Always consult the official Owners Manuals first

Delay block

From Fractal Audio Wiki
Jump to navigation Jump to search

Available on which products

  • Axe-Fx III: 4 blocks
  • FM3, FM9: 2 blocks
  • Axe-Fx II, FX8, AX8: 2 blocks

Channels or X/Y

  • Axe-Fx III, FM9, FM3: 4 channels
  • Axe-Fx II, FX8, AX8: X/Y

Delay block diagrams

Delay block - Mono.PNG

Delay block - Mono Tape.PNG

Delay block - Stereo.PNG

Dual Delay.png

Delay types

Firmware 17.01 and later, and corresponding FM3 and FM9 firmware:

  1. 2290 W/ Modulation
  2. Ambient Stereo
  3. Analog Mono
  4. Analog Stereo
  5. Deluxe Mind Guy
  6. Diffused Delay (not available on the FM3, because the FM3's Delay block doesn't support Diffusion)
  7. Digital Mono
  8. Digital Stereo
  9. DM-Two Delay
  10. Dual Delay
  11. Ducking Delay
  12. Graphite Copy Delay
  13. Lo-Fi Tape
  14. Mono BBD
  15. Mono Tape
  16. Pan Delay (Axe-Fx III only)
  17. Ping-Pong
  18. Reverse Delay
  19. Stereo BBD
  20. Stereo Mind Guy
  21. Stereo Tape
  22. Stereo Trem Delay (Axe-Fx III only)
  23. Sweep Delay
  24. Vintage Digital
  25. Wandering Delays (Axe-Fx III only)
  26. Worn Tape
  27. Zephyr: FAS original [1] (not available on the FM3, because the FM3's Delay block doesn't support Diffusion))
  28. Dual Head Tape Delay (added in firmware 23 and corresponding FM3 and FM9 firmware)

The following blocks are also capable of providing a delay effect:

Looper functionality (also a delay) is provided through the dedicated Looper block.

Additional information is in the Owner's Manual.

Deluxe Memory Guy


[2] The model is matched to an original, not reissue, DMM.

Vintage Digital Delay


[3] It uses virtual crappy converters with low bit depth.

[4] The whole point of the Vintage Digital type is the artifacts. It's modeling an early, primitive digital delay w/ an 8-bit converter.

[5] Just modeled it after how devices in that era worked.

DM-Two Delay


[6] […] the DM-2 has a pre-emphasis/de-emphasis topology. The input is pre-emphasized by over 15 dB. The pre-emphasized data goes into the delay lines. When you switch to a different type with no de-emphasis, which is most of the types, you are getting pre-emphasized echoes out which will sound louder and brighter for a brief moment.

The fix is to use two delay blocks and switch between them.

True Tape Echo

The Tape Echo mode type is an authentic model which works in mono. It's based on a different algorithm than the Stereo Tape Delay.

Firmware release notes:

Added Tape Echo algorithm to Delay and Multidelay blocks. This algorithm simulates a tape echo where modulation occurs due to tape speed variation. In the Delay block the algorithm is implemented as a two head monophonic tape “deck”. The Time/Tempo parameters set the distance between the record and first playback head. The Ratio parameter sets the relative distance between the record and second playback head as a percentage of the first playback head. The Multidelay block is implemented as a monophonic deck with four independent heads. See the updated manual for full details.


[7] The algorithm is a single tape with one record head and two playback heads. The signal from each head can be fed back (Feedback) and panned and leveled.

Head 1 Time sets the delay time of head #1 when the Motor Speed is 1.0. This is analogous to the distance between the record head and the playback head in a tape deck. As you increase the time you move the heads apart and vice-versa. Whatever the time is set to is the delay time when the Motor Speed is 1.0.

As you increase the Motor Speed the tape moves faster so the delay time decreases and vice-versa. The relative position of the heads do not change.

Ratio sets the delay time of head #2 as a percentage of head #1. If you set Ratio to 50% then the delay time of head #2 will be half that of head #1. This is analogous to positioning head #2 exactly half the distance as head #1 from the record head.

In brief: Delay Time = distance between heads, Motor Speed = tape speed.

