spring reverb analogue predelay "design"

[JohnRoberts]

[quote author="bergmann disney"][quote author="JohnRoberts"]This clearly needs to be filtered out of expander sidechain.JR[/quote]
do this mean : BBD -> expand -> low-pass ???
[/quote]
The expandor can not only be after the LPF, but it could be after the spring reverb if you choose to play with reverb tail modification. My consideration was that the rectifier input that is used to read the level and reconstruct the 1:2 expansion needs to have the same frequency response as the rectifier signal used to perform the 2:1 compression before the delay other wise any frequency response errors will also be expanded 1:2. In your schematic the gain cell and rectifier are shown connected to the same circuit node but they don't have to be.

anyway, i found this NE570 companding schematic :
http://twin-x.com/groupdiy/displayimage.php?pos=-1003
and checked it regarding to the NE570 datasheet.
it seems allright, providing 2:1 compression and 1:2 expansion.

the only thing i am not fixed about is the rectification caps value : is 1µF providing attack/release times correct for audio signals? (n the datasheet typical test circuit they put 2,2µF)

2 uf will be slower than the 1 uf, I'd be inclined to start with 1 uF. I also notice that the rather simple schematic has some spare parts. The 10M dumping current into pin 16 of the compressor is reducing the amount of max gain during compression. In addition the 10 M pulling current from pin 1 will act like a low level noise gate/downward expander in the expander section. The way the two resistors are connected they are not complemetary but additive. So as drawn they will both cause manipulation of low level noise downward. I would be inclined to remove both of these, and maybe experiment with adding them back, one at a time, later to see if you like their effect.

note that datasheet says : G(t)=(Ginitial-Gfinal)e-(t/T) ; with T=10KxCrect, but i don't understand anything to this formula...

That equation describes the voltage at pins 1 and 16. The cap will decay like a simple RC to ground from a mostly charged loud signal level (e^(-t/RC), but from low signal levels the 10k will look likem ti has a diode (or two) in series. Actually the 10k is feeding a current mirror but you don't need to worry about those interior details to use the chip.

IMO start with 1 uf caps and lose the 10M resistors.

JR

 

[bergmann disney]

[quote author="JohnRoberts"]The expandor can not only be after the LPF, but it could be after the spring reverb if you choose to play with reverb tail modification. [/quote]
what i usually do when mixing is expanding and compressing reverb return in order to redraw completly a sort of ADSR to the signal (expanding for A, and R and compressing for D and S), wich i cannot do if i have only the companding expander left... i need a comp after it.
and i also like the kind of vintage noise the spring adds to the signal, i am not shure that i wan't to get rid of it...
so i was thinking about a NE570 chip fully dadicated to companding for the BBD noise problem, and then maybe another NE570 unit for the "ADSR"

My consideration was that the rectifier input that is used to read the level and reconstruct the 1:2 expansion needs to have the same frequency response as the rectifier signal used to perform the 2:1 compression before the delay other wise any frequency response errors will also be expanded 1:2.

do my circuits look like they don't have the same frequency responce?? it is the same NE570 chip, and with the same rectifier capacitance, so i thought it was allright.

In your schematic the gain cell and rectifier are shown connected to the same circuit node but they don't have to be.

i understand that key in and deltaG pins do not necessarly have to be connected, for the gate.
so what you suggest me is to definitly connect the deltaG after the output low-pass, am i right?
then, for the key in pin, do you mean i should connect it to the same point as compressor in? because like this, yes they will act exactly the same at the same time, but i think there will be a problem considering the circuitry latency (wich will be very long as it's a delay line), isn't it?

IMO start with 1 uf caps and lose the 10M resistors (and maybe experiment with adding them back, one at a time, later to see if you like their effect)

ok, i will do so.

 

[JohnRoberts]

The comp delta G can be before the input anti-alias filter and expander delta G after the anti-image filters. The comp rectifier or detector input can be connected to just before the BBD and exp rectifier to just after. You want the rectifier input to be roughly time consistent with the signal at delta G to perform a first order correction of LF envelope distortion and response to signal dynamics. If the rectifier input was out of sync with the audio you'd get audible gain modulations.

Note: taking the expander rectifier from immediately after the BBD means it will contain clock frequency artifacts. Further since this input needs to be low impedance you are probably better off taking the feed from after at least one stage of filter. It might be advisable to add an additional LP pole in just the rectifier circuit to reduce the impact of these frequency response differences. Optimal is to add a filter to the comp rectifier that matches the expander audio path but it's common to use a lower pole frequency in both rectifiers so they will track correctly. Simple filter could be 1-2k R in series with C to ground. If this filter is too severe the compressor may underrespond to pure HF signals.

JR

 

[bergmann disney]

so do you think this would be the right chain :
http://twin-x.com/groupdiy/displayimage.php?pos=-1007
(instead of 9.5Khz rect in filters cuting frequencies i am thinking of 8.8KHz. is it to low?)

 

[JohnRoberts]

Depends on what your lowest clock frequency is and tuning of other filters.

