TZF Electric Mistress (sample and layout)

[markusw]

Since Mark Hammer mentioned that every Flanger can be converted to a TZF I thought I'll give theTZF Mistress a try.
And I must say, it really sounds cool  Turning it back to normal Mistress gives you the strong feeling something is missing.
The clock trim pot is set to a fixed delay of about 4 ms. This way the TZF point is about in the middle of the LFO travel. Don't know whether the Mistress is the ideal Flanger to be converted to TZF but it's definitely worth a try.

One tiny drawback: there is a bit of "pseudo" clock" noise. It's not really problematic but I would like to know it's source. The weird thing is: the frequency of the "clock noise" (at least I initially thought it was clock noise) doesn't correspond to LFO. It sounds like it's  frequency is modulated but this modulation simply is completely different from the LFO "travel". Also interesting: this "pseudo" clock noise increases if nothing is connected to the input and with the input gnded noise it is gone. This noise hasn't got anything to do with the TZF board since it is also present if the TZF board is not connected to the main board. The level is just really low. Therefore I didn't realise so far that it's there.

In addition there is a bit of hiss once the TZF board is connected to the main board. Again, nothing dramatic but I don't know where it is derived from. The hiss is also present if the input is grounded, btw.

Do you think that one reason for these noises could be that the two boards are wired together with about 10 alligator wires and the whole mess isn't boxed yet??

Below the schems.




The fixed delay:



I know there are some things that could be simplyfied, but hey... it works E.g. the two boards both have their own PS filtering caps + Vref opamp (I just thought that having it this way might help to prevent noise). R12 (10R) is replaced by a jumper. Also C15 (15n) is not present.

Comments are highly appreciated. BTW, in case some one is interested I will try to put together a project file within the next days.

Regards,

Markus

[Mark Hammer]

Do you think that one reason for these noises could be that the two boards are wired together with about 10 alligator wires and the whole mess isn't boxed yet??

Well, DUH!!
A "classic" DMM has a switch for selecting between flanger and filter-matrix.  The filter matrix is essentially a fixed delay.  The DMM also has a separate dry and wet output.  If you have two DMM units, you can feed your guitar to the first one in filter matrix mode.  The clean-dry output of that goes to the second DMM and the fixed-delay wet output goes to a mixer.  The second DMM is set to flanger mode but you take the wet-only output and feed that to a second mixer input.

You now have a mixer with a fixed-delay signal, and a swept-delay signal.  Set the fixed delay on the first one "just right" and the second one will sweep ahead of the first one, producing TZF.

Is this any "better" in some way than what Markus has described?  Technically, no.  The practical advantage is that there are layouts available for the non-TZF DMM over at GGG (now THERE is an abbreviated mouthful!  ), but there are no layouts available for a TZF version.....yet.  The other practical advantage is that at least *some* of the potential risk posed by "dueling clocks" is reduced by having them each in a nice shielded enclosure.  The disadvantage, obviously, is that you need a mixer, and you have to run patch cables.  Of course, one of the other neat things about having an outboard mixer is that you can produce a sum and difference mix of the fixed and swept delays.  The Line 6 Liqui-Flange does very nice stereo TZF, and having the TZF move around across two amps is little short of mesmerizing.  Stick a decent fuzz ahead of the flanger so you get lots of harmonic content to comb filter and its damn near paralyzing!

I salute your spirit of adventure, and commitment to this however.  I need a thumbs up emoticon here!

[markusw]

Hey Mark,

thanks a lot for your explanations!

Well, DUH!!

OK, will put it in a box Any idea what this pseudo clock noise might be derived from?

Re the dual DEM approach: isn't the wet signal of the DEM already a mix of dry and wet (i.e. notches are already present)? Now if you mix this wet signal (of a DEM e.g. in Flanger mode)  with the wet signal  of a DEM (in Matrix mode) that has notches too, wouldn't it sound different from just delaying the dry signal for a fixed time?

Regards,

Markus

[A.S.P.]

quod licet iovi, non licet bovi 

[Mark Hammer]

I could have sworn I saw three jacks on the back of my buddy's pair of DEMs.  The schematic at GGG only shows one output jack, though, and none of the schems I have shows more than one output jack.    So, it may be issue-specific, OR I may simply have a memory flaw.  Apologies for any false alarms.

