Chorus hiss - source?

[herrdoktor]

I know this topic has already been discussed in numerous threads,
but I would like to understand some things about hiss.
I bought a chorus that is a clone or almost of a BOSS CE except that it  uses a MN3207 supplied to 5V.
I have a loud noise when the pedal is turned on so I changed almost everything (I'm an engineer)
except for the BBD and its clock looking for something that justifies it (voltage regulator, OPAMPs, electrolytic caps etc..)
I read a lot about the  filter anti-aliasing and  reconstruction filter ...  and I also changed the caps and resistors
like original BOSS scheme, without consequences.

The question is:
Is the MN3207 BBD in its new version against the MN3007 adopted in BOSS CE Pedal  
that produce this noise and so it's impossibile to avoid?

[Mark Hammer]

The principle difference between the 300x and 32xx series is their ability to run on lower supply voltages, and the flipped ground-polarity.  Just how that would affect S/N is beyond me.  In theory, the frequent use of 15V supply voltages for MN3007-based circuits might facilitate better S/N ratios....if only it meant such circuits exploited the greater available headroom and goosed the input.  The fact is, however, that chorus, and BBD-based devices in general, run at unity gain, and the BBD itself has much lower headroom than the supply would support, so any change in noise levels is unlikely to be a result of supply voltage IMHO.

The history of BBD-based circuits has been one of eternally balancing off bandwidth against delay time against noise.  You can always get more bandwidth, if you're willing to sacrifice either noise-control or delay time.  You can always get more delay time if you're willing to sacrifice bandwidth or put up with noise.  Some circuits sit riiiiiiiiiiigghhht on the edge of that perfect balance.

It is also worth noting that one of the advantages of the MN3102/3101-BBD arrangement is that the clock can be placed very close to the BBD to keep the clock lines mercifully short and out of the way of everything.  BBD-based circuits are susceptible to clock noise bleeding, and need good layout.  If you used something other than an MN3102 for your clock generator (not a bad thing in itself), it is possible your layout has somehow placed audio lines precariously close to clock lines.

I might also popint out that many chorus pedals, Boss included, use a sort of "poor man's Dolby", by providing pre-emphasis of treble before the BBD, and complementary de-emphasis at the final mixing node/stage.  You can see it in the 6800pf/10k network ahead of the first op-amp and in the feedback loop of the last op-amp here: http://www.generalguitargadgets.com/boss/BossCE-2.gif   It is technically possible that the tolerances of the components involved can be providing more pre-emphasis than is counteracted by the de-emphasis.

Finally, while not every circuit does so, it is considered a "best practice" to use a trimpot on the BBD output to perfectly balance the two complementary outputs so that the clock signals coming out of each sub-line cancel out with the other.

[herrdoktor]

Hi Mark,
I would to thank you for your comprehensive answer.
I've only a doubt but I find another post in which you explain the same "best practice".
I would only a confirm about that, from you

I can balance the 2 output of my MN3207 with a trimpot like showed in figure?

The overall impedance remains the same, so it seems ok....

[Mark Hammer]

Unfortunately, I can't see the picture (work firewall), so maybe somebody else can confirm.  The other option is to simply use 1% resistors for the pair we usually see on the output of designs that do NOT use a trimpot.

[Fender3D]

Hi Mark,
...trimpot like showed in figure?

Yes

...The other option is to simply use 1% resistors for the pair we usually see on the output of designs that do NOT use a trimpot.

Or you may tie pins 7 and 8 together

BTW
fine adjustment is needed with low clock freqs, if you're well above audio you won't notice the difference (with or without trimmer)

[herrdoktor]

The Clock is a 88KHz....
I'll put the trimpot only to do the LAST TEST!

[Mark Hammer]

Or you may tie pins 7 and 8 together

yes.  Zero ohms are usually within 1% of each other.  

[Fender3D]

My "well above audio" er.. above ^^ should be "well above twice the max allowed audio frequency"
With a 88KHz clock you can safely get a 20KHz signal bandwidth 
You might as well change LPFs' cutoff freq, maybe a simple RC filter is enough, thus you might even avoid active LPF (I found single transistor filters noisier than op-amp filters).

[herrdoktor]

Today I have made the mod (for the output balancing) at the MN3207 and this evening I'll test it.
Yesterday I added a capacitor in the filter stage before the BBD and nothing changed but If I add a cap in the filter after BBD
the hiss obviously is reduced but I'm reducing the bandwith of my audio signal....

The hiss is modulated by the LFO.
Did you experiment this beahviour?
With a battery, I see a ripple on the battery voltage syncronized to the oscillations of LFO.

[Mark Hammer]

If the hiss is modulated by the LFO, then that would suggest we're dealing with clock-related noise and aliasing, rather than simply a "hissy" circuit.

