FV SPIN - NOISE

[AndyF]

Hello, I tell you I have tried different schemes, more than 50 algorithms, my own and others that I got out there.
The conclusion is that there is noise in all of them, in some more, depending on the type of effect, you can hear a lot.
The solution to filter the output of the fv spin, the problem is that I cannot find a balance between what is filtered and the sound is darkened, with the elimination of the blowing of the fv spin itself.

So, has someone made something interesting that they can share with a schematic, or some tips to solve this?

Ldo regulators do not influence anything.
The pcb design is, but it is not decisive.
Different types of mixes make the fv spin noise more or less heard, but when you want to achieve wet / dry the noise is a lot. There are brands on the market that launch their products with tremendous noise, I can't explain how they don't care.

ok, if some genius can give a solution I will appreciate it for all eternity :-)

thanks

[vigilante397]

Welcome to the forum

What have you tried? It's hard to know what to tell you without knowing where you're starting from. The FV-1 isn't a hi-fi chip so you can't expect noiseless, but plenty of people have gotten great results with it. At the end of the day you have to know how much noise you're willing to accept though, hopefully you aren't expecting crystal clear hi-fi sound.

[AndyF]

Welcome to the forum

What have you tried? It's hard to know what to tell you without knowing where you're starting from. The FV-1 isn't a hi-fi chip so you can't expect noiseless, but plenty of people have gotten great results with it. At the end of the day you have to know how much noise you're willing to accept though, hopefully you aren't expecting crystal clear hi-fi sound.

Exactly, the noise is always there, I can't explain why the manufacturer doesn't solve it. This circuit is the one that I like the most but the one that gets a lot of noise and I can't find how to modify it.

[ElectricDruid]

Welcome!

I can't explain why the manufacturer doesn't solve it.

Probably because designing noiseless ADCs and DACs on the same silicon as a processor is *extremely* difficult. It's one of the reasons codec chips are always separate from DSPs (otherwise it would be very handy to have the ADC/DAC on the same chip, but apart from the FV-1 hardly anyone does).

This circuit is the one that I like the most but the one that gets a lot of noise and I can't find how to modify it.

You could get all "old school" with it and add a 571 compander circuit like a BBD delay. That would give you a better signal level going in and squash the noise coming out.

One other trick I've seen suggested is to feed a differential input signal to the FV-1's inputs, and then combine the two outputs at the other end. That should cancel any common mode noise, at least. How much improvement that makes, I can't say - I haven't tried it myself.

HTH

[AndyF]

Welcome!

I can't explain why the manufacturer doesn't solve it.

Probably because designing noiseless ADCs and DACs on the same silicon as a processor is *extremely* difficult. It's one of the reasons codec chips are always separate from DSPs (otherwise it would be very handy to have the ADC/DAC on the same chip, but apart from the FV-1 hardly anyone does).

This circuit is the one that I like the most but the one that gets a lot of noise and I can't find how to modify it.

You could get all "old school" with it and add a 571 compander circuit like a BBD delay. That would give you a better signal level going in and squash the noise coming out.

One other trick I've seen suggested is to feed a differential input signal to the FV-1's inputs, and then combine the two outputs at the other end. That should cancel any common mode noise, at least. How much improvement that makes, I can't say - I haven't tried it myself.

HTH

ok thanks, among so many things tried, you should try this.
Hopefully someone else contributes other ideas.

Now it would be much easier if the manufacturer indicated how to solve this, when you write in their forum, they only describe what is the design, the regulator, the algorithm, the problem is never the fv spin ...
that is wrong! The fv spin is the noise generator, which of course, depending on what you input, will come out better or worse, but always with noise.

[vigilante397]

The fv spin is the noise generator, which of course, depending on what you input, will come out better or worse, but always with noise.

I was afraid you would say that Have you looked into the FX-Core from Experimental Noize? Similar-ish to FV-1 but more powerful and uses external CODECs like Tom mentioned, so it would be much more suited for a low noise solution. It sounds like you're expecting a lot from an inexpensive all-in-one package that was designed more to be user-friendly than to be perfect.

[AndyF]

The fv spin is the noise generator, which of course, depending on what you input, will come out better or worse, but always with noise.

I was afraid you would say that Have you looked into the FX-Core from Experimental Noize? Similar-ish to FV-1 but more powerful and uses external CODECs like Tom mentioned, so it would be much more suited for a low noise solution. It sounds like you're expecting a lot from an inexpensive all-in-one package that was designed more to be user-friendly than to be perfect.

