9V BBD Flanger Layout problem/question with video of issue

[bushidov]

Hi Guys,

I have another question about Flanger circuits that is more based on layout than in schematic. So, a while ago, I took apart a Mooer/Donner/Stagg/Rowin 9V Flanger. It works great. I traced every trace and identified every component and built up a schematic, layout and BOM.
Stagg E-Lady Base Board (Has all 3 pots, most of the power regulation, the input/output op-amp 4558 and quad op-amp LM324 which is the square/triangle wave generator, 4.5V bias maker, and Range Controller)


Stagg E-Lady Daughter Board (Has the MN3207 BBD, the LM311 Diff Comparator, CD4013 Flip-Flop, and CD4049 Inverter, as well as both BJT's)


So, after tracing this whole thing out, I noticed that it was kind of close to MadBean's Current Lover, but had a few things different with it. BUT, it is almost exactly like the EM3207 v1.1 that Thomeeque came up with a while back. The only differences being the surface mount components, and such to make it into two boards (headers, LEDs, stuff that crams it into a 1590A enclosure)
Here is the schematic of the Stagg E-Lady


So, I decided to lay the board out similar to how Thomeeque laid his out, but with out the want of jumper wires and such that his project demanded.


Well, this made a ticking noise, but not one that is a straight ahead ticking, but an irregular ticking that follows a repeating pattern. I asked my friend Mickey to take a look at it and he stated I should get my input lead further away from my output lead. So I did. Still the same problem.

I then redid the layout of the board, expecting that the issue was something akin to the old Zombie Chorus, where you need to keep the PLL and BBD chip as far away as you can from the input op-amp. I noticed this layout method is what Mooer/Stagg/Rowin/Donner did on their E-Lady (Keep the op-amps on one board and the BBD and CD4XXX chips on the other). So I made this layout.


Still is making this ticking noise. I recorded a video of the issue when I was Skyping with Mick to help me.
https://www.youtube.com/watch?v=jKQtqsFc_6Q

Anyone got any ideas? Paul, Rob Strand and Mark Hammer, I am definitely looking at you folks. LOL! Anyone even got a hint, I'd gladly take it (although, I'd love a solution). This thing is driving me nuts.

I can also add that the frequency of the pattern of ticking increases and decreases with the turning of the Rate potentiometer and the intensity of it increases and decreases with the turning of the Range potentiometer. The "Color" pot has no bearing on it. When I throw the switch from normal mode to filter, if I have my Range all the way up, it just makes a solid faint whistle (same volume as the ticking was), and if I move the Rate pot lower, the whistle changes its frequency to lower. If I move the Range down, the whistle starts high pitch and then goes down to low, then starts back right at the top again and then goes down low.

[DrAlx]

The tweets sound like heterodyne noise.  Suggests something else in the signal chain either before or after that pedal has a clock of its own or is sampling the signal somehow.
If you have just the guitar, the pedal powered by either a battery or a linear supply (not switched-mode), and an analogue amp, do you still get the tweets?

Reason I mention an analogue amp is that if you have an amp with built-in digital effects then the amp could be sampling the signal (e.g. at 41 kHz) and that would give heterodyne noise due to sampling residual clock signal on the pedal's output.

When I was working on my EM3207 build, I audio-probed the circuit at various points and found I could hear tweets (like yours but VERY MUCH louder) when I audio-probed the clock lines directly. Don't recommend you try it.  I didn't realise till much later that I was only getting those tweets because I had my probe plugged into a digital recorder.  If I plugged the probe directly into a simple (non-digital) amp the tweets were gone.

[bushidov]

Hi DrAlx,

Interesting. So I am just going from guitar, to that pedal, to my amp and it happens whether I am using a power supply (which I built, which just takes a 12V switch mode and drops it down to 9V via LM7805's with some passive filtering caps) or just a 9V battery.

Now, the amp is interesting, as it is a Joyo Bantamp, which is basically an analog amp. It has a 12AX7 preamp tube circuit to some vanilla TL072 op-amps and then through a class D solid state power amp. So that would be all analog... HOWEVER... It does have a Bluetooth option, which goes through a Renesa's microcontroller of some sort, so that falls into the digital category.

I have a "Smokey" amp clone which is just an LM386, a 9V battery, and some discrete passives (resistors and caps), so I'll check that out.

