Theo Verelst Local Diary Page 32


The same holds as for my tripod and other diary pages: nothing on this page may be copied or changed and distributed except that the page as a whole may be printed or otherwise transferred in unchanged form,  mentioning the original URL and global page reference, and that holds for everyone on the globe and beyond, including so called 'christians' and self-appointed 'Secret Services'.

Home (alt)   Tripod Home   Local Diary list (Secure)   Tripod Diary List   Previous Diary

  Thu 23 22:10,  2006

   Well, a lot more diary pages lately, measured by averaging per month of so. I thought I'd write some more about interesting subject, at the risk that I compensate with interest in myself for the lack of interest from many others which would be reasonable.

The server did more work already in february than in the previous 3 months, but a quarter of the hits is my own, because of the cgi script mentioned on the previous diary page.
Summary by Month

Month Daily Avg Monthly Totals
Hits Files Pages Visits Sites KBytes Visits Pages Files Hits

Feb 2006 488 426 53 16 1398 3149462 389 1239 9799 11240
Jan 2006 338 272 51 21 1737 667423 662 1590 8441 10493
Dec 2005 351 274 56 13 1335 1656608 412 1764 8508 10886
Nov 2005 442 308 31 10 437 324448 114 343 3389 4870

Totals 5797941 1577 4936 30137 37489




   Great work: the Yamaha S90 was crafted with a Digital Output!

I had the spirit last sunday (march 12, the date above is the start date of the page being written) to open up a years old S90 sound synthesizer which is (in a newer version ?) probably the or amoung the top selling synthesizers of the last years, and figured out wether I would be able to tap the signal going to the main output DA converter and figure out the digital format, so as to use the digital synthesizer internal signal for my own purposes, for instance to be able to record music from it without having the AD convert the (analog) synthesizer sound output to get it in a computer or other digital machines.

I had not too much trouble guessing what the various electronics in this high tech machine would have to do in the output section, though some chips were not main stream, and the DA converter chips (up to 96 kHz, 24 bit) I had to look up on a japanese server, but luckily they had (english!) free datasheets available, so I could figure out the pin connections, the internals, and also the signal formats to drive these digital to analog converters.

Don't get me wrong: this HIGHLY specialistic electronics work as in that your average computer hardware guy wouldn't even know where to begin, and your TV repairman wouldn't either, I'm quite sure... So this is a theover game when he wants to do something a bit more exiting than some average electronics making, and as we know theo is an (official) university level electrical engineer wit a lot of electronics experience. So this is a hobby project, but not for starters, you shouldn't try this at home unless you really know what you're doing! Breaking a synth like this is not funny, so this project had to be done with great care, because all those digital chips and electronics on the board of the synth included soldering wires to essential signal lines, tapping internal supplies, and getting (live, though of course very low voltage) signals to the outside of the enclosure, while some of the wires being tapped could already break to attached chips by static voltages from me wearing a static sweater or so when only touched by a finger!

So I had to ground myself and the machine and work very carefully, I included resistors in the signal lines coming out to prevent accidental shortcircuiting of the outputs of important signal processing chips in the machine, and I made a stabelized supply and a buffer chip in between the outputs to make the flatcable which I slid through an extension slot lid in the back at its external connection relatively safe, and included resistors to make it even shortcircuit-current safe to use in like I have now first tried on a breadboard, to connect it over a voltage divider to a Xilinx FPGA demo board, which to my pleasure has now been used to extract the correct bits from the outcoming bitstream to reconstruct the 20 bit stereo sound samples produced by pressing keys on the keyboard of the synth.

To check wether those samples are indeed sensible and error free the signals as are also fed to the internal DA converter, I connected a self made fast 8 bit DA converter to other pins of the FPGA board, and fed it with the most significant 8 bits of the 20 output bits of one digital audio channel, which, after changing also the two complements representation into a positive number one, indeed makes that DA converter output sound like the pianos and other sounds from the synthesizer!

Great, isn't it!

So now I have studied S/PDIF a bit, I think I know the protocol, and I happened to have optical transmitters and receivers lying around from Conrad for normal optical link connections, and looked at an open source spdif fpga implementation to make this project into a spdif output for a S90 synthesizer. Yep, thats right. Should be cool.