I understand what you say, what I do not share that things are done more or less, for me they are good or bad, and I believe that something can be achieved well, but I don't know how to do it.
The fx core, I saw it but for now I am not interested in getting involved with it, I do not see it so simple.
thanks for your answers ...

[MetalGuy]

Never noticed FV-1 was noisy. Ears are still good though.

[AndyF]

Never noticed FV-1 was noisy. Ears are still good though.

with respect friend, you should try to listen better, in all the circuits that contain the fv spin there is noise, unfortunately.
because it is very good to do effects.

But hey maybe you have a schematic designed that overrides this, and it would be great if you help me.

[ElectricDruid]

I think there might be a degree of "expectation management" required here.

If you buy a simple chip for not a lot of money, it is highly unlikely to do the same job as a more sophisticated and expensive group of chips that (on paper) do a similar thing. You can do a basic delay using an FV-1. You can do it using a STM32-F4 with bags of RAM and a serious codec attached. One gives a better technical result than the other. Examples of 570/571 companders are all over the internet (

Which do you want to do? How much money and development time are you willing to spend?

Those questions affect the quality you can expect to get.

Sure, you can do stuff to improve the FV-1, but if it's not in the ballpark that you want to play in, you need to adjust you ideas (and prices!) accordingly.

[potul]

One other trick I've seen suggested is to feed a differential input signal to the FV-1's inputs, and then combine the two outputs at the other end. That should cancel any common mode noise, at least. How much improvement that makes, I can't say - I haven't tried it myself.

Interesting one. Will this require to code the program in stereo, so that both signals are processed equally before merging them?

[free electron]

I think you can just read the adcr adcl and use the difference as the input signal, processed further as mono.
At the end, do the opposite: write sig_out to dacl, sig_out*-1.0 to dacr and use differential amp + some lowpass filtering.
The key would be to cancel out the internal ADC+DAC noise.
The question is, tough: does the ADC sample the L and R signal exactly at the same time?
Also nothing forbids using separate 3.3V regulators for the analog and digital part.

[ElectricDruid]

One other trick I've seen suggested is to feed a differential input signal to the FV-1's inputs, and then combine the two outputs at the other end. That should cancel any common mode noise, at least. How much improvement that makes, I can't say - I haven't tried it myself.

Interesting one. Will this require to code the program in stereo, so that both signals are processed equally before merging them?

As free electron says, you could invert one of the inputs and merge internally. That should cancel common mode noise in the ADC. Then you could process in mono, and then flip one channel again before sending to the DAC. Then the analog side uses a differential amp to regain the signal and cancel any DAC noise.

Or as you say, you could process the two entire chains identically. That might even start to cancel stuff like rounding errors in the algorithm, if they're consistent. Dunno.


Has anyone actually done a proper test on the chip? If you feed a string of zero values to the DAC, how much noise does it produce?
If you ground the inputs and read the ADC value, how much noise is there on the readings? I've never tried, but it never seemed noisy to me.

[octfrank]

If you ground the inputs and read the ADC value, how much noise is there on the readings? I've never tried, but it never seemed noisy to me.

Actually you want to run a Vcc/2 signal in as the converters are signed so ground is actually -1.0.

Also keep in mind these are delta-sigma which means they do not really like DC so best to use something like a 1KHz signal at -6db.

Most noise issues are related to PCB layout (poor ground plane, bad routing, etc.), power supply noise (if it goes away on battery then it is the power supply) or a low input signal so they are adding lots of gain then complaining that there is noise (run a hotter signal in).

[ElectricDruid]

Actually you want to run a Vcc/2 signal in as the converters are signed so ground is actually -1.0.

Doh, yeah, 'course. 3V3 single supply. Silly me.

Also keep in mind these are delta-sigma which means they do not really like DC so best to use something like a 1KHz signal at -6db.

Ah, ok. So time for a sine wave test signal then!

Most noise issues are related to PCB layout (poor ground plane, bad routing, etc.), power supply noise (if it goes away on battery then it is the power supply) or a low input signal so they are adding lots of gain then complaining that there is noise (run a hotter signal in).

Clearly for best quality, you need a decent level going in. 3.3V isn't a lot of headroom, so it all counts. That was why I suggested a compander - and for the retro-cool factor too, of course Some regulators are a lot quieter than others, so I imagine that makes a difference too. Do you know how important it is to use separate analog and digital supplies, Frank? A lot of circuits I've seen don't, but I'm guessing it makes a noticeable difference or you wouldn't have bothered including the option.

