EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: jazzalex on March 22, 2012, 09:43:45 pm
-
Hello all,
I have been working a lot with audio equipment and for certain reasons I need to apply an analogue delay to an audio signal. In terms of musical performance with my guitar a Roland Space echo works great ( http://www.rolandmusik.de/produkte/RE-20/index.php (http://www.rolandmusik.de/produkte/RE-20/index.php) ), however, regarding my lab experiment it exhibits unacceptable drawbacks (supports not more than line level / does not support full audio bandwidth up to 20 kHz).
So, what I essentially need is an analogue (really not digital) device that adjustably delays my input audio signal of 5 Vpp by up to 250 ms. In the sound engineering domain nobody can help anymore -- maybe in the EE-world ?
Any pointer appreciated,
thanks in advance
Alex
-
I don't know what's inside the Space Echo, but if you want to remain in the analogue domain I think you need to use something old and traditional like a tape loop. You put the recording and replay heads a certain distance apart on a closed loop of tape and run the tape at different speeds to get different delays.
Otherwise, since the speed of light is 300 000 km/s you are going to need a length of wire 75 000 km long to provide a 250 ms propagation delay...
-
Otherwise, since the speed of light is 300 000 km/s you are going to need a length of wire 75 000 km long to provide a 250 ms propagation delay...
A 'bit' more, 3E8 m/s is the speed of light in vacuum. In a coax it is something like 0.7c or so ???
-
I initially thought of PT2399 but it's analog style only works up to 5kHz , beyond that all digital craziness sets in but i believe that's just a hunk o' crap .
I see PT2399's in reverb/echo/delay commercial things everywhere .
One warning though , original dip PT2399's can be had for 30cents per piece in 100 off quantities but pray hard that you can find a SOIC package .
-
I don't know what's inside the Space Echo, but if you want to remain in the analogue domain I think you need to use something old and traditional like a tape loop. You put the recording and replay heads a certain distance apart on a closed loop of tape and run the tape at different speeds to get different delays.
Otherwise, since the speed of light is 300 000 km/s you are going to need a length of wire 75 000 km long to provide a 250 ms propagation delay...
Actually the space echo is a "handy" modern representation of such older devices (the old one's name was also "Space Echo"). So theoretically it's great apart from some additional low-pass filtering and the line level limitation. I might check other approaches, however, the 75 000 km cable might not fit into my office ;-)
Thanks, best
Alex
-
I initially thought of PT2399 but it's analog style only works up to 5kHz , beyond that all digital craziness sets in but i believe that's just a hunk o' crap .
Thanks for that hint -- it has to be "real" 20 kHz but this is at least a good way to start the search.
Alex
-
I've seen people hack floppy drives to do something similar to what you're talking about.
-
What you want to look for are bucket brigade delay chips. They were out of production for a long time but with the increased interest in analog devices for the music industry they're back in production as of a few years ago at least...I want to say it's designator is SAD1024. It gives about 200mS(I may be off by a little so don't quote me directly) of delay on a chip and they can be multiplexed together for a much longer delay time. 500mS should be very easy to achieve depending on what you want to do. The thing to remember is that the more delay you design in with these the more signal degradation you'll get, which is part of the vibe of the old delay pedals from the 70's and why they're still in demand.
As for your application, google analog delay pedals, and dig around the diy stompbox forums for bucket brigade delay pedal schematics. Then just hack the circuits to your needs. They're pretty straightforward designs and shouldn't be too hard to cook something up you can work with.
-
What you want to look for are bucket brigade delay chips. They were out of production for a long time but with the increased interest in analog devices for the music industry they're back in production as of a few years ago at least...I want to say it's designator is SAD1024. It gives about 200mS(I may be off by a little so don't quote me directly) of delay on a chip and they can be multiplexed together for a much longer delay time. 500mS should be very easy to achieve depending on what you want to do. The thing to remember is that the more delay you design in with these the more signal degradation you'll get, which is part of the vibe of the old delay pedals from the 70's and why they're still in demand.
As for your application, google analog delay pedals, and dig around the diy stompbox forums for bucket brigade delay pedal schematics. Then just hack the circuits to your needs. They're pretty straightforward designs and shouldn't be too hard to cook something up you can work with.
Nope . BBD's can do funny things to your analog sound and they are really expensive .
http://www.generalguitargadgets.com/diagrams/pt80techinfo.pdf (http://www.generalguitargadgets.com/diagrams/pt80techinfo.pdf)
-
Nope . BBD's can do funny things to your analog sound and they are really expensive .
http://www.generalguitargadgets.com/diagrams/pt80techinfo.pdf (http://www.generalguitargadgets.com/diagrams/pt80techinfo.pdf)
[/quote]
I checked the PDF. Interesting but also problematic indeed. Nevertheless, it seems to be the best way to start a custom-hack.