Just for an experiment, try using a smaller value clock cap.  That would move any aliasing artifacts upwards in the spectrum, and possibly above the corner frequency of the lowpass filtering.

[Fender3D]

According to datasheets MN3007 has 80dB s/n whereas MN3207 has 73dB...
You'll just hear those 7dB as added hiss...

ADA Flanger has a minimal bandwidth reduction at BBD output, and I wouldn't say it's hissy...
But maybe it's just me, I have more a "vintage" approach when building my sound, and my noise tolerance might be somewhat higher... 

[Mark Hammer]

Unfortunately, we never know the circumstances under which those S/N ratios are obtained.  I'm certain it is nothing as deceptive as the conditions under which $10 plastic computer speakers get rated as "180W PMP".  Still, what sort of signal, what supply voltage, and what support circuitry?  It may be that typical use negates 307/3207 noise differences; it may be typical use exaggerates it.  I have no idea, and I imagine most of us here don't either.

Part of what allows the A/DA to have lower audible hiss is:

a) very high clock frequency (i.e., aliasing up where only your dog can hear); and

b) a gate that cuts out the delay signal when it goes below a level threshold.

[Fender3D]

Right, but
a) goes as low as ~35KHz and
b) may be bypassed
and it is not hissy, at least not so hissy...
what is different from every other flanger (and chorus) design is the lack of LPF at BBD output. Just pre/de emphasis blocks...

[herrdoktor]

I did the test for the balancing with no changes.


Fender3D Would you suggest me something about LPF at BBD output?
A 5th order or a lower corner frequency to despite of bandwith?

thx

[Scruffie]

I'd try upping the BBD voltage from 5V, i've found them to run best around 8V2 (set with a shunt diode) regards S/N.

[Mark Hammer]

Sensible and do-able.

The MN3207 can be run at 5V, and many chorus/flanger designs DO run it at 5V, but that doesn't mean it HAS to be run at 5V.

The MN32xx series, as near as I can fathom, was intended to solve an enduring problem regarding batteries; namely, how do you set the bias reliably if the supply voltage you are dividing down keeps changing as the battery ages?  If the BBD can run off, and be biased from, a stable 5V supply, then a simple LM7805 5V regulator can yield that stability until the battery drops down to just over 7V.

But again, that is a challenge for a purely battery-operated circuit.  If the circuit is intended to be used with an external supply, then there is no requirement whatsoever to stick with the regulated 5v, and every opportunity to opt for a different supply voltage that might provide better performance.

I personally don't know if BBD supply voltages above 5V sound better.  But they might.

Good call, scruffster.

[Scruffie]

Sensible and do-able.

The MN3207 can be run at 5V, and many chorus/flanger designs DO run it at 5V, but that doesn't mean it HAS to be run at 5V.

The MN32xx series, as near as I can fathom, was intended to solve an enduring problem regarding batteries; namely, how do you set the bias reliably if the supply voltage you are dividing down keeps changing as the battery ages?  If the BBD can run off, and be biased from, a stable 5V supply, then a simple LM7805 5V regulator can yield that stability until the battery drops down to just over 7V.

But again, that is a challenge for a purely battery-operated circuit.  If the circuit is intended to be used with an external supply, then there is no requirement whatsoever to stick with the regulated 5v, and every opportunity to opt for a different supply voltage that might provide better performance.

I personally don't know if BBD supply voltages above 5V sound better.  But they might.

Good call, scruffster.

Cheers 

If I recall from the Datasheet it does actually show that S/N improves as voltage increases and as does THD but that begins to climb again after a certain point (I read this after i'd experimented and had found 8V2 to be the best balance of distortion/noise).

So yes if memory serves correctly, it wasn't just me hearing things. And by using a Zener as a shunt, it'll still run as voltage goes down unlike a regulator... just not quite so steadily.

You can see this applied in the Ibanez AD-9.

[herrdoktor]

I'm sorry, but the power supply at 8.2V
I could bet doesn't affect the issue.
The minimum THD at 8V and the SNR a little lower at 5V don't affect the hiss coming out.
73dB or 80dB...no differences.

[Fender3D]

At this point, after your tests, I think you might stack up several little improvements (out balance, reduce bandwidth, raise power supply...) to achieve one major improvement.
Since you tried lowering input LPF cutoff with no effect, you might as well raise BBD input signal by tweaking pre-emphasis gain and then lower BBD out signal on final mixing...
Watch out for distortion though...

[herrdoktor]

If I have time today I'll test the 8.2V
in the circuit (I hope it will be easy to do)

I bought few days ago an MN3207 and a MN3102 so
I'll test these 2 new IC when they will arrive.

After that I could consider to throw out the pedal...