Echotapper

I never understand these discussions what is the best echo?

Developing patches that are fit to play "The Shadows" music is key as those patches are instrumental so to speak in the sound creation process. The objective of this is information sharing is, to an extend, that the interested guitar player can use this information as a basis to create his own patches.

The information provided is based on the collaboration of Jacob and Piet over the last years building JPatches for a range of echo devices including the design and build of free VST modules emulating echo devices.

1. Introduction

There are some discussions ongoing that there should be a correlation between the head timings, amplitudes and equalizing curves is a correct one from an analyst point of view. Somewhere else we have read a message that clearly stated that there was no difference between a Meazi Echomatic and other devices. We do believe that this is does not make sense as both character and exact head configuration matrices (timing , volumes including feedbacks) are key characteristics for our music.

As a minimum requirement to emulate the type of echoes wich H.B. Marvin uses the minimal requirements of the unit you would like to build patches for should have:

Up to 6 replay heads providing 100% coverage or 4 replay heads providing <90% coverage. All heads are to be capable of variable timing and last but not least 3 feedback loops from the last 3 replay heads in an logical AND/OR configuration. Although this information is intended to cover all tracks we know that early recordings are done with other equipment including tape manipulations on normal studio recorders. Obviously providing patches in relationship with recorded tracks and used echogear is not fully addressing this "non tape/disc echo era" but more towards a correct life performance quality.

Charlie’s EFTP patches are, as an example, a combination of an exact echo device and some additional seasonings to make it “as the recorded track” or “towards his point of reference” being his famous ears including exact measurements on the record. The same applies for some other echo device creators also!! In this write-up we will address this in more detail.

A good point, made by one of the guitar players in that forum, was made to limit the importance of a correct echo against playing skills. As the maximum echo pulse of the first replay head is normal 3-6 dB less in volume then the dry signal his observation is correct. However, as the timing of the echo patterns from the vintage machines are not really having a full relationship with the BPM (Apache is good example) they are therefore more distinct then when “in sync” hence of instrumental importance so to speak for our music.

A highly recommended discussion on various echo systems quality is, to my knowledge, not adding too much value as all parties are not touching base on the inability of providing patches on gear that’s not fit to be used to produce what’s in the mind of each programmer to produce perfect echo patches.

We have noticed a policy developing in the use of the Q20/EFTP as the point of reference of all what’s perfect. I am not sure if that’s already putting too much water in the wine as this device it’s just not capable producing genuine echo's ala disc/tape echo’s (ala TVS3) as the available “pre- programmed routines are just not covering the aspects you would like to get in or loading the device to an extend that they cannot being left in due to overloading! On the same token, assuring I have covered them all, the TVS 1 and 2 are depending about correct timing amplification of the “host sequencing device” hence are candidates to be referred to as a collection of echo gear.

Let’s leave this discussion of no winners an focus on an assumed correct method to attack the problem of creating/developing optimised patches for equipment/software capable of simulating the old vintage units.

2. A general model

We guess showing the basic inner workings of a digital echo device provides some insight towards the elements required when considering tasking yourself with building patches or even an echo device like this:

Your mental model could be obstructed by the way the heads are placed in sequence in the real world. In the model presented they are put in parallel so they will "only" accept the signal from one input. The connection to the actual delays is created from a direct signal and a signal that's routed through some flangers and tremolo. The flangers are there to mimic the wow & flutter and the tremolo to mimic the slow volume oscillation due to miss alignment of the tape/heads. All is kept very rudimentary the real world requires some more here.

After mixing those two signals back to one a shelving filter is used to assure that at each feedback of signals already treated by this filter acquire a slightly more narrow bandwidth (the shelving filter will slightly amplify the centre frequency so this part gets some slight amplification over the other frequencies)

After the signal passes the "replay heads" the wave shaper is modifying the signal so it's getting saturated like a tape device. I guess it's now time to jump in to the details of the patch building steps?

