I've built an original 18V EM with SAD1024 using the schematic on the Electric Mistress Mystery page. Even allowing for the volume drop, there is a noticeable difference in sound compared to the EM3207. I know the EM3207 is a clone of the 9V EM and that there are differences in the audio path compared to the 18V EM. I like the sound of the EM3207 but I prefer the sound of 18V EM so I wanted to see if I could modify the EM3207 to achieve that sound. I think I now understand why the original EM sounds like it does, and will describe how the EM3207 can be modified to make it sound much closer to both the original 9V and 18V versions.
Here is how the relevant section of the EM3207 compares to the 18V version. I've highlighted the key differences in component values.
The 18V EM uses smaller cap values for C8 and C10 and this reduces frequency content in the lower midrange (a couple of hundred Hz). That's higher than the frequency of the low open strings on the guitar, and those strings are indeed quieter than the high strings on the 18V version.
So I changed C8 and C10 to 47 nF. The side effect of reducing the low-mids like this is that there is a noticeable volume drop overall. This reduction of low mids is one of the main contributions to the volume drop in the 18V EM. So the next thing I did was increase R9 in the pre-emphasis section from 5k6 to 6k8 (as in the 18V EM). This increases the overall gain and recovers most of the volume drop lost to reduced low-mids. Increasing R9 didn't give me any problems with headroom or clipping.
In the end, I ended up changing all cap and resistor values in the audio path to match the 18V EM values (i.e. all the values in red on the above diagram apart from R14 and C7 on the BBD output). These changes made the EM3207 sound more like the 18V EM but I could tell that something still did not sound the same. This was when I decided to take measurements given that I had already made the necessary component value changes to make the two circuits "match". I only have a handheld scope that is not lab quality but it is still better than making subjective comparisons based on my ears.
I started by measuring what I'll call "BBD gain" for the two circuits. To be more precise, for each circuit I measured 20log10(Vout/Vin) where Vin is the pk-pk voltage measured at the BBD input, and Vout is the pk-pk voltage measured at the smoothing cap C7. I took measurements in filter matrix mode with no feedback, using 7 different clock rates covering the whole sweep range (using doubled clock rates for EM3207 of course), and using 21 audio frequencies from 100 Hz to 11 kHz. The following graph shows the measured BBD gain.
The EM3207 shows a mainly flat BBD gain of roughly 2 dB, in agreement with the MN3207 datasheet.
The 18V EM shows a significant roll-off in BBD gain in the audio band.
If we subtract one curve from the other ("18V EM BBD Gain" minus "EM3207 BBD Gain") we get the following curve that gives the relative BBD gain between the two circuits. The point of doing this is to use the EM3207 BBD gain as a baseline for comparison, effectively subtracting out any drop in BBD gain that is common to both circuits.
This last graph shows that the BBD gain in the 18V EM has suffered a sort of "shelving" effect. In other words, there is a roll-off in gain that flattens out at low and high audio frequencies. The reason for this is that the 18V EM does not just have a 10k resistor on its BBD output. It has the rest of the mixing circuit hanging off it too. This means that for high audio frequencies there is not 10k on the BBD output but a much lower impedance instead. Lower impedance on the BBD output means lower BBD gain.
In the EM3207 circuit, Q1 shields the BBD output from the rest of the mixing circuit and so prevents the roll-off in BBD gain. This makes the EM3207 a more "pure" flanger but on the other hand it stops it emulating a key part of the original circuit's tone. (Note that the original 9V EM by EHX will also have a roll-off in BBD gain but it will be "shelved" differently to the 18V EM because it has "trim resistors" in series with the drain-source resistance of the BBD output.)
I believe Thomeeque included Q1 because the outputs of the MN3207 have much higher drain-source impedance than the SAD1024. Something like 4k5 compared to 1k for the SAD1024. So the MN3207 chip needs 47k for R14 in order to give similar gain to the SAD1024 with 10k. Including Q1 allows all the other caps and resistors in the mixing/feedback circuitry to use the values of the original EM circuit. The other option would have been to leave out Q1 entirely and instead rework the mixing/feedback circuitry by scaling up resistor values by a factor of 4.7 and scaling down cap values by the same factor (just as Thomeeque did for the components on the BBD output). I am not sure why that approach was not taken. I could be missing something
The overall output impedance would go up by doing that, but that could be corrected with an output buffer.
Anyway I was too lazy to change out any more components, and wanted to see if there was a simple modification that would give the EM3207 a roll-off in BBD gain matching that of the 18V EM. I worked out I could get a good approximation to the shelving effect by putting (15k + 6n8) across the 47k resistor at the BBD output of the EM3207 as follows.
Note: A similar sort of trick could be used to mimic the BBD gain roll-off seen in the original 9V EM, although I didn't calculate the values needed for that as I don't have a working reference circuit of the 9V EM from which to take measurements.
I took another set of measurements after my "shelving" modification, and the following graph shows what happened to the relative BBD gain.
The relative BBD gain between the two circuits is now pretty flat across the audio bandwidth and varies by less than +/-0.2 dB. Overall the EM3207 still has a BBD gain that is about 0.5 dB higher than the 18V EM. That could be fixed in several ways (e.g. by reworking the R and C values of the things hanging off the BBD output pins) but I didn't want to change any more values and think I will live with the 0.5 dB difference in BBD gain in my build.
Summary of overall result with the above modifications ...
1) The circuit now sounds much more like an original 18V EM. Maybe I'm cork-sniffing but I think the difference is significant. You would think that throwing away high frequencies in the wet path would make things sound less clear but I found the opposite to be true. My EM3207 now has the same sort of watery clarity as the original 18V EM. That could be related to point 3 below though.
2) The EM3207 ends up with a small volume drop. It is not as bad as the original 18V EM though because of the boost at the input.
3) Noise is noticeably lower after the mods
Connecting/disconnecting the "shelving" mod at the BBD output makes this obvious. Maybe that explains the improvement in clarity. For the same overall volume, there's less noise present, especially in the high frequencies because those get to the output more through the less noisy "dry" path than through the noisy "wet" one. I'd say the noise reduction alone makes the shelving mod worthwhile.
3) Feedback/Color pot is less harsh after the "shelving" mod.
What I've learnt from all of this is that the original EM (probably my favorite pedal) sounds the way it does mainly because of all the signal that it throws away. It reduces low-mids during mixing, and it reduces highs in the wet path due to the "shelving" in BBD gain. (The sweep giving linearly increasing delay is also important).
Based on what I've learnt above, my plan is to do a clone of the 18V EM but using MN3207 chip running at 9V. My aim is to make it a "sound-alike" and "control-alike" rather a straight copy of the circuit. So basically I am going to rip out the LFO/VCO used in the 18V EM (which is horribly noisy) and replace it with something along the lines of the LFO/VCO of the 9V EM but with better VCO values and different filter matrix switching and range pot behaviour.
I think I can reduce the part count a bit too.
One other thing that I will do is redo the BBD gain measurements for the EM3207 but without the clock buffers. I know the circuit will work without the buffers but getting a quantitative measurement of how much they affect the BBD gain (if at all) will be useful.
