Passive analog LC ladder audio delay

[mr coffee]

Hi All,

I have been fascinated with the prospect of adding a very short delay to the input of a compressor to give the side chain enough time to reduce gain before the signal hits the variable attenuator\gain stage.

I started reading up on L-C ladder filter delay lines, which look like they would have a minimal noise penalty, but I am not much of an electronics math jock, and I can’t seem to find any examples from Google or other resources to help me get even a vague idea how much delay at audio frequencies can be achieved with a reasonable component count. I get the impression from what I've read that using inductors and caps with the same values (except for first and last sections) would make a audio band delay line, right? ???

I have some 18 uh inductors and some nice 2 mh toroid inductors laying around and have thought about just putting a string of LC sections together, but I’m not really knowledgeable enough to have a clue what general value of capacitors to use with either value and what impedance levels would result (as in what drive and load impedance levels would result). And if these aren’t suitable, I’m not opposed to hunting up some other values.

Does anyone have some schematics laying around of something of this nature or perhaps somebody who is knowledgeable who can pull this out of their hat without much effort?

I would really appreciate any help anyone might be able to offer.

mr coffee

 

[Twenty Log]

Maybe an all-pass filter too can keep rotating the phase (delaying) but ultimately it was the analog bucket brigade chips in the analog delays (but distortion was immense)...  hopefully not to create the TARDIS of time travel though  

But another thought thinking of vinyl..  For vinyl mastering "in the old days" they have 2 signal paths, one is delayed (the playback) and one is not (the preview path)....  This was done with a tape machine with both a preview and playback head and a long-ish tape loop in between the 2 heads.... This was to allow the cutting lathe computer to calculate the groove spacing / pitch (big bass requires bigger groove spacing) one revolution before cutting with the playback delayed signal..

But it sounds like you might be printing the compression to the Tape/DAW live and want to do it live?

If not, I suppose you could "simulate" that vinyl process mentioned above with 2 stereo DAC outputs from said DAW (or if you are using tape, get the 2 heads or a tape loop; have used a mic stand as part of a tape loop... yikes...)... Just copy the track in the DAW and slide it, then output to stereo DAC #2...

There are also be mercury delay lines: http://en.wikipedia.org/wiki/Delay_line_memory#Mercury_delay_lines 

but I think the Gin delay line would be best  ;D

 

[mr coffee]

I'm familiar with the various delay stuff you mentioned, but I'm really looking for a very short delay (less than 1 ms.) that doesn't add much noise (less than an A-D conversion digital delay anyway), and it would seem that does require a multiple stage LC ladder delay, sort of like the kind of thing oscilloscopes use to slow down wideband signals to the deflection plates so you can see the event that triggered the sweep.

And yes, I'm looking for real-time (well, a millisecond) delay, not look-ahead compression for use on recorded tracks.

mr coffee

 

[Twenty Log]

Indeed... Now that I think of it, I have used a transmission line "simulator" during my undergrad tenure at university for one of the labs and courses... It was essentially exactly what you speak of and was about 5 feet (~1.5 meters) long of just large-ish inductors and capacitors, creating the typical RF 50 ohm system.  We were able to measure time of flight through the "transmission line"...  Maybe approaching a millisecond?  I can't remember...  I seem to recall that it was "dumbed down"; we certainly were not trying to measure the transmission line properties up at GHz frequencies (probably more like below the AM radio band)...

So I imagine it is available through electrical engineering laboratory supply houses... This was almost 15 years ago for me....

 

[PRR]

> even a vague idea how much delay at audio frequencies

Your 2mH could work as a loudspeaker crossover coil (though perhaps not at speaker power). Look up 2nd order crossovers. Near 8 ohms you can use 2mH and a few uFd to get crossover at a few KHz. And a 2nd order low-pass has 180 degree phase at the corner. How much time is a half of a few-KHz wave?

Obviously you want to shift the low-pass to a higher frequency. This will need a smaller cap. This will raise the working impedance up from speaker impedance to a couple hundred ohms. Phase is still 180 degree at corner freq, but freq is higher, so delay is shorter.

With a "reasonable number" of sections you can only get a few cycles shift of your highest frequency. If you expect major trouble in your top octaves, and can NOT clip, that may be useful. This might include disk-cutting and FM broadcast under strict regulation. Both use large pre-emphasis so with "bright" program are prone to top-octaves trouble.