The LFOs modulate the motor speed. Modulating the motor speed (which changes the tape speed) is a very different effect than modulating the delay time. You can hear the difference between by varying the Motor Speed manually and the Head 1 Time manually.

In a typical tape deck the tape speed is not constant. In the days of tape recorders much effort was expended on trying to keep the tape speed as constant as possible. This included periodic maintenance. However, when applied as a delay effect, the variation in tape speed could create a desirable modulation. Tape Echos, in particular, were built to cost points and suffered from considerable tape speed fluctuation.

The two primary components of speed variation were termed "wow" and "flutter". Wow arises due to low-frequency variations in the tape speed. This is typically due to variation in the speed of the motor itself and/or eccentricity of the pulleys. Flutter is a high-frequency modulation and arises due to eccentricity of the capstan and pinch roller.

javajunkie attributed this to Cliff:

The LFOs actually modulate the tape speed. It’s the same as attaching a modifier to the tape speed.

This is because in an actual tape deck the tape speed is what is typically subject to variation. Wow occurs due to variation in motor speed (or belt/pulley, etc.) and flutter is due to capstan and pinch roller. Wow varies the tape speed slowly, flutter is a high-frequency variation in speed. The actual delay time is very stable since the heads are stationary and the tape is taught between the heads. Delay modulation due to tape stretching is negligible.

[8] […] Originally all the delays "chirped" when changing the tempo (like many other effects processors). I rewrote the primary delay types so it cross-fades between tempos when you tap a new one to avoid this. The tape delay type doesn't do this so that you can use it as you would an actual tape delay and get all those cool effects by changing the head position or motor speed.

[9] The tools are already there for wow and flutter. Wow is modulation below 6 Hz. Flutter is above 6 Hz.
Wow is low frequency modulation. It occurs with records when the hole in the middle isn't perfectly centered or when the record has a warp or due to motor issues. It occurs with tape decks due to the tape being stretched or the motor not being regulated well. Flutter occurs in tape decks primarily due to the capstan being eccentric.

[10] That algorithm does not support bit reduction.

[11] LFO 1&2 modulate motor speed in that type.

The FM3 factory preset CLOUD DELAYS has three tape delays.

Reverse delay

Also known as: backwards delay. It's a type in the Delay and the Multitap Delay blocks. You can also use Pitch Crystals or Multi Delay Plex Shift, with Pitch set to 0, and Direction set to: Reverse.

A little Diffusion, low Time value and increased Crossfade help to smear the initial attack points.

Set Feedback (number of repeats) to 0 for best results. To get multiple reverse-only repeats, send the wet signal through a separate forward delay. [12]

Search the Factory presets page for reverse or reversed. Check out the presets BASS ACKWARDS and REGNAL EREVER.

2290 w/ Modulation

The 2290/mod delay type has the Phase Reverse parameter set to Right. This will make the delays disappear entirely if the signal is summed to mono after the Delay block (for example when placed before the Amp, or when using a mono cab), due to phase cancellation.


[13] The original 2290 was a “bit slice” processor. The sample rate was necessarily high because it was a one-bit converter.

Delay block position on the grid

Just like in analog rigs, there's a difference between the delay effect positioned before the amp and after the amp, especially when using distorted amp tones. It's up to you. Most people place a delay after the Amp block, but delay-before-amp certainly has its own charm.

Before the Amp block 
 This will distort the delay trails and make them louder because of compression.
After the Amp block 
 This results in cleaner-sounding delays.


[14] Sounds like you have the delay before the amp. This is normal behavior when putting a delay before a distorted (and therefore compressing) amp.

Regarding the position of the Delay block relative to the Reverb block:

[15] If there isn't distortion or modulation in the delay/reverb then the order is irrelevant since they are then Linear Time Invariant (or shift invariant in digital parlance). If there is a small amount of distortion or modulation then the order is probably still irrelevant. If there is a lot of distortion or modulation then the order may make a difference. However, typically the biggest difference, as noted above, is series vs. parallel since h1(t)*h2(t) is not the same as h1(t)+h2(t). If LTI h1*h2 = h2*h1. It may seem counter-intuitive that the order doesn't matter but try it and you'll be surprised.

