EM3207 (v1.1) - MN3207 based EHX Electric Mistress (9V) clone

[Scruffie]

First, for doing all the work to redesign the circuit and board to accommodate a 3207.  Very much appreciated.  Not just by me, but by many others, I'm sure.

Second, the layout is very compact.  Compact enough, that a person could consider shoving two of them in the same pedal.  This creates some interesting possibilities.

First, there is the obvious possibility of splitting a mono input signal and having two asynchronous flanger outputs.
Second, there is the possibility of having a mixer and combining the two flanger signals into a single more complex mono output.
Third, there is the possibility of lifting the dry signal from each flanger outputs, and running them "stereo-out" to separate amps or mixer channels.  At slow sweep speeds, this can produce both apparent movement as the phase relationship between left and right keeps changing.  It can also produce "flanging in-air" (though that likely requires standing in the sweet spot).  At faster sweep speeds, you get stereo vibrato, which can be interesting.
Fourth, with dry lift, and setting one of them to "filter matrix" (non-swept) mode, combining flanger A and B gets you variations of through-zero flanging.

Clearly, throwing a pair of these into a suitably-sized chassis, that can accommodate both boards, a splitter/mixer board, additional jacks and toggles, can yield something truly wonderful.

I tried running an Electric Mistress in Filter Matrix into another Flangers With the Same Audio Path as the Mistresses Send & Return and to be honest, it didn't work all that wonderfully... I think it'd need some refinement.

[Fender3D]

To achieve what Mark said, you must connect the 2 flangers in parallel, not in series,
then you'll need to find a way to lift the dry signal

[Scruffie]

To achieve what Mark said, you must connect the 2 flangers in parallel, not in series,
then you'll need to find a way to lift the dry signal

I had them in parallel, but now that you mention it... i'm not sure I did kill the dry signal, it was a while ago, i'll have to go back and try it again.

[trat rödenback]

WOW! Great Project! I allready got all the Parts for the EM with mn3007 (Jorge's Layout) but this Layout is so much more compact. I'd definitely like to build the EM3207. How does it sound compared to the mn3007 version? Is there a big difference?

[Mark Hammer]

To achieve what Mark said, you must connect the 2 flangers in parallel, not in series,
then you'll need to find a way to lift the dry signal

I had them in parallel, but now that you mention it... i'm not sure I did kill the dry signal, it was a while ago, i'll have to go back and try it again.

Through-zero requires that the delayed signal be able to achieve a "negative" delay; i.e., a delay time that is less than the dry signal.  If your dry signal is simply a buffered input, then clearly this is not physically possible, even for Dr. Who.

HOWEVER, if you are able to delay the dry signal, by just a couple of milliseconds (and often, as little as 1-2msec), the swept-delayed signal will, at some point, achieve a delay time which is actually less than the fixed delay of your "dry" signal.  If those two are produced in parallel and mixed, you get through-zero flanging.  If they are placed in series, then all you do is add the delay time of the one to the other, which only gets you poorer flanging.

[oldschoolanalog]

...then clearly this is not physically possible, even for Dr. Who.

[Keppy]

I just discovered the scarcity of the SAD1024 last week and was lamenting that the 3207 version was not complete. Turns out it was, but I just noticed today. Thanks Thomeeque!

[HarryBBD]

RE: Mike Irwin TZF
(where is Mike these days, BTW ?)

Its possible to use a fixed, analog delay (dome filter) for one audio path and the BBD for the other. A dome filter is a bunch a stagger-tuned
all pass filter stages (more commonly known as 'phase shift' stages) Choose the right values and there is a slight audio delay that is reasonably
flat for a wide audio range. Juergen Haible of synth-diy has some schematics online for his frequency shifter. The dome filter gives two
outputs that are 90 degrees apart (sine-cosine). You only need half of this circuit for it to work, but probably you'd need at least six opamp
stages to make it work well enough. Each stage is three resistors, an op amp, and one capacitor.

[oldschoolanalog]

http://www.diystompboxes.com/smfforum/index.php?topic=25168.0
 
Also:
http://www.diystompboxes.com/smfforum/index.php?topic=77987.0
 

[Mark Hammer]

I think the most trivial and easy access to through-zero flanging these days is to simply buy a Behringer DC-2 clone ( http://www.behringer.com/EN/Products/CC300.aspx ), and lift the dry signal from the two delay paths like Scott Stites did:  http://www.birthofasynth.com/Scott_Stites/Pages/dimc_main.html

[Thomeeque]

 Hi fellas,

 thanks again for your compliments, ideas, questions, discussion etc.