Insanity? Nope, the digital signal is already there, and works fine, and without glitches or doing something wrong with the synthesizer (like heating up chips because of wire loading the outputs).

Here's  a picture of what is working thus far, without any damage to the synth, its kept functions throughout the whole operation, which mainly makes use of some clear soldering points and 4 test point connections on the leftmost synth board, which are very small, though:


   The synth mod/extension with a 7805 chip and a 74hc14 schmitt trigger to buffer all output signals

The signal is a 64 bit frame length, 2 subframe (stereo) of 32 bits each msb first, lsb left aligned 20 bits effective , 44.1 kHz frame rate bit stream, where the word clock ( left/right state change at the beginning of each subframe) is available and the bit clock is 2.844 and then something MegaHerz. Then I have two buffered signals for the main out and the assignable output, so 4 20 bit signals at CD rate.

For people interested it seems the DA converter chip is probably running in mode 4. I'll have to carefully measure someof its pins to find out wether it has a emphasis on the signal on. I improvised a frequency counter in the the fpga (the demo board form the xilinx page) to give me 4 digits on its display indicating frequencies of signals on one of its input pins in kiloherz, which soon gave me the word and bit clock speeds which showed up nicely and immedeately when the synth is switched on.

Remember, unless you want to risk unrepairably blowing up your (or the owners) S90 synth, DON'T TRY THIS YOURSELF, unless you really understand what you're doing and know how to work anti-static. I'll make the fpga programming project available as open source, probable when I have the SPDIF working, which is not yet...

Sketch diagrams of the buffer circuit and the fpga board number connections:

 

 

The 12 volts is taken from the synth boards 7805 input pin, and the ground from a big capacitor nearby. The word and bitclock are taken from the assignable output DA nearby test connection pads, which I corrected after I took the picture of the board from above, liek I indicated, and the main output DA has one test point: the digital data bitstream.

As I said: the signals are gitchless stable and can be read well by the fpga and the interpretation after deserializing thus far seems completely correct as a two complement binary digital signal of 20 bits.

These are dumps from the FPGA "program" schematic diagrams to drive the self made DA converter (8 bits) to test the circuit and the decoder of the data which takes the information from the synthesizer internal circuit board and buffers it to be fed to the external FPGA. First, the block diagram with the as yet three signals (1 bit) in, and the byte out to the DA converter connected to external pins of the FPGA:

 
This is the actual test circuit:

 

The serial data comes in from the left top, and is shifted for 4x8=32 bits where 4 stages of shift registers are feeding eachother from the most significant bit, then from the word which has been formed at the transition moment of the word clock, the relevant 20 bits are taken, and converted from two's complement into linear positive offset codes, of which the highest 8 bits are connected to the DA converter.

Of course I'm now working on a matching SPDIF (optical) link converter which fits this output, that should even be a nice product.

  Early Telecommunication

   How early was telecomunication happening after royal mail with seal, horses over roman roads, and smoke and fire symbols?

Well, we all know Morse code and the Telegraph already exist a long time, and telephone existed in the early 1900s, but for instance what happened between Europe and the US?

In fact in EIGHTEEN hundred 60 there was already a telegraph connection!

But an automated phone connection had to wait until the 1960s.

Lots of transmission line properties and heavy science are in that field, including the ampification along the thousands of kilometers of wire, without accumulating a lot of noise. Currently I'm sure the same holds for internet lines: how much bandwidth would there be in total ? Most of the internet cross atlanctic traffic used a cable I conclude because ping times are too short for normal satellite connections.

There also was the automated telegraph: the telex, which in the beginning of the 20th century was centered around London and spun most of the world.

In fact probably the most interesting line was the line Marconi has pioneered: the wireless transmission across the atlantic! That was a major branch of electronics sports in the early 1900s. Hundreds of kilowatts of transmission power and very big and long antennas to get messages around to europe. Thats something else than wireless phones or walky-talkies.

   Just a word

I read in a Sound On Sound magazine article on the web that analog synthesizers were though to sound splodgy, which is a great word I think.

splotch Pronunciation (splch)
n.
An irregularly shaped spot, stain, or colored or discolored area: "spectacular splotches of color and beauty in the blossoms" Wendy Lyon Moonan.

Ha ha, funny. Splodge.