Read PRE and POST routing for more information.

CPU usage

Some Delay types, such as Tape and Ambient, require more CPU than other types.

On the FM3, the Delay blocks share a core DSP with amp modeling. To avoid overloading that CPU, the Delay blocks on the FM3 do not support diffusion.


[16] We had to make a management decision. Under certain circumstances (using an Amp block and two Delay blocks with Diffusion on) the second DSP could get overloaded. The easiest solution at this time was to remove the Diffusion parameter. If we are able to improve the code efficiency we will reinstate the parameter. You have to realize the FM3 has only about 1/5 the DSP power of an Axe-Fx III. Trying to fit all those goodies in there while not sacrificing audio quality is a balancing act. I ALWAYS insist on quality over quantity. That's what differentiates our products. It takes a LOT of horsepower to do it right.

On the FM9, the two Delay blocks run in a dedicated DSP core, without much of an impact on the the overall CPU usage.

But features added to the Delay and Reverb effects can overtax the effects’ processor in extreme use situations. Therefore, when viewing a Delay or Reverb effect editing GUI screen on the FM9 hardware, a CPU meter will appear on the display if the cpu usage moves past 75%. As with the primary effects core, the recommended maximum cpu use value is 80% to ensure best performance.


Read this: Spillover



Adjusting the Mix parameter has no impact on the dry signal, unless it goes beyond 50%. This only applies to the Delay block. It's referred to as the Mix Law for the Delay block.


[17] […] Changed mix law for Delay block. The dry signal now stays constant at unity until Mix reaches 50% then decreases linearly to zero. Conversely the wet signal starts at zero and then increases linearly to unity when Mix reaches 50%. This eliminates having to compensate for decreased dry signal when increasing the mix.

Time and Tempo

There are two ways to set the delay time:

  • Enter a specific delay time in milliseconds, or
  • Set the tempo in relation to the global preset tempo (Tempo button), say 1/8, 1/8 Dot, etc.

Important: if you use the Tempo method, the menu won't let you enter a delay time in milliseconds anymore. To do that, first set Tempo to None.

Maximum delay time is 16 seconds on the Axe-Fx III and FM9, and 8 seconds on the FM3.


The Delay block contains a parametric EQ for the wet signal. Low cut and high cut have adjustable slopes. The EQ stuff is on a separate GUI page with a graphical display.

Forum discussion about EQ-ing repeats


[18] IMO, what people like about analog delays are the narrow bandwidth. The noise, aliasing and crud is debatable. The early analog delays typically used fixed anti-aliasing and reconstruction filters. These filters were designed for the worst-case scenario: maximum delay time. Typically at maximum delay time the clock frequency was only 4-5 kHz IIRC which means the filters need to be < 2 kHz. One popular pedal had the filter at 1.75 kHz. They filters also have a very steep cut-off, typically at least 4th-order but usually 6th order. The slope is given by order x 6 so for 6th-order you would set the slope to 36 dB/oct. For those who are interested there's a paper on DAFX about it.


L/R Time Ratio

Tip: Try 61.8, which is also known as: The Golden Ratio.

For more information see Wikipedia's Golden ratio article or watch Leon Todd's video below.


The Drive parameter determines the amount of distortion added to the echoes.

Turning up Drive will decrease the level of the repeats.


[19] Some blocks have clipping simulation, the delay blocks among others. This simulates the input of the delay line being overdriven (like a real analog delay). It also prevents runaway if the feedback is >= 100%.

If it didn't do that the repeats would get louder until they started clipping.

Time Offset

The Time Offset parameter (Mono Delay only) allows adding up to 100ms of delay to the right wet signal, which can be used for widening effects.


When HOLD is activated, the wet input to the block is muted and feedback is set to infinity. This can be used to achieve pad sounds and drone notes/chords.

When set to STACK, incoming audio is stacked on existing audio and held. When set to HOLD existing audio is held.

The Stack Feedback and Hold Feedback parameters allow adjusting the decay time independently for the stack and hold modes.

