The Sandwich Echo discussion

[expanoncolin]

Lots of people like the sandwich echo, some of them want to learn how to make their own out of PB&Js.  Eventually I will stop making them out of PB&Js and design my own circuit as parts of the circuitry are pretty bad:
The input and output buffers are powered by 5v, not 9v... this means less headroom.
The output buffer is inverting, so that it can sum the delay signal and the dry signal, but you have an inverted version of your input, which is a bad idea...
There is no compression or expansion, but let's be honest, does it really need it?  It sounds good as-is.
There is much room for improvement...  which leads me to the attachment.  I've pointed out the extraneous parts of the circuit, like the switching, and pointed out the mods I've done and the mods I'm going to try.  I need to mess around with the addressing and the RAM-I have experience with analog and CMOS but limited experience with highly integrated situations, like RAM.  So feel free to build your own sandwich echo, and try new things...  I will keep this thread updated with what I do with them, IE more info about the expression pedal system.  Please post your experiences.

UPDATE The commented schematic has been updated with hold, some verifications, and more info on TBing.

PB&J Schematic with no comments:
http://www.experimentalistsanonymous.co ... 0DJ-17.PDF
PT2395 Datasheet (if anyone has the more complete version, let me know!):
http://www.experimentalistsanonymous.co ... PT2395.pdf

-Colin

[SpectralJulian]

Yeah, there is some lack of info on the PT2395 chip datasheet.  If I am reading it right, it looks like you can switch between 3 different delay times, where with the PB+J you can only do 2.  

WE permanently high for hold seems about right to me, looking at the 21256 datasheet.  I think that the other method of cutting the data in wouldn't work quite right because the chip is still going to be writing, just writing nothing, as you said in my thread where I suggested cutting the input, which I guess would be right after 106.  Still, if holding WE doesn't work for some reason, if you put the pedal into self oscillate, play something, then turn your guitar volume off, how long will it last at a similar volume?  It might not be perfect and have a little drop-off, but there would be the upshot of being able to pitch and speed shift the "loop." by changing the speed.

In other news, apparently IC 2395 is a cluster of stars 4500 lightyears away.

[expanoncolin]

Yeah, there is some lack of info on the PT2395 chip datasheet.  If I am reading it right, it looks like you can switch between 3 different delay times, where with the PB+J you can only do 2.  

WE permanently high for hold seems about right to me, looking at the 21256 datasheet.  I think that the other method of cutting the data in wouldn't work quite right because the chip is still going to be writing, just writing nothing, as you said in my thread where I suggested cutting the input, which I guess would be right after 106.  Still, if holding WE doesn't work for some reason, if you put the pedal into self oscillate, play something, then turn your guitar volume off, how long will it last at a similar volume?  It might not be perfect and have a little drop-off, but there would be the upshot of being able to pitch and speed shift the "loop." by changing the speed.

In other news, apparently IC 2395 is a cluster of stars 4500 lightyears away.

But DRAM is not static, it must be rewritten constantly, or the data will leak out the storage capacitors inside of it.  So something must be rewritten constantly to it.  It  stays at comparable quality/volume for about 3 repeats, max.  It's really not good enough...  If you were to get the ram to hold you could still pitch/speed shift, as the RAM is being clocked by the IC in the same way regardless (hopefully).  I guess if the long/short/mid pins work, you could actually switch between six modes, two for each... the PB&J is just switching between the two "short" ones.  If the datasheet had more info on it, we could know if the pins were simply making the clock lower, or if they were accessing more RAM (but it seems like the former)....

-Colin

[SpectralJulian]

What if you put a compresser in the effects loop and cut the input?  Sort of a half-assed way of doing it, and there would probably be some quality degradation but it would atleast hold each repeat at a similar volume.

[expanoncolin]

What if you put a compresser in the effects loop and cut the input?  Sort of a half-assed way of doing it, and there would probably be some quality degradation but it would atleast hold each repeat at a similar volume.

The degradation is so bad pretty much in any situation that you wouldn't get even more than a few repeats... I mean, this thing is very much like an analog delay... if you take a signal and put it through a lo-fi amplifier and a compressor and a filter, it won't sound the same a few times around!  But the RAM has a perfect copy of it, so it's better to try to get it to just keep playing that signal.

-Colin

[expanoncolin]

An larger version of the PT2395 datasheet is now available-thanks to Ken Stone for alerting me to the bigger datasheet and sending it my way.

http://www.experimentalistsanonymous.co ... PT2395.pdf

-Colin

[expanoncolin]

Hold is now a reality as far as I can tell.  Connect write enable (a low-enabled input) to pin 8 (which is just 5v) and it will hold indefinitely.  I will look at what's happening on the O-scope before committing this to a pristine mod, but for now, it's hold damnit!

(EDIT-Update)

I think it will work best with an SPDT switch, the center being the WE pin, one end being where it normally is on the PCB, and one end being +5V (such as pin .  This switch will probably pop unless there is some sort of pullup resistance-I am using 47k right now...  so an SPDT with a 47k resistor from the center lug to the +5v lug.  When WE is connected HIGH as such, write is *not* enabled-so simple, but it works!