Mickey said it sounded like Heterodyne noise too, so I think you may be onto something.

The thing that doesn't make any sense is that the Mooer/Donner/Stagg/Rowin circuit, which schematically is identical to this one, doesn't chirp at all. So I am confused to what they did differently. Unless the surface mount chips make everything magically work better (which I do know is the case with antennas and RF noise, but usually in the MHz levels, not the kHz levels).

I am still betting on distances or something related to grounding. However, with grounding, it has a whole ground plane, so that too would confuse me.

[DrAlx]

Class D works by switching. If it's switching in the range of the EM clock you can get heterodyning.
Strange the micro pedals didn't have the same problem.

[bushidov]

Yeah, and you further threw me a curve ball. I can't find my Smokey Amp anywhere, so I just opened Amplitube on my PC and created a chain of guitar->Flanger->Interface (Scarlet 2i2)->Amplitube 3 VSTi and like you called it... no ticking or noise.

So, you are onto something. But you are correct that the odd piece to this puzzle is the micro pedal, with the same circuit, had zero ticking/whistling/noise on my Bantamp.

However, I am starting to wonder something. In a non-related issue, at my day job, I was working on a switching power supply for a pumpboard controller I am working on and on my prototype, I used ceramic capacitors that were rated X5R or worse for the coefficients, which means they were "slow". I spec'ed X7R, which are faster, but in my haste and only having X5R's on hand and no X7R, I rolled with what I had. Because of this, my microcontroller flipped out when creating a PWM signal to the MOSFET driver that makes the motor run. The problem was that the caps where not fast enough and caused enough noise on my circuit to tick off the microcontroller. Took the X5R's off and put on X7R's when they came in and voila, noise went away and the microcontroller was happy again.

I say all of that to say this: My through hole design uses electrolytic capacitors for the multiple bulk capacitors in the power supply line of this flanger, some of which, are smoothing out DC ripple (large bulk caps after the small value resistors, not the ones before). On the micro-pedal, they are Tantalum capacitors, which do have a much faster switching rate. Could that be the part I fudged? Would that remove that heterodyne noise?

[PRR]

> X5R ...they were "slow". I spec'ed X7R, which are faster,

Not really slow/fast. The high-salt ceramics vary in value a LOT with voltage, which  goes along with "my microcontroller flipped out". If it expected 0.01u, or 'some steady value', but really saw 0.007u to 0.019u as the cap-voltage varied, it might be confused and do a wrong thing.

[DrAlx]

Use a scope to compare clock noise voltages on your build with the micro pedals.
I put my EM3207 on slowest widest sweep with no feedback (color) and nothing plugged into the input jack or output jack.
If I measure directly on the emitter of the 2N3904, I see clock wave (i.e. ripple) leaking through, and it goes from around 40 mV pk-pk at the high end of the sweep increasing to 97 mV pk-pk for the bottom of the sweep.
If I measure on the pedal output, I can barely see the waveform on my cheap handheld scope, but it is still there and scope gives a reading of about 6 mV to 8 mV pk-pk.

[bushidov]

Well crap... I let out the magic smoke on my working good Stagg/Donner/Mooer/Rowin pedal. When probing, I slipped and bumped something and now it won't turn on... I'll have to tear into it and see what pooped, but now even the LED won't come on.

I'll measure the through-hole ones on my scope when I get a chance.

What I did noticed was on the two square wave forms that the CD4013 was creating. I see what appears to be the problem on the oscilloscope on the CD4013, pin 2, where it goes to pin 5 of the CD4049. On pin 1 of the CD4013, where it goes to pin 7 of the CD4049, I get a square wave that shrinks and grows in frequency (and not amplitude), like it is supposed to be. On pin 2 of the CD4013, where it goes to pin 5 of the CD40490, I get the inverted square wave that shrinks and grows in frequency (and not amplitude), like it is supposed to be... BUT on the pin 2 square wave, I get a "horn".


That horn is only on pin 2 of the CD4013 and not on pin 1. This is what happens on my through hole version.

On my micro-pedal clone, both square waves lacked that horn.

Not really slow/fast. The high-salt ceramics vary in value a LOT with voltage, which  goes along with "my microcontroller flipped out". If it expected 0.01u, or 'some steady value', but really saw 0.007u to 0.019u as the cap-voltage varied, it might be confused and do a wrong thing.