On the FM9 and Axe-Fx III, Stack/Hold repeats are infinite (or nearly, may degrade over many minutes/hours) unless:

  • Compander is enabled (Stack/Hold is not available and will be disabled), or
  • Bit Reduction is greater than zero, or
  • the delay type is Tape (tape algorithm is inherently lossy like a real tape).


[20] Note that if the Compander is enabled Stack/Hold is not available and will be disabled.

[21] (Repeats degrade using Hold on the FM3) The FM3 is not an Axe-Fx III. To implement this in the FM3 would increase CPU usage which is not worth the trade-off.

Hold can be controlled with an External controller. When attaching a pedal to Off/Stack/Hold, Heel is Off, Stack is middle and Toe is Hold.



[22] To gain basic understanding try extreme settings in an experiment. Start with a "hot" delay mix of about 50%.

The first way you might do this is to set DUCKER ATTENUATION to 80 dB and then adjust the DUCKER THRESHOLD, from the top down—starting at 0.0 and lowering it to hear how ducking kicks in. This will teach you where your threshold is most dynamic.

Another good experiment is to set the THRESHOLD all the way down to -80dB and then increase the DUCKER ATTENUATION from 0-dB upwards. This helps you understand what a reduction of "X" dB sounds like.

Once you have extreme ducking dialed in, you can play with RELEASE to see how it works.

After this, you should be able to set all three controls so they suit your needs.

Master Feedback

This goes up to 200% to create oscillating delay effects.


Increasing Diffusion smears the separate delay trails, creating a reverb-like effect. It increases CPU usage a little.

Diffusion has its own LFO modulation parameters, demonstrated by the Zephyr delay type.

The Delay blocks on the FM3 do not have Diffusion parameters to save CPU usage (they share a DSP with amp modeling). This means that certain delay types are not available on the FM3.


[23] We had to make a management decision. Under certain circumstances (using an Amp block and two Delay blocks with Diffusion on) the second DSP could get overloaded. The easiest solution at this time was to remove the Diffusion parameter. If we are able to improve the code efficiency we will reinstate the parameter. You have to realize the FM3 has only about 1/5 the DSP power of an Axe-Fx III. Trying to fit all those goodies in there while not sacrificing audio quality is a balancing act. I ALWAYS insist on quality over quantity. That's what differentiates our products. It takes a LOT of horsepower to do it right.

[24] Dug my PCM70 out of the closet and compared the diffusion to the Axe-Fx III. The Axe-Fx III is vastly superior. The PCM70 sounds metallic and comb-filtered. The Axe-Fx III sounds like a little reverb is added to each delay.

[25] I designed the Axe-Fx diffusers to sound like little reverbs. I like a little, 10-20%, which makes the echoes slowly turn into a wash. To me the PCM70 goes "brzzzzpppp" on the attack which I don't find pleasant at all. I've experimented with diffusers for years. I hate all the examples in the literature (i.e. Schroeder allpass sections). I've implemented quite a few out of the papers and they all sound like crap IMO. I tried the allpass cascade in the JCRev algorithm again today and, yup, still sounds as bad as I remember.


Current Axe-Fx III, FM9 and FM3 firmware provides virtual companders in the Delay blocks. A compander compresses the signal at the input and decompresses it again (expands) at the output. This results in better SNR ratios.

Disabling the compander saves CPU.

Because of the compression due to the compander, spillover is affected.

If the Compander is enabled, Stack/Hold is not available.


[26] A Compander is a compressor followed by an expander. If the time constant of the compressor and expander are the same (and the "channel" is perfect, i.e. linear and lossless) then a Compander is transparent. If you change the time constant of one vs. the other the transients will become distorted.

The "Transients" knob controls the time constant mismatch. Negative values smooth the transients, positive values enhance the transients.

A Compander with transient enhancement can be used on kick drums, bass, etc. to add punch. Transient reduction can be used to reduce plosives.

For guitar you can use the Compander to add or remove pick attack.

[27] […] The first step in adding compansion to the delay block was creating a compander algorithm. So I created a Compander type first in the Compressor block.

