Hi all,
I've been flapping my gums about this one for about 2 years now, but finally it's about to see the light of day.
Many of you noticed that I recently took a couple of months away from forum. While I didn't talk much on-line during that time, I did work on my DSP project every single day.
The designs have gone through a lot of territory, but now it looks like they're finally as stable as they're going to get. So much of the cost with these things is in the DSP support and not the DSP itself. The box, knobs, microcontroller, memory, A/D/A, opamps, etc. make up so much of the cost that it was looking like it would make little sense to build an underpowered effects box when at the hacker level it could be three times as powerful for another $15 in chips.
In the end... we now have the beginning. If this first box takes off, there will be three, possibly four more in the series. I'm going to start out with a design that I have titled "Lotus".
The Lotus is actually a simplified user interface version of another design called "Soma" so I'll refer to the Lotus as using the Soma DSP architecture. Lotus is composed of 5 computers crammed into a 1590B. Four of them are DSPs and the fifth is an Atmel ATmega165 which controls the show. The DSPs are a single Wavefront AL3102 1KS and a trio of Wavefront AL3201B Digital Reverb Engines. A 16 position rotary switch selects the effect program in use and parameters are adjustable via four pots, a 3 position switch, plus an expression pedal input. An RGB LED indicates status. One true-bypass stompswitch handles signal chain switching for channel 1 and channel 2 is wired as always on (it's possible to do stereo, but the Lotus is really designed as a guitar input device with a line level effects loop). The interface to a host PC for programming is done via an opto-isolated serial port which may also be used to provide other things in the future.
Why am I telling you this?
Because I want you to be asking yourselves, "What would I do with 2 seconds of digital delay, 16 LFOs, and enough DSP horsepower to run over 220 2nd order parametric filters?" ;)
Info on the chips is available at http://www.wavefrontsemi.com/
PCB layout for the prototype is almost done. 8)
im thinking!
(this is going to be unreal!)
good work!
:mrgreen:
Very cool!! So, where's the schematic ??!! :mrgreen: :mrgreen:
Mr. *low-tech* here w/a simple suggestion: We've all heard the sounds runaway echoes make w/different EQs. If there is a low pass filter....it degenerates to mud...using a high pass turns things into screechy static.
It would be nice to have a VERY gentle roll-off on both ends...so that a long delay...w/many repeats...would not turn into noise...but simply become *mid-rangey* and stay out of the way of new notes. I'm an old Frippertronics Fan. I suppose with them reverbs you get both a nice tight room sound and blend in a hall as well eh? Sounds cool bro!
That sounds great Peter, can't wait for the next "installment" !!
Marty.
That sounds very interesting indeed - and a pretty darned impressive achievement!
CONGRATS!!!!
Looking forward to see it!
Looks like a powerful machine. It was a very complicated build, I presume. Do you have a photo of the board of this "child" to show us?
Cool! About how large is the board going to be?
I don't think we need to be "afraid" of digital anymore. At the same time, for tinkerer-level DIY people, there is this particular challenge to meet head-on: http://diystompboxes.com/sboxforum/viewtopic.php?t=35168
Thanks for the compliments. :D
Squidsquad: That's a great idea! Can I call that algorithm the Squidverb?? :o I'm a big Frippertronics fan too. Those two Fripp/Eno albums made a big impression on me.
I hope this device will allow people to get an idea like that, quickly prototype an algorithm based on modular functions, and to be able to tune the parameters in real time via a PC until they want to store it as a preset. The range of the pots and what they do is under programmer control.
Here are a few more specs:
Total processing capacity: >144 MOPS (Million Operations Per Second); 133 MFLOPS (Million Floating Point Operations Per Second)
Real DSP Speed: >66.5 DSP MIPS
Controller speed: >11 RISC MIPS
Sample Rate: 48KHz
Sample width: 24 bits
Conversion style: Sigma-Delta
In/Out Channels: 2
Opamps: 5532 (3) & 34078 (1)
Coupling caps: Film
ADC: Wavefront AL1101
DAC: Wavefront AL1201
Control CPU: Atmel ATmega165 @ 12.288MHz
DSP1: Wavefront AL3102 1KS DSP
DSP2-4: Wavefront AL3201B Digital Reverb Engines
Flash: 512K x 8
E2PROM: 8K-32K x 8
Control A/D inputs: 5 (4 pots + 1 expression)
Control A/D resolution: 10 bits
Program selector: 16 position rotary switch
Control switch: 3 position toggle
Host interface: 5V serial via 1/8" stereo jack. (externally isolated)
Expression input: 1/8" stereo jack
Input: Stereo 1/4" jack
Input Impedance: 47K/2.7M
Output: Stereo 1/4" jack
Output Impedance: LOW
Output short circuit protected: Yes. Indefinite length.