Both XR5's and XR7's were rated at the same voltages (50V), were the same surface mount MLCC package (1206), and both at 10% tolerances. XR5's couldn't keep up with the switching power supply and allowed for the noise to inject on the 5V power rail while the XR7's kept up with the switching power supply and rejected the switching noise into the 5V power rail. The noise on the 5V rail made the PIC18F242 jumpy. I figured that would be an issue of speed, as the cap-voltage didn't vary on a capacitance meter and didn't reflect that on a 'scope. It seemed like it acted different under different frequencies, like how electrolytics act different than MLCC, than film capacitors, than Tantalum.

[DrAlx]

I'll check the CD4013 waveform on my build.  I don't recall it having "horns".  Hard to say if its significant or not.
I would start by measuring clock noise in the two places I mentioned to first see if it is much larger than expected, and if it is, then try to explain why.

Filtering on the EM is very light and it's part of its sound.  There are a few things you can do to minimise clock noise though.

1) Replace the two 4k7 resistors on the BBD output with a single 10k trimmer.  You might be able to mimic that by adding a large resistor in parallel to one of the 4k7 resistors to tweak its effective value.  Just use trial and error to get the right value to add in parallel.

2) You might not be able to hack this on your board but I'll mention it anyway.  Instead of using a 680pF smoothing cap before the 2N3904, it is better to use a 3n3 smoothing cap like the original 9V EM but it needs to be put after the 2N3904 and separated from it by a 1.2k resistor (1.2k is the value of "Rds||10k" in the following picture). 

https://1drv.ms/i/s!AvrH61utWEtEiAbzPDPv5UK-geVK

This will not only make the effective frequency response of the "wet" path a closer match to the original SAD1024 based 9V EM, but it will lower the noise too.
Described in a lengthy post I did on the EM3207 thread here:

https://www.diystompboxes.com/smfforum/index.php?topic=91981.msg1085896#msg1085896



EDIT: I have checked the CD4013 waveforms on my build.  No horns.  But then I only have a simple handheld scope.  One thing I did on my build was have a 100nF ceramic soldered directly between supply pins on every IC.

[bushidov]

One thing I did on my build was have a 100nF ceramic soldered directly between supply pins on every IC.

Tried this first, as it seemed the easiest to hack onto the board. Sadly, didn't change a thing.

1) Replace the two 4k7 resistors on the BBD output with a single 10k trimmer.  You might be able to mimic that by adding a large resistor in parallel to one of the 4k7 resistors to tweak its effective value.  Just use trial and error to get the right value to add in parallel.

Tried that. I removed both 4.7K resistors and inserted a single 10K trimmer, tying the BBD's pins 7 and 8 to one end of the trimmer and the other to the R19 47K pull down. Throughout the span of the 10K trimmer 0-10K resistance: no change. So that isn't it either.

You might not be able to hack this on your board but I'll mention it anyway.  Instead of using a 680pF smoothing cap before the 2N3904, it is better to use a 3n3 smoothing cap like the original 9V EM but it needs to be put after the 2N3904 and separated from it by a 1.2k resistor (1.2k is the value of "Rds||10k" in the following picture). 

I was able to pull this off kind of easily. Desoldered the 680pF and desoldered the 82nF cap that connects the Q1 transistor back into the in & out op-amps. Then I made a "tripod" of the 82nF cap, that 1.2K resistor and the 3.3nF film cap. I connected the 1.2K to the emitter pin of the transistor, which then connects to the 82nF cap. The other end of the 82nF cap goes back into the line that reconnects it back into the in-out op-amps. I then took that 3.3nF, which is tripoded to where the 82nF and 1.2K resistor meet and put the other leg of the 3.3nF to ground.


No difference. I tried changing the 1.2K to 10K just for kicks. Still, no go.

I didn't think any of this would work as it still doesn't explain the primary problem of with another board with the exact same schematic, how is it working and this not? The 1590A didn't need to change the schematic or values of items in the schematic to work. BUT... DrAlx, you were correct on it works on an analog amp, which I didn't expect to be the case. So you are onto something.

I am waiting for my 100uF tantalum caps to get here as I do realize that was a difference between the two circuits. I hope that is it. Otherwise, it still has to be some sort of layout issue.

Any other ideas? I am open to them.

[DrAlx]

The changes I mentioned would not eliminate clock noise, just reduce it.

