Very short delay circuit (from 0 up to 50 ms)

[HeaD]

Hey everyone,

I'm looking for a simple circuit that provides a very short delay to a line signal (from 0 up to 50 ms). I need to output the delayed signal only.

Is there an easy way to achieve this goal? Thanks!

[Fender3D]

Ciao Gus,

since you asked for 0 to 50 ms, the easiest way would be a BBD circuit, search for "EHX full double tracking", then just swap BBD for desired delay.
Otherways you must dip yourself in FV-1, or similar DSP, programming

[HeaD]

Ciao Federico,

thank you for your answer. I see that this nice pedal uses some BDD ICs like MN3005 and 3205. Looking at the datasheets, I figured out that these ICs can manage a variable signal delay from 20.48 ms to 204.8 ms.

I see that the MN3007 has a variable delay from 0.5 ms to 50 ms, but I fear that even in this case I'd lose an important range for my purpose (0 to 0,5 ms).

I will investigate even on the FV-1 stuff. Do you have some more infos about it? Thank you.

[Fender3D]

Having a great resolution between 1 and "0.something" won't be easy...
You'll always reach the limit set by BBD's stages number...

You may have a switch and 2 or more BBDs, on eBay you may still find MN3009 or TCA350 for lowest delays;
you may as well "overclock" chips if you can live with level attenuation...

Otherwise, on the other site, I saw a 0.5ms analog delay line; perhaps it may be edited to achieve different delays, though it won't be a continuous control...

[samhay]

If you want really short delays then you may have to look at all-pass filters, which can get pretty big.
However, if you are only useing the wet signal, do you really think you can hear a delay of less than 0.5 ms?

[Mark Hammer]

The MN3207, and the Beiling and Coolaudio replacements for it can give you delay in the range you're looking for with acceptable bandwidth.  The application notes for the MN3207 show a circuit for doing this.  

A lot of the more elaborate circuitry that comes with analog delays often have to do with keeping the noise under control when you are aiming for the longest delay possible.  If you are using the chips well within their capability - so high clock rate for short delay - you can get away with much less.  A good example of that is the Zombie Chorus.  Replace all of the modulation circuitry in the Zombie with an MN3102 to achieve a short fixed delay, and there isn't a lot of parts there.

If you want to aim for a delay of less than 1ms with an MN3207/equivalent, THEN it starts to get more complicated.  The capacitance on the clock input pins of the BBD can change the square waves you feed them into something more like sine or triangle waves at high clock rates, unless you use a buffer for each clock line to provide more current.  This is why the datasheets for Panasonic BBDs usually only shows performance up to clock rates of 100khz.  They assume you will be clocking the BBD directly with an MN3101/3102.  They are a very conveient and compact solution to producing complementary clock outputs, but neither the 3101 or 3102 can provide enough current to overcome the input capacitance in the BBD, so Panasonic only guarantees their performance to around 100khz.  They CAN be clocked up to 1-2mhz, but you need to provide suitable buffering of the clock outputs.

So, if you are happy with 1-50msec, then it is a simple circuit and not a problem.  If you want to go below 1msec, then you will need to make a more complex circuit.

[earthtonesaudio]

How much timing resolution do you need for times shorter than 1ms?  Can it be in 1ms increments?   Cause then it's Mark Hammer's suggestion plus a bypass switch around the delay line.

[HeaD]

If you want really short delays then you may have to look at all-pass filters, which can get pretty big.
However, if you are only useing the wet signal, do you really think you can hear a delay of less than 0.5 ms?

Hey Sam, thank you for the all-pass filter tip.

Is it possible to arrange an all-pass filter to get a desired phase change? Could you please share some stuff?

I have to couple the wet signal to a clean signal coming from another microphone during the same recording session.

[HeaD]

So, if you are happy with 1-50msec, then it is a simple circuit and not a problem.  If you want to go below 1msec, then you will need to make a more complex circuit.

Hey Mark, thanks. Unfortunatly, I really like need to go below 1 ms. What do you think about the SAD1024?

How much timing resolution do you need for times shorter than 1ms?  Can it be in 1ms increments?   Cause then it's Mark Hammer's suggestion plus a bypass switch around the delay line.

I'd really love to get an analog control to change time. Thank you.

[Mark Hammer]

So, if you are happy with 1-50msec, then it is a simple circuit and not a problem.  If you want to go below 1msec, then you will need to make a more complex circuit.