On the other hand, you can always re-direct the dry signal to an added output jack in one/both of them.

[A.S.P.]

adding an inverted dry signal to a "wet" signal-mix of dry+delayed,
should subtract the dry,
and leave the delayed...

[Processaurus]

Do you think that one reason for these noises could be that the two boards are wired together with about 10 alligator wires and the whole mess isn't boxed yet??

Yes   Its a good sign that the noise goes away when you ground the input, that means its clock noise getting getting picked up the input wire, and getting in there because the guitar is a high impedance (low current) source and goes to a high impedance (sensitive to tiny current) input stage.  Grounding it makes a very low impedance source, so all the current the noise puts on the input wire can't compete.  I bet if you precede the dual mistress contraption with any buffered pedal, the noise will be reduced, because of the low output impedance of the buffer. 

Basically I'd re-route the input wire, or this may be a good opportunity to use a shielded cable, with the ground attached at one end. 

A while ago I had two small clones lashed together with a fistful of alligator clips to make a DC-2 kinda thing, and was suprised there wasn't any heterodyning whine from the clocks, given all the reading around here made me believe it was unavoidable.  I guess it was from having seperate power supply filtering for each board and some distance maybe?

Nice work by the way.

[RaceDriver205]

Sound samples, sound samples! Got to here this! 

[markusw]

Thanks for all your replies!

On the other hand, you can always re-direct the dry signal to an added output jack in one/both of them.

I think the DEM could be easily modded to remove the dry signal that is mixed to the  delayed (probably with a switch). So with this modded unit in Matrix mode your approach should be identical to to just delaying the dry signal, if understood everything correctly

Yes   Its a good sign that the noise goes away when you ground the input, that means its clock noise getting getting picked up the input wire, and getting in there because the guitar is a high impedance (low current) source and goes to a high impedance (sensitive to tiny current) input stage.  Grounding it makes a very low impedance source, so all the current the noise puts on the input wire can't compete.  I bet if you precede the dual mistress contraption with any buffered pedal, the noise will be reduced, because of the low output impedance of the buffer.

Yup, putting a buffered box in front of the whole mess helps Thanks for the explanation! Will use shielded wire for the input and output once I put it in a box.

I guess it was from having seperate power supply filtering for each board and some distance maybe?

That's why I decided to have separate ps filtering for each board. I just wanted to reduce the chance of clocks interfering with each other. Basically, I'm sure that there are more elegant approaches but I'm not aware of any

Still I don't have any idea why the frequency modulation of the "pseudo" clock noise I hear does not correlate with the LFO travel of the variable delay. Spooky Any comments on this observation?

Sound samples, sound samples! Got to here this!

Will try if I manage to find some time

Regards,

Markus

[markusw]

Re noise:  I just put a Hammond box over the fixed delay board and this already clearly reduced noise although the wires to gnd and to the main board as well as the main board it self are still uncovered :-) Intentinally, moving audio signal wires towards the clock stages doesn't increase noise. Even putting the fixed delay board on top of the main board doesn't change anything. Therefore, I'm pretty sure that noise won't be an issue once the stuff is boxed :-)

I love playing CDs through the TZF Mistress. It's so nice to hear that swoosh over and over again :-)

Regards,

Markus

[nelson]

Hi Markus

You could do all this with one IC if you used a PT2399, minimum delay time, I think, is 10ms though.

[markusw]

Hi Markus

You could do all this with one IC if you used a PT2399, minimum delay time, I think, is 10ms though.

Hi Paul,

thanks for the hint! Didn't even think about this 

On the other hand 10 ms might be too long. The Mistresses LFO travels from about 0,5 to about 10 ms maybe a little further. At 12.8 ms (if my calcs are correct) the VCO frequency reaches 20 kHz.  So the TZF point wouldn't be symmetric. Don't know if this would hurt though

Markus

[Mark Hammer]

From one vantage point, the PT2399 suggestion is smart.  If I've understood Ton/Puretube's comments about heterodyning with multiple clocks, it is chiefly a problem when the difference between two clocks (we'll assume their sum is out of hearing range and well filtered) falls in the audio range.  So, a fixed clock of 90khz, and a variable clock that ranges from 100khz to 10khz (as a plausible example), will have a point in the sweep when the difference between those two clocks is well within hearing range, and more importantly, well below the corner frequency of any onboard lowpass filters.  Given that the PT2399 and MN3x07 will have vastly different clock frequencies, that audible heterodyning is unlikely.