What pk-pk clock-noise voltage levels are you getting using the setup I described (i.e. no inputs/outputs or feedback, and with slow wide flanger sweep)?  Measured both at the top of the 0.082uF  (i.e. top left corner point in your diagram with the mods) and at the circuit output?  I'm interested to know if it is in line with my measurements, or if it is a lot larger.

EDIT: I wonder if a half working BBD would explain things? Unlikely, but if one BBD output is fine but the other does not output anything and just floats, then the balance trimmer would do nothing and you would have a lot of clock noise.  That would show up in the measurements.

[bushidov]

Well, now I am afraid to go further. I noticed when doing the whole "tripod" thing and splicing up the board, it now barely flanges. I mean, I can kind of hear it, but it isn't well defined anymore. I tried messing with the bias and trim pot on the PNP transistor section, but no go. And, I still get the ticking/whistling, albeit not quite as much. The BBD isn't dead because I swapped it out with another BBD from another 1590A board. No issues when in that pedal.

So, I am using my second board I made which is just using my schematic, which again, I know works because of the 1590A.

I did change out all my capacitors with a value of 1uF or higher with Tantalums to perfectly match the 1590A board that works and I am still getting that whistling and ticking. It is driving me nuts, but what the heck did they do that is not having this problem that I am not doing? I mean, yes, I am using through-hole components and they are using SMD, but is that really creating that much of a difference? Or is it that I really need to make 2 PCBs to separate the "problem" that I cannot seem to identify?

Being as I am using this schematic now:


Where should I get my oscilloscope info from?

[DrAlx]

I can understand the eagerness to get things fixed, but changing parts and layout without first taking measurements is not the best way to go about it. Even if that approach happens to work, we won't be able to get an understanding of what was actually wrong, or quantify what effect those changes had.
You are in a fortunate position if you still have a working 1590a pedal since you can take measurements from that and compare them against your own build.

So...

No inputs or outputs to the pedal except for power. Ideally use a battery or linear supply.
Feedback (color) at minimum.
Range at maximum.
Filter matrix on.

Put clip of probe on the ground at the output socket.
Then probe the amplitude of the residual clock wave (i.e. AC measurement) at

1) The line labelled BBD-OUT.
2) The point labelled OUTPUT, i.e. the output jack.

The frequency will be around 40 kHz (25 microsecond period) but it depends on how the VCO trimmer is set.
Report the period of the wave along with the pk-pk voltages.

You should at least be able to take measurement 2 on the 1590A without frying it since it is measured on the jack.

[j_flanders]

All the debugging so far revolves around the noise being generated or injected or modulated 'internally', electrically like the clock noise or switching power supply heterodyning for example.

Could it be that the whistling noise is being 'broadcasted' 'through the air' and picked up somewhere along the signal chain by an 'antenna component'? (induced noise?)

The last three times I had this ticking or whistling issue (tremolo, phaser, modulated bbd delay) each time it was broadcasted from the shape or rate control in the LFO section and picked up by a 'sensitive', susceptible part in the circuit nearby.

So, could it be a lead dress or layout issue?
How about you disconnect the output wire from the pedal and hover it above and around the board to find out which part or component is broadcasting the whistling.
When and if you locate it, you could check which nearby components or traces in the signal chain could be susceptible to this noise or which could serve as an antenna.

Maybe a long shot but very easy to try.

If that is the case, another thing I would try is to desolder the three pots and move them away from the pcb (besides instead of above)  and see if the whistling is gone or at least a lot less.

In fact, this was your own first idea (referring to the Zombi Chorus problem)
And this would only confirm it:

I didn't think any of this would work as it still doesn't explain the primary problem of with another board with the exact same schematic, how is it working and this not? The 1590A didn't need to change the schematic or values of items in the schematic to work.

Actually, even the title says it: layout problem

[bushidov]

DrAlx, the problem now is I have a dead (no power at all) on the 1590A. When taking a measurement a few post ago, I accidentally shorted something out and now it won't even power on. I am in the process of seeing what that component (components) are, but until then, I cannot get anymore measurements on the 1590A board. I have in addition to that, 2 PCBs I made with the exact same schematic, but with two different layouts, one of which is from my original post on the top. Both make the exact same whistle noise and both, at first, were flanging perfectly.

I decided to start splicing into the one with your tests, but as I mentioned earlier, after I did the who thing where I removed the 680pF cap and injected a 1.2K resistor and 3.3nF cap, it still ticks and whistles, but it barely flanges anymore.