An interesting dutch word which I cannot just form the top of my head translate is: "Schallen" Consider it a homework assignment and mail me the answer.. ("blare" ?)

  The imperialist IT motherf*s

   All of them ? No. I'm sure many "real" programmers and Li/Unix users aren't that way anyway. But I think the software beast is all too real, especially in Europe, which messes up more than can be tolerated by reasonably sensitive human beings.

The crudity, the raw lust for power and even control, the ruthless Gestalt thinking without humanity, mercy, grace, space, respect, sensitivity, professional ethics (much), and, the hardest word of all:

DEMOCRACY (from Greek δημοκρατία (demokratia), δημος (demos) the common people + κρατειν (kratein) to rule + the suffix ία (ia), literally "the common people rule") is a system where the population of a society controls the government. It may be narrowly defined as that of nation-state government specifically, or more broadly to describe a society as a whole, which can also exert political power and social power.

Just as a recent finding on the wikipedia, this is from the 1917 encylcopedia brittanica:




Because the copyright expired, it appears to be free on the internet, I've scaled and processed the next a bit:



Would it be because of the army EE nature of the early beginnings that thpse concepts didn't exactly make it in the IT world?

 
   The Eniac Vacuum Tube computer built in the 40s (!)

No, I don't think so.

I guess in the time of the 5th (6th ?) generation databases and Artificial Intelligence, talking robots, and not to forget (though I rather would immedeately) adventure and role games it would be a good challenge and nursery level education for computers and their users to learn such concept and get their brains and processing cores around that concept. And proof in tests that the lesson has been learned, or I'd have to believe that their IQ is not sufficient, or persons want to be bad. Ough, that would have to be noted as a criminal organisation.

A world wide one. Brr, spooky. Maybe the international socialists must then take over. Nationalistic socialism has proven a bad idea.

Most IT are quite unaware of the real powers behind their little gross beast, to begin with: the computer makes of the first hour, the writers of the Operating Systems and their parts which are copied into all "new" things, except seldom intelligently, and most of all: the computers themselves, the hardware, the processor and its microprogram which are all made by only one singular party: Electrical Engineers.

Recently I voted for the ACM SIGGRAPH new director, for which a paper (snail mail) card was sent with information to vote on voting-site. To keep the jesuits from taking things over at siggraph, I voted for Sue Gollifer as new director, which should normally be ok: a woman art person is probably an ol angle for a change event though the conference is more technology and especially science oriented.



Well, hopefully the new democracy of the protestant kind is working...

   Self-making High Quality Computer Graphics

   Well, it's done on a supermarket-bought machine with only Free Software, in this case Cygwin, which is great to compile professional programs outside of Linux on a windows machine.

I've looked a bit at this (good) book:

 

And for instance downloaded and compiled the following basic texture example:



The program can be compiled with cygwin (on a windows machine) and run:



Resizing the window works, then the pattern doesn't interfere so clearly with the dot pitch:



The above is a simple texture demonstration where two three-dimensional polygons (not flat but in 3D world) are seen through a virtual camera with a checker pattern made on them. On the other side of the ocmplexity spectrum, the following is a (not self compiled) demo of a pond with grass around it, which is real time with moving water waiving grass and moving clouds. Click on the image for a larger version with higher resolution.

 

Click on this link for a mpeg4 movie which plays in the quicktime movie player, or on the next image to play a regular (but much bigger) mpeg-2 movie:




The scene can be rendered in real time with a NVidia 6700 card, which in such event turns a bit hotter (up to 70 degrees or so), and has an incredible number of polygons of 150,000 for all the moving grass pieces which still is rendered up to 30 frames per second in almost high definition resolution. There were high priced top workstations in the past which could only dream of such quality.

I made the movies in real time using a hardware mpeg-2 encoder which reads from the tv-output, which isn't optimal, but not bad either, and gives a fair impression of the result in a short time.

The next example of OpenGL 3D graphics is the following sphere which is adapted from a opengl example from the web (check opengl.org) which can from source code be compiled on windows using VC or Cygwin (prefered by me, VC appears to not work good). I've added a image save routine which writes an animation frame in compressed jpeg form to disk for each step in the animated sphere rotation. Thats great to make animations of 3D scenes, and I used ffmpeg to create movie files from autotmatically generated stills in mp4 and mpeg(2) format, in this case both can over a fast dsl line be viewed without waiting, the mp4 quality (for instance viewable with the free quicktime movie viewer) is better.

