4 second digital delay project

[ElectricDruid]

No, no typo. Cloudscapes is spot on.

Think what happens when you recirculate the delays. You add two 12-bit numbers together, and you get a 13-bit result. You do that a couple more times, you get a 14-bit result, etc etc. It keeps adding up.
It helps quite a bit to have more resolution than your input format, and if you can maintain that extra resolution at the output, so much the better. The alternative is truncation or rounding, and that introduces more errors.

Tom

[bluebunny]

I bit.  $$$

Can't wait to build this . . .

        .
        .
        .

. . . onto the end of my backlog.   

[ElectricDruid]

Aww, go on, it'll only take an old hand like you a couple of hours to put together! You can start after breakfast and be done by coffee time!

Seriously, the slow bit is the enclosure, since there's hardly any off-board wiring. Or maybe I'm just a bit c**p at drilling.

T.

[Ice-9]

What can I say Tom, outstanding project. Nice of you to bring it to the forum and I look forward to building this one when I find a little spare time.

[Jamforthelamb]

Mine came in the mail yesterday! Can't wait for the next Tayda code to come out  .
One question. What does the tap tempo set the timing of the delay to? I think I read further up that it isn't a typical 1/4 not or 1/8th note type deal (meaning where you can pick the tempo).
Thanks, looking forward to building it!
-Kevin

[ElectricDruid]

No, it's straight tap time. You tap, the echoes go at the speed you tap.

Good luck with your build. Let me know if you have any problems or find any mistakes in the build docs.

Thanks,
Tom

[Cozybuilder]

Mine arrived in the post today- it looks great! Now I just need to get some free time. Consulting work with travel, am busier than ever......

[ElectricDruid]

Just to say that I updated the construction guide on the website. Someone pointed out to me that I'd completely messed up the film caps section! The BOM and schematic were correct, but the description of where to find/fit them was way out.

Anyway, it's fixed now. If anyone spots anything else whilst they're building, do let me know.

Thanks,
Tom

[mth5044]

Got mine built up but I'm having a bit of noise. Voltages seem to check out (solid 3.3v off the regulator and around 4.2v Vb off a steady 9.14v supply), and all my parts/values are correct.

I audio probed around to see what I could find. Some noise and whine out of pin25 which seems to be exacerbated by the filter section (probing at their outputs) so when it gets to the circuit output it's pretty loud. At first I had it plugged into my testing rig which includes a charge pump for different voltage selection (9, 15, 18, -9) so I bypassed that and went straight to the One Spot. That definitely did help with some noise but it's still pretty bad. Also swapped out for a different batch of TL072 just in case. I have not tried plugging into another power supply but my One Spot is pretty solid for testing (and obviously the build isn't boxed yet). Should I expect some noise level with this build or are there any suggestions on where to look next? Delay itself is working great.

Any updates on this beanman?

[bean]

Got mine built up but I'm having a bit of noise. Voltages seem to check out (solid 3.3v off the regulator and around 4.2v Vb off a steady 9.14v supply), and all my parts/values are correct.

I audio probed around to see what I could find. Some noise and whine out of pin25 which seems to be exacerbated by the filter section (probing at their outputs) so when it gets to the circuit output it's pretty loud. At first I had it plugged into my testing rig which includes a charge pump for different voltage selection (9, 15, 18, -9) so I bypassed that and went straight to the One Spot. That definitely did help with some noise but it's still pretty bad. Also swapped out for a different batch of TL072 just in case. I have not tried plugging into another power supply but my One Spot is pretty solid for testing (and obviously the build isn't boxed yet). Should I expect some noise level with this build or are there any suggestions on where to look next? Delay itself is working great.

Any updates on this beanman?

Have not gotten back to it yet. There are a couple spots I still want to audio probe though. I'll have an update maybe today, if it helps.

[ElectricDruid]

Whine is definitely not right. Some hiss is to be expected, especially with the repeats up, since it's only a 12-bit ADC.

Someone recently had problems with bad distortion and the filters oscillating. The problem in that case was synth levels going in. Turning the input down to guitar level helped no end. It's possible that if you feed it a very hot signal, you might get some weirdness.