Alex
-
Nope . BBD's can do funny things to your analog sound and they are really expensive .
http://www.generalguitargadgets.com/diagrams/pt80techinfo.pdf (http://www.generalguitargadgets.com/diagrams/pt80techinfo.pdf)
IKR, which is why I said you're going to get noticable signal degradation. Those things are fraught with problems, which is exactly why people love them. They have a definite "sound" to them. Tape delays are just as bad. They're just different problems. I recommended BBD chips because I thought he wanted to retain a lot of those artifacts. Alex might be well served to expand a bit on why he's set on staying in the analog domain and what he's trying to accomplish. At which point a solution may become obvious.
-
Looks like you will have to go digital. A stereo ADC with some output, probably SPI or something, a DSP chip with a 256M DDR ram on it for the sample delay ( gives an arbitarily long delay easily) and a DAC to give the output. Clock the ADC and DAC with a 96kHz sample rate ( easy on current chips and gives a full 20kHz with minimal filtering), and use the ram to hold the samples, with a simple recycling loop to write data and another to read out. Difference between pointers is delay, you can get a long delay and will not have to worry about ram refresh in this case at all. With this you can get over 45 minutes delay or near zero.
-
Alex might be well served to expand a bit on why he's set on staying in the analog domain and what he's trying to accomplish. At which point a solution may become obvious.
It is too difficult to explain the whole intention and details of the experiment but with respect to my question I will try to be more specific:
I am comparing two channels of audio. On each channel I am transmitting a single sine wave of 10 kHz. Path A has a cable length of 1 m, Path B has a cable length of 21 m. Hence the propagation time differs by approx. 100 ns. An essential part of the experiment requires that both paths must be delayed by approx. 150 ms and after that delay line I still want to be able to measure the time shift of 100 ns.
So far I assumed that an ADC/DAC converter (as it is the case with a digital delays) is not able to "see" a 100 ns timeshift and would consider both signals as equal (even with 96 kHz sample rate we have less than 10 us time resolution). As a consequence I thought of analogue delays in order to prevent the time shifts. Please correct me if I am mistaken.
Best
Alex
-
Sampling preserves the phase information, even if the time shift would be much less than one sample interval. Phase is encoded in such way that sample values are a bit different for the phase shifted signal. However, analog filtering circuit tolerances around ADC/DAC and tolerances in the ADC/DAC itself may prevent measuring that difference.
Regards,
Janne
-
Alex might be well served to expand a bit on why he's set on staying in the analog domain and what he's trying to accomplish. At which point a solution may become obvious.
It is too difficult to explain the whole intention and details of the experiment but with respect to my question I will try to be more specific:
I am comparing two channels of audio. On each channel I am transmitting a single sine wave of 10 kHz. Path A has a cable length of 1 m, Path B has a cable length of 21 m. Hence the propagation time differs by approx. 100 ns. An essential part of the experiment requires that both paths must be delayed by approx. 150 ms and after that delay line I still want to be able to measure the time shift of 100 ns.
So far I assumed that an ADC/DAC converter (as it is the case with a digital delays) is not able to "see" a 100 ns timeshift and would consider both signals as equal (even with 96 kHz sample rate we have less than 10 us time resolution). As a consequence I thought of analogue delays in order to prevent the time shifts. Please correct me if I am mistaken.
Consider what you are asking. A difference of 100 ns in 150 ms is about 0.7 parts per million. That's a very high precision. Any extra elements you introduce into a circuit must be more precise than this by one or more orders of magnitude or they will distort and destroy your intended measurement results.
Logically you could mix both signals together, put the combined signal through a delay path, and then separate them at the other end. This way any errors will be applied equally to both signals and cancel out (you will have common mode rejection). But if you did this, we could ask what is the point of introducing the delay at all? It adds nothing to the experiment.
In the real world, the only way you are going to see 150 ms propagation delays is with long signal paths like satellite relays. If you want your experiment to make sense you would need to use a real satellite link. If you want to simulate a satellite link you should ask the question in a different way, for example, "I want to simulate a satellite uplink/downlink path, because using a real satellite is too expensive". Then at least you may get some relevant answers.
-
In the real world, the only way you are going to see 150 ms propagation delays is with long signal paths like satellite relays. If you want your experiment to make sense you would need to use a real satellite link. If you want to simulate a satellite link you should ask the question in a different way, for example, "I want to simulate a satellite uplink/downlink path, because using a real satellite is too expensive". Then at least you may get some relevant answers.