Once that was working I used that algorithm in the Delay block. The Compander allows you to set the Ratio, Time, Transients and Level. Turn Transients down and the attack is reduced, turn it up and the attack is enhanced. You can use it with kick drum to enhance the punch, put it on vocals to reduce plosives, etc.

[28] The DMM, DM2, and Carbon Copy have it on by default.

[29] […] The HIGHER the compander threshold the more prominent the effect. It's the threshold that makes the compander do what it does. An ideal compander would have a threshold of negative infinity dB. A real compander suffers from a finite detection threshold which is what causes the softening of the transients.

The ubiquitous NE570/SA571 has a threshold of about -60 dB so that's the default.

[30] That's unavoidable without losing spillover. When the compander is active the data is compressed before being written to the delay line and decompressed when being read. So if you switch from not using the compander to using the compander uncompressed data is expanded which makes it louder.

Config (legacy)

This parameter has been removed in firmware Ares and later.


[31] The Config parameter selects the base algorithm. The Type parameter selects the default parameters. If you selected the Deluxe Memman type and then changed the Config to dual you got the dual delay algorithm with the Memman default values.

Tips, tricks and troubleshooting

Factory presets

The Delay block is showcased in many factory presets. Search the Factory presets page for "delay".

Change pitch when adjusting delay time

Only the Mono Tape delay type changes the pitch of the delay trails when adjusting the delay time.

To force other types to same behavior, attach a modifier to the Delay Time parameter.


[32] […] The original algorithm simply slid to the new time which would cause pitch shift. People complained vociferously since tapping in a new tempo would cause a pitch glitch. The new algorithm does not do this. It treats time/tempo changes as "requests" and cross-fades between two delays running simultaneously.

If you want pitch shift the use the Mono Tape type. There are other ways as well but I'm not at my desk now so I forget the details. I believe if you attach a modifier to the Time parameter in the other types it will pitch shift when you change the time. I seem to remember I made it so you could do the pitch shift thing with a modifier but not when tapping in a tempo or setting the time by hand. I.e., you can do the Eruption dive-bomb by attaching an expression pedal to the Time parameter.

Prevent cutting off tails when using a pedal

By attaching an external controller (connected to a pedal or switch) to the Input Gain parameter, you're controlling the delay level at the input stage. This will make delay trails fade out nicely when muting the delay.

Use Ping-Pong delay to test stereo output

The Ping Pong delay type lends itself well for checking the stereo output of the unit. The first repeat should be heard at the LEFT side.

Delay settings for lead guitar

For a lead tone with ambient delay, try this: Stereo Delay, Tempo 1/4, Right Time Ratio 61.8%. You can use these settings with a stereo or mono rig.

Fixed number of repeats

If you need an exact number of repeats, use the Band Delay type or Tap Ten Delay type in the Multitap Delay block.


The delay can be set to create a micro-delay for enhanced stereo separation.


[33] Another way would be to use the Delay block. Set the Mix to 100%, dial in the desired amount of delay and set Balance to full right.

Or use the Enhancer block instead.

Effect on delay trails only

To add an effect to the delay trails only, not to the direct tone, place the Delay block in a parallel row (Mix at 100%) and add the effect after it.

If you want reverb on the delay trails only, try the Diffusion parameter in the Delay block. This smears the separate delay trails, creating a reverb-like effect.

Engage a delay automatically by setting tempo

If the parameter Auto Delay in the Tempo menu is set to “ON,” any delay blocks that are bypassed will become active whenever a tempo is tapped in. This allows you to set the tempo and un-bypass your delay block(s) from a single footswitch. Read this: Tempo

Rack effects processors from the 80s

Watch Leon Todd's video below for recreations of various rack effects processors from the 80s, including SPX90, 2290, PCM780 and H3000.

Turn the reverb into a delay

Firmware 20 for the Axe-Fx III (and corresponding FM3 and FM9 firmware) added the ability to use Pre-Delay in the Reverb block as a simple echo. Pre-Delay provides Tempo, Feedback and Mix parameters. The pre-delay time has also been increased to 1s. Several Reverb types demonstrate the capabilities: Echo Plate, Echo Hall, Echo Room.

More information