On the other hand, a 10ms latency for the straight signal would make this a psychologically difficult device to play through.  Not a problem for studio work necessarily, but definitely not the sort of thing you want to jam through.

Scott Stites (DIYFreque) has some followup discussion on his site, prompted by our discussion here, about symmetrical vs asymmetrical TZF.  he also has some samples.  It is well worth revisiting that content.

[nelson]

A simpler and perhaps cheaper way to add the TZF option would be to use a CD4047 + MN3007 combo.

I just have a problem of using an SAD1024A as nothing more than a fixed delay.


I would def have the delay as a panel mounted control.

[StephenGiles]

I saw this thread just as we were turning in last night and my printer went pear shaped. Then after work today we lost our internet connection which foxed me completely, because it was fine this afternoon. Before I kicked the modem, I unplugged it from the mains and replugged it in. Then it was fine. Now, this should be an interesting thread.

My question is how critical is it to use the same type of BBD for both the fixed and variable delay? Could we get a short enough delay from an MN3005 for the fixed delay if it was bufferred a la ADA Flanger mod by Mike Irwin? If so surely any noise generated would be less.

Must do some gardening - we have weeds up to 4 feet high!

[Mark Hammer]

I saw this thread just as we were turning in last night and my printer went pear shaped. Then after work today we lost our internet connection which foxed me completely, because it was fine this afternoon. Before I kicked the modem, I unplugged it from the mains and replugged it in. Then it was fine. Now, this should be an interesting thread.

My question is how critical is it to use the same type of BBD for both the fixed and variable delay? Could we get a short enough delay from an MN3005 for the fixed delay if it was bufferred a la ADA Flanger mod by Mike Irwin? If so surely any noise generated would be less.

Must do some gardening - we have weeds up to 4 feet high!

Smart! (The BBD suggestion, not the "landscaping strategy")

For the baffled, what Steve is suggesting is that if the short fixed delay needed for TZF was achieved by "superclocking" a larger-capacity BBD (a 4096-stage MN3005, or maybe a 2048-stage BL3208, since they are more available), then the sum and difference of those two clocks would be well out of hearing range, and much less problematic or risky as far as heterodyning goes.

[markusw]

Re the PT2399: if I understood the data sheet correctly, the minimum delay is around 30 ms with an external resistor from 0 to 500 ohm. So 30 ms would defintely be too long for live playing. Oscillator frequency would be around around 20M Hz though 

Scott Stites (DIYFreque) has some followup discussion on his site, prompted by our discussion here, about symmetrical vs asymmetrical TZF.  he also has some samples.  It is well worth revisiting that content.

Thanks for the hint!  

A simpler and perhaps cheaper way to add the TZF option would be to use a CD4047 + MN3007 combo.

I just have a problem of using an SAD1024A as nothing more than a fixed delay.

I know it really hurts to use a second SAD for the fixed delay. But the "aim of the study" was to get a working TZF ;-) I m also sure that a lot of things could be simplified. Using another SAD just relieved me from redisigning the delay. Also I supposed that adding just another active 2-pole lp filter should be OK to suffiently reduce clock noise of the fixed delay to avoid this heterodyning thingy (if I understood correctly what Mark mentioned )

I would def have the delay as a panel mounted control.

Already planned how to include the fourth pot on the panel ;-) Again some pot wiring...aargh. Anyway, don't think there is any box I built that ended up the way I initially planned it.

My question is how critical is it to use the same type of BBD for both the fixed and variable delay?

I don't know whether the SAD has any advantage over cheaper (in-production) BBDs. Does running the the two 512 stages in parallel like in the Mistress somehow improve sound (or decrease noise) when compared to a cheaper BBD? According to the data sheet running the two stages in "differential operation" allows the "most effiecient cacellation of clocking glitches". On the other hand the SAD might still be noisiser than e.g. a BL2307 ??

was achieved by "superclocking" a larger-capacity BBD

The MN3005 had to be run at 400 kHz to get down to 5 ms (specs say 100 kHz max). Would this be OK??