 I'm sorry I don't respond very much, but I'm still kinda busy..

 Anyway, I have some news (both good and no so good)..

 I've got finally my hands on the real vintage Electric Mistress with SAD1024 chip! It's a Deluxe not the Standard version, but since 9V EM is SEM/DEM hybrid actually ("standard" audio path, rest is deluxe, more or less), it is pretty useful too.

 

 It was an "eyeopener" for me in many ways and it made me start a new round of EM3207 R&D (actually I did plan to do some tweaking of rev1.1 yet anyway, but comparison with the real piece moved it to another level).

 Biggest difference is definitely sweep range (and sweep character). EM3207 1.1's maximal range hardly touched DEM's minimal.

 I did measure control voltage range at clock VCO and it's V-f characteristic:
 
 

• Blue values/line represent measured clock frequency,
• Red values/line represent calculated BBD delay time (=512/(2*f_CLOCK) here)
• 1.69 to 9.24V is Filter-Matrix mode range
• 1.26 to 10.31V is maximal sweep range (triangles).
• (V_CC = 15V, R_CLOCK = 18.43k)

 So the delay time range goes from 8.83 to 0.20ms which makes max/min ratio 44:1!!

 And btw., since V_CTRL sweep are linear triangles, frequency sweep is HYPER-TRIANGULAR! Oh yeah - that's why we love EM so much, I just did not see it that clearly up to now

 I've made two samples with maximal range, lowest rate and no feedback (color):

 DEM15_noise.wav
 DEM15_shine.wav

 EM3207 v1.1's maximal ranges were pathetic compared to this (e.g. 271~1245kHz at one particular setting), max/min ratio was cca 5:1~7:1 depending on C17 and clock trim setting, and maximal frequency I was able to get out of it was cca 1250kHz (above this limit 4013 failed to read 311's output and 4013's outputs were frozen) which equals only to 625kHz of DEM (0.41ms, we have BBD twice longer!), plus sweep did not sound that pleasant when clock were adjusted to this range, like if it would slow down at the top (now I know that sweep curve was loosing it's hypertriangularity).

 So heavy R&D started again Since then I am tweaking my EM3207 prototype:

 

 emulating VCO in LTSpice:

 

 and I have even built the standalone replica of the clock to be able to study more comfortably and safely influence of all possible factors including e.g. V_CC (you know, EM3207 cannot be fed by more then 10V and I've wanted to see what's going on with VCC = 12 or 15V):

 

 

 All of this to get the EM3207 clock working as close to DEM with doubled frequency as possible I did break original 311 to 4013 transition top limit relatively easily (actual "record" for f_CLOCK is 2.1MHz, eventually 2.7MHz), but just to find that lowering value of the clock capacitor is not the way, that other parts have it's limits too (including LM311N, those 2.7MHz I did achieve with old Tesla MAB311 equivalent) etc., it is really long story and it's not over.

 Here's the best what I've achieved so far (Tesla MAB311 used, same setting as for DEM15):

 EM3207_noise.wav
 EM3207_shine.wav


 Just to calm down all those who already started to build original EM3207 v1.1 - all those tweaks should be posible using v1.1 PCB, it should be (so far it is) just about different values of some components mostly. And actually, even original EM3207 v1.1 sound pretty good


 Cheers, T.

 Note: pics are clickable for hi-res.

[Mark Hammer]

Thanks, Tomas, for all the hard work, and especially your conscientiousness and willingness to continue.

Yes, when it comes to flangers, the sweep ratio is VERY important!  Let's hope that somethng in the right direction can be achieved with a 9v supply.

[12Bass]

FWIW, while flanging using short delay times is audible on a full-range signal, I've found that delay times shorter than ~ 0.5 ms aren't very audible with an instrument like a guitar, making it of questionable value to push the clock much higher.  From the clips, I think that I might prefer the original DEM15 over the EM3207 version - sounds a bit fatter and more dramatic; it could be a matter of calibration, however.  Perhaps the DEM15 is adjusted closer to 50/50 and getting more complete cancellation?  Both sound as if they are accelerating as they sweep up, though the LFO waveforms might sound a little different.  I've spent some time trying to "smooth out" the sweep on my A/DA clone, as it tends to go through the upper part of the sweep more quickly than I'd prefer.  

Nice work!  I've got some MN3007s around which might be good for a similar project.  Curious to see/hear how this progresses....

[Thomeeque]

Thanks, Tomas, for all the hard work, and especially your conscientiousness and willingness to continue.

 Thanks, Mark

Yes, when it comes to flangers, the sweep ratio is VERY important!

 Ratio and sweep course, I'd say..