Power supply: 9-12VDC via 2.1mm coaxial jack (center minus)
Reverse polarity protection: Yes. Series Schottky Diode.
Board size: 2.2" x 3.65" (sized to fit into a 1590B or 1590N)
Construction: Mostly SMD SOIC & SMD 0603 packages. A couple are TSSOP and one is PQFP with 0.8mm pitch. All easily hand-solderable if you skip the triple cappuccino. I think it's actually much easier and faster to solder SOICs than DIPs. With tweezers, 0603s are very easy to solder too.
At some point I will offer some sound samples from the prototype and then start taking advance orders to cover 50 boards with solder masks, legend, and plated throughs. They'll probably be $10 each.
All my software for this thing will be all open-source and I would encourage other authors to make their creations open too, but there is capacity for somebody to create proprietary algorithms too. This is where everyone here comes in. :)
BTW: The effects loop is where you put the analog effects. ;)
Um... WOW! :shock:
Can't wait to check this out, Peter.
~ Charlie
(just slide a little food under my door occasionally, thanks)
With specs like that, I doubt it will be cleared for export! (just kidding.... I think!)
I'm not a programmer, but it has to be good to see the DIY field opening up to people who are into maths & DSP etc!!
over 12 hrs online, and nobody asked about tap-tempo yet... :?: :?: :?:
I'd be looking forward to a rackmount version w/ LCD ... Very Cool!!
Something else for your interest Peter that I stumbled on ...
http://www.elecdesign.com/Articles/Index.cfm?ArticleID=9876
Quote from: puretubeover 12 hrs online, and nobody asked about tap-tempo yet... :?: :?: :?:
LOL! :lol:
The really funny thing is that it won't do tap tempo without something external connected. There are plans for a "remote" with more buttons, knobs, and switches on it. ;)
I guess a little programming and a switch that would send the expression input all the way to one extreme would do it too.... hmmmm....
Scratch,
Thanks for the link. 8) There was somebody who came through here a few months back who was talking about building effects with those chips but I don't think I every heard more. That's pretty cool that they have a guitar phaser example! Those chips work using switched cap filters IIRC. I would love to hear how it sounds. 8)
That is surely a company to watch. They quote 10,000 qty. pricing which is nice if you happen to be Roland. I'm guessing the low quantity price is 3 to 4 times the range they show. Cool chips.
xctly my thought (the switched xpr-input...) :wink:
Can you run VST plugins on this thing?
And it's in a 1590B did you say?
-- T. G. --
Once again...congratulations!
Please add me to the mailing list to know when there are updates...i WILL build one (or more)
maneco@gmail.com
Thanks!
How about MIDI? Seems like a simple enough solution for external control. I've got a couple of MIDI floorboards, sure would be sweet to be able to use them!
As an added bonus, there are lots of DIY midi controllers on the net. You can pretty easily put together custom control surfaces.
As far as effects, what about an algorithm that acts as a pitch-dependant sort of EQ, that has its center frequencies on specific harmonic overtones of whatever note you're playing? if done right, you could really alter the fundamental character of your guitar/bass. Of course, it would only work right for monophonic playing, but you could sure get some freaky-ass sounds out of it, especially with some time-based morphing parameters. Like, say, delay the third overtone 20ms, ramp down the first while ramping up the second, whatever.
Awesome project! Can't wait to see this develop!
Through Zero Flanger anyone? 8)
What architecture are the DSPs?
VST plugins? Nope.