3. Defining the date of recording

For defining the exact date of recording the book “That Sound” written by Roberto Pistolesi, Malcom Addey (studio engineer) and Maurizio Mazzini provides on pages 102 - 115 exact information (publisher Vanni Lisanti 2000). By defining this exact recording date one can also define what equipment was used since this would be part of a grouped recording (one echo device -> many recordings/day)

4. Analysing the record and deciphering the timing

Using a sound editor like CEP/AA or Steinbergs WaveLab
First, define the area where the echoes are very clear. Look and listen for mute type of sounds where the echo can clearly be heard:

When distinctive echo's are found,select that area to obtain only those detail:

Zoom to this area and select from the [VIEW menu] Spectral view to make reading and measuring easier.

Now you would see the echo trails as light pulses in your view. Use the cursor and drag from the original pulse to the first echo and write down the value. Then use the cursor and drag from the original pulse to the next echo and write down the value.

Continue this process for all the echo's you can find, belonging to the original pulse. Obviously these pulses are not only representing the replay heads but also the signals that are reintroduced through the feedback loop(s).

5. Define head timings and vintage device

Here is a table showing replay-head timings obtained from actual measurements using the Creating fingerprints process as the toolset.

6. Using defined amplitude rules as per devices head configuration

Besides the replay-head timings the amplitude per replay-head in conjunction with the recorder head (our dry/wet ratio that is!!) is required. For the Meazzi Echomatic #1 and #2 the replay-heads are configured with paired resistors as per table and schematics.

	Meazzi Echomatic #1 partial schematics	Meazzi Echomatic #2 partial schematics

The information on resistors used is from the actual schematics of the Meazzi device. Our general understanding is that the paired replay head attenuations should be fully reflected in the patch designed.

• For the Meazzi#1 would that be 2K2/2K2/4K7/4K7/10K/10K/ and the last 10K could also be 15K in a particual configuration.
  For the Meazzi#2 this would be 2K2/2K2/4K7/4K7. A 10K potentiometer is part of this configuration impacting the first 3 replay heads.

• A good initial value when building a patch is to keep close to those values ( adding a wet dB value that's acceptable like about -6dB as an initial value) to assure it is configured like a vintage system.

• One should also take into account the tolerances of the heads being around +/- 20% due to tolerances and the wear and tape contact of the heads directly contributing to a different attenuation. As a general statement the build-up of the electronics is done with off-the-shelf- items so the tolerances are rather high.

Note: measurements can be done by feeding a toneburst of 440Hz (tone -A-) through the loaded echo patch and readjusting the received signals by selecting the dry signal as a reference to reflect 0 dB first.

7. Using defined frequencies curves per device

For each device a specific curve is required to create that specific sound. Care is to be taken to adhere to this as it directly affects the sound and some specific behaviour in the feedback.

8. Using defined wow&flutter characteristics per device

First the definition: A form of distortion caused when a tape transport or a turntable is subject to rapid speed variation. Irregular motion of the recording medium sometimes induces forms of frequency modulation. "Wow" usually refers to the range of fluctuation frequencies between about 0.1 Hz and 10 Hz and is perceived as pitch fluctuations.

"Flutter" usually refers to fluctuation frequencies above about 10 Hz. To provide a common understanding the following is to show the basic components under discussion here:

W&F Calculations done based on the assumption that the rubber pulley, the capstan and, if available, the other pulley are part of a spring model that will introduce vibrations to the tape related the actual frequencies generated by the part itself. The main objective of this exercise is to define if a set of 2-3 frequencies is sufficient in modelling W&F.

It seems that there are 3 basic parameters in the tape echo's to emulate W&F.If we limit ourselves to the tape echo we can simulation to prove that sufficient random behaviour was there using only those 3 variables is shown in the picture below. Although we recommend one flanger to simulate this effect it is advisable to use at least 2. Those two frequencies will also "generate" a third frequency. In the Alesis Qxx family, a LFO function is also available to produce this:

For the disc echo's the complexity is more towards not available information on the pulley/disc interface. When we assume that the pulley is driving the disc is having a thickness of say 6 mm. then the speed difference on the disc would be +/- 3mm over the point of the actual position of this pulley below the disc. The disc itself is exact 100mm and it is assumed the average speed 12 IPS.

If we would assume 12IPS for the tape speed (part of the disc) and a disc diameter of 100mm and a connecting area from the pulley to this disk of +/- 3 mm the theoretical "Wow" would be around 0,212 Hz. Obviously there are more effects in that area of contact but value provided should provide a start that's based on reasoning.

