long digital delay build log

[PRR]

> Previously used 9 opamp stages... down to 5.

I see two more to weed out: image

The tone control does not need two buffers. Or any. The DAC filter should easily drive the tone control.

The 8k2 does nothing. (It would, on the other side of the 2200pF.) Assuming the DAC Out is near Vref, the C-R into the post-tone buffer is not needed?

Check my thoughts against your actual build before you burn PCBs!

[cloudscapes]

Thanks for checking it out, and your input!

Maybe its because it's late, but I don't see how I can get rid of the 8k2. The tone control isn't just a low pass, it's hipass/lowpass, and the 8k2/2200pf part is a hipass rc filter. I'll have to try out some of your other mods tomorrow.

The tone control is largly unchanged from the one I used two years ago, I seem to recall I needed two buffers or else I wasn't getting good hipass/lowpass "seperation". My memory may be rough though, something I'll also try out tomorrow.

[PRR]

It was late here also.

> the 8k2/2200pf part is a hipass rc filter.

Not the way I see in the image "New schematic" in Reply #137. I strongly suspect the 8200 is drawn in the wrong place. It should be "after" the cap, not before.

2200p into 1Meg is a 75Hz low-cut; hardly any cut at all.
2200p into 8200 is a 9KHz low-cut; major slope over the whole audio band.

Is the PCB layout based on this schematic, and may have the same error?

For completeness: 6800 into 0.1u is a 250Hz high-cut; slopes most of the audio band. Reasonable.

> I wasn't getting good hipass/lowpass "seperation".

Passive cross-fade like this would go-to-the-end more solidly with buffers after the R-C networks and before the pot, so the pot comes down to a "zero" (<1r) impedance instead of several Kohm. Then the pot could be 10K instead of 1Meg.

[cloudscapes]

You know what, you're absolutely right, the hipass rc filter is on backwards.

I didn't really use any math when I chose those values for both the hipass and lowpass. I did it by ear. A little more, change a value, a little less, change a value again. Except I chose values for a backwards rc filter.

I'm going to have a tinker with the values, this time not working while looking through a mirror.


The bad part of this issue is it sounds great "wrong".

[cloudscapes]

To be honest, I'm tempted to leave the tone stack be, it sounds great as-is. I "corrected" the hipass rc filter and I can't get the hipass I want even after trying a few values.

I agree that the math looks wrong, but it sounds right.

I made a minor modification in the routing. I'm now routing the last opamp's output of the tone stack to the wet/dry mixer, instead of routing "DAC out" to the mixer. That allows me to tweak the tone while in loop/hold mode.

[cloudscapes]

I'm getting flashbacks from the last time.

I had a perfectly-functioning breadboard prototype, but to get the PCB version working, I had to remove some vrefs/resistors, and add others where there were none. To this day I still don't understand how that's possible. But yeah, flashbacks.

[PRR]

> I can't get the hipass I want

Math (or a math idiot) is sometimes helpful.

image

Top is the "wrong" way-- the 8K loading the opamp does nothing and the "10M" is a do-nothing placeholder 'cuz this special idiot won't take parts out. It is in fact a mild treble boost, a near-flat, and a heavy treble cut. Useful.

Middle is the "right way" with a true low-cut C-R. Heavy dip in all mid-knob settings, going to extreme all-over slope (up or down) at the ends. Also the dip is rather high, above 1KHz.

Bottom is where I "think" a music dip should be (I get this from Fender and others), 700Hz. Dip is still extreme for general use, knob shifts it around but still extreme.

So maybe you DO want some form of the top "wrong" plan. It isn't wrong to mix an all-over with a very-dull to get a shelving filter. There is only one real filter action so you don't get the cancellation dip.

As said, separation when knob is turned to "0" or "10" is a lot to do with impedance mis-match at the pot ends. With few-K filters you had to go to 1Meg for ample separation. As I still believe both filters (or filter and all-pass) can be driven from the one DAC buffer, you can put two opamps buffers just to drive the pot, and absorb "ALL" sneak-back from the other end of the pot.

[cloudscapes]

I'm afraid I don't follow your example software picture, I don't see the signal source (is it AC?) and I'm not familiar with circuit sims in general.

For the rest, something more along the lines of this?



The structure I mean, not the values, which I'll tweak.