1590B? Yep. ;)
MIDI? There's no reason why it couldn't be done with an adapter. I didn't want to list MIDI at first because I don't know how it's all going to shake out but the serial port will handle it no problem. The only real issue is powering the MIDI interface. One more 2.1mm plug I guess.
The coefficients in this system may be computed either by the DSPs, or via a basic scripting language that runs in the ATmega. This script allows you to get a value (from preset, table, pot, expression pedal, switch, or serial port), test it, perform basic math on it (add, subtract, multiply, divide, reference in a table), and then send that value into the DSP subsystem. There's no reason why you couldn't wrap universal effects DSP code around whatever control devices you have available.
How about a dual through zero flanger? ;)
The 1KS DSP has four stereo in ports and four stereo out ports, three of these pairs are connected to the singular stereo in and out of the three reverb engines while the fourth pair is connected to the ADC and DAC. There are not full DSP CPUs.... they're more like MAC cores with sequencers attached. Datasheets for the chips is available from http://www.wavefrontsemi.com/ . They used to be Alesis Semiconductor before being spun off totally.
I'm served...send me up a pm when the boards are ready!
Cheers,
Nish 8)
Peter,
You ever used the OhmBoyz VST plug in?
Because I do...and I think it's the software equivelent of what you are going for. Pre-delays....multi taps....LFOs...rsonance....filters a-gogo. I love the thing...and it would be SO fun to have in an ax pedal. If you're unfamiliar...look here:
http://www.ohmforce.com/
and in particular:
http://www.ohmforce.com/ViewProduct.do?p=OhmBoyz
Not to plug anything...but even reading the ads may give you ideas.
And you could DL a demo & play w/it.
And if you're already well aware....uhhhh.....sorry!
got a vero layout? ;)
I'd also be very interested in this board, I'm a programmer by day job and would love to get into DSP.
This may not sound too exciting to some but, a high quality reverb in a small package is something I've never seen presented in a DIY package before.
What do the reverbs sound like?
Is this really more exciting than a fuzz box???? :lol: :lol: :lol: :lol:
Stephen
The reverb chip is the one Alesis uses for their units; I'm pretty sure it's the same one you get in a Nanoverb or Picoverb. If that's correct, you get some good reverbs in addition to chorus, auto-wah, flange, delay, and a rotary sim right out of the box (on the 3201 chip).
http://www.alesis.com/products/nanoverb/index.html
The 3201 has these ROM programs as well as some user-programmable memory. The wavefront site has some pdfs that have some example code for simple custom reverbs, "off-center 45's" effects, and pitch shifting (using the circular buffer approach - nothing too special but a good starting point for some crazy stuff). The 1K datasheet area includes some EQ code as well as examples of a compressor and more.
Definitely looking forward to building one of these :D
Quote from: squidsquadYou ever used the OhmBoyz VST plug in?
I've never used
any VST plug-in. Now who's mister low-tech? :P
I don't do any serious computer recording and don't have any software that will run a VST plugin. :) PS: I compute on an 800MHz P3 machine, at least I've got two monitors and a RAID.
I took a peek at it and it sure looks cool! 8) I'm hoping that the SOMA DSP architecture can get used as a starting point for creating similar things... up to the limits of the hardware. It should be a playground for easy and high quality DSP effects.
Quote from: bryantabuteaugot a vero layout? ;)
I'd also be very interested in this board, I'm a programmer by day job and would love to get into DSP.
Yes I have a vero layout, just grab some 0.8mm pitch SMD Vero.... :lol:
You can download their assembler from http://www.wavefrontsemi.com . Be sure to download the assembler and all the app notes for the 1K and the DRE DSPs. SOMA uses both.
Quote from: RobBThis may not sound too exciting to some but, a high quality reverb in a small package is something I've never seen presented in a DIY package before.
What do the reverbs sound like?
Once somebody writes them I'll point you to some sound samples. ;) This is just hardware with nice reverb capability. There's enough power to create some really nice reverbs. :) There is a total of 2.04 seconds of delay memory to play with along with nice filter capability.
Quote from: StephenGilesIs this really more exciting than a fuzz box???? :lol: :lol: :lol: :lol:
Stephen
Fuzz... Yes.