Good luck!

Tom

[bean]

My vRef voltage is on the low side (4.2v on a 9.4v supply) so I pulled all three TL072 to see if one might be causing a problem. No change on vRef with that so I went ahead and replaced the two 10k voltage dividers and 47uF decoupler. Still sitting at 4.2v. So, maybe that is what it should be here.

Anyway, here is a quick clip to show what I am experiencing. It's starts with hard bypass on my testing rig then with the effect engaged. Using the circuit bypass does not change the noise level. Noise level also does not seem to be effected by the delay time, either.

https://soundcloud.com/madbeanpedals/digidelay-hiss

PS I'll start a separate thread for this if I come up with anything else - I don't want to pollute this one with a debug.

[ElectricDruid]

I've just posted an addition to the project page over on my site. The schematic as shown accepts up to nearly 3.3Vp-p input level. This is good for some stuff, but not so great if you're running it with a low-level guitar output.
By modifying four resistor values, you can tweak the levels in the circuit and improve S/N for low level signals by 13dB. This reduces the maximum input level to 680mVpp.



If you're finding the pedal hissy, or you're building it and don't have a guitar with a hot output, it might be a good idea to make these changes.

HTH,
Tom

[cloudscapes]

Ever think of implementing a compander circuit for hiss reduction/more effective bandwidth?

[roseblood11]

... or to design a new version with better converters? I know nothing about digital circuits, but would it be too difficult to make a delay with at least 16 bit / 44.1kHz?
Nice add-ons would be modulation and the possibility to add an effects loop just for the repeats.

[ElectricDruid]

Yes! Both of the above!

Of course I thought about adding a compander circuit, and it would definitely help, but it also adds more circuit complexity and takes up more PCB space and adds another large chip. Is there room? Barely.
I've been thinking of a rack-mount version along the lines of the Ibanez DM2000, though, and that would definitely include the compander. It'd up the S/N and provide a bit of vintage character at the same time. Win/win in my book.

And yes, of course you can build a pedal with 16-bit 44.1KHz converters. These days it's simple and cheap to build one with 24-bit/96KHz converters. So what's the problem? Well, they're designed to be soldered by robots. There are no through-hole decent quality converters that I can find (someone prove me wrong! I'd love you forever!) Don't expect to put one together yourself. Some brave souls do, but it's not what I'd call "DIY friendly". And anyway, if I went down that route...what's the point? There are ten pedals like that in the market already from international manufacturers who can build 10K units as an opening offer and consequently offer finished units at a price I could never hope to meet even for bare boards and chips. There's really no point me trying to compete with the likes of Boss and others. If you want a pedal like that, buy one off the shelf. It'll be cheaper and a lot easier than building one. Sorry, but that's how the economics work.

So, how important to you is the "DIY" aspect? Is there any interest in "delay modules" a bit like the Belton brick (except unpotted, most likely) that could provide all the surface-mount parts to do a really serious digital delay? Would building such a thing into a pedal be considered cheating?!?  How much is the cost an issue?
I ask because I could probably design and program such a thing (I'm thinking of SMD codec, processor, and RAMs on a mini-PCB module) and then I'd have to get a (few?) hundred built up. Is it worth it? That's the big unknown.

If any of this changes in the future, I'm all ears. Always open to a new idea and a new project!

Tom

[cloudscapes]

I totally get keeping the design minimal, chip-wise. If your chip can handle it, maybe you could try coding an envelope follower, and use that to gate noise digitally! Won't get rid of the DAC noise, but it'll help with the ADC. At least i ntheory. I haven't tried it myself, but I'm thinking of it.

I had also considered throwing an ne570 in my own delay, but decided not to much for the same reason as you. Board realestate.

