Long before audio engineers had fancy digital delays, or even crappy analog delays, there were tape delays. Running a tape around in a loop with a record and play head is the basis of the Echoplex and Space Echo, and both of these machines are incredible pieces of engineering.
Microcassette recorders are not, in general, incredible pieces of engineering. They do, however, have a strip of magnetic tape, a record head, and a play head. Put two of them together, and you can build your own tape delay.
The basic principle of a tape delay is simple enough – just run a loop of tape round in a circle, through a record and playback head, record some audio, and send the output to an amplifier. In practice, it’s not that simple. [dogenigt] had to manufacture his own tape loop from microcassettes, a process that took far too long and was far too finicky.
For a control circuit, [dogenigt] is using four audio pots and one linear pot for speed control. The audio pots are responsible for input gain, feedback, the amplitude of the clean signal, and the output of the signal after it’s been run through the delay.
Apart from being one of those builds that’s very dependent on the mechanical skill of the builder, it’s a pretty simple delay unit, with all the electronics already designed for a stripboard layout. You can hear an example of what it sounds like below.
Epic sounds in the demo.
Add some mixers and feedback (perhaps shorten the loop) this could make for an interesting reverb or echo. Perhaps it could be dampened slowly by mixing feedback with silence or simply adding a magnet a set distance away.
This reminds me of that instrument that used loads of tape loop samples for each note.
Perhaps you could put code on them and use each loop as a microcode routine in a home brew computer, thus saving on seektime.
You could use one master and say eight slaves with read heads all the way around the loop. They all read the code from the master but out of phase. Enabling bit rotation and manipulation.
Were you thinking of a Mellotron? Those were cool.
Mellotron wasn’t actually a loop but a bit of tape that rewound when you let the key back up, I like the “data loop” concept though.
great sounds indeed! and looks great also!
check here a Mellotron with one walkman: https://www.youtube.com/watch?v=6jSnSV3SSxE
My favorite type of delay line memory has to be those solid state blocks of quartz you used to get in old tvs and video recorders that stored only one scan line acoustically by bouncing it around in a grid like pattern inside the crystal. It almost seems like alien technology these days.
Paddy Kingsland – eat your heart out… ( https://en.wikipedia.org/wiki/Paddy_Kingsland )
A little known device from the Vietnam War era was developed by Motorola secretly there in the field. It’s spinoff is used in cellphone technology today. One of the key components was a audio delay line. Of course you could just design a millisecond electronic bucket-brigade circuit or just have your PC’s sound card provide the audio delay. But this is a way to go too.
Ideally it was a 2-way radio transceiver that could transmit and receive on the SAME frequency apparently simultaneously without an expensive RF diplexer or duplexer.or even a split site setup involving some sort of long phone line or microwave link. The system would just switch back and forth from transmit to receive at such an extremely high rate using a old-fashioned power vibrator circuit. The human ear would notice nothing but a almost imperceptible hum. The receiver’s power would never occupy the same slot in time as the transmitter’s power hence the receiver was de-sensed from it. Even while using the same antenna attached to the same transceiver.
Now this configuration will only work as a remote base/mobile as stated above and was patented by the railroad industry in the 1950’s to interrupt the train engineer mid-sentence if he was holding down the mic button on his quasi-simplex radio. But how did the Motorola make it into a same-frequency-repeater in Vietnam? How could the transmitter repeat what the receiver was hearing if they never shared the same power-slot in time and still sound like it was 100% full-duplex? A AUDIO DELAY LINE!
Essentially, the delay line stored the human voice comment (bridging the time gap) for a millisecond and repeated it into the transmitter circuit during it’s power time slot, and on and on. Humans can not notice the difference. It appears that the transmitter is repeating on its own frequency in real time BUT it’s actually not. As far as the reciever is concerned it NEVER hears it’s own transmitter but it still functions normally. This is not the well-known Simplex Repeater which is a store and forward system (still pretty handy system though). This is much more complex than that. It is very strange to listen to. Cell phones (i.e. CDMA – code division multiple access) does that today. That’s why you notice a very small imperceptible time delay in time when talking to another person on your CDMA cell phone.It appears to be 100% full-duplex but it’s actually not.
Well Motorola decided to send this impromptu ad hoc SFR repeater aloft on a helium balloon to test it on an active military com-channel. In a matter of seconds it started repeating a LRRP Team that got lost in the bush and was frantically calling for a helo-extraction. For some reason their RF signal was buried in the hills and valleys and could not be heard otherwise. So this audio delay line could be used to make a personal SFR radio repeater. It could apply to any frequency of operation even CB or FRS radio. However, it’s FCC legality is in question though.
(NOTE: FM works better as the FM “capture effect” would mute out the distant FM mobile unit while the FM SFR repeater transmitter was repeating. That way your not hearing two almost simultaneous signals like an echo or reverberation. With AM you might hear a heterodyne tone or reverb if both signals where close to you like someone “double-keying”?
That would make it difficult to listen to. CB – HF 11 meters is AM but FRS is FM. Not sure what SSB would sound like.)