Distortion... Maybe. ;)
Quote from: travisskThe reverb chip is the one Alesis uses for their units; I'm pretty sure it's the same one you get in a Nanoverb or Picoverb. If that's correct, you get some good reverbs in addition to chorus, auto-wah, flange, delay, and a rotary sim right out of the box (on the 3201 chip).
http://www.alesis.com/products/nanoverb/index.html ....
The Reverb Engines come with a bunch of programs built-in, but SOMA makes no provision to use them. That stuff is Alesis code and it's not available for viewing, manipulation, or basic adjustment for that matter. Allowing the use of the built-in programs would have also taken 15 additional outputs which would have required a different microcontroller or an additional I/O chip and 3 outputs that the microcontroller still can't spare. That would take space that the board can't spare as a 1590B sized board with a single side populated.
The reverb engines are the same chip used in the Picoverb, but this is like three picoverbs in parallel/series/you decide.
I think the nanoverb uses a different chip set because it claims 18 bits rather than 24, and it states the sample rate as 46.875KHz while processing at 3 MIPS. That would equate to 64 instructions per word clock and the DRE provides 128 (with 124 accessible to the programmer). The nanoverb also lists 64K x 16 as the delay memory. That's twice the RAM length of an AL3201B but the 3201B RAM is wider.
DRE code for Wide Sterero Chorus Example: (note: this chip allows programs up to 123 instructions long. LFO frequency assignments not included.)
Alesis comments in RED Mine in BLUE
____________________________________________________________
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;
; File: AN320103.ASM
; Description: Wide Stereo Chorus Example
; Authors: Jeff Rothermel
; Copyright 2001 Alesis Semiconductor
;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;
LFO0 SIN AMP=10000 FREQ=2 ; f = FREQ * 0.029Hz for Fs=48kHz set LFO 0 (for left channel) to sinewave, amplitude=10000, speed=.058Hz
LFO1 SIN AMP=10000 FREQ=3 set up LFO 1 (used by right)
;
MEM chorusmeml 8192 ; 8192 big enough for full AMP LFO left channel delay line memory allocation for assembler
MEM chorusmemr 8192 ; right chorus memory right channel delay line memory allocation for assembler
;
;NOTE: memory locations are referenced by:
; name Start of memory block
; name' End of memory block
; name" Middle of memory block
;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Process Left
RZP ADCL K=.5 ; Read left/2 into accumulator
WZP chorusmeml ; Write acc to start left chorus mem
RZPB chorusmeml+400 ; Read delayed left to B reg
;
CHR0 RZP chorusmeml" COMPK LATCH ; Read middle of chorus memory
CHR0 RAP chorusmeml"+1 ; Read middle+1 chorus memory
;
WBP OUTL K=.999 ; Write dry (B) + chorus (acc) to OUTL
;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Process Right
RZP ADCR K=.5 ; Read right/2 into accumulator
WZP chorusmemr ; Write acc to start right chorus mem
RZPB chorusmemr+400 ; Read delayed right to B reg
;
CHR1 RZP chorusmemr" COMPK LATCH ; Read middle of chorus memory
CHR1 RAP chorusmemr"+1 ; Read middle+1 chorus memory
;
WBP OUTR K=.999 ; Write dry (B) + chorus (acc) to OUTR
; That's all there is to a dual-mono chorus! 12 instructions + LFO setup + DRAM refresh (16 instructions)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;Add 16 extra reads for refresh These instructions refresh the buffer which is made of DRAM
RZP 0x00
RZP 0x40
RZP 0x80
RZP 0xc0
RZP 0x100
RZP 0x140
RZP 0x180
RZP 0x1c0
RZP 0x200
RZP 0x240
RZP 0x280
RZP 0x2c0
RZP 0x300
RZP 0x340
RZP 0x380
RZP 0x3c0
one important point...where to buy those wavefront chips? is any distributor (digikey,mouser) carrying them?
http://www.wavefrontsemi.com/products.html
http://www.wavefrontsemi.com/salesinfo.html :D
The rest of the parts come from either Mouser or Digikey. I'm only designing with parts that are currently in stock. The wavefront chips are the only odd-ball parts.... oh yes... and the stompswitch which comes from Aron. ;)
QuoteAllowing the use of the built-in programs would have also taken 15 additional outputs which would have required a different microcontroller or an additional I/O chip and 3 outputs that the microcontroller still can't spare.