Markus

[StephenGiles]

My back is suffering now!

Interesting answers. Another point that struck me is whether it is sensible to have analogue and digital delays together. Is it worth throwing caution to the wind to use a basic compandor here? I think there was a simple compandor in the ETI Playmate (not the magazine!!!) using an LM 13600, without going the more usual NE570/571 route.

Personally, I would be inclined to say stuff the noise, because the EM sound is unique in itself, and I would not want it to be watered down and sound like a Japanese flanger.

[Mark Hammer]

I know it really hurts to use a second SAD for the fixed delay. But the "aim of the study" was to get a working TZF ;-) I m also sure that a lot of things could be simplified. Using another SAD just relieved me from redisigning the delay. Also I supposed that adding just another active 2-pole lp filter should be OK to suffiently reduce clock noise of the fixed delay to avoid this heterodyning thingy (if I understood correctly what Mark mentioned )

I would def have the delay as a panel mounted control.

Already planned how to include the fourth pot on the panel ;-) Again some pot wiring...aargh. Anyway, don't think there is any box I built that ended up the way I initially planned it.

My question is how critical is it to use the same type of BBD for both the fixed and variable delay?

I don't know whether the SAD has any advantage over cheaper (in-production) BBDs. Does running the the two 512 stages in parallel like in the Mistress somehow improve sound (or decrease noise) when compared to a cheaper BBD? According to the data sheet running the two stages in "differential operation" allows the "most effiecient cacellation of clocking glitches". On the other hand the SAD might still be noisiser than e.g. a BL2307 ??

was achieved by "superclocking" a larger-capacity BBD

The MN3005 had to be run at 400 kHz to get down to 5 ms (specs say 100 kHz max). Would this be OK??

Markus

The specs on all the Matsushita chips are for an unbuffered MN310x clock driver into a Matsushita BBD.  The input capacitance of those BBDs is high enough that those nice crisp square-wave clock pulses from the MN3101/3102 get seriously rounded over a certain upper frequency boundary.  THAT's what the 100khz thing refers to.  Should the user/builder/designer include suitable buffering, however, that upper bound changes from that of the circuit in the Matsushita appnotes.  Mike Irwin told me he was able to get 2mhz clocking on an MN3007 without difficulty by providing suitable buffering of the clock lines to overcome that input capacitance.  The basic idea is to pump up the current drive of the clock signal.  A typical means of doing this is by paralleling 3 invertor sections and sticking them between the clock and the BBD, like what you see on the Hollis Ultraflanger.  There is nothing magical about '3'.  It's just an extension of the fact that more invertors and current drive is better, hex invertor chips have 6 invertors on them, and there are 2 clock lines to buffer.  Do the math.

Datasheets for the MN3007 and MN3005 show input capacitances of 700pf and 2800pf, respectively (2800 = 4 x 700).  So let's say that the most an MN3007 could be sucessfully clocked to, with buffering, would be 2mhz.  Okay, divide by 4 to compensate for the extra capacitance,and we'll say that locking to 500khz with buffering should pose no great difficulty.

This is partly why I suggested the BL3208.  You can get BL3207 (1024-stage) and BL3208 (2048-stage) devices relatively inexpensively and superclocking the 2048-stage device should provide enough of a stagger in clock rates plus: a) short enough fixed delay, and b) decent sampling rate for the clean signal.  By being a smaller capacity chip, shorter delays should be easy to get.

[markusw]

Mark, thanks a lot for this explanation!! Your answer definitely improved my knowledge of clocking BBDs a lot

Re assymetric versus symmetric TZF: personally I prefer to have the TZF points evenly distributed. However, moving the clock pot of the fixed delay to the pannel won't hurt for sure.

How does the SAD1024 compare to let's say BL3207 or BL3208 BBDs regarding clock noise??

Forgot to mention before: coupling caps and the 2-pole active lp filter I added for the fixed delay were designed to keep exactly the same frequency response (up to about 20 kHz) as for the "dry" signal in the Mistress. Don't know whether this is important though. I just wanted to change the frequency response of the Mistress as little as possible when switching to TZF.

Regards,

Markus