Click on the images to see the movies:

 


 

A little adaptation to the graphics program made the sphere into a ellipsoid:

 

In all these renderings, note the smooth highlight on the ball, which is a reflection of the virtual light source, and is also rendered in real time, including the jpeg saving (which is first preceeded by a raw image save in ppm form, taking some disk bandwidth), because it is made with the hardware shader from the graphics card, driven by a Cg program. Interesting new possibilities.

A non-shader OpenGL example is billiardgl, which is opensource, but I couldn't easily compile it on Fedora Core 4, so this a dump from the windows pre-compiled (free!) version:



Looks great, and plays good, too.

A real piece de la resistance is the windjammer demo from NVidia which also runs in realtime (at least 25fps at 1200x1600 resolution) on the 6700 card, see the movie below, click on the screendump to download mpeg format movie, with sound effects I added:



The screen dump shows a Linux mplayer running over a network to a windows X server, taking about 90 percent of a 100BaseT ethernet bandwidth, but works. Here is a mp4 version of the same movie, which loads realtime from the server provided you have enough download bandwidth and not everybody tries at the same time... To start viewing immedeately, copy the link and paste it in the Open Url of the viewer you're using, mostly that gives you immedeate picture while downloading, possibly after pushing play after a little bit has downloaded.

What are we planning here, Verelst?

   Audio Amplifiers

   As most readers will know, I've made various audio amplifiers and of course many people have looked at HiFi equipment and folders. Back in the end of the 70s I was into HiFi as a concept because I had well read my library books, Philips magazines, electronics store catalogs (Contact in the Hague, Tandy later on) and HiFi flyers and folders.

HiFi was like a holy grail. Not like people or religion, but for music. Absolute spitzenklasse was out of financial reach possibly for half a lifetime, but a good obere mittelspitzenklasse or so was a good target for dreaming about or so, in the field of audio. So I had my interesting radio, one of the first Philips cassette recorders, with seperate microphone, a stereo tube amplifier with sony speakers, an old tape recorder, a pick up, and a connection to the stereo in the living room when I was 11 and started making my own mixers. At first mono devices in a large cigar box, later I made my own wooden mixer enclosures, which I liked to design to look and work good and which held more and more electronics the next years.

I found this poster in delft, the mixer drawing looks like they found my old notes from when I was a kid and used them, disgusting.

       

Anyhow I at some point made a working and good sounding powered mixer and then I had a reasonable cassette deck, converted a car radio to a tuner and made my decades lasting stereo 4 channel good quality mixer in a neat wooden enclosure. I made good recordings from my tuner and started to work on better and more powerfull amplifier such a s a good sounding 2x10W tip31/32 based end amp, even with stabelized supply and I tried some powerup amps and ways to make them work with good frequency range and stable,too. Finally when I guess I was 15 I made myself a speaker set with quite good Philips 10" woofers and middle top range tweeters, which were large pressure enclosure with very good damping and thick wood and to drive those up to real loud sound levels, I built with adaptations a 2x75 Watt (!) RMS end amplifier from a book example, which served as my main amp for years, and was able to at school a;so blast nearly the soft tiles from the ceiling when really put to the test... I could use an oscilloscope at home, which showed that the clipping point really corresponded with an actual 75 W RMS per channel, which especially in 1980 or so was really really loud.

At the time I was in the school technical commitee, which I later led for years, which was about loud band amplification, with only an 80 watt amp, but at home and with friends I was into good quality, from source to speaker a good frequency response, phase linearity, low distortion, low noise, low wow and flutten for tapes and rumble for the turntable, more those propeties than raw power. When I at 18 had some synthesizers and a home studio (teac tascam 244) to use, I still was also as musician more into high quality sounds than what many guitarists for instance are into: more sound.

