SMF-1
Analog Dry Signal for
Digital Effects



Kaz Kylheku <kaz at kylheku dot com>

Date
Notes
December 2, 2012
Initial revision.
January 2, 2013
Picture of custom faceplate.
February 19, 2013
Old transformer is gone.
March 19, 2013
Added missing schematic.


Introduction

In the 1980's, early rack-mount digital effects processors for live guitar or studio use, such as the Yamaha SPX-90, had a nice feature: a true analog dry signal. Mixing the effect, such as delay or reverb, was not performed in the digital domain. Rather, it was done in the analog domain by mixing the reconstructed effect signal, with the original analog signal.

The early boxes obviously needed this, because they had poor sample sizes and low bit rates. The aforementioned Yamaha SPX-90, took only 12 bit samples, at about a 31 Khz sample rate, well below CD quality. Yet, this unit is still traded on the vintage gear market. This is because thanks to its dry analog signal, it has a good frequency response and dynamic range. Just not in the effect path. But these issues are not so noticable if the effect is not mixed in in ridiculous quantities. If the effect is mixed in relatively quietly with respect to the main signal, its inherent noise also becomes proportionally quieter. Moreover, the effect does not need the full frequency spectrum of the dry signal. Real-life reverberations and echos are not "hi fi". Only very hard, smooth surfaces reflect high frequencies. Irregular and absorptive surfaces attenuate high frequencies. Digital reverb effects usually incorporate high frequency attenuation to make the simulated "hall" sound less "live". If a time-based effect such as delay or reverb has the same frequency spectrum as the original signal, it can mask the main signal and distract from it.

As the sample sizes and sample rates improved, effects started to appear which did away with the analog dry path. For instance, in 1989, Yamaha introduced the small half-rack-sized FX-500, a 16 bit 44.1 Khz unit with no analog dry. This unit is digital even if it is put on bypass. I have one of these units (purchased new from a retailer in 1989!)

A few effect units continued to provide analog dry. For instance, Rocktron Intellifex units from the late 1990's have analog dry, at least depending on how the noise gate topology is configured. I think that this feature has by and large disappeared from the market.

However, I have come to believe that a true analog dry will always have the best sound quality, and some other advantages. So I decided to build a rack-mounted box which can provide the guitarist with a true analog dry signal regardless of what effect processor is used. Furthermore, this rack-mounted box also has presence controls, which independently operate on the dry and wet signal. This unit is called the SMF-1.

The SMF-1: Block Diagram

Note: this is an SVG image, requiring an up-to-date web browser:

How it Works

The input signal from the guitar preamplifier is split into two buffered filter stages. One carries the dry signal, and one carries what will be the wet signal. These stages each contain a variable-frequency, active low-pass filter which can be swept from a knee frequency of about 3.8 Khz to about 20 Khz (thus, from a sigificant roll-off of high end to essentially complete transparency with respect to human hearing). The filters are second order, and so roll-off at -12dB per octave.

The output of the wet signal's filter appears on a mono send jack. This can be connected to an effect box which provides stereo outputs. There are matching left and right returns on the SMF-1.

The dry signal and the effect returns are all mixed together, and the combined signal is sent to the stereo output. The mixer provides two mix level knobs for the two channels of the wet signal. The knobs can completely cut out the wet signal, or mix it in at its full volume (which depends on the output volume of the effect box).

In stereo mode, the SMF-1's mixer duplicates the dry signal into the left and right outputs in equal amounts, so that the dry signal is panned dead centre. The wet return signals are mixed into the respective outputs, creating a stereo effect around the signal. The mix knobs control how much of the left wet signal to mix to the left channel, and how much of the wet right signal to mix to the right channel.

There is also a mono mode, selectable with some jumpers on the circuit board (which could be easily re-routed to an external stereo-mono switch). In mono mode, the right channel output is disabled. Both channels of the effect box, and the dry signal, are mixed down into just the left output channel. The mix knobs control how much of the left and right effect signal is mixed in.