16 bit audio is pretty undoable in DIP unless you go for discontinued parts. Sometimes I come accross NOS 16bit ADCs in DIP, but they're often really low samplerate, or don't have the filters that new audio chips have. You have to do the aliasing filter front end and back end yourself. PARTS EVERYWHERE

I've been playing around with codecs from time to time. I like the ak4552. Cheap, small package (sop 16 or something, not dip), 24/88 quality, and uncomplicated (no i2c configuration, just pump clocks into it and you're set). I can pipe audio in and out of it just fine, but I haven't quite figured out yet to use it along with how I usually code my pic effects. One interrupt that's too long making it miss a clock, and the codec glitches. And my programs rely on a LOT of interrupts. to make it work I'd have to fit absolutely everything I do inside of 20-30us or something, per sample. Including sampling pots from time to time. Not all pots every time obviously.

Pros use stuff like DMA and buffering audio samples in blocks of 128 or something. Pros who are way smarter than me. 

[markseel]

Don't expect to put one together yourself.

If I'd have followed that advice I'd have never given it a shot and wouldn't be where I'm at today.  I'm not a hardware engineer - I do software.  But after years of wanting to empower myself with a bit of hardware design (I use FreePCB) and assembly (OSH Park rocks! and Digikey/Mouser has everything) I can now make boards for cheap and even solder TSSOPS, 0402 and 0603 SMT passives, and smaller LQFP chips.  I ended up buying a used microscope for SMT work off of eBay, and a cheap soldering iron with a fine tip.  You wouldn't believe how much easier an old scope makes soldering small stuff.  You even learn to be more steady.  A bit of solder flux and decent/fine soldering wire and you're off and running - with a bit of practice.  But you can do it!!!!

There are ten pedals like that in the market already from international manufacturers who can build 10K units as an opening offer and consequently offer finished units at a price I could never hope to meet even for bare boards and chips. There's really no point me trying to compete with the likes of Boss and others. If you want a pedal like that, buy one off the shelf. It'll be cheaper and a lot easier than building one. Sorry, but that's how the economics work.

I think that sentiment may be missing the point, especially in a DIY forum like this.  DIY isn't necessarily an economics driven decision - sometimes it is - but now always!  Sometimes it's not about getting the cheapest unit but rather building one yourself to your specs and design and learning (and sharing) your ideas, progress, and results with other folks.

I've been playing around with codecs from time to time. I like the ak4552.

Totally agree - that's a real simple device with respectable performance.  It's easy to hookup (just a handful of caps and resistors).  Check out the AK4556 - still easy as you can just hook it up (no need to configure via I2C/SPI) to an I2S MCU/DSP - and have 24-bit up to 192 kHz.

Pros use stuff like DMA and buffering audio samples in blocks of 128 or something. Pros who are way smarter than me.

Oh I don't know about that!  And as smart as the pro's may be (I'm a career audio/wireless FW engineer myself) they don't always do things the easy way (the hard way might be more impressive, provide more job security, and is often over-designed for obscure features, etc).  If someone showed a clear/simple example of audio/I2S DMA and block-based buffering I bet it would look plenty easy.  Side note; 128 would be pretty big - which creates latency.  Maybe something like 32 or 16 if the ISR overhead is acceptable?

I ask because I could probably design and program such a thing (I'm thinking of SMD codec, processor, and RAMs on a mini-PCB module) and then I'd have to get a (few?) hundred built up. Is it worth it? That's the big unknown.

Well as the creator of the (unfinished) XMOS-based FlexFX stuff I'm biased a bit.  But that Teensy board with a Cortex M4 looks like a solid platform.  Add your yet-to-be-designed audio board with an AK4556 and some high impedance input buffers and you'd have a nice platform for delay effects.  Hey if you want and other are interested I can design a Teensy compatible AK4556 board and make it available at OSH Park - wouldn't take long :-)

[Strategy]

I'll probably order and build one of these anyways, but to satisfy my need for instant gratification...any video or audio content of this online yet? I searched youtube (no luck, lots of good electricdruid projects posted though) but I have not checked Soundcloud yet.

[bsoncini]

I built one of these. At first it had a bunch of hiss. Did the mods that Tom suggested and it is much quieter now. Still a bit of noise but hardly noticeable now.

I don't have the time to make a video. But I can say. I've built numerous pt2399 delays and the quality of the delay in this pedal sounds better (cleaner?). The tone controls are nice to have. With dirt pedals most pt2399 sound very muddy. With this they sound really good playing around with the tone knobs. But I really love the echo splash function.