[octfrank]

Clearly for best quality, you need a decent level going in. 3.3V isn't a lot of headroom, so it all counts. That was why I suggested a compander - and for the retro-cool factor too, of course Some regulators are a lot quieter than others, so I imagine that makes a difference too. Do you know how important it is to use separate analog and digital supplies, Frank? A lot of circuits I've seen don't, but I'm guessing it makes a noticeable difference or you wouldn't have bothered including the option.

All your comments above have been great, especially the one about "expectation management", every design from chips to pedals is all about that.

As to split supplies, it really depends on the product. For most pedals there will be little to no advantage, most pedals are simply designed to have just the FV-1 as the only digital component (oh, and the 24LC32A) so good bypassing right at the pins is all that is needed. But if it is a more complex device with a micro doing USB and MIDI, a display, etc. then yes, get all that digital noise off the analog supply. In this case you may choose to run the FV-1 as split supplies or completely off the analog supply. In fact the only time we run split supplies on the chip is in test as there are a ton of digital components on the tester so having an isolated analog supply is necessary to make clean measurements of the converters.

[Sweetalk]

Welcome to the forum

What have you tried? It's hard to know what to tell you without knowing where you're starting from. The FV-1 isn't a hi-fi chip so you can't expect noiseless, but plenty of people have gotten great results with it. At the end of the day you have to know how much noise you're willing to accept though, hopefully you aren't expecting crystal clear hi-fi sound.

Exactly, the noise is always there, I can't explain why the manufacturer doesn't solve it. This circuit is the one that I like the most but the one that gets a lot of noise and I can't find how to modify it.

One of the issues with this circuit is that the input signal is low in amplitude. For better signal to noise ratio you want the input to be as hot as possible without clipping, so you need another to amplify the signal before hitting the FV1. You want to have it as close as possible to 3vpp.

Another thing is that diodes clampling the ADC to 3v3/2, is not necessary because the ADCs are internally biased to Vcc/2. Don't know if this adds noise, but never used in my designs. At least you gain in lower the parts count.

Layout REALLY matters, I have some boards with the FV1 that has poor layout and compared with others that haves a proper layout the difference is HUGE. Ground every pin with it's own via to a ground plane and add decoupling caps to every voltage pin, 0,1uF will do the trick.

Also, in that circuit you don't have any output HPF-LPF to get rid of the out of band noise. That is a must!. I always use a 2nd order LPF on the output.

[MetalGuy]

There's an FV-1 forum project thread which is a good starting point for an FV-1 project (pedal).

[pruttelherrie]

Excuse my ignorance, please educate me if I'm wrong:

One of the issues with this circuit is that the input signal is low in amplitude. For better signal to noise ratio you want the input to be as hot as possible without clipping, so you need another to amplify the signal before hitting the FV1. You want to have it as close as possible to 3vpp.

Fair enough, the input buffer could/should be made 'boostable'.

Another thing is that diodes clampling the ADC to 3v3/2, is not necessary because the ADCs are internally biased to Vcc/2. Don't know if this adds noise, but never used in my designs. At least you gain in lower the parts count.

As I understand it, these diodes are there for protection? What's the difference between protecting and clamping? (Note that they are Schottky's, not zeners. I don't agree on the symbol used though.)

Layout REALLY matters, I have some boards with the FV1 that has poor layout and compared with others that haves a proper layout the difference is HUGE. Ground every pin with it's own via to a ground plane and add decoupling caps to every voltage pin, 0,1uF will do the trick.

Agree 100%!

Also, in that circuit you don't have any output HPF-LPF to get rid of the out of band noise. That is a must!. I always use a 2nd order LPF on the output.

Strictly speaking, the cap on the output buffer is a 1st order LPF? Corner frequency of 1 / (2 * pi * 0.00000068 * 15000) = 15 Hz?!?!? Huh?

[Sweetalk]

As I understand it, these diodes are there for protection? What's the difference between protecting and clamping? (Note that they are Schottky's, not zeners. I don't agree on the symbol used though.)

Yeah, Schottky's!, my bad

Also, in that circuit you don't have any output HPF-LPF to get rid of the out of band noise. That is a must!. I always use a 2nd order LPF on the output.

Strictly speaking, the cap on the output buffer is a 1st order LPF? Corner frequency of 1 / (2 * pi * 0.00000068 * 15000) = 15 Hz?!?!? Huh?
[/quote]

Yes, that cap acts like a LPF but dependant of the gain of that stage. Also, when your signal reaches that stage you already mixed the "noisy" signal from the FV1 with the clean in the mix control. You want to do all the reconstruction filtering right after the DAC and then keep processing the signal, mixing, etc.