CD4069 For The MXR Envelope Filter

[smallbearelec]

I have heard reports that the Texas Instruments version of this chip doesn't work in that pedal. The one I carry is the TI CD4069UBE, UBE meaning "unbuffered." Is it the manufacturer or the UBE version that is the problem with this build? Can anyone shed some light on the problem? Should I stock a few of the Fairchild CD4069UBCN?

SD

[mdh]

In my experience, the ones that work are unbuffered, just from a different manufacturer than TI.  My working version of this circuit has Toshiba CD4069UBP's in it, but my sense of the build reports really is something like "anything but TI works," strange as that may seem.

[jefe]

SD - I built the MXR Envelope Filter about a year ago from the tonepad layout. I'm going by memory at the moment, but I seem to remember the build reports at tonepad warning against using the TI chips. I found the CD4069's listed at your site, but noticed that they were made by TI, so I went elswhere to find them.

I was such a newb at that point (and still am to a large extent) that those build reports scared me away from the TI chips. FWIW, I don't know if there is any actual difference between the Fairchild and TI chips, but I can tell you that I would have bought the Fairchild CD4069UBCN's from you if you had stocked them.

[slacker]

Steve, I think the circuit needs unbuffered chips, it's just that there seems to be an issue with TI versions.
It's a bit of a mystery why the circuit doesn't like TI chips, there's been a lot of discussion about it and some fairly extensive debugging, but no one seems to have come up with an explanation.
In all the cases I've read about replacing the TI chip that makes up the high frequency clock part of the circuit with another manufacturer's has solved the problem. The chips seem to work fine for the audio and envelope follower parts of the circuit.
It could well be that with a simple tweak the circuit could be made to work with the TI chips but as far as I know no one has ever tried.

[Paul Perry (Frostwave)]

I wonder how high the frequency of the PWM clock is?
Because, the first generation of 4000 series CMOS was only good to 1 MHz or so, which doesn't leave much to spare in that particular clock circuit.

[mdh]

I wonder how high the frequency of the PWM clock is?
Because, the first generation of 4000 series CMOS was only good to 1 MHz or so, which doesn't leave much to spare in that particular clock circuit.

Mine runs about 40 kHz or so, IIRC.  Definitely somewhere in that ballpark.  This somewhat lengthy thread contains a pretty complete record of people's thoughts on this problem, I think.  And no, I never really got around to solving this definitively.  If it really is just the clock, it seems that changing some of the RC details in that part of the circuit might do the trick.

But I think what Steve really wants to know is, "Will these Fairchild ICs work consistently in this circuit?"  I looked in my stash of 4069's and didn't come across any Fairchilds, so I can't say for certain.  However, there are three build reports on Tonepad specifically stating that Fairchilds worked, so I think the answer is probably yes.

[brett]

Hi
there is a weird thing about part of the circuit that I don't fully understand, but seems relevant to this topic.  

There is a voltage divider (R14/R16 on the tonepad layout) attached to the IC2 4069.  Presumably, it influences the bias.  But the derived voltage, connected to IC2, is offset from Vsupply/2 (it is 100k/(62k+100k) x Vsupply, from memory), and this seems to stop some circuits from working.  Certainly, my TI-equiped circuit didn't work.  Fortunately, changing the 62k resistor to 100k resulted in instant success.  I believe that others have had similar success.  So the TI chip might have a tighter spec or be more sensitive to bias (if the MOSFETs have higher transconductance than the ones in old 4069s, or if the relationship between the N and P channels has changed over time it might lead to "locking up" more easily if the bias is "out" or different).
cheers

[smallbearelec]

Brett--

So you are saying that the change in the resistor value nmade the TI chip to work?

[R.G.]

(a) The circuits must be the UBE version (unbuffered) to work correctly. The normal (buffered) logic chips are not only buffered in the sense we usually think, with a high current gain stage, but also have about another 30-40db of gain after the logic section to ensure that the signal that comes out is fully banged against the power supply or ground. This is better behaviour for a logic circuit, but no for misuse as an analog circuit. This is true of all manufacturer's versions of CMOS.
(b) Each manufacturer has their own secret sauce procedure for making CMOS, and this changes as they scale down their semiconductor processing for faster, smaller, cheaper logic chips. The MOSFETs in today's CMOS are at least an order of magnitude smaller than the original ones were. As that happens, the side effects of how the circuits work in an analog fashion change slightly. This matters not at all to logic guys and a lot to us. It is not at all surprising that the ideal bias setup on two different manufacturer's CMOS 4069 chips would be different; I'd be shocked if it was the same for two chips from the same manufacturer after they changed fabs and processes for building their CMOS.

[Mark Hammer]

Hi
there is a weird thing about part of the circuit that I don't fully understand, but seems relevant to this topic. 