Mocking up 9 sections of 2mH 0.1uFd driven and loaded with 200 ohms, I get sorta-flat to 20.6KHz and 136uS of delay. I believe you can do somewhat better. I don't believe you can reach a millisecond of delay with 20KHz bandpass and a "reasonable number" of L-C sections.

> give the side chain enough time to reduce gain before the signal hits

And why? You don't hear short clipping. Depending who you ask, 1mS clipping is not a problem, certainly not when you are monkeying-around with (modifying) the signal (limiting). Larger clipping (few mS) is musically useful. If you can not pass the original peak due to non-infinite dynamic range, you can approximate most effects of "louder" by adding a dash of higher harmonics to tickle the ear beyond the actual amplitude.

 

[peter purpose]

Any use?  BBC AM6/17 delay.

 

[JohnRoberts]

If you step back and think about what a compressor is doing, the result of any gain manipulation is waveform distortion. In general slowing down the rate of change shifts the energy content to lower frequency but lengthens the perturbation, perhaps making it more audible. A few decades ago when digital dynamics processors first became available, capable of pre-delay, they were met with a yawn (perhaps useful for very clip sensitive applications).

There is some success using pre-delay in combination with noise gates to allow the gate to open fully before transients but this is more of a live sound application than recording (RANE makes a unit). Note: the leading edge of most transients is not the cleanest sounding waveform out there. I did a lot of work with gated tone bursts (decades ago), and there is a sound characteristic from the gated burst, independent of the dynamics crunching you perform on it. 

Good luck and let us know how this turns out. A full bandwidth all pass circuit is not trivial.

JR

 

[mr coffee]

WOW,

Thanks for all the helpful ideas, and the BBC schematic. <thumbs up>

Even 100us. might be enough, and I may be overdoing it, as PRR points out, since short clipping distortion isn't normally objectionable or detected by the ear.

I'm interested in this for more than just a single application, and I don't have the formal EE background to trust my decimal points trying it from formulas, so I especially appreciated the scale-it-like-this intuitive approach from PRR - I was a "new math" experimental subject in the 60s. <grin> Get in the ballpark and breadboard is more my style.

And I'm not entirely opposed to using a somewhat unreasonable amount of inductors for a one-off, let's-see-what-it-sounds-like-if-I-do-this circuit. (After all, I can tear it down afterwards and use the inductors for something else)

As in a "Slow Gear" kind of circuit without the sound of the bump at the beginning of the note. John, your ideas about the gated tone burst sound are very pertinent. I'll check out what Rane has over in their schematics, and see if it was based on an L-C delay line or just a short A/D and D/A digital delay line.

Thanks again to all for the great help!!!


mr coffee

 

[JohnRoberts]

AFAIK, RANE uses a digital delay in their modern gate. 

Coincidentally RANE used some active all pass filters for short delay within modest bandwidth bandpasses for time aligning individual drivers in a loudspeaker crossover back in the '80s(?).

JR

 

[mr coffee]

Thanks, John - I'll check it out

mr coffee

 

[squib]

the SSL5000 compressor module uses a passive LC network to achieve a delay in the main path whilst the sidechain path shoots ahead. Strangely the min attack time according the front panel legend is not "0 ms".
I'm on the road at the moment and away from my schematic library so can't provide much detail at the moment. In two days time i'll be working on a 5000 series console and so will have the opportunity to look at the front panel again and indeed once back home can get you a electronic copy of the schematic.
It is similiar to the BBC one posted. From my knowledge to date, Audix / BBC and SSL all made compressors with LC delays.

 

[abbey road d enfer]

Coincidentally RANE used some active all pass filters for short delay within modest bandwidth bandpasses for time aligning individual drivers in a loudspeaker crossover back in the '80s(?).

JR

BSS and E-V did the same, probably many others. As you mention, this is much simpler because the time alignment is concentrated on a very narrow band around the cross-over frequency; in that case the time delay can be considered almost constant, even with the simple one or two-section all pass circuit they employed.

 

[mr coffee]

Squib,
I'd love to hear more about it (and see a schematic, if it's not too much trouble). And if the minimum attack time is longer than "instant", I like to hear your subjective impressions of what the delay accomplishes somically.

Thanks,

mr coffee