Isn't there a bank-select switch that allows you to toggle between presets and the one memory set? I don't know how you have it set up but it seems a simple toggle and a 4-line decoder would allow you to use the internal sounds?
The sample code on the Wavefront site are not much use until one figures out how to load the program into memory via the serial port.... which you obviously have managed! :)
regards, Jack
Quote from: Peter Snowberg
Quote from: StephenGilesIs this really more exciting than a fuzz box???? :lol: :lol: :lol: :lol:
Stephen
Fuzz... Yes.
Distortion... Maybe. ;)
What, no overdrive?!?! I knew it was too good to be true... :(
:wink:
Seriously, fantastic work, Peter!
Jack,
I saw you were bumping around over at Wavefront recently. :)
The internal/external effects selection requires one line and then you have to give each DRE a four bit code to select the algorithm you want. The serial port is already connected to two of the select lines so actually it's only requiring 9 more lines that are not available. These additional lines must be hard-wired. They are not available via the serial interface. The built-in effects are cool for reverb, but the Leslie sim, chorus, flanger, and delays without speed control limit the utility drastically. The reverb programs are all stereo and are made to be used with external wet/dry mixing which is OK seeing as how the 1K DSP will do that, but this is really a mono box first.
I figure the DIY crowd will have the reverbs replaced with open source code by the end of day 1. 8)
The PCB is pretty packed. I could make it much smaller with 4 layers and parts mounted on both sides, but I want some shot at being to build commercially so 4 layers isn't going to happen. ;)
Tim,
Thanks. :D
The effects loop is where the overdrive goes... that way you can do:
In -> 20 band EQ -> loop send -> analog overdrive -> loop receive -> 20 band EQ -> Out
Rather than switch the overdrive in and out totally, you could blend it away and/or modify the EQ around it.
IMHO this is the most trascendent project/thread /topic in the history of this forum,a breakthrough in what can be acomplished by DIY ...
I vote to make it sticky
all my respect and admiration
Maneco
bump
Hey Peter, is this project still in the works? Any news?
I'd love to build one of these!
Bump! Where is this going?
Right after posting that my life went somewhat to hell. Actually it was a long decline, but that's another story. :)
Last year my available time went from abundant to about zero overnight while stress went from mild to excruciating at the same time. That really stopped things almost cold and I went a couple months without even posting here. Things are much better now but I still don't have the time I want yet.
There were a couple problems with the wavefront designs. My use of +5V logic next to audio made for too much noise (it was just too small a form factor) and that was solved mostly through layout and a switch to 3.3V digital stuff. The next problem was not so easy... that was the lack of good modulation waveforms. Triangles... easy. Sawtooth... no sweat. Sinusoid anything? Nope. The LFOs are provided by hardware so you get what you get and it isn't what I wanted. :( The chips are really cool for some effects like delays, reverbs, and EQ. They're also capable of lots more but not with the flexibility that would cause me to jump for joy for too long. Reverb and EQ were the two top uses so it was a good compromise. The ability to access these chips on a DIY basis was what pushed things in their favor.
This basic design has existed for years before the Wavefront chips were made, so do look at that stuff as the "specification de jour" :). Chips have advanced and the weaknesses of existing parts have been flushed out. The bulk of the DIY DSP stuff is still very much alive, evolving, and consuming 4 to 6 hours a day for me, but it's a bit different now. Details will follow some time in the future, but no comments at this time. :)
I may still release a Wavefront based design to the DIY world, but it's going to be fairly limited as far as firmware support goes, in fact I've going to leave it all up to somebody else. If it happens, the chip complement would probably drop to one DSP1K and one DRE and would be functionally very similar to the Ms. Parker+ project. I do keep thinking about a digital routing box around a single DSP1K with 8 inputs and 8 outputs as well as a simple programmable DRE based reverb, but other projects are taking priority right now.