I didn't have the powers to achieve supreme sound walhalla, but an expensive headphone on a good preamp an recordplayer got pretty far, and I sure didn't have modest means, like I wrote, and of course I listened to sometimes expensive enough stereo sets from friends or even at Hifi stores, and a bit later I had a very good spec Phillips amplifier and Jamo 504 (I think) monitors for the home studio, which with the first and a half generation of CD players sounded pretty good, and also recordings form the only Phillips 3 head cassette deck I had worked for sounded pretty good. I lacked a nic big sub-woofer for the lowest notes, though I had experimented with old radio speakers I had, of which one big one I put on top of a wooden blanket crate, which was the best low I'd had ever, but not up to 70 watts of power...

This picture I processed from a setup at a big concert, where a bunch of Crown amplifiers are used to fill a big hall with a love performance:



Some of those type of amplifiers (each single 19 inch rack unit) can drive speakers with a few kilowatts of real output power, where they even reuse energy to keep the speakers under control to prevent excessive heating up of the case, which for 'normal' purposes is a low of power, especially if its real and continuous. The crowns are well known and of reasonably quality, certainly for the power and power effectiveness they offer, and appear to have a switching output design, which is like a special Class-D type of design, also costing less heating of the amps. Distortion is not so good for HiFi norms, depending on the model, and I'm not sure what the damping at higher frequencies is like, so this is a compromise for making a stadion filled with loud sound, which isn't too bad.

On the consumer side of things, this is a current Conrad offering: a HiFi amplifier with pre-amp and remote control, and a nice display:



The screws on the side and a few details 'spoil' the super high end look, and indeed it isn't super high end quality we're talking about, but it's interesting just the same, the specs aren't bad:



About 10 times more distortion and 10 times more noise then my recent end-amp (which the mos chips) (...), and less than half the power, but not too bad, and it includes a pre-amp.

It's almost orthogonal (--> "tangent" , nodge nodge, know what I mean) with the design traks of my latest self made end amp, which is relatively few parts (power-chip based), no pre-amp, though I use a measurent-type opamp preamp from the synth enclosure, which is very un-dressed up, no input switch, even, DC coupled output, very low noise, very low distortion (the pre-amp even extremely), which is also by absence of electronic (also digital) volume and tone controls, which usually generate rather horrific distortion by products, and the case is heavy duty enough: black metal 19 inch, metal frontplate, it sat under all kinds of luggage in the back of several cars, even to france and back, and performed without a glitch.

Here we have a lot of elecronics parts, inputs, electronic and digitally controlled volume and tone controls:



and a discrete component end amp:



with elcos in the signal path, and a relais (for switch on delay or short circuit/DC offset protection ?), a discrete (!) stabelized pre-amp supply, a plastic enclosure (from the looks of it) with fragile display and "modern" knobs, which I rarely like better, though. And it costs only 80 euros, remote included.

Would this all be like in this rather expert level book:



Well, parts of it, I'm sure. I enjoyed making amplifiers in the beginning of highschool, and I must have made all kinds by now: I started with a tube amp and some tube radios, then I started building basic push-pull and comarible ones with parts I could get my hands on, 2n3055's were expensive on allowance at age 12, but I had those, and 2n2901s from radio shack iirc, I had good results with chip amplifers back then: good quality, nice sound, though oscillation control was an issue with one of them, and power only up to 2x6 Watt. Then I built a tip32/33 (or 31?) based amp from a book which was 2x 10 Watt RMS, which was in fact a real breaktrough: good power (10 real watts in my small room was loud!) and also good quality for the time and good high-response, those tips aren't sluggish. Then the 75 Watt beast of pleasure, that was great, I used that with pleasure for many purposes for years, when I read about in the library book under the blankets (when I had to sleep) I hadn't dreamt much of actually making it and using it a number of years later. I also made headphone amps, and more recently I experimented with (self designed) power FET end amp stages with tranformerless voltage multipliers (for 12 V supply)  and even with switched designs. And then of course I saw the 100Watt Fet Chip Amp at Conrad Rotterdam (now gone) and bought more than a few to built in, and even sell the result.

I knew that the place where the ground is attached and where the feedback lines are soldered matters for quality, really, it does, see the section from the below as an example, which could be measured like this:



A difference measurement with a linear transformed version of the imput signal and mixed with the end amp output signal in the right ratio to end up with the added distortion products.








