over 0 dBFS is okay, MiniDSP says

[skyfly]

While a retailer suggested that I try to avoid over 0 dBFS (try below -1.5 dBFS) on any RMS meter on MiniDSP Device Console for MiniDSP Flex Digital (with Dirac Live), the manufacturer MiniDSP wrote to me that over 0 dBFS is fine.

(By the way, in sine sweep, the Dirac Live filter (PEQ and Crossover off) gives the largest output somewhere in 100 - 300 Hz. It seems that it is due to the compensation for dips in the measurement locations. "Wide Imaging" was chosen.)

Below is what MiniDSP wrote. Is it correct?

"For the DSP to "lower" the gain based on a 0dBFS input + Processing = clipping, it's not going to work. Why because we/the DSP can't know what signal you're going to feed. e.g. 99% of content is for example -10dBFS, i.e. you're going to have lots of headroom even with more positive EQ. You only CLIP "based" on your the strengh of the input signal. Looking at the RMS meter with a realistic signal (i..e NOT a test tone at 0dBFS = completely unrealistic) is the way to start."

I think a 0 dBFS sine test signal should never be above 0 dBFS (indeed, I try it below -3.5 dBFS for the bad case: 0 dBFS 200 Hz square wave test signal, around 600Hz (the third harmonic that reduced the peak in source signal) is sharply reduced at some processing stage, then the output peak is over 0 dBFS).

I asked MiniDSP the 'digital headroom' of each processing stage in MiniDSP MiniDSP Flex Digital (with Dirac Live). (I mentioned, for example, the processing stage, input level control, has a control range from -12 dB to +12dB. Hence, that processing stage seems to have 12dB or more digital headroom.) MiniDSP did not answer this question. This is strange because I saw Benchmark or RME talking about digital headroom in their products. I do not think the (rough or exact) amount of digital headroom is a trade secret. MiniDSP mentions "lots of headroom even with more positive EQ" most of the time in music material.

 

[skyfly]

DeepSeek response

Question: The manufacturer MiniDSP wrote that over 0 dBFS with near 0 dBFS sine test signal is okay. I think it is wrong. Could you check. This is the copy and paste of what MiniDSP Dev Team wrote to me: "For the DSP to "lower" the gain based on a 0dBFS input + Processing = clipping, it's not going to work. Why because . . . .

A part of the answer:

Conclusion: MiniDSP is Wrong​

• A properly designed DSP should handle a 0 dBFS test tone without clipping (either by internal headroom or gain adjustment).
• Real-world signals with peaks near 0 dBFS + processing (like EQ boosts) will clip if the system doesn't manage headroom properly.
• Their response suggests their DSP lacks proper headroom management, which is a limitation.

 

[Curvature]

There is a difference between peak and RMS levels, even for sine waves, called crest factor.

Your post is confusingly written. I would not try to overinterpret measurments without a solid grasp of the fundamentals. Based on your multiple other threads on this topic, I'll write this very simple answer, and I hope that will be enough. You really need to read a book on the basics of digital signals if you want to understand the details.

Intersample overs can be as high as 10dB in extreme cases. I think the majority fall within 3dB or so. These are a problem only when your DAC is outputting a full scale (0dB) signal because the recorded signal is clipped and to reconstruct the waveform a level exceeding full scale is required. Be clear that we are talking about the digital domain here. It's highly unlikely you will set your DAC at that level. Everyday use of my RME DAC has me setting levels at -30dB or so, maxing out at around -10dB occasionally for quiet content. That attenuation is more than good enough to address intersample overs.

When using a MiniDSP or any other digital pass-through that allows control of levels, you have to be careful not to clip the signal. However, even if you do clip it, it's possible that the distortion will be low level and inaudible. That's suboptimal, but not a real issue. If it's an issue, you best believe you will hear it. Bad clipping is obvious.

The way to use MiniDSP, or any device allowing EQ, is to use a combination of digital gain adjustment and EQ. Generally, the advice is to only pull down peaks and leave dips alone to ensure effectiveness of the filters and prevent clipping. However it's possible that some dips will respond to EQ, or you may want to boost certain frequency regions for taste, or apply a positive shelf filter. In this case, to prevent clipping, you set the gain to inversely match or slightly exceed the highest value of your positive filter.

That's all you really need to know.

As a reminder, AI is a probability machine that, in real terms, is a mix between a search engine and a random string generator. It's helpful but don't rely on it to do your thinking for you. In all cases always read the source material for the actual answer.

 

[skyfly]

There is a difference between peak and RMS levels, even for sine waves, called crest factor.

Your post is confusingly written. I would not try to overinterpret measurments without a solid grasp of the fundamentals. Based on your multiple other threads on this topic, I'll write this very simple answer, and I hope that will be enough. You really need to read a book on the basics of digital signals if you want to understand the details.