EDIT: tried this with both filters having a 700hz corner frequency. I'm getting a good lowpass filter but that's it, no hipass (or way too mild) when I crank the pot to the "top". I don't think this new one is working.

In your last example picture, is the green line frequency? Assuming it is, both top and middle are "right", for me. Top is when I I have the pot at 100%, middle is when I have it at 0%, and when I set it at 50%, it would have been around neutral. As "wrong" as my last schematic was, those were the filter characteristics I was hearing. Which is why I kind of want to revert back to what was working for me.

I appreciate the help you've been giving me, but if it sounded great before, I kind of want to keep what I had.

I don't know much about Fender filters/music dips. I also use this delay for much more than guitars, so I'm kind of working to make the filter work with a large range, rather than tuned for guitar.

[PRR]

> signal source

"Zsrc", circle with a squiggle.

Green purple blue lines are arbitrary (random) colors for response with pot set different ways.

Amplitude up the side. "0dB" is unity gain, "-20dB" is a pretty deep cut.

Frequency across the bottom.

Top picture, White happens to be the Treb setting of the knob. 10Hz-100Hz is about -6dB, which is about "half loud". 1KHz-10KHz is 0dB or full loud. So deep bass is somewhat cut, all treble is full, and the mid-bass area slopes up. Green is knob turned to one end. Everything below 100Hz is "full", but sloping down for higher frequency. (This is "very very muffled", like inside the blanket closet.) Violet is "5" on the knob and gives an intermediate between light top-rise and heavy top-cut.

[cloudscapes]

> signal source

"Zsrc", circle with a squiggle.

Green purple blue lines are arbitrary (random) colors for response with pot set different ways.

Amplitude up the side. "0dB" is unity gain, "-20dB" is a pretty deep cut.

Frequency across the bottom.

Top picture, White happens to be the Treb setting of the knob. 10Hz-100Hz is about -6dB, which is about "half loud". 1KHz-10KHz is 0dB or full loud. So deep bass is somewhat cut, all treble is full, and the mid-bass area slopes up. Green is knob turned to one end. Everything below 100Hz is "full", but sloping down for higher frequency. (This is "very very muffled", like inside the blanket closet.) Violet is "5" on the knob and gives an intermediate between light top-rise and heavy top-cut.

I thought the Zsrc might be a source, but it looks like AC since it has two outs and neither are ground, so I'm still not reading it right I don't think. Again, no experience in circuit sims.

Thanks for the info on the colored lines. The middle graph does look like the right way in that case.

With my old "wrong" schematic, the middle graph is also what I'm hearing, thought slightly shifted up in frequency. Pot at 0% is pink, 50% is grey, and 100% is green, roughly.

[PRR]

> it looks like AC

Audio IS "AC".

Just not rock-solid 50/60Hz Power AC. Our guitars make 82Hz to >5KHz, multiple tones at once, at any level from about zero to nearly 1 Volt.

For measurement sanity, we test with a variable frequency source (signal generator). Or we do a similar thing with just calculations. The TSC sweep runs 10Hz to 30KHz.

[cloudscapes]

> it looks like AC

Audio IS "AC".

Just not rock-solid 50/60Hz Power AC. Our guitars make 82Hz to >5KHz, multiple tones at once, at any level from about zero to nearly 1 Volt.

For measurement sanity, we test with a variable frequency source (signal generator). Or we do a similar thing with just calculations. The TSC sweep runs 10Hz to 30KHz.

I mis-worded it, I meant Zsrc looked like an AC socket with two prongs, not including ground.

[PRR]

> Zsrc looked like an AC socket with two prongs

Ah. It is a circle with a sine-wave on it. A very traditional representation of an AC generator (alternator). The circle represents the fact it is a rotating machine. The sine shows that the output is AC, not DC.



Yes, we don't usually put alternators into amplifiers. (Though a Hammond Organ is a rotating AC alternator.) The *concept* extends to "vibrating generators" like electric guitar. The sine-in-circle has become a generic symbol for "any" AC source.

[anotherjim]

The modified tone control 8k2 in the HP should return to Vref? If the amps run from 9v, the 5v supply could provide a Vref (as I don't see another Vref nearby).

As drawn, the load on the tone pot wiper is negligible while Paul has a 1M load in the simulator (the Duncan tone stack calculator or TSC). If that load is necessary, add a resistor returned to Vref.
TSC is a free Windows software, very easy to use and you don't have to produce multiple traces unless you want that.