfying tone cards back then.</P>


David</P>

[radagast]

Re: Help with 1957 Allen C-1 for a New Player

[quote user="arie v"]


Bill,</P>


I don't think the early standard multiple computer organs had computers or waveforms that were any different than single computer organs. In fact dual computer organs had 2 computer boards in them with exactly the same data in them, just that they were slightly detuned. Usually the 3rd computer handled Mixtures and some other functions. It was only by the late 70s that their custom organs, I think they were called SDDS organs, had tonal improvements in them.</P>


I think there are 3 reasons why Allens sold as well as they did in the 70s and 80s. First they did sound different, technically had a more "accurate" tone than non digital organs. Second, it was promoted as the latest in high tech, the way of the future, maintenance free etc. and thirdly, they were very well built, which at least some others were not. So, for a lot of churches anyway, Allen became the "safe" choice to buy.
</P>


Anyways, even from organs of the 70s, I preferred organs from AOB, Shaw, and larger Rodgers to at least dual computer Allens. The Allens to me had a stainless steel kind of sound.</P>


In fact, listening to early Allens and then a newer one say a Rennaisance, makes me wonder why Allen hasn't progressed a whole lot more than they have. The latest Allens to me don't sound near as pipe-like as a number of other vendors of digi-organs.</P>


Arie V</P>


[/quote]</P>


I was speaking mainly of MOS-1 organ design. Some of the things you describe are MOS-2. While it's true that the MOS-1 600 series 2 manual organ simply doubled each stop and offered celesting and entire division, the 632 3 manual, which only had 2 computers also, split the stoplist up between the two computers. The result was more audio channels for different stops. And eachcomputer could be slightly detuned from theeach other, and the phase starting points for the waves were different.</P>


The design of using a computerjust for mixtures happened in the MOS-2 series organs. What I was referring to, was when an organ had three computers for the rest of the stops (non mixtures)like the900 series and higher MOS-1 organs and the 1100 series and higher MOS-2 organs.</P>


In those cases the stops were spread out amongst three computers. They could design a principal chorus and mixture on the 1100 MOS-2 that had four separate audio channels. That made the sound much more pleasing than a single computer or double computer organ could do.</P>


SDDS was late in the MOS era and it was much more advanced . Late in the ADC era, Allen developed computers that could cycle through 4 waveforms simulating attack sounds for stops. In addition there were four multi-wave ranges across the keyboard so that low notes could have a different waveform than the middle registers and the high notes different waveforms as well. Certain stops using this technology then would have 16 waveforms to use - four sets of four waves doing four "multi-samples" across the keyboard with four steps of attack with the final wave being the sustain wave.</P>


Allen does a lot of mixing in the digital domain these days. They would sound better if they had more externalaudio channels. That is one reason, I think, that Walker and Marshall &amp; Ogletree sound better. Not the only reason, but one of them.</P>


Allen was a safe choice as they were a healthy business. Elsewhere is a report from a church organist whose church bought a Baldwin. It broke down and Baldwin was out of business. What is the use of buying a better organ if the company isn't going to stay in business? That is a good reason for churches to choose "safe" companies.</P>


Bill</P>