 Let's hope that somethng in the right direction can be achieved with a 9v supply.

 I run all my analog pedals at 10V (it's closer to fresh 9V battery ), but I did try to run EM3207 at 9V too, there's almost no difference.

FWIW, while flanging using short delay times is audible on a full-range signal, I've found that delay times shorter than ~ 0.5 ms aren't very audible with an instrument like a guitar, making it of questionable value to push the clock much higher.

 Well, IMO (!) the trick is in synergy of the big range (going from very reasonably low to very high clock) and hypertriangular sweep, which sweeps the highest clock range quickly (as you can see from V-f graph, more then 90% of the sweep time is above 0.5ms). It's not necessary there, especially if you want just decent light flange, but for "jet-plane" effect it would be missing..

From the clips, I think that I might prefer the original DEM15 over the EM3207 version - sounds a bit fatter and more dramatic; it could be a matter of calibration, however.  Perhaps the DEM15 is adjusted closer to 50/50 and getting more complete cancellation?

 DEM15 sounds fatter because of different audio path, 9V EM cuts lot of lows, so it sounds much more midrange (which can be actually better in full band mix), it could be adjusted to taste by using bigger caps in the output mixer (C10, C8).

 About more dramatic I'm not sure (could be partially equalisation metter as well). But yes, cancellation may play the role here - even IMO not that major (I have 8k2+trimmer there now for R18 and there's not that much difference, if it's adjusted to 8k2 or to e.g. 12k).

 Btw. absolute cancellation is something you cannot achieve in full clock range (even with SAD1024), gain of BBD is f_CLOCK dependant. But in the case of EM3207 more as we go to higher clocks - it is another issue which I may try to address yet later (those 512 cells more really suck ).

 Btw. I did measure GAIN-f_CLOCK characteristic for EM3207 v1.1 as well:

 

 GAIN here represents ratio of voltage at Q1 emitter and voltage at BBD input (200Hz 2V_PP sine).

 I'd say that used clock buffer does good job, I wonder how would Ultra Flanger's buffer behave here..? I did choose used configuration over UF long time ago based on resume of some discussion I've googled (probably at this forum), but it could be interesting to dig again..

 For DEM15 I did just quick measurement so far, BBD gain was 1.06@33kHz and 0.92@663kHz (plus I have note there, that gain drop starts at 130kHz).

 Both sound as if they are accelerating as they sweep up, though the LFO waveforms might sound a little different.

 Both samples are made in the FLANGE mode (though the LFO waveforms)!

 And yes, "Both sound as if they are accelerating as they sweep up", that's the goal = that's what is hypertriangularity about

 I've spent some time trying to "smooth out" the sweep on my A/DA clone, as it tends to go through the upper part of the sweep more quickly than I'd prefer.  

 Looks like we have oposite tastes But actually, with lower range it's not that dramatic plus there are ways how to simply lower this effect in case of EM3207.

Nice work!  I've got some MN3007s around which might be good for a similar project.  Curious to see/hear how this progresses....

 Thanks!

[12Bass]

The 4049 buffer in your circuit appears to be very similar to the one used in the A/DA flanger clone.  From what I recall in earlier threads, measured performance was worse when using three inverter stages (like Ultra Flanger).  Even with a 4049 buffered SAD1024, IIRC, there's a roughly 2dB delta in gain between the shortest and longest delay settings; MN-series may be somewhat worse (or perhaps better, if your graph indicates less than 0.4 dB and not 0.4 V difference).  I also noticed that it is impossible to adjust for maximum cancellation throughout the entire sweep, so RTA was used to set the BBD output so that maximum comb filtering was centered in the "sweet spot", somewhere in the upper-middle of the sweep.  Some designs incorporate a transistor buffer at the BBD output which is supposed to improve linearity at higher clock rates. 

I don't have measurements handy at the moment, but I seem to recall from earlier testing that the flanging effect was mostly lost when the cancellation of the fundamental rose above around 2 kHz, which I thought was more than 0.4 ms.  Above that point, even with the extended bandwidth in my build, there simply wasn't much signal left to audibly flange, except if the source was full-range with a fair amount of high frequency content and when listening through a full-range playback system.  Then there's the issue of less dramatic comb filtering due to the loss of BBD gain at higher clock speeds.  From what I recall, my A/DA can sweep to less than 0.3 ms; however, I've found that when sweeping that high, the flanging effect disappears pretty easily, so I usually tend to reduce the sweep range to less than half of maximum in order for it to be musically useful.  Put another way, I've found that a huge sweep range is cool to have at your disposal... but it can easily be too much unless you're listening to the flanger sweep through a full-range signal in a controlled setting.  It's easy to set the sweep so high that you can barely hear any flanging in the top range. 