There is a voltage divider (R14/R16 on the tonepad layout) attached to the IC2 4069.  Presumably, it influences the bias.  But the derived voltage, connected to IC2, is offset from Vsupply/2 (it is 100k/(62k+100k) x Vsupply, from memory), and this seems to stop some circuits from working.  Certainly, my TI-equiped circuit didn't work.  Fortunately, changing the 62k resistor to 100k resulted in instant success.  I believe that others have had similar success.  So the TI chip might have a tighter spec or be more sensitive to bias (if the MOSFETs have higher transconductance than the ones in old 4069s, or if the relationship between the N and P channels has changed over time it might lead to "locking up" more easily if the bias is "out" or different).
cheers

We had this discussion a while ago in another thread.  My sense is that you are correct, and that apparent differences between chips that "work" and chips that don't is really the difference between those that will work with no further fine-tuning of that bias voltage, and those that will work once you find the specific bias required for it.  Keep in mind that all those instances where people said their chip didn't "work" were instances where they heard a muffled sound that did not sweep.  If they heard something, then clearly the invertors serving as filter sections were working fine.  The holdup seems to be whether the circuit that generates a HF clock with a changing pulse width is working correctly, and certainly a misbiasing at the point Brett describes can easily be the culprit.

[TELEFUNKON]

A look at the voltage (gate) transfer characteristics in the datasheets of Fairchild and National shows the vertical line close to 6V,
while TI shows a typical 5V with a min-max spread from 3V to 7V (whatever that wants to tell us)  for Vdd=10V.

This probably influences the thresholds of a trigger more (bias!), than differences in propagation delay times harm the functioning of
such circuits at ~50kHz.

[StephenGiles]

It certainly works with any old 4049 - I only breadboarded it though.

By Cristina Kirshner's legs, perhaps there should be a pcb for using 4049s.

[rhys]

Hey--so maybe that's the problem with my MXR EF.  I built it a few months ago with TI CD4069s.  I get some guitar sound, but no funk; it just sounds like the tone knob is rolled back.  I sat it aside to trouble-shoot later.  Sounds like I might should try some other kind of 4069s. 

[brett]

Sorry if I was unclear.  The answer is a YES, changing the bias seems to bring most (all?) of the non-working circuits to life.

The evidence seems to be backed by RG's theory about differences due to construction.  Coupled with the differences in specs in the datasheets for the zero-crossing point, it seems that we now understand this bug quite well.  Could there a single (compromise) set of R14 and R16 that works for all chips? Somewhere around 100k and 82k?  Even if there isn't, it wouldn't seem too hard to add a note to the projects saying "If using the Fairchild chip, Rx = 62k, if using the TI chip, Rx = 100k".
cheers

[Paul Perry (Frostwave)]

The moral of the story is, if you are using a chip for a purpose that the manufacturers
didn't suggest.. then you are kind of on your own.

[smallbearelec]

Thank you, all who responded to this post! I have confirmation from the customer who wrote to me that the change to 100K fixes the problem and fires up the pedal.

I have always appreciated, very much, being able to refer people to the Forum for advice and support. The depth of knowledge here is amazing, and the willingness to share it freely helps the community, and the businesses that depend on and support it, to grow and prosper. 

SD

[rhys]

Yeah, I also changed out the 62k for a 100k in mine today and the pedal works now too.  Glad I saw this thread. 

[bigp7099]

realize this is an old thread but i can also confirm that changing r14 from 62k to 100k got it working (and it is pretty awesome) i am using SCL4069UBC and MC14066BCP

if there is another more current place to post this agreement just point me in the right direction

this site is an excellent resource, can find an answer to almost any pedal build right here just need to dig a little 

[RodSerling]

The tonepad.com website has been hacked and no build reports are available anymore.
Does anyone have an old build report for the MXR Envelope Filter?
I'm building it now and could really use that help file.
If so, that would be greatly appreciated 

Thanks!

[Mark Hammer]

Soooooo.....we meet again. 

One of the early errors in Francisco's layouts - long since fixed - was that changing the "range" of the filter involved changing the value of two caps (C6 and C7), not one.  The more recent versions of the board fixed that.

Another issue that came up concerned the reference voltage needed for the high-frequency clock to work.  The schematic will show a two-resistor divider between V+ and ground (likely shown as R14 and R16, 62k and 100k, respectively), which provides a reference/bias voltage to the input of IC2a.  Depending on things like the specific make of the chip, and resistor tolerances, the reference voltage provided at the junction of the two resistors may be insufficient to set the chip into oscillation (and a non-oscillating clock chip is just a piece of silicon wrapped in epoxy). People have reported/found that varying one or the other resistance can bring the reference voltage up to the appropriate threshold.

Finally, there is a 3k6 resistor in series with the output, and a 0.1uf cap to ground on the output.  People have reported that adjustments to the value of each can make the volume-level balance between bypass and effect more even, if it isn't already.  One doesn't want to make the value of the cap too small, since there is the matter of that pesky HF clock you don't want as part of your audio signal.