Rationale

The need for the SMF-1 can be summarized by the following rationale points:
  1. What is going on in an all-digital effects processor is actually pretty silly. First the original pristine signal is digitized, and then it is split. A copy of the signal is turned into a delay or reverb signal. Then the two are digitally combined, and the result is converted back to analog. It is obvious that the operation of splitting and combining can be done in the analog domain, so that only the effect signal takes a trip to digital and back to analog. Analog circuits are very good at splitting and combining: no digital signal processing is required to do these operations. By moving these operations outside of the digital domain, we can preserve the purity of the main signal. I believe that even modern effects boxes, which boast 24 bit A/D converters, and 48Khz (or higher) sample rates, simply do not preserve the sound quality of the dry signal. At least in gear that is affordable for the average guitarist, the hardware is simply not up to snuff. The numbers lie. You do not achieve anywhere near the 24 bit resolution, if the A/D converter is based on poor quality sample-and-hold circuits based on silicon that is not suitable for audio, and if the sampler has a jittery clock. "24 bits" only states the granularity of the digitizer, and the width of the data path. It is not a statement about the accuracy of the digitizer, and not a statement about the dynamic range. If the least significant 12 of those bits are poor-quality garbage riddled with jitter distortion and noise, then it's effectively a 12 bit unit.
  2. It is not only about purity. The analog domain has better dynamic range. Digital boxes cannot match the headroom of high quality, quiet, virtually distortion-free analog op-amps running on +/- 15 Volt power rails.
  3. If frequencies are going to be rolled off in the wet signal, it makes sense to do this before the sampling into the digital domain: why not implement an analog presence control which applies to the signal which goes to the effect box. This helps sampling: the nyquist filter does an even better job if the input signal is already low-pass filtered at well below the nyquist frequency. If the effect box exhibits even a slightest hint of aliasing, it should disappear when we low-pass-filter the input.
  4. It is useful for the guitarist to have a simple global control over the wet/dry mix, the presence (high frequency content) of the dry signal and the presence of the wet signal, without having to reprogram effect processor patches. The reason is that these quantities are sensitive to the volume level. If they are controlled with simple analog knobs on the front panel, there is no need to have a separate set of MIDI patches for live volume, and for practice volume.
In any case, the SMF-1 sounds fantastic and provides exactly the "last missing piece of the puzzle" in my guitar rig: terrific control over high frequencies, and a clear, punchy analog sound with digital reverb.

Schematic

Note: the 5 uF coupling capacitors (C11 and C12) are unnecessary and are replaced by jumpers on the circuit board.

Build and Pictures

The SMF-1 is fantastically quiet and transparent.

It is 1 built on dual-supply op-amp circuits, on a custom two-layer circuit board which I designed myself and had professionally manufactured. For the audio paths, four dual op-amp IC's are used, all of them excellent NE5532 parts from ON Semiconductor.

The rackmount chassis used to be an early 1990's DOD Stereo Crossover, which was a freebie. I threw out its circuit board, but kept the power switch, power connector and fuse. I also desoldered the transformer header and connector from the unit's original circuit board, and transplanted it into mine. I built a rectifier with four 1N4001 diodes. Voltage regulation is provided by a pair of positive and negative linear voltage controller IC's: a Fairchild Semicondutor 7815, and an ON Semiconductor 7915.

Initially I worked with the unit's original transformer. It was causing voltage regulators to run hot though because it put out +/- 30V into each one. The original unit had this problem also (I know because I took voltage measurements before pulling it apart). Not only did the regulators run hot, but 30V is the very maximum input voltage allowed by the datasheet for these regulators. This was a gaping design flaw in the original unit, which I only went along with because I somehow "suspended my disbelief" and rationalized it like this: if that transformer is okay for a commercial device, maybe it's okay to go with it? But that design flaw negatively affected the subjective sound quality of the unit, though. Probably, the hot, over-voltaged regulators were limiting current to the NE5532's, impacting the integrity of the signal swing. With the new transformer, the regulators drop from only around 18.8V and are just slightly warm, and the sound quality noticeably improved.

Some of the pictures show the front panel knobs as standing off from the front panel. That was early in the construction; they are now snug with the panel, as the later pictures show.





Custom Faceplate

I made a faceplate out of 1/8" thick aluminum. It still needs artwork. This picture shows labels around the knobs, experimentally created by heat transfer of laser printer toner. The method is not producing satisfactory results, as you can see. I will likely have it professionally silk screened. That is why I did not completely clean the transfer paper from the toner.
I replaced the mounting screws also, opting for more attractive stainless steel ones. The original screws were not threaded right, by the way! They were difficult to turn (not just in the original chassis, but also in 6x32 nuts that I tried at the hardware store). But the thread size is 6x32 for sure. The new screws fit perfectly and turn with ease. Also note how I turned the rocker power switch the right way up: you push upward to turn it on, not down.

Meaning of "SMF"

This is a reference to a certain Dee Snider, of Twisted Sister fame.

Those who are not comfortable with that expansion of the acronym might prefer to think of it as standing for splitter-mixer-filter.