From the look of it, the DEM does not have the lower (~15 ms) range of the A/DA, though that may help keep it from sounding too metallic.  That delay lower range can be nice for chorus sometimes, or to make robot voice regeneration sounds.  The SAD1024 also gets noticeably more noisy when the delay exceed ~10 ms.  If I had to guess, I'd say that the most musically useful flanging range lies somewhere between 0.5 and 5 ms.   

Wish I had the actual figures handy.  I could go back and record some proper measurements if it would be helpful.  I think that I did the clock/frequency calculations previously out of curiosity, but didn't write anything down.

In case you missed it, here's an earlier thread with RTA plots showing the comb filtering at various settings (after setting the output trim for maximum cancellation):

http://www.diystompboxes.com/smfforum/index.php?topic=91378.0

 

[Mark Hammer]

FWIW, while flanging using short delay times is audible on a full-range signal, I've found that delay times shorter than ~ 0.5 ms aren't very audible with an instrument like a guitar, making it of questionable value to push the clock much higher.  From the clips, I think that I might prefer the original DEM15 over the EM3207 version - sounds a bit fatter and more dramatic; it could be a matter of calibration, however.  

I find the "drama" of a flanger is a function of its transition from no apparent/audible notches to being "infected" with notches.  So, you are quite correct that the naturally occurring bandwidth of a guitar signal can make a flanger sound inoperable at the short-delay end of the sweep, if you aim for spectacularly short delays.  At the same time, I suspect the impact of those short delays is that the sweep may appear to spend a conspicuously long time "not doing anything audible".

There are two ways you can look at this.  One is to calibrate the delay-range so that it doesn't go beyond those audible inflection points.  The other is to alter the LFO waveform so that the sweep doesn't spend an inordinate amount of time on the "wrong side" of that inflection point.  The schematic shown here - http://hammer.ampage.org/files/hypertriangleclock.gif - is a simplification of an earlier circuit that used the FET+diodes arrangement to produce a sine wave approximation from a standard LFO triangle output.  That original had used diodes running to both drain and source from gate to derive a sine wave, and FWR'd that to produce a hypertriangular wave.  The McConkey circuit took a shortcut and uses just a single diode with the FET to produce a sinusoidal shape at one peak of the waveform, and retain the triangular shape at the other.  I wonder if one could use some arrangement of that to adjust the degree of sinusoidal shape at that end of the sweep, such that if the width/sweep was going to go beyond that ideal inflection point, you could shorten how much time the sweep spent on the other side of it.

Does that make sense?

[Govmnt_Lacky]

Phewww! I'm glad I waited on this one. I was just gearing up for this build 

Just a thought... would an investigation into the Ibanez FL9 lend any help to this? It uses the MN3207/MN3102 combo in the flange circuit. 

Here is the info:

http://www.diystompboxes.com/smfforum/index.php?topic=68966.0

Just finished building this and I must say... I am impressed with it! It does a great job... ya know..... not being an MXR M117 and all 

[Mark Hammer]

Note that the Boss BF-2 doesn't have a sweep range much wider than what Tomas was disappointed to see in V1.1.  And from the little I can tell, the BF-2 has a straightforward triangular sweep.  So even if this 3207-based EM does not eventually manage to attain the grand sweeps of its 1024-based ancestor, building it will still probably get you a reasonable BF-2 workalike that  sounds good.

[12Bass]

Judging by the samples, the MN3207 EM is sweeping higher than a BF-2 can.   

What I gather is that the audibility of the flanging effect rapidly decreases when the lowest notch is greater than the highest fundamental frequency of the source signal.  With a guitar, once the comb filtering starts above the fundamental frequency of the the highest note (1175 Hz), the sweep begins to effectively disappear, as it can only effect the higher harmonic overtones.  At 2350 Hz, the flanger would be cancelling the second harmonic of the highest note.  So, the size of the "black hole" at the top of the sweep would seem to be dependent upon the highest fundamental of the source, plus its upper harmonic content (e.g. a square wave fuzztone, which could have harmonics going to infinity and beyond...).  If there's no appreciable upper-harmonic content, the flanger is effectively flanging "air".     

[Mark Hammer]

And this is precisely why the benchmark flanger sounds over the years have been post-production flanging applied to multi-source mixes, where the bandwidth is wide and full.  People hear those and thnk somehow they"re going to get the same magic throwing a flanger at a single measly guitar.  Sorry, folks, doesn't work that way.