NEW CIRCUIT DESIGN: NZF Flanger

Started by DrAlx, May 27, 2014, 05:26:49 AM

Previous topic - Next topic

DrAlx

Update on the LFO ticking....
The ticking is definitely not coming through the power lines.
It's being caused by the the rate pot wires being too close to the audio section.
I wrapped the rate pot wires in aluminum foil and when I grounded the foil, the ticking got a lot quieter but it was not completely gone.
I can see why, and that is because there is coupling between strips on the bottom of the board too.

I have therefore redone the layout of the LFO section.  I basically rotated the whole section by 180 degrees so the rate pot wires are on the edge of the board instead of next to the audio section.  I also managed to include the anti-ticking measure mentioned by Larry, and I've managed to do it without increasing the board foot print  :)

Unfortunately it will be a real mess to try and take apart the vero build and recut all the tracks in the LFO section and short out unwanted breaks.
I think I'll start trying to do that, and if I screw it up then I'll have to redo the board from scratch.


armdnrdy

Quote from: DrAlx on July 11, 2014, 09:09:26 PM
I have therefore redone the layout of the LFO section.  I basically rotated the whole section by 180 degrees so the rate pot wires are on the edge of the board instead of next to the audio section.  I also managed to include the anti-ticking measure mentioned by Larry, and I've managed to do it without increasing the board foot print  :)

Great Alex,

You're moving right along.

I'm eager to hear the end result!

The first bunch of boards that included LFOs and/or clocks that I routed worked well out of dumb luck.

It wasn't long before I managed to back myself into a corner having to hack on prototype boards to solve ticking/noise related issues.

Now I take great care in keeping clock and LFO traces as far away from the audio path as possible. There are times when you think that you provided enough distance only to find that a LFO trace is screaming when you fire it up.

I used to use solid ground planes as well. Now I use multiple ground planes for different sections that all come back to the main power filter caps.

I don't have to hack up prototype boards quite as much now.

I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

DrAlx

I think I know why the ticking is louder at higher LFO rate (it's definitely not subjective on this build).
There are two wires that go to the 1Meg Rate pot.
For slow LFO rates, the pot resistance is 1Meg and only one of the wires has a big square wave voltage on it.
For fast LFO rates, the pot resistance is zero so both wires have a big square wave voltage, and so the amount of
interference on the audio part of the circuit increases accordingly.

anotherjim

Hi,
I've been an avid lurker on this thread since I happened to be half way through a BBD project of my own when this started.
It's complete now but not perfect - there is a lot of noise/heterodyne at some settings via the power wiring, but the way I've done it, it would be back to square one to fix!
So I'm afraid I don't have design that's good enough to publish as it is. I really was just looking for a use for an old SAD1024 I've had for over 30 years. I still have the original Archer application sheet that came with it too. But I wanted to do more than a simple delay project.

I do think it's general layout has merit...

2 individual delays. Separate clocks but delay 2 can be switched to delay 1's clock.
Series or parallel routing.
2 triangle LFO's. Delay 1 has depth control for LFO1 and inverse LFO 2. Delay 2 has LFO2 depth only.
LFO2 has rotary type speed fast/slow inertia control with footswitch.
Regen from Delay 1 or 2.
Output mix of Dry, Delay 1 and Delay 2
Mix output jack and Delay 1 only output jack (disconnects Delay1 from mix out when used)
AC mains power! No wall wart! I got a cheap SMPS and put it in the case and fitted a standard "kettle" lead socket. This doesn't even get warm and is not responsible for the noise issues.

I wasn't setting out to make a thru-zero flanger, but it can be set up that way. I was more interested in an ensemble chorus effect. Hence 2 LFO's.

I think my main source of trouble was in putting all the oscillators in 1 chip! I used each gate in a single 4093 quad schmidt for each VCO and LFO. The VCO controlled by varying the timing cap (a fixed cap and another modulated in and out via a BJT). The LFO Triangle obtained from buffering the LFO timing cap.

I did put multiturn trim pots on the BBD outputs. It's certainly a help trimming out glitches but I never compared it to simply using fixed 1% resistors.

The noise is very much improved by switching to use only Delay1 clock. Obviously, there's no heterodyne noise (since this stops the Delay2 VCO). In this case I also phase shift the clock to Delay2 by 90deg. The Reticon data recommends spreading the clocks of multiple devices by a formula (2Pi/n Radians, where n is number of devices). This was easy to do by inverting the clock feeding Delay2's f/f (4013). Each BBD is clocked from it's own half of a 4013 and no extra buffering. Obviously that's not relevant if there's always independent clocking.
The Reticon data also suggests clock voltage can be lower than the BBD supply - maybe this is a way to reduces the strength of clock glitches.