Am I misunderstanding here? If you have, say, a 50/50 split on the time between transmit and receive and you record speech t/2 of the time during receive how do you transmit t speech in t/2 time?
Or was it a precursor to digital sampling, slicing out extremely small segments of audio while still retaining intelligible speech?
I know its a difficult radio concept to imagine. It is thinking “outside the box”.
Humans can not perceive things at such switching speeds. The audio is NEVER switched at all. It is continuous but the start time is delayed a few milliseconds by the audio delay line. The audio from the receiver is recorded continuously then dumped a millisecond later into the microphone circuit of the transmitter. It is time-shifted just a bit to bridge the gap of no audio being recorded during the transmit cycle. But that gap is so small no human could notice it at all.
The recorded audio from the mobile unit (to be repeated) is NOT choppy or distorted. Yes it is sampling the audio at a high rate like over 10,000 samples per second (or more). It only samples the receiver when the transmitter is off and the receiver is on. When the receiver is off and the transmitter is on it plays the audio into the mic circuit. One would think this arrangement would sound awful with horrific choppiness. However, the sampling and playback rate is so fast a human can’t hear any choppiness or even clicks. Maybe a small hum but that was due to older mechanical power vibrators that could only work at lower frequencies in the 1960’s. Today it’s all solid-state opamps and gates.
All of this sampling and stuff would apply to a all-electronic-version. But this slower tape recorder version really doesn’t need all that fancy sophistication. You just set it to record and playback a millisecond later in a continuous tape loop as if it were an echo (but an echo-chamber effect for this application is not desired). Anything spoken into it would just continuously dump into the transmitter’s mic a millisecond later continuously apparent choppiness and all (if any).
You’re ideally only looking to bridge the time audio time gap between switching states. The tape loop could be several minutes long but an additional erase head could remove old audio so it doesn’t repeat the same words again minutes later. A VOX (voice operated switch) circuit for the transmitter would be nice for radio channel etiquette but not really needed as the receiver’s squelch circuit could be polled for RF carrier presence (COR or carrier operated relay) to enable or trigger the overall repeater-mode event. A delayed-drop-out-circuit could be employed to stop repeater-mode event after several seconds of radio silence, awaiting next COR event. For radio channel etiquette you don’t want a free running transmitter with dead-air (no apparent signal).
So yes the duty cycle is 50/50 but the audio is not on any duty cycle it’s a continuous non-stop loop.
I understand the short times involved here, that’s not what I was asking.
If there’s a 50% duty cycle (however fast) and you’re recording audio for t/2 and replaying that audio *with* the live audio, you’re creating overlapping audio, are you not? With such short segments being discussed here would it not be simpler not to play the afforementioned recorded audio at all? I imagine human speech might sound “flatter” but it would still intelligible.
Think of it like decimation of digital samples. With the analog sample rate being fast enough, you could remove every other sample and still have audio; you’ve just effectively lowered the sample rate by half. A millisecond wouldn’t be fast enough, that’s only 1ksps, and you need a Nyquist limit of around 4kHz to 8kHz for human voice to be recognizable. Assuming that they just needed to get a point across without being able to recognize the person by their voice, a 4kHz Nyquist should be fine, so 8k samples per second are needed. So, the radio broadcasts for 1/16000th of a second, then receives for 1/16000th, and repeats. That’s faster than the human ear can hear the switching, so as the OP says it just sounds like a hum.
Now, let’s look at tech from the era, and realize that repeating that at 16Khz would be tough, but 1kHz would be doable. So, instead of recording and broadcasting 1 sample and then receiving 1 sample, the system the OP describes would do something like record and broadcast 1ms of audio, then switch to receiving for 1ms. Yes, detail is lost; but apparently it was little enough detail that the technology found a purpose.
Now I want to grab Audacity and see what it would sound like. Could make an interesting vocal effect.
Oh I’m sorry I didn’t mean to say that you would be “replaying live audio”. That would be two audio sources, the delayed and the live. That would not work out well for this repeater. The overlap would sound echoed. I agree using a tape loop for millisecond repeat is a bit counterproductive in lieu of better electronic methods available like electronic delay limes, quartz crystals like one poster said, or using a PC/MAC sound-card audio repeat routine. In any case you have to get the audio from the receiver’s speaker to the transmitter’s microphone somehow. It just wont happen in real-time (i.e. a simple audio patch cable attaching both) as the actual received audio in real-time is missing during the transmitter’s duty cycle as the receiver is effectively muted due to not having power to it’s circuits for a split second. So you’d be repeating nothing at all. Audio needs to be delayed somehow for some arbitrary amount of time in the milliseconds (experimental figure?). When this thing is operational the audio would not be so flat as you could get the switching rate as high as the transceivers components can handle. In 1960’s it was 10,000 switching cycles per second. That would handle a 10Khz bandwidth which is well within standard voice bands, Telephones have 300-3,000 Hz bandwidth and you can hear the caller pretty good.