Thanks for asking. :)
Peter,
Sad to hear about the problems you ran into w/ the wavefront design and life in general. Still, I'm glad to hear that your new project is moving along well and that you have ironed out some of the problems with noise on the digital logic side.
Right now I don't want to press you on your new project, since you have declined to comment on it in two recent threads and I respect your privacy, but I was curious about the amount of labor and parts costs in general for making a do it yourself DSP system like this as opposed to trying to find an evaluation board. Can you give me a ballpark in these two areas? The miss parker project looks hella complicated.
i and dann green{SSSHHHHAND OTHERS} are gonna start a new clubhouse.
ooooooooooohhhhhhhh no, i hear someone coming!
- tom
this is very intriguing! often it's the last 10% that takes 90% of the time/effort.
for the waveforms, what about external chips (XR2206) to generate sinewaves, fed into a cheap ADC (even an 8-pin AVR)? I need to spend more time with the datasheets to see if external modulation is possible.....
I'm all ears, keep us posted to progress, and/or if there's anything we can do to help!
Hi Dan. :icon_biggrin: It's great to see you posting here! 8)
Unfortunately the Wavefront chips are not true DSPs; they're more like MAC cores with a sequencer tied onto the side so they don't even have things like branching. The registers that provide buffer offsets are auto-incremented and not available for program viewing. You can add an FLO to index those pointers, but I don't know if you can get the audio data in channel 1 to affect the index of channel 2 without a serous kludge like presetting an LFO with input data on each cycle and then indexing with that LFO.
In the end I don't think it's a nut worth cracking as there are too many other options in the same price range that give much more performance and flexibility. These chips are great for EQ and delay/reverb applications and that's just where their strength lies. Ask the guy who just picked up the back of the truck what the square root of 4 is and you may not get an answer. ;)
Fun DSP stuff is around the corner and 2006 I think will be the year digital control and DSP finally break into DIY effects. 8)
I know you and Tom will have a part in that process. :icon_cool:
Primus, eval boards are a great way to go. I've got three of them from Motorola and they are often worth their weight in gold. This isn't always the case. I did my own boards for the wavefront chips (rather than just buying their boards and figuring out months ahead that it wasn't the direction I wanted) and now I'm counting all that layout time as "practice". Hmmm... maybe they are all worth their weight in gold. :icon_neutral:
If you've never built a DSP system or high speed digital hardware in the past, I would go the eval route. If you have ample time and budget to spare, have fun with DIY.
Quote from: danngreen on February 21, 2006, 11:11:17 AM
this is very intriguing! often it's the last 10% that takes 90% of the time/effort.
for the waveforms, what about external chips (XR2206) to generate sinewaves, fed into a cheap ADC (even an 8-pin AVR)? I need to spend more time with the datasheets to see if external modulation is possible.....
I'm all ears, keep us posted to progress, and/or if there's anything we can do to help!
that 10% thang you stole from mike both ......... i'm tell'n :icon_mrgreen:
your bro,
- tom
thanks Peter!
OK I checked out the AL3101 and the reverb chip datasheets. I see what you mean about the simple straight instruction RAM, basically it seems to be set up for effects that have a straightforward algorithm... hmm unless another processor could load instructions in as it reads/responds to the digital out? ayyyy, by then it's best to go with a real DSP.... Which you must have already figured out! ;) but how's the pitch shifting? A (moderately) accessable pitch shifter with 4 LFOs is enticing!
Let's just say Eventide has nothing to worry about. ;D The pitch shifting you can do on more capable processors is just too superior to use the 1970s technology available here.
I still do have some plans based on Wavefront chips, but they're more special use devices; one mainly for routing and EQ, and another for reverb.
Have you looked at http://www.axoris.be/ ? 8) 8) 8)
Download the AgALag effects ZIP (actually download everything) and check out the AL3101 code for sine generation. It can be done in math so I was really copping out earlier, but the flexibility just isn't there that I would like for modulation effects.
They have drag and drop effects generation happening. The only downside is that coefficients are computed in the Wavefront chips which takes serious overhead since they're made for in-line processing. I also worry about any PC interface with music gear where you could easily get a big ground loop. Still, very cool stuff! 8) They have also mentioned working with hardware other than Miss Parker :D.