JR

 

[bergmann disney]

i want to use the whole range of delay times that MN3205 can provide, so my lower clock frequency is going to be 10KHz :?

this means, regarding it as a "sample" frequency, that my anti-alias and anti-image filters must have a cuting frequency of 5kHz, am i right?
if yes, what about key input filters frequency?

 

[mobyd]

I realise it is not analog, but by far the simplest way to skin this particular cat is this :
www.princeton.com.tw/website/downloadprocess/downloadfile.asp?mydownload=pt2399_1.pdf
M

 

[enthalpystudios]

not to hijack, but the answer may be relevant to a spring design thread, just curious if more or less all springs will work electronically in a spring reverb circuit.... wondering about getting a replacement spring for a fairchld reverbertron, and was going to just try out one of the guitar amp style springs, biggest i could find.... just wondering if they all work just as well electronically

 

[JohnRoberts]

[quote author="bergmann disney"]i want to use the whole range of delay times that MN3205 can provide, so my lower clock frequency is going to be 10KHz :?

this means, regarding it as a "sample" frequency, that my anti-alias and anti-image filters must have a cuting frequency of 5kHz, am i right?
if yes, what about key input filters frequency?[/quote]

If you sample at 10K your anti-alias has to kill the audio by 5kHz, not just start attenuating at 5kHz. The anti image filter is just dealing with 10k clock, once aliases are created they can't be filtered out.

The issue for the rectifier inputs is that compress and expand have same bandwidth. There are sundry ways to accomplish that but if you're LPF the audio path at <5k perhaps use a sharper knife and just tweak comp to be same as expander while not << audio LP.

JR

 

[bergmann disney]

well, 2309 is maybe going to be my solution, if my bbd cicuits does nothing good.
but there is no reason.

 

[JohnRoberts]

[quote author="enthalpystudios"]not to hijack, but the answer may be relevant to a spring design thread, just curious if more or less all springs will work electronically in a spring reverb circuit.... wondering about getting a replacement spring for a fairchld reverbertron, and was going to just try out one of the guitar amp style springs, biggest i could find.... just wondering if they all work just as well electronically[/quote]

Springs mechanically are similar just like all speakers are similar but drive and pickup transducers can vary in sundry details. Springs are even designed specifically for vertical or horizontal mounting (clearance for spring hang) so even physical orientation matters.

My recollection was that we had at least a half dozen or more different springs in use at Peavey. The springs with sound character good for guitar amps were different than springs acceptable for use in vocal PA heads (more diffuse reverb).

JR

 

[bergmann disney]

[quote author="JohnRoberts"]If you sample at 10K your anti-alias has to kill the audio by 5kHz, not just start attenuating at 5kHz. The anti image filter is just dealing with 10k clock, once aliases are created they can't be filtered out. [/quote]
ok.
how much dB loss is kill?? -80 dB? -100dB?
but what are going to be the aliases sounding like? i mean, yes they will be there at extreme delay value, but are they really a problem on the sound?
because a bandpass of 5KHz is already very low, and i don't want to loose that much hi-end...

[quote author="JohnRoberts"]The issue for the rectifier inputs is that compress and expand have same bandwidth. There are sundry ways to accomplish that [/quote]
this, i clearly understand.

[quote author="JohnRoberts"]but if you're LPF the audio path at <5k perhaps use a sharper knife and just tweak comp to be same as expander while not << audio LP.[/quote]
can you explain it in other words, because i don't understand
(i do not speak english very well... )

--------

anyway, john, i would like to thank you a lot for the help you are providing me, because i know i may look a bit heavy with my strong help needs.

 

[JohnRoberts]

How much kill depends on how much audio in problem range and your tolerance for aliases. Aliases are created by signals sampled less than 2x per cycle. A 6 kHz signal sampled at 5 kHz will create an alias tone at 4 kHz. A 7.5kHz signal would alias down to 2.5kHz.. This is unnatural and doesn't sound good.

I would expect knocking it down a few tens of dB would be adequate but you need to be careful about sending it hot HF signals.

=======

My comment about the rectifier is if your whole bandpass is only 5 kHz, I wouldn't add any additional LP to the expander side beyond maybe the one filter stage serving as a buffer and clock freq filter. Then just tweak the compressor rectifier to have the same freq response as expander rectifier.

JR

 

[analogguru]

You should have a look at the Fostex 3180 schematic.

Everything included, Peak limiter, analog pre-delay....

IMHO a pre-delay of up to 100mSec gives the impression of a bigger hall.
I did this in the late 70´s with 70 mSec from an AKG TDU-7000 into an AKG BX-20 - nice sound.

analogguru

 

[clintrubber]

[quote author="analogguru"]You should have a look at the Fostex 3180 schematic.[/quote]
Link seems broken.

Thanks,

Peter

 

[clintrubber]

[quote author="clintrubber"][quote author="analogguru"]You should have a look at the Fostex 3180 schematic.[/quote]
Link seems broken.[/quote]
Tried again, it works now, thanks.

 

[clintrubber]

[quote author="mediatechnology"] were PVC sewer pipe.[/quote]
FWIW, I recall having seen a reverb using that as well, IIRIC it was called something like the Great British Spring.