The second board, I left alone, except I finally got my tantalum capacitors for all the 1uF, 10uF, 33uF, and 100uF values. I hoped that would have done it, as that would, component-wise, make it identical to the 1590A working unit (except for layout and the SMD vs through-hole thing)

I will grab those oscilloscope measurements this morning and post from my last working PCB.

j_flanders, I absolutely agree, it has to be a layout thing.

How about you disconnect the output wire from the pedal and hover it above and around the board to find out which part or component is broadcasting the whistling.

I tried roughly doing the whole floating the output wire around till I got the noise pin-pointed, but I didn't seem to pick up anything when trying that. I am not sure I was thorough enough, though. I will check again.

If that is the case, another thing I would try is to desolder the three pots and move them away from the pcb (besides instead of above)  and see if the whistling is gone or at least a lot less.

Haven't tried that yet. I'll give that a go next.

The part that DrAlx nailed, and further confuses me about the "layout" possibility, is he is absolutely correct: It only whistled when attached to my Joyo Zombie Amp. If I plug it into my computer's interface and I open my DAW or even just a VSTi pluging (Amplitube, Guitar Rig, etc), the ticking/whistling doesn't show up at all no matter how loud I crank it. But, that Joyo Zombie Bantamp is a Class D, and there is switching going on, on that amp's PCB. So there may be whistling because of clock-mismatching. But then again, the 1590A version of the circuit din't have that problem with the Joyo. So, can a layout issue do that? I don't know.

At this point, I really want to know, as there is something to be gleaned from this that I may need to know and apply to further designs I may do in the future.

[j_flanders]

It only whistled when attached to my Joyo Zombie Amp. If I plug it into my computer's interface and I open my DAW or even just a VSTi pluging (Amplitube, Guitar Rig, etc), the ticking/whistling doesn't show up at all no matter how loud I crank it.

I don't know if this is related but with my OneSpot and some other switching psu's I have loud ticking when there's nothing in the signal chain (guitar, pedals, amp, laptop, active speakers) that is directly connected to mains earth. (Floating ground)

Connected to a laptop (DAW) with a two-prong: ticking
Connected to another laptop (DAW) which has a three prong: no ticking
Connected to an amp with a two prong (adapter): ticking
Connected to a (tube) amp with a three prong: no ticking

This could also explain why some people complain about noise while others say it's dead quiet using the same power adapter.

When there's ticking I usually also get 'zapped' or shocked lightly when sensitive skin touches metal parts anywhere in the signal chain.

I solve this issue by simply running a wire (with alligator clips at each end) from the grounding pin in a wall socket to a metal part of a pedal, amp or laptop. This completely eliminates the ticking and the zapping.

This issue is different though from the LFO ticking and whistling I got in the tremolo, phaser and bbd. Wich I think DrAlex wants to rule out with the scope measurements.

I see the Joyo Amp is using an 18V adapter instead of a 'regular' three prong cord and is probably not directly connected to mains earth.
Is your laptop using an adapter with a three prong, or is something else in that setup connected to earth?

[bushidov]

Laptop and joyo have 3 pronged power supplies. 1590A had no problem with either.

[DrAlx]

Sorry to hear the 1590A is fried.  I got the impression you had 2 of them for some reason.

"Ticking" and whistling (i.e. heterodyne noise) are two different effects and have two different causes.
Your problem seems to be whistling rather than ticking.

"Ticking" is down to the LFO part of the circuit.  One of the op-amps in the LFO outputs a square wave.  That, and the associated rapid change in current draw by that op-amp can affect the audio in the circuit.  Remedies for ticking typically include:
  1) Adding extra parts to the LFO to round-off the square-wave corners.
  2) Giving the LFO circuit its own it's own RC filter direct to the main power source.  i.e. do not have the LFO daisy-chained onto the audio chips as it is in the EM3207, since large changes in current will manifest themselves as changes in the voltage supplied to the audio chips.
  3) Keep the wires to the square wave output as short as possible and as far away from the audio as possible.  In this circuit, that means the wire to the rate pot, since that connects directly to the square wave.  If you have a long wire to the rate pot and turn the pot resistance down to zero, then the long wire will  radiate like an antenna and broadcast the tick all over your circuit.  Tick noise typically gets louder with faster rates since that corresponds to the situation when the current in the wire is strongest (because rate pot resistance is at minimum).