Intersample overs can be as high as 10dB in extreme cases. I think the majority fall within 3dB or so. These are a problem only when your DAC is outputting a full scale (0dB) signal because the recorded signal is clipped and to reconstruct the waveform a level exceeding full scale is required. Be clear that we are talking about the digital domain here. It's highly unlikely you will set your DAC at that level. Everyday use of my RME DAC has me setting levels at -30dB or so, maxing out at around -10dB occasionally for quiet content. That attenuation is more than good enough to address intersample overs.

When using a MiniDSP or any other digital pass-through that allows control of levels, you have to be careful not to clip the signal. However, even if you do clip it, it's possible that the result will be low level and inaudible. That's suboptimal, but not a real issue. If it's an issue, you best believe you will hear it. Bad clipping is obvious.

The way to use MiniDSP, or any device allowing EQ, is to use a combination of digital gain adjustment and EQ. Generally, the advice is to only pull down peaks and leave dips alone to ensure effectiveness of the filters and prevent clipping. However it's possible that some dips will respond to EQ, or you may want to boost certain frequency regions for taste, or apply a positive shelf filter. In this case, to prevent clipping, you set the gain to match or slightly exceed the highest value of your positive filter.

That's all you really need to know.

As a reminder, AI is a probability machine that, in real terms, is a mix between a search engine and a random string generator. It's helpful but don't rely on it to do your thinking for you. In all cases always read the source material for the actual answer.

Is it possible that we see over 0 dBFS on RMS meter, but the peak value is not over 0 dBFS?

 

[Curvature]

Is it possible that we see over 0 dBFS on RMS meter, but the peak value is not over 0 dBFS?

What is an RMS meter?

 

[skyfly]

There is a difference between peak and RMS levels, even for sine waves, called crest factor.

You really need to read a book on the basics of digital signals if you want to understand the details.

Is it possible that we see over 0 dBFS on RMS meter, but the peak value is not over 0 dBFS?

You don't need to read a book. Just tell us all the book you already read long time ago. Which book says that we see over 0 dBFS on RMS meter, but we can safely assume that the peak value is below 0 dBFS? What is the books' suggestion on avoiding 0 dBFS when only RMS meter is available?

Tell us the books and chapter titles, section titles, right a way. Those foundation should be already in your mind, when you wrote the post.

 

[skyfly]

What is an RMS meter?

You do not need to fool people this way. Your post already proved your lack of competence in audio.

 

[Curvature]

You do not need to fool people this way. Your post already proved your lack of competence in audio.

Is an RMS meter:

The digital meter in your media player?
An RTA on your DAC?
An analog SPL meter?

What kind of device or software do you mean and what kind of signal is it measuring?

 

[skyfly]

There is a difference between peak and RMS levels, even for sine waves, called crest factor.

Your post is confusingly written. I would not try to overinterpret measurments without a solid grasp of the fundamentals. Based on your multiple other threads on this topic, I'll write this very simple answer, and I hope that will be enough. You really need to read a book on the basics of digital signals if you want to understand the details.

Intersample overs can be as high as 10dB in extreme cases. I think the majority fall within 3dB or so. These are a problem only when your DAC is outputting a full scale (0dB) signal because the recorded signal is clipped and to reconstruct the waveform a level exceeding full scale is required. Be clear that we are talking about the digital domain here. It's highly unlikely you will set your DAC at that level. Everyday use of my RME DAC has me setting levels at -30dB or so, maxing out at around -10dB occasionally for quiet content. That attenuation is more than good enough to address intersample overs.

When using a MiniDSP or any other digital pass-through that allows control of levels, you have to be careful not to clip the signal. However, even if you do clip it, it's possible that the distortion will be low level and inaudible. That's suboptimal, but not a real issue. If it's an issue, you best believe you will hear it. Bad clipping is obvious.

The way to use MiniDSP, or any device allowing EQ, is to use a combination of digital gain adjustment and EQ. Generally, the advice is to only pull down peaks and leave dips alone to ensure effectiveness of the filters and prevent clipping. However it's possible that some dips will respond to EQ, or you may want to boost certain frequency regions for taste, or apply a positive shelf filter. In this case, to prevent clipping, you set the gain to inversely match or slightly exceed the highest value of your positive filter.

That's all you really need to know.

As a reminder, AI is a probability machine that, in real terms, is a mix between a search engine and a random string generator. It's helpful but don't rely on it to do your thinking for you. In all cases always read the source material for the actual answer.

If you read my other post, you knew the context or situations. I wanted to avoid 0 dBFS. However, MiniDSP Device Console does not have a peak meter. It has only RMS meters. So I did sign sweep near 0 dBFS as the slowest speed possible on REW signal generator. 20 - 20000 Hz, 60 seconds. Then 100 - 350 Hz, 60 seconds, etc. That was how I could say 0 dBFS occurred.

Which book says this is a wrong approach in detecting over 0dBFS? Please tell us right way.

You do not need to pollute this thread with the definitions of peak meter and RMS meter with the assumption that only you know the definitions and others do not know.