Hey Mark, thanks. Unfortunatly, I really like need to go below 1 ms. What do you think about the SAD1024?

I think you will have a difficult time getting one, and that, unless what you have planned is brilliant, then it is a waste of a good chip.

Let me be clear, though, the complexity of buffering the clock lines is not THAT hard to do.  It increases your parts count by maybe 20%.  See the Hollis Ultra-flanger and the A/DA Flanger for some examples of how it is done.

[samhay]

An allpass filter passes all frequencies (amplitude is flat) while introducing a frequency-dependent phase-shift, which causes your signal to lead or lag the 'clean' signal by some fraction of a wavelength - e.g. with a 90o phase shift, you are looking at 1 ms for a 250 Hz signal.
There is a big problem with this - the delay is frequency dependent. When I said that the circuit would get pretty big, this was probably a significant understatement.

I suspect that BBDs are the way to go for the sort of times you are looking at. If you absolutely must have a delay <1ms, you could build 2 delays; one with a fixed delay of 1 ms and the other with 1-50ms.

[earthtonesaudio]

For VERY short delays, you can also: convert the signal to PWM, delay rising and falling edges separately through one-shots, then recombine at the end with SR latch.
Drawback to this method is the max delay per stage is less than one PWM period (e.g. PWM @ 20kHz, max delay per-stage = 50µs)
Benefit: easy voltage-control, see here.

[slacker]

Re the FV-1 this will only give you a delay down to zero for one channel relative to the other. If you're only interested in the delayed signal then the minimum is a milli second or so due to the inherent latency of the chip.

[midwayfair]

The shortest delay the human ear is capable of hearing is .1ms. It's almost impossible to express how fast that is, or even how fast 1ms is. So an MN3007 should be fine, right? Have you looked into just modding an MN3007 chorus or vibrato, and switching off the LFO?

[samhay]

The shortest delay the human ear is capable of hearing is .1ms. It's almost impossible to express how fast that is, or even how fast 1ms is. So an MN3007 should be fine, right? Have you looked into just modding an MN3007 chorus or vibrato, and switching off the LFO?

My guess is that the OP is having problems mixing signals from a mic (with the inherent delay due to sound waves travelling through air) and e.g. a guitar. You can't hear at 100 us resolution, but you can hear the weird comb filtering that happens when you mix 2 signals with a slight delay.

Edit - well unless you are really old/broken you can hear at 100 us resolution - its 10 kHz - but you can't hear a delay of 100 us.

[midwayfair]

The shortest delay the human ear is capable of hearing is .1ms. It's almost impossible to express how fast that is, or even how fast 1ms is. So an MN3007 should be fine, right? Have you looked into just modding an MN3007 chorus or vibrato, and switching off the LFO?

My guess is that the OP is having problems mixing signals from a mic (with the inherent delay due to sound waves travelling through air) and e.g. a guitar. You can't hear at 100 us resolution, but you can hear the weird comb filtering that happens when you mix 2 signals with a slight delay.

Edit - well unless you are really old/broken you can hear at 100 us resolution - its 10 kHz - but you can't hear a delay of 100 us.

OOOOH. Okay. You're right, I misunderstood the problem.

[HeaD]

[...] you can hear the weird comb filtering that happens when you mix 2 signals with a slight delay. [...]

that's the point...

[Fender3D]

that's the point...

If you have 2 different deparate signals, you may as well build 2 delays, one for each signal, then fine tune delays so they are 0.x ms each other...

This will be easier than trying raising BBD's clock.

@Mark
when raising clock signal level attenuation is worse than clock current issue...
ie. MN3007 @1.6MHz is 10-12 dB less than @100KHz with or without clock buffers...

[HeaD]

If you have 2 different deparate signals, you may as well build 2 delays, one for each signal, then fine tune delays so they are 0.x ms each other...

Yeah, as Sam suggested before that seems the best solution so far...

[Mark Hammer]

@Mark
when raising clock signal level attenuation is worse than clock current issue...
ie. MN3007 @1.6MHz is 10-12 dB less than @100KHz with or without clock buffers...

That's an interesting piece of info.  I never knew that.  Do you have any sense of what the critical point is with respect to where such attenuation becomes more apparent?  For example, would a clock rate of 300khz only lose you 1.3db, or whatever?