RF discrete components in most RF transceivers today are rated much higher than the transceiver’s actual operating frequency. So today you’d probably be able to surpass 10,000 switches per second by maybe 100 times or more. But the audio bandwidth is regulated pretty heavily in FM transceivers with companders, pre-emphasis, and de-emphasis etc, So if you could achieve 20,000 SPS you’d be doing pretty good in avoiding flat audio in this repeater. That would translate to 20Khz bandwidth but I don’t think the standard FM transceiver could achieve that on it’s own. I think 5Khz is the best you could achieve so the 10K SPS seems enough here.
The actual Patent by the Motorola engineers is right here ( http://www.google.com/patents/US4475246 ). They waited 20 years to patent it due to it being classified for awhile. What they used for an audio delay was a long “spring” meant for audio delay. Imagine how that must have sounded.
This SFR repeater was manufactured in UK by a UK Defense Contractor. I saw a data sheet on the Internet somewhere but I can’t remember where now. It is still being used by UK military. So I guess it still classified in UK at least. I know amateur radio enthusiasts are using it in UK and call it OFR (On-frequency-repeater).
Using quartz crystals for audio delay is interesting. Dr. Ron Mallet of UCONN in Storrs CT (time machine fame?) actually uses “photonic crystals” to “slow down light” – huh? He is trying to do frame dragging to make a gauge boson go back 1-second in time. He is trying to make a time machine messaging system. He claims he has achieved it. Wow! I don’t understand any of it.
I understand your question as being, how do you extract audio from an analog delay line in bursts, twice as fast as it went in? At least with a bucket brigade style delay, you can double the clock rate and get the audio out in t/2, but that will also mess up the audio going in. It seems impossible.
I’m not familiar with the system [sonofthunderboanerges] describes. It seems it simply uses a delay line to fill the gap during TX with a repeat of the prior RX timeslice. Half the audio is actually lost. Ever played a CD with an unreadable sector, so the player fills the gap with a repeat of the prior sector? Imagine what would happen if every other sector was unreadable, and you’ve almost got it. Just faster, with each sector containing much less audio; so that it’s no longer perceivable as a repeat, but as distortion. Although it wouldn’t sound great, with sufficiently fast timeslices it would be intelligible, and certainly better than the audio cutting out altogether. Good enough for the day.
But assuming one had access to more modern parts like bucket-brigade delays, you could do this better, with virtually no audio loss, while still keeping it totally analog. These may be closer to what you were originally imagining:
Method #1: Use stereo transceivers, and two delay lines per transceiver; one each for TX and RX. On the first transceiver during its TX timeslice, transmit live audio as the left channel, and delayed audio as the right. On the second transceiver during its RX timeslice, play the received right channel live. The received left channel goes to a delay line to be played during the upcoming TX timeslice.
Method #2: Use mono transceivers, and four delay lines per transceiver; two each for TX and RX. The microphone goes into both TX delays simultaneously. During the first TX timeslice, audio is clocked out of TX delay #1 at double rate, which corrupts the incoming audio to that delay. But delay #2 continues recording audio properly, so it is used for the next TX timeslice, and it continues switching between delays lines on TX timeslices thereafter. On the RX side the inverse occurs, with speaker playback occurring from whichever delay isn’t being used to receive data.
“That’s why you notice a very small imperceptible time delay…” If it’s imperceptible you can’t notice it. ;) Sometimes I hear my own voice echoed back a fraction of a second later on my cell phone. Then there’s all the times it **eak* u* l*k* **is, all one can do is hang up and try your call again.
run the delay fast enough and you can emulate steeo.
run it slow enough and you can make a censoring box for radio where you can easily lose your license or get fined if offensive content is aired.
Cool idea! Now go listen to one of my all-time fav records, the 1974: Evening Star by Fripp&Eno. They used two revox77 reel to reels and the distance between the machines governed the delay.
Very cool and fun project :)
Is there any way to use this system as a secure voice communicator? It brings to mind a system once known as Fast Talk I think the USN used it back decades ago. You’d record your message at normal speed. Then tell the recipient to start recording on his end. Then you’d play your tape in reverse and at high speed. It would be so high it almost surpasses the comm system’s bandwidth limits. So keep the playback speed down to compensate for that. The recipient would play it back in reverse and slow it down too hear the normal message.
It’s not very secure but it sounds like noise when played. Not like Donald Duck but a Chipmunk on steroids! People monitoring would not recognize it as human communications. It would sound like bursts of noise.
I wonder if it would work?
The demo would be far more effective if recognizable audio like music or voice or pure sine wave tones was run through it instead of noise where it’s difficult to tell if there’s any delay effect at all.
Robert Fripp and Brian Eno would be proud, this is such a sweet hack!
It seems like you could just wrap the tape around the circumference of a disk. Place the record / play heads around the circumference, and change the speed of the disk to change the delay. Additional effects could be had with multiple play heads / multiple record heads / mixer(s) / direction of tape disk rotation / and changing the spacing between record / play heads.
i did it before with a 5.25 floppy
I must have missed this the first time around. I have junk reel to reel machines and a scanner slide to mount the head on like an Echoplex. I need to hack.
I could not hear a thing in the video but faint noises. Re shoot and turn up the volume.