My LP filtering is quite harsh @7Khz. Each delay has a 2 pole active going in, 1 pole passive and 2 pole active going out. The input/regen mixer also has 1pole active on it. All op amps are NE5532. and all the audio (Except BBD) is on a separate perf board with pots attached. The VCO/LFO and BBD are all on another board, again with pots attached so I didn't have any wires carrying fast stuff. Audio connections to the BBD are screened cable (Screen terminated only on audio board) and the BBD bias is via the pre-filter opamps. I did use a star supply scheme in general but there's still quite a bit on the digital board where the IC supply is daisy chained. I was a good boy and fitted 100n ceramics close to each chip as well as scattering 47 to 100uF electros in spare spaces.

There is no LFO ticking though!

The annoying thing is that the old string synths like the Solina had an ensemble chorus of 3 BBD's with separate clocks. I never heard any Het noise on those.

Anyway, good luck with your projects DrAlx. I'm sure your hard work will pay off.

Jim

DrAlx

That sounds like a really flexible circuit Jim. As far as getting rid of heterodyne noise goes, I think Puretube's advice (he posted a link to another of posts) is invaluable.   It sounds like you've read that and know the cause of you problem.  I suspect having two VCOs on the same chip is a problem, but have never tried that myself.  Clocking two halves of the SAD1024 with different rate clocks appears to be similar to having two separate BBD chips sharing the same supply.  That's what I had on my previous build (where the BBDs were using the audio supply) so I reckon that is OK.

Sometimes redoing a layout is the only way.
I decided it would be too much work to try and fix my last board with the new LFO layout, so I have decided to rebuild everything from scratch.
I've just spent 90 minutes cutting breaks in the vero tracks.  This will be the third time I've built the circuit (and desoldered the previous one which takes almost as long).  I'm hoping this will be the final build.



anotherjim

I think it's telling me something that if I turn both delay taps off on the output mixer, the Dry only signal is as clean as I'd want. I could split the clocks onto 2 chips on a little daughter board, and it might even be possible to do the same with 2 BBD's in place of the single one.

I started off building this on a proto board but it soon became evident that all the coupling between the socket strips was causing havoc - I don't think this kind of device can reasonably be worked out that way - you have to go straight to copper and solder.

What I've built is usable I think. I haven't done any recordings yet, but the noise would probably get lost in a mix. It's capable of a very liquid chorus, so it's a shame that it probably couldn't be allowed on a solo part. The noise has some use -  If I turn up the regen into feedback, the 2 clocks and 2 LFO's can set up a mighty weird load of noise all by itself :)

All the Best

JIM

StephenGiles

Quote from: anotherjim on July 14, 2014, 11:33:38 AM

What I've built is usable I think. I haven't done any recordings yet, but the noise would probably get lost in a mix.

All the Best

JIM


I think that is sometimes forgotten - as I've said before, if you are doing a demo for your girlfriend who wants to know what you get up to for all those hours when you are not at her beck and call, well yes you want to be able to brag that you've cracked the noise problems and prove it! But in a mix or worse (better) in a live band situation, you simply will not hear it.
"I want my meat burned, like St Joan. Bring me pickles and vicious mustards to pierce the tongue like Cardigan's Lancers.".

DrAlx

#107
Just finished the rebuild.  No LFO ticking now :)

The de-ticking mod mentioned by Larry needed adjustment though.  
The suggested values (33 nF + 68k) smoothed off the square wave too much and caused the sine wave amplitude to vary with sweep rate (which is bad for this circuit).
So I changed the values to 22nF + 10k and that stopped that problem,

I also made some simplifications.

1)  I removed the 100nF smoothing caps that I put on the bases of the current-source transistors in the VCOs.
    They were no longer necessary with the new layout.
    I kept the discharging VCO diodes under the comparator ICs though.

2) The LFO and audio share the same ground (so I could get a nicer layout).
    The LFO and audio sections are still star-grounded with respect to each other.  

3) I tried removing the 4.7uF Vgg smoothing caps.  The circuit works without them, but when I probed the Vgg
pin I saw large clock noise there.  So I ended up deciding to use 100nF ceramics instead of the 4.7uF electros.