-Colin

[expanoncolin]

I've pretty much got my designs for the delay time control via H11F3 sorted.  Making the perfect H11F3 system is a pain because there are so many resistances you can tweak, but you only want one "input" resistance range (0-10k) and one output resitance (0-20k or so), and you want it to be done in a linear fashion.  You can turn an LED on and off in so many ways, because it's current controlled...  and then you can vary the delay resistance itself directly with parallel and series resistors!  But I think these methods are best, depending on what you want to do.  The H11F3 I am pretty confident is the best method.  If you want to do an expresssion pedal without it, I have found it's impossible, because the cable's capacitance messes with the clock oscillator and makes the pedal work very strangely/poorly, so you need to isolate the cable from the varying resistance.  I may edit this picture numerous times, as it's so hard to get just the right range.

-Colin

[expanoncolin]

I've got it reversing.  It's really reverse-forward-reverse-forwarding, but there is reverse there.  The reverse is distorted, however.  I think it's because the gates I'm using are not high enough bandwidth, but I'm a bit stumped.

What you need is 3 quad XOR gates, and a flip flop.   For each of 9 XOR gates, one input should be a address output from the PT2395.  We will get to the other input in a secdond.  The output should go to the respective pin on a 41256 chip.  The rest of the connectinos for the 41256 should be left in tact.

All other inputs of the XOR gates should be connected together, and finally connected to the center lug of a SPDT switch.  One side lug should go to ground.  This allows the addresses to pass through un-inverted.  For the other side lug, you have to first take the address output for A8 from the PT2395 and connect it to another XOR gate input.  For the other input of this XOR gate, connect to ground so it's never inverted.  This is just a buffer now.  At the output, wire a cap to ground.  Value isn't very important (well, I don't think it is...), aroudn .01uf will work.  Connect the output to the clock in of a flip flop.  Connect the set and reset to ground.  Connect data to Q bar (the Q with a line above it).  Connect Q to the other lug of the SPDT.  This is now a signal which goes up and down at each repeat, so that all of teh addresses are inverted, then not inverted, then inverted... and so on.  I wish I knew why it didn't work perfectly, as it does, but something is buggered.  I think the flip flopping wave's position matters more than I think.  Below are a few pictures, the first is just testing to see if I took out the first 41256 right, then it's with 2 quad XOR gates, t hen there's the whole system.  What's funny is that all it is is ICs, and one capacitor.  I should probably add power conditioning caps.  Maybe that will make a difference.  The final picture is the filtered A8 after the buffer, and the subsequent flip flopping signal.  My apologies for the low picture quality-my old camera broke, now I am using a camera phone.  At least it's convenient...

EDIT-schematic added.  It's still not perfect.

-Colin

[Simonetta]

Hello,   I downloaded the PB&J schematic but it doesn't display in either Acrobat 4 or 5.  It says that I 'need Traditional Chinese language support'.  Adobe Acrobat fails when it comes to actually delivering a program that can display text and images across multiple computer platforms and operating systems.  They talk a smooth line, but the code inevitably fails should you forget to give them large sums of money.  Adobe would never just add a button that exports the PDF image to a JPG file even though tens of millions of users would like that feature.
  Nevertheless, if the schematic actually displays correctly in your version of Acrobat, I  suggest doing an ALT-PrintSCR on the image.  This copies the current screen into the clipboard buffer (Windows).  Then do a paste (Control-V) of the clipboard buffer into an image editing program like PaintShop Pro or GIMP.  Then use the select tool to highlight the actual schematic section of the Acrobat screen image.  Save this highlighted section as a GIF file if it is bit-mapped black and white, or as a JPG file if it has any color or greyscale components.  This may require multiple passes if the Acrobat schematic does n't fit on the screen all at one time.
   I do this with an old DOS version of OrCAD schematic capture because it is the only way to get the schematic image out of the OrCAD undocumented and unsupported image format into a standard GIF file.  It works very well but it is rather cumbersome.

[expanoncolin]

Eek-that is a rather cumbersome method!

The incompatibility probably has to do with the fact that I made it in Acrobat 7, on a Mac.  If you are on a PC, sometimes the fonts do get funky.  But Acrobat as I have it has export to JPEG (among many other exports).  So I would either do taht, or simply open in Photoshop and save as a JPEG.  But I highly prefer vector-based schematics in pdf format because you can zoom in with no loss of quality!  I really like focusing in.  Sorry that you can't view the file.  I've attached a JPEG of the current version.

[expanoncolin]

I have finished the preliminary designs for the new device, probably named the "Space/Time Continuum".  As you can see from the schems, not much is changed in the digital component-the PT2395 can only be used in one way, after all.  But most of the possible modifications are there.  The input and output buffers are slightly changed for better fidelity, and the feedback buffer is no longer a transistor (which I may end up changing after I perf this).  There is also an added volume pot for dry level, and a buffer after it.  The H11F3 circuitry is the same as above.  The only other mod I think not included here is the reverse mod, which is outlined above, and the modulation input above, which is also above.

-Colin

[expanoncolin]

So, unsurprisingly, I'm not done yet

-Colin

[chokeyou]

any chance of hearing what the forward-reverse-forward-reverse delay sounds like with that mod? also, is there a part list/layout for that? (or will there be?)

[expanoncolin]

any chance of hearing what the forward-reverse-forward-reverse delay sounds like with that mod? also, is there a part list/layout for that? (or will there be?)

If I wire up the fake reverse again, I will do a soundclip.  Sorry, I have no plans to make a parts list or layout available.

-Colin