ChatGPT 4 is helpful as always
>"What can be the reason of elevated noise floor when PEQ filters are applied? It's especially high for low frequencies."
<"The elevated noise floor when PEQ (Parametric Equalizer) filters are applied, particularly at low frequencies, can be attributed to several factors:
(...)
Digital Processing: Digital processing can sometimes introduce quantization noise, which can raise the noise floor. This is more noticeable when dealing with lower frequencies because they have more energy and thus more potential for quantization errors."
>"The effect is in the digital domain so probably factor in point 3 is responsible for such behavior. Can you explain it further?"
<"Sure, I'd be happy to explain further.
In the digital domain, signals are represented as a series of discrete values, rather than a continuous waveform. This process of converting a continuous signal to a discrete one is known as quantization.
Quantization noise is a type of error that occurs during this process. It's the difference between the actual analog value and the quantized digital value. This noise is typically random and has a uniform distribution.
When you're dealing with low frequencies, these signals have more energy and are represented by larger values in the digital domain. Therefore, the potential for quantization errors is greater. This is because the difference between the actual analog value and the quantized digital value can be larger.
When you apply a PEQ filter, you're altering the digital representation of the signal. If the filter isn't designed or implemented correctly, it can introduce additional quantization errors, which can raise the noise floor.
Moreover, if the bit depth of the digital audio is not high enough, the quantization noise will be more noticeable. This is why professional audio processing is usually done with a bit depth of 24 bits or higher, as it allows for a much lower noise floor.
In conclusion, to minimize quantization noise, it's important to use high-quality filters and a sufficient bit depth when processing audio in the digital domain."
>"Can it be the effect of quantization errors/noise amplified by the recursive part of IIR filters? Due to insufficient precision or bit depth during the signal processing?"
<"Yes, absolutely. IIR (Infinite Impulse Response) filters are recursive, meaning they use previous output values in their calculations. If there are quantization errors or noise present in these previous output values, the IIR filter can indeed amplify these errors each time the filter is applied, leading to an elevated noise floor.
This is particularly problematic when dealing with low frequencies, as these signals have more energy and thus more potential for quantization errors, as mentioned earlier.
The precision or bit depth during signal processing plays a crucial role here. If the bit depth is too low, the digital representation of the audio signal won't be accurate enough, leading to quantization errors. These errors can then be amplified by the IIR filter.
In order to minimize this issue, it's important to use a sufficient bit depth when processing audio signals. This will ensure a more accurate digital representation of the audio signal, reducing the potential for quantization errors. Additionally, using high-quality IIR filters that are designed to minimize noise can also help reduce the noise floor."