 

[Curvature]

Is it possible that we see over 0 dBFS on RMS meter, but the peak value is not over 0 dBFS?

You don't need to read a book. Just tell us all the book you already read long time ago. Which book says that we see over 0 dBFS on RMS meter, but we can safely assume that the peak value is below 0 dBFS? What is the books' suggestion on avoiding 0 dBFS when only RMS meter is available?

Tell us the books and chapter titles, section titles, right a way. Those foundation should be already in your mind, when you wrote the post.

Principles of Digital Audio by Ken Pohlmann. The book will explain the fundamentals. It will be up to you to apply these to specific situations.

 

[skyfly]

What kind of device or software do you mean and what kind of signal is it measuring?

This is a contradiction. You wrote that you read my other threads. You lie again?

 

[skyfly]

Principles of Digital Audio by Ken Pohlmann. The book will explain the fundamentals. It will be up to you to apply these to specific situations.

You cannot say that the book says my method of avoiding 0 dBFS in my situation is wrong. What I did is probably consistent with the fundamentals in Principles of Digital Audio by Ken Pohlmann.


" I'll write this very simple answer, and I hope that will be enough." Wow. You boldly believed that you are correct and others are wrong. Amazing attitude!

 

[amirm]

Below is what MiniDSP wrote. Is it correct?

"For the DSP to "lower" the gain based on a 0dBFS input + Processing = clipping, it's not going to work. Why because we/the DSP can't know what signal you're going to feed. e.g. 99% of content is for example -10dBFS, i.e. you're going to have lots of headroom even with more positive EQ. You only CLIP "based" on your the strengh of the input signal. Looking at the RMS meter with a realistic signal (i..e NOT a test tone at 0dBFS = completely unrealistic) is the way to start."

This answer has little to do with what you are asking. They are saying that if you are applying EQ, it is not necessary to create headroom for pathological cases. That you should play your music and see if clipping really occurs. I do this all the time in my EQ development as often filters are above bass where input amplitude is quite a bit lower than max level.

 

[amirm]

This is a contradiction. You wrote that you read my other threads. You lie again?

Watch your language. We don't get personal in this forum.

 

[Curvature]

MiniDSP Device Console

miniDSP Device Console - miniDSP docs

docs.minidsp.com

So this is your screen:

Understanding exsctly what that meter represents will depend on your signal chain, if digital or analog, but as Amir wrote with normal content you will likely not experience issues.

If you really want to check, you can create a loopback into REW and use the RTA to look for clipping. Make sure you aren't outputting any actual audio to your speakers or headphones while doing this since you could damage them and your hearing with full scale signals.

 

[skyfly]

There is a difference between peak and RMS levels, even for sine waves, called crest factor.

Your post is confusingly written. I would not try to overinterpret measurments without a solid grasp of the fundamentals. Based on your multiple other threads on this topic, I'll write this very simple answer, and I hope that will be enough. You really need to read a book on the basics of digital signals if you want to understand the details.

Intersample overs can be as high as 10dB in extreme cases. I think the majority fall within 3dB or so. These are a problem only when your DAC is outputting a full scale (0dB) signal because the recorded signal is clipped and to reconstruct the waveform a level exceeding full scale is required. Be clear that we are talking about the digital domain here. It's highly unlikely you will set your DAC at that level. Everyday use of my RME DAC has me setting levels at -30dB or so, maxing out at around -10dB occasionally for quiet content. That attenuation is more than good enough to address intersample overs.

When using a MiniDSP or any other digital pass-through that allows control of levels, you have to be careful not to clip the signal. However, even if you do clip it, it's possible that the distortion will be low level and inaudible. That's suboptimal, but not a real issue. If it's an issue, you best believe you will hear it. Bad clipping is obvious.

The way to use MiniDSP, or any device allowing EQ, is to use a combination of digital gain adjustment and EQ. Generally, the advice is to only pull down peaks and leave dips alone to ensure effectiveness of the filters and prevent clipping. However it's possible that some dips will respond to EQ, or you may want to boost certain frequency regions for taste, or apply a positive shelf filter. In this case, to prevent clipping, you set the gain to inversely match or slightly exceed the highest value of your positive filter.

That's all you really need to know.

As a reminder, AI is a probability machine that, in real terms, is a mix between a search engine and a random string generator. It's helpful but don't rely on it to do your thinking for you. In all cases always read the source material for the actual answer.

I read this from beginning to the end. There is zero enlightenment. I usually do not speak out such a comment, but I decided to do so in this special case because you wrote this: "Based on your multiple other threads on this topic, I'll write this very simple answer, and I hope that will be enough. You really need to read a book on the basics of digital signals if you want to understand the details."

 

[amirm]

I read this from beginning to the end. There is zero enlightenment.

Definitely not in anything you are writing. Your OP is completely confusing. You didn't even bother to post what you asked Minidsp. Members are trying to help but you are getting personal. I warned you to not do that but you are continuing. I am closing your thread. Watch your manners in the future.