Dear Ian,
as I wrote in my previous message I cannot explain the details of what I am doing. It is not that I am naive or do not know what I am doing here. The experiment is complex and related the an experimental physics-background. Further info would lead to a discussion for which I might open another thread sooner or later. So far I am pretty happy with the answers since they give me inspiration how to possibly apply alternative approaches of what I actually want to achieve. Your suggestion to mix both channels to one and separate the two signals afterwards sounds interesting. Mixing them is obviously easy to do but I am wondering how you would separate them afterwards (there is no carrier or such -- just the two single sine waves).
Alex
-
Sampling preserves the phase information, even if the time shift would be much less than one sample interval. Phase is encoded in such way that sample values are a bit different for the phase shifted signal.
Oh -- that I didn't know ! I had thought that such small phase shifts would lead to useless results. Well, in fact my digital scope obviously shows that it does work but I assumed that here additional features are being applied.
However, analog filtering circuit tolerances around ADC/DAC and tolerances in the ADC/DAC itself may prevent measuring that difference.
So I could give it a try with a digital delay and do some trial-and-error to figure the difference between the two channels or applying Ian's suggestion of mixing them to one channel (and separating them afterwards if that works out).
Alex
-
Dear Ian,
as I wrote in my previous message I cannot explain the details of what I am doing. It is not that I am naive or do not know what I am doing here. The experiment is complex and related the an experimental physics-background. Further info would lead to a discussion for which I might open another thread sooner or later. So far I am pretty happy with the answers since they give me inspiration how to possibly apply alternative approaches of what I actually want to achieve. Your suggestion to mix both channels to one and separate the two signals afterwards sounds interesting. Mixing them is obviously easy to do but I am wondering how you would separate them afterwards (there is no carrier or such -- just the two single sine waves).
Alex
Cannot explain or will not explain?
You claim to know what you are doing, yet the facts as you have given them make little sense. That makes it difficult and frustrating to have any meaningful participation in the thread. You are keeping relevant facts secret that might have a huge bearing on possible solutions.
A propagation time difference between two signals will introduce a phase shift between them proportional to the time difference on arrival at the sensor, the 100 ns in your example. If you now introduce an equal delay in both signals of 150 ms the difference between them will remain equal. In other words a delay followed by a difference is the same as a difference followed by a delay. So you may as well measure the difference between the two signals (subtractive mixing), transmit the difference over the 150 ms delay, and then examine the difference signal at the other end for whatever purpose it is needed.
If you should say this is not a suitable solution, then there are more relevant experimental details that you have not conveyed that apparently have an impact on the outcome.
-
You claim to know what you are doing, yet the facts as you have given them make little sense. That makes it difficult and frustrating to have any meaningful participation in the thread. You are keeping relevant facts secret that might have a huge bearing on possible solutions.
OK - I got it and I understand that it is frustrating and for that reason I will provide the whole idea/setup:
One of the domains I am working in deals with superluminal data transmission. I am not sure if anyone here has ever heard about Günter Nimtz ( http://en.wikipedia.org/wiki/G%C3%BCnter_Nimtz (http://en.wikipedia.org/wiki/G%C3%BCnter_Nimtz) ), who is a quantum physics professor in Cologne/Germany. I have been working with him for approx. 2 years and I revisited his tunneling experiments. Please find my latest paper here (I will present it end of April at the AES in Budapest):
http://dl.dropbox.com/u/53934415/CarotAichmann.pdf (http://dl.dropbox.com/u/53934415/CarotAichmann.pdf)
It describes what I have done so far. These effects I have also reproduced with single pulses, however, the majority of scientists still denies the existence of superluminal signal transmission. For that reason I now want to increase the observed time shift by increasing the physical distance in order to generate an audible (and hence more obvious) effect. This, however, results in physical problems since I'd need to have a cable length of 10 000 km or more in order generate clearly audible shifts. As a consequence I decided to still use my 20 m waveguide, apply amplification and feed the output back to the input (or better the inputtrigger, which generates an new single wave). The same thing I do with channel B, which exhibits 100 ns propagation delay. What will theoretically happen is a time shift increase of 100 ns with every "feed back round" to be displayed on the scope and sonified with a speaker.
That's the idea in short. The problem, however, is that the actual pulse length is approx. 25 us or more, which is significantly longer than the propagation time. As a consequence the fed back pulse would arrive at the input before the signal has been completely transmitted and this would mess up the experiment. For that reason I have to apply a delay line, which delays the fed back pulse in order to prevent this overlap. I consider something between 100 and 200 ms useful for now -- just in case but it should be adjustable anyways.
Well - that's the whole story. I am looking forward to discuss further.
Best
Alex
-
No analog device will do what you want. The 100ns is 10MHz; you delay line needs bandwidth that is a significant portion of that, preferably over. You want to have video AD and DA, that go to DC (not hard to find), an SDRAM chip and some control electronics. You might even find this off-the-shelf, search for FPGA video prototype boards.