9. Add tape saturation curves

A magnetic tape recorder's harmonic-distortion specification is very important. It usually determines where the record level of a recorder's electronics should be set. The record level is also used to determine the signal-to-noise ratio and frequency-response specifications. Moreover, The primary harmonic distortion in magnetic tape recorder systems is third-order harmonics. If the level of third-order harmonics in a recorder increases, the level of distortion will also increase.

Tape behaviour is linked to the BH curve of magnetics (see picture above). so we are using this curve as the reference for the saturation curve. We do not take into account the other artifacts related to head/tape spacing/clearance here. Our effort to emulate this saturation we use a tangent function tanh(a*x) to achieve this. In building an echo device wave analysis measurements are to be taken to assure this function is not overdone.

10. Add valve characteristics

Valve saturation is key to assure that even-order harmonics are generated as those harmonics are sounding as musical chords (notably octaves), which subjectively makes the sound "richer".

A wave shaping function in the echo device, when available should be used to obtain this effect.

11. Build the patch

So we have arrived at the core of our objective, building the patch with the information we have obtained through all the process steps:
Let's take one example to go through all the steps covered:

Example: The Stranger

Initial process steps for building a patch	Results
Step 1: Defining the date of recording	Recorded 06-10-60
Analysing the record and deciphering the timing	Analysed: 123mS,424mS and 596mS
Step2: Define Vintage Device by using the head timings	Meazzi Echomatic 1 Disk Echo
Step3: Using defined amplitude rules as per devices head configuration	-2dB -2dB -4dB -4 dB with variable heads 1,2,3 as one set through 10K potmeter. Remember to add an initial wet value of about -6dB to signal feeding the heads. So the Wet_in + Head would be possible around -8dB as an initial value before the actual measurements are started.
Using defined frequencies curves per device
Step4: Using defined wow&flutter characteristics per device	Use flangers. Although 3 are required it could provide exessive loading to your device so use two as being explained in this write-up.
Step5: Add tape saturation curves	Pending capabilities of device. A model would be a signal measurement based on 20Log and the output controlling a soft limiter (who creates the required harmonics). Saturation curves are always related to specific tape characteristics. Where the signal is entering the "RED" domain of that particular tape.
Step6: Add valve characteristics	Pending capabilities of device

12. Try the patch

Obviously, the ear is very important to establish if the patch produced is close to what's expected. We are using a process that's covered under the creating fingerprints chapter that allows for measurements to be taken from the patch.

As we have focused on the echo part of the total signal chain the other items should also be fit to be used to establish if the echo is correct. It is therefore recommended to use a correct speaker/amplifier and to apply a compression between 1:2 and 1:4 so it's more towards the way it was recorded.

13. Improve the patch

Finally, we have arrived at the core of the activity. Building the patch was completed. A written record of the exact settings of that patch should be kept and when small alterations are done this record should be updated to reflect that.

Interesting is to know that it is possible to use a mother child relationship between the patches. So, there are basically only about 8 patches and the rest is derived from them directly. The derived patches only differ in different heads used and different feedback(s). The actual task seems at first a big undertaking with endless alterations. When using the mother <=== child relationship the light will eventually shine at the end of the tunnel sooner!

14. Creating fingerprints

Understanding measuring devices and calibrating resulted patches is the main objective of this write up. Obtaining fingerprints from each echo device that’s popular for the music we like to play is key to understand it’s capabilities and possible improvements. Also, having those values will help to create the initial settings of the designed patches including the fine adjustments.

The wave we have developed (part of http://www.echotapper/) is optimised for all factors required to build/calibrate patches. It provides two ranges of short tone bursts to check the throughput characteristics of the device. Each range will fit a particular equalizer to ease adjustments. Also a sweep is added to obtain initial frequency curves and, finally, a long tone burst to measure wow&flutter artifacts.

Here is a total overview of that calibration wave with resulted waves obtained from eTAP2:
This plot of the results obtained by applying the mentioned test wave file part to the echo device under investigation or test. The dry pulse is here the upper channel and the resulted pulses are in the lower channel as processed. In this particular case we have used the eTAP2 device as the device under test.