4) I added more 100nF smoothing caps to the layout.  I haven't put any directly between IC supply pins on the bottom of the board.

5) The feedback trimmer was increased from 100k to 200k (so the optimal setting is close to the middle of the trimmer range).


I just need to write all this up now. New build instructions soon.

armdnrdy

Great work Alex!

Did you say sound sample soon?  :icon_wink:
I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

anotherjim

Great news Alex.
When you've put this thing to bed, you can start writing a BBD design manual  :icon_eek: :icon_eek:

DrAlx

New build instructions are here: http://1drv.ms/1nZwx1i

There are actually two PCB designs in the doc.  A simpler one (the one I built) and another with all the extra trimmers for BBD bias and output balance.
The layouts are based on a grid structure, so you could use them to make a vero or pad board build (which is what I did, but it is rather too much work).

Now waiting for someone else to build this  ;)

armdnrdy

Nice work Alex!

You seem to be the master of conservative component design!  :icon_wink:

What is the chance of you posting the "heavy" PCB PDF file as actual size? It looks like it's ready to go but...I have no clue how to, or if it's possible, to resize a PDF and keep the high resolution.

I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

DrAlx

#112
Quote from: armdnrdy on July 25, 2014, 05:45:46 PM
What is the chance of you posting the "heavy" PCB PDF file as actual size?
I'm not sure how to do that either.  Doesn't the size of the printed image depend on the size of paper that you use for printing? i.e. Isn't "A4" different to "Letter"?

If you click on the pictures below you'll get the raw images for the heavy PCB.
You can download them and paste them into something like Word, and resize them there.
The images should be scaled so they are 3.9 inches tall and 3.3 inches wide.  Does that help?

EDIT:  The resolution on these pics is too low. See posts below..
       



armdnrdy

#113
What program did you use to create the original board files?

The images in your project PDF print out at almost full page even if I choose "page scaling" "none" on my onscreen printer menu.

The files are very large.
I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

DrAlx

Quote from: armdnrdy on July 25, 2014, 06:37:16 PM
What program did you use to create the original board files?
Forget those last pictures. They got messed up when I uploaded and lost some resolution.  Use these links instead and choose "View Original" option and download them.  They are higher resolution.

http://1drv.ms/1nq0sV3
http://1drv.ms/1nq0Anu

I wrote my own software for making all these layouts (working on a grid system of colored squares) because I couldn't find anything out there that would let me build veros the way I like.
Unfortunately it's still a work in progress, and although I can output the pictures to PDF (i.e. in vector format) I can't get the size at all correct.
So all those pictures in the build doc are actually just screen captures.  Same goes for the links I've just posted to the PCBs but I got higher resolution by  gluing together lots of zoomed-in pictures using a graphics package.

I haven't tried printing them yet !!!

I'd suggest copying a very small section of one of the images (to save toner) and do a test print to see if the gap between tracks and holes look OK.
I don't have a printer at home, or I would do the test myself.





armdnrdy

Will do.

Thank you.

If I get these resized correctly, I'll repost them on this thread.
I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

DrAlx

#116
I loaded those images into a graphics package (Gimp) and it has the option of setting the print size of the image and printing to pdf file.  I did that and set the paper size to A4.  So if you have "Letter" size paper, remember to switch off page-scaling when you print.

Choose option to download as pdf.

Here are the "Heavy PCB" pdfs
http://1drv.ms/1l1Xa60
http://1drv.ms/1l1XNfG



Here are the "Light PCB" pdfs
http://1drv.ms/1nqk9MB
http://1drv.ms/1l1Xem7








armdnrdy

There we go!

I measured it with an digital caliper and checked IC pin spacing with a DIL14 IC. We're all good.  ;D
I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

armdnrdy

#118
I was just reading through the BOM and saw that you suggested a 1590D or better yet a BBDD enclosure.

A 1790NS will work as well allowing plenty of room for jacks, switches and the like.

https://www.google.com/search?sourceid=navclient&ie=UTF-8&rlz=1T4GGHP_enUS450US450&q=1790NS


Edit:
The 1590XX is another option.

https://www.google.com/search?sourceid=navclient&ie=UTF-8&rlz=1T4GGHP_enUS450US450&q=1590XX#q=1590XX&start=10
I just designed a new fuzz circuit! It almost sounds a little different than the last fifty fuzz circuits I designed! ;)

DrAlx

My BBDD arrived in the post, and there is a sticker on it saying 1790NS  :)
So they are the same box, but marketed differently for some reason.