In my EM3207 build, I did not make any circuit changes.  i.e.  I did not address points 1 or 2.
I only made sure that my layout satisified point 3) and I have zero ticking.
Your layout seems to satisy that requirement also, so I would not expect it to tick.
I did not hear ticking in the video you posted, only whistling.


The "whistling" (i.e. heterodyne noise) occurs when you have two similar frequencies "beating" with each other.  There is a single clock in the pedal (~40kHz to ~400kHz) which is how I knew you must have had another clock elsewhere in your signal chain to get those whistles.
I have built several EM based flangers, and they ***all*** have a measurable clock signal on the output. 
If I listen to those flangers through a digital recorder with headphones, I can hear very faint whistles.  On an analogue amp they are dead quiet.
If your computer (which I assume also samples things) is not giving you heterodying, then it must be either sampling at a much higher rate, or more likely filtering the audio before sampling.

To reduce heterodyne noise you need to minimise how much clock signal leaks to the output. 
It is very unlikely that layout is the cause of your problem and there's a very simple reason for that.
Remember that the clock signal operates on the audio ***directly*** in the BBD, and that's why you see a large unwanted clock signal on the BBD output.
That unwanted clock signal is easily measurable with a scope, and it will be many orders of magnitude larger than any sort of parasitic coupling through the air that you might get from your layout.


I can see two possibilities at the moment:

Either the clock leaking through to the output...

1) ... is much stronger than measured on my build.
That would explain the noticable heterodyne noise, and from that point we could progress by using my circuit as a reference to try and find out why.

Or

2) ... is the same as on my build.
I couldn't give much further advice in that case.  It is the more interesting situation though, since it suggests
the Mooer guys have done something different to the EM3207.
You claim it's the same circuit as the EM3207, but is it?  I can see the surface mount resistors in the photo have the same sort of values, but what about the caps?  How can you tell they used the exact same cap values?  If the clock noise at the output of the Mooer is much lower than the EM3207, they are more likely to have achieved that by filtering with heavier caps than by layout.

[bushidov]

So, good news and bad news. I did notice a slight difference in my schematic. The 10K pot that goes from the bias to 9V instead of the one going to ground, well, I have it going to 9V after the 10 ohm resistor, so I moved it to pull direct from the 9V to match the 1590A. Also, the 1590A had an inline Schottky that I missed on mine, so I added that as well. Still, no change.

I did get my 1590A working again. I found out that when I was taking measurements earlier and accidentally bumped something, I shorted out the 10 ohm 0603 resistor on its power supply. After I replaced it, it began working like a champ. So, now I can make comparisons against that.

You claim it's the same circuit as the EM3207, but is it?  I can see the surface mount resistors in the photo have the same sort of values, but what about the caps?  How can you tell they used the exact same cap values

I did a post here a couple months back where I took hi-res photos of my tear down. I literally desoldered all 3 PCBs in the 1590A Flanger. Desoldered all the components one at a time and measured them. I have a capacitance meter that goes down to the 0.1pF, so I am pretty confident in the readings I got by doing that. Then I traced the board out with a ohm meter and drew up a clone in EagleCAD. I still have the files if you want them. The only problem I had there is that I have slightly different footprints for some components and can't draw slotted holes to make the DPDT switch fit, but everything else is spot on.

But yeah, that's how I know it's the same, minus what they had to do make it surface mount (the use of SMD parts instead of through-hole ones) and headers and such for connecting it all together.

As of the 1590A, as you'll see in this video, I re-attached with wires instead of headers to try j_flanders' idea of moving some components closer and further away. But so far, the 1590A doesn't care how close or how far that top board is from the bottom board with pots.

Here's a video to show the comparison:
https://youtu.be/EN8XwPA1jkE

[DrAlx]

Wow. I didn't know you did a whole strip down.

There is one thing I just noticed about your schematic vs the EM3207. Your BBD input in the schematic is not connected directly to U1B pin 7. It is on the wrong side of R10.  So your BBD input is only separated from the circuit output by a cap rather than by R10 and the cap.

Missing direct connection to U1B pin 7 means you will always have some output fed back to BBD input, which is not the case in the EM3207.  Is that a schematic error, a tracing error, or a known difference?

And did you measure the clock noise voltages?