-
Hi Alex,
Certainly you have an interesting area you're working in, and no doubt made more difficult due to pushing against the mainstream views.
From reading your paper (and my previous experience in academic research albeit limited), the question really relates to how you're going to produce an experiment with inherent unquestionable timing accuracy to prove the proposed theory.
Straight away the idea of delaying a signal by 100ms, with an accuracy to be able to measure, and more importantly prove beyond question a 100ns shift is a very tall ask. Even in your paper (playing devils advocate here), I'm immediately brought to question how you ensured the delay path through the two channels of the oscilloscope were calibrated, along with the delays of the attenuator, the stability of the oscilloscopes clock source etc.
All electrical instruments have clock jitter, aging drift, short term instability and so on. Analogue devices like those that have been discussed so far will be many orders of magnitude too inaccurate for what you're asking. Even the use of an A/D, digital delay, D/A would need to be clocked by some form of precision oscillator (like maybe one of the Rb sources in Daves video), but you'd have to deal with possibly amplifying the heavily attenuated signal and any delay paths introduced in that, then any effects the analogue path from the D/A back to the wave guide.
Hope I'm not going off on a tangent here, but I see this as a study of timing accuracy in all the elements of the experimental setup.
Good luck with your paper!
-
From memory, BBD delays didn't have much bandwidth.
They were great for ADT and Chorus effects though.
The colouration was an asset for those type of effects.
-
Hi Harvs,
Straight away the idea of delaying a signal by 100ms, with an accuracy to be able to measure, and more importantly prove beyond question a 100ns shift is a very tall ask. Even in your paper (playing devils advocate here), I'm immediately brought to question how you ensured the delay path through the two channels of the oscilloscope were calibrated, along with the delays of the attenuator, the stability of the oscilloscopes clock source etc.
sure -- this is all pretty tough and indeed hard to "fight" with the mainstream opinion. However, for me it is one of the most interesting research fields I have ever worked in and after having verified this effect for more than a year of intense calibration work and various double-checking techniques I am 100% sure that this effect is not a measurement fault. My current motivation is to present the effect in a more descriptive way -- rather than displaying time shifts of a scope, which range in areas people do not have an imagination for.
With respect to my "loop"-experiment I first compare two conventional paths (path A will have 1 m cable length, path B 21 m cable length) in order to verify if my assumption of drifting pulses works out. Also here I will have to carefully calibrate the components.
I truly don't want to convice anyone here, I simply ask for constructive feedback or maybe even better suggestions how to make this effect more descriptive and understandable to the world.
For now I'd be more than happy if anyone else can confirm that theoretically (assuming idealized conditions) a two-channel digital delay of 100 ms maintains a time-shift of 100 ns between two signals or if an analogue device might be the better choice.
Thanks
Alex
-
Maybe you could use Grace Hopper's wire technique to implement your delay.
Grace Hopper - Nanoseconds (https://www.youtube.com/watch?v=JEpsKnWZrJ8#)
-
From memory, BBD delays didn't have much bandwidth.
They were great for ADT and Chorus effects though.
The colouration was an asset for those type of effects.
Yes, this is actually the reason why my Roland Spaceecho does not work. Well, but it's still a nice sounding guitar delay.
-
Maybe you could use Grace Hopper's wire technique to implement your delay.
Grace Hopper - Nanoseconds (https://www.youtube.com/watch?v=JEpsKnWZrJ8#)
Very cool -- but since the delay shall have approx. 100 ms we will again end up with our 10 000 km cable ;-) -- BTW: She was a great person !
Cheers
Alex
-
Sorry, I just skimmed the thread and saw 100nS.
I couldn't help wondering why you would want 100nS delay for Audio.
-
Sorry, I just skimmed the thread and saw 100nS.
I couldn't help wondering why you would want 100nS delay for Audio.
Certainly nobody would be able to hear 100nS of delay when it audio there is at least a few hundred nanoseconds delay going on . At least i say .
-
Sorry, I just skimmed the thread and saw 100nS.
I couldn't help wondering why you would want 100nS delay for Audio.
Certainly nobody would be able to hear 100nS of delay when it audio there is at least a few hundred nanoseconds delay going on . At least i say .
Right - you need at least 5 ms (for a trained ear) to notice that there are actually two sounds and 25 ms (also for a trained ear) to tell which one is first. Nevertheless, the nanosecond discussion is only related to accuracy/precision. The challenge is to have the 100 ms delay line as precise as possible and to make it calibratable. At least the thread has given me inspiration to try it with a digital delay. I will post intermediate results soon.
Alex