The above also speaks about grounding of the pre amp and the end amp, which is also an issue with the power opamps I used recently: a  soldering point a few centimeters away can make a difference, for instance in hum or feedback accuracy, see above. Imagine making a potent 100W RMS (300 Watt peak per channel) and very wide frequency range MOSFET amplifier oscillate supersonic without noticing, that could fry some perfectly good tweeters, so I've been quite on the alert to prevent that, used short feedback wires and loops, carefull with parasitic PCB capacitances which could include small poles in the feedback lines, etc.

And especially: use the right grounding points: 10 amps per channel through ANY normal wire is bound to creat voltage differences, even over 10 cm of wire which can easily be amplified and lead to (audible!) distortion and even instability. It really can, I found out with my first powered mixer with mic input. It's called 'motorboating' and it wasn't a funny result of all my work. But after having redone most the soldering and a lot of shielding, I got that machine to work at the time, too!

A great enhancement to audio amplification is whats oftn called Surround Sound, or reverberation.I programmed a dual core DSP to generate that effect, and to work around the same problems that plague end amplifiers with switched supplies I gave the DSP + analog circuitry board a torroid transformer instead, which a conventional stabalizer chip:

 

It gets pretty hot at 1 amp because the smallest tranformers are 12 volts while I only need 8+ the ripple and the chip voltage drop. "There must be some kind of way out of here"

On the driving side the below shows a digital volume contol which is up to standard, this high grade version chip could possibly even be used for a SSL mixer console quality level, in this case it is driven by a FPGA which is serially linked to a linux PC.

 

The pcb with the leds normally contained a ARM7 microcontroller which reads a potmeter and drives the stereo voume control chip from TI, the microcontroller has been lifted from its socket and 3 test wires from a self made connector to the fpga prototying board are slid in the 1/10" connector holes instead, which worked good when driven over a TCL script.

The little board just under the white inset lines is a self made 8 bit DA converter with a R-2R network on a HC series chip, which is driven by the FPGA which connected to the signals from the above addition to a S90 synth so that in 8 bits the DA sounds like that instrument!

I visited a lecture at the hilversum visual art academy from a A&R manager of some relevance in holland, this is after the lecture:

 

   Musical equipment and software

   I got the sweetwater catalog from the US for free by asking for it, which is cool, though it was a bit damaged by the mail men.

 

The plugin prices are the opposite of the synthesizer prices: quite high! For instance these sony sequencer plugins, which in the sony demos don't sound very special, but then again those are mpeg files:

 

a thousand bucks, well well. I make reverbs, too, which are NOT Open Source (pretty much the only thing at the moment) and wich can sound great on practical examples, even in surround setup very good.


    Designing with the latest Xilinx software

Well, it's not a complete improvement only to go from ISE webpack 7.X to 8.1, because the hierarchy list fully expands automatically (I use not small designs), and the schematic editor window can't be detached to use full screen effectively...



The above shows a blackfin DSP interface and a serial port interface and a memory interface (which seems to make the blackfuin nterface with physical 100Ohm series resitors mess up it seems. I recently saw a 'to many signals switches simultaneously' per group warning somewhere deep hidden in the ucf erros or so, maybe that doesn't help much either... Some work I could do.

I played around a bit, it never hurts to get free software like block design dialog boxes, for istance a clock switcher, but inf fact a lot of blocks, even things like cordics and such:




    So what happened to Sci Fi?

Well, I didn't read it after about 1981 or so, so I don't know!

 

Sure know the difference between the 70s captain Kirk and evertything that followed it being soooo much lower quality and pleasure.


I have looked at this though, with some great music (banjo and such):



So a motor company supplies kitchens for television shows? Aha, in the sixties



   Mathematics and Experiments

When I was at Cern (1991 or so) for my work at the time, about distributed computing, I met one of the makers of the Root package, which exisits today:



At the time one had to use expensive UNIX workstations to do serious computations with, like the HP720s, Digitals, Suns etc, nowadays one can use a run of the mill 64 bit motherboard with a not expensive AMD on it and get a free download of a package like ROOT which can then run prety well and without much installation work on a perfectly decent linux workstation.

Another favorite of mine is texmacs, though I'm sure it is not perfect, and maybe there are even better ways to run Maxima with a real time formula layout interface:



The above is a solution to the well known 2d order differential equation of harmonic motion. I test this because I have considerable interest in solving electrical network equations symbolically, and preferably using symbolic excitations and solving system responses.