This wave file contains all that’s required to measure/adjust all what is required. The first pulse is a 440Hz pulse (tone = A) then 63Hz, 250Hz, 1kHz, 4kHz and 16kHz. Together with the 440Hz pulse you have 6 different pulses who are directly compatible with various 4-5 pole equalizers on the market so adjusting filters is rather simple (but time consuming). The log sweep, as the next burst is used to create the 3D plots and to fine-adjust filters. The next set of pulses is for other frequencies so 3pole filters and others can also be correctly measured. The last burst is a long burst so measurements on W&F can be achieved.

When the 440Hz+5EQ+Sweep+4EQ+WF_Burst wave file is used to measure vintage gear the resulted file should be normalized so the dry pulse should be exactly 0 dB. This makes recording the exact required attenuations simple and correct. Let’s have a complete overview from an echo device obtained from the described sweep part of the test file:

Although at first difficult to read and understand on this 3D plot ( based on the Log Sweep 63Hz- 16kHz) for a vintage echo you will notice that the bandwidth is just 5kHz with the centre-frequency as shown. As each heads transfer function is diminishing the throughput each time the echo packet passes the bandwidth will be smaller and smaller again. The top line shows the decay over time and provides a measure towards the used amplification (if you put the head timings on the time axis). On 0 to about 70mS there is no echo yet so that area gives a perfect view of the dry curve of the echo device.

The lower channel is in stereo with the upper part the dry+wet part and the lower part is just the wet results.
If we look in more detail the first sample we use to calibrate the patch is the 1kHz sample:

This sample together with the other samples are measured to calibrate the throughput curve of the echo device we would like to emulate. This sample is also used to define exact timing and amplitude information from the echo device under test. Although not shown here, the actual dB measurements on a pulse set like above is always done by calibrating all pulses. this is done by measuring the first dry pulse and adjust the amplification to all pulses to make this first pulse 0 dB. This next curve is the sweep we use to define the way the filter should be configured this together with the mentioned measurements. In most cases the course adjustments are made just with this sweep and the fine adjustments with the obtained pulses

When we design a new set of JPatches the measurements we have archived are then mirrored against our work so it’s the main workbench to produce accurate patches. Obviously the ear also comes in for fine tuning for the final adjustments.
The modern way of fine tuning echoes towards music is only BPM so it’s always related to the beat. I believe as a non musical guy that it’s the creative process that defines the best use of heads in that resulted series of timings including the amplification. Obviously, from a taste point of view, a head that’s too loud or even more then the dry signal is not palatable by most people.
We do not believe that echo devices are to be used to emulate reverberation as this domain relies totally on multifaceted room/reflection/absorption parameters (reference: the Sabine formula) where the characteristics are mostly coloured by the requirement of totally blocking the lower frequencies while the echo devices throughput (wet) would be of a complete different nature.
So, to our knowledge, echo devices are there to fill a specific need by in-sinc behaviour with the music or even better to extend the instrument’s capabilities in the macro/micro domain of timing and sound.

As a sidestep: reverberating is a normal (additional) step in the sound chain. If it’s to be done ala abbey road there is always a pre-delay involved of about 80mS- 100mS this to mimic the tape recorder setup at the studio. By doing so, the reverb will be moved away from reinforcing the direct sound (early reflections < 80mS) making it more distinctive.

Most of the vintage echo devices are with fixed amplitudes or attenuation of the tape playback heads are normally controlled by fixed resistors in series with the signal that feeds the preamp valve see Head amplitudes on http://www.vividinteract.com// for details.
As you possible could agree, it’s a domain that’s limited by the capabilities of the gear used and last but not least the artist himself (again, limited by the gear),

As we have already touched base on the signals and derived echo pulses it is possible also a good idea to shortly touch base on the general frequency characteristics of an echo device in particular eTAP2: For vintage gear the throughput curves as measured on the actual gear or through Microcap’s simulation of a working model are shown here. The curve’s pivot point is around the 1kHz for the Meazzi’s while other devices are slightly different: Using defined frequencies curves per device provides the curves key for each covered device.

So, there you go!! This document is a living document and will be augmented over time. Let me know through known channels if there are area's that should be elaborated more upon.

(c)2009 echotapper.nl

-Piet

aa