PCM Signal From Roon

Roon offers users the ability to limit PCM digital bit rate output to my DSJ. After upsampling (to facilitate my use of Roon’s EQ), Roon downsamples bit rate before sending the signal to my DSJ. In the absence of my affirmatively limiting bit rate output to 24 bits, the signal sent to my DSJ is 32 bits. In that case, my DSJ recognizes the digital input as 24 bits, nevertheless…

Alternatively, I can limit Roon’s digital output to 24 bits, in which case Roon itself will downsample a 64 bit file to 24 bits, rather than a default of 32 bit. In that case, my DSJ again recognizes the incoming signal as 24 bits.

So…should I require Roon to downsample a PCM signal that has gone through its DSP to 24 bits? Or should I just let it downsample to the default 32 bits and let that signal go into the DSJ?

The DS and DS Jr don’t care about the sample width being 32 or 24 - they act identically in the actual FPGA signal processing (hence the report of 24 bits on the DS screen.) There are some differences per input about whether 32 bits is accepted, but if it is it’s fine.

Using the words downsample, upsample or talking about bit rates when referring to narrowing the number of bits per sample is confusing. Normally downsampling, psampling and rates apply to the sample rate. And bits per sample, sample width or bit width is used to talk about the number of bits per sample. (Some programs report the total bit rate which is the sample rate * the bits per sample, but that’s not relevant here.)

Assuming there’s no difference in the sample rate and actual downsampling isn’t going on, either 32 bits per sample or 24 bits per sample are fine if the DS or the DS Jr accepts them.


I appreciate the guidance and apologize for the mix-up in my terminology. Chalk it up to my technical naïveté. Henceforth:

Bit Rates can be “converted”


Sampling Rates can be “upsampled” or “downsampled”


I didn’t intend to chastise, I simply wanted to make sure I was understanding what you really wanted to ask. I tend to err on the side of too much info :slight_smile:

Ted Smith said I tend to err on the side of too much info :)
A wonderful thing. Many of us like to learn and you are of great help.

I’m using the Parametric EQ in Roon to compensate my room acoustics. The frequency spectrum is compensated by 12 band’s. Arround 50 Hz I had a gap of 9db which is the biggest problem. With the compensation of this 9db gap most music files will clip when there’s no headroom adjustment for at least 6db. I can hear a decrease of SQ by this headroom adjustment.

My questions are is this clipping dangerous for my system (speakers)?

Can I create headroom by converting 16bit files to 24bit? Or maybe to DSD? So I can stream and play the new 24bit files without a headroom adjustment.


I don’t know the Roon interface, but no going to 24 bits won’t give you headroom, but if you are currently sending 16 bits to the DAC you may notice an increase in sound quality going to 24 bits because of the increase in resolution.

Almost always louder sounds better - when you give a headroom adjustment you will have to turn up your preamp to make up for it since your digital signal is now quieter. A digital headroom adjustment shouldn’t affect sound quality if you are using 24 bits to your DAC and you turn up your analog output to make up for it.

In general clipping is bad, it generates a lot more power at the clipping points which may indeed stress your equipment more, especially if you are anywhere near the maximum levels for your preamp, amp or speakers.

Usually with room correction you don’t want to try to compensate for nulls in your room. Low frequency nulls are caused by standing wave cancelations and no amount of boosting will fix them. And, as you have noticed, the other downsides of trying to fill those nulls can cause other worse problems. You are much better off adjusting the seating position or the speaker position, etc. to find the flattest frequency response, etc.

Hi Ted,

Thank you for the answers and explanations on my questions. I will measure my room again and correct only the those frequencies which sound to loud. Maybe the frequencies that are suppressed by other freqencies at this moment will find their way back.

Unfortunately I don’t have many other posible position left for my speakers. They are already on good positions, but yes that would be best!

Do you mean sending the native 24bit files to the dac will have more resolution or with the 16 bit as well?

If I’m right when I create headroom in music files then a few bits will be shifted out at the least significant side of the word. So I thought that the louder a music file sounds (less headroom) the more information is still in the file?


In digital audio everything is relative to the full scale signal so we tend to think of samples representing a fraction of the maximum or full scale values. So samples always represent a value between -1 and +1. The number of bits in each sample tell you how accurately you can represent that value. If you only have one bit you can only say +1 or -1, if you have 8 bits you have 256 different levels between -1 and +1, if you have 16 bits you have 65536 levels between -1 and +1. More bits don’t get you more loudness or more headroom it gets you more resolution.

Putting it another way, you don’t get to choose whether more bits show up on the top or the bottom of samples, they always show up on the bottom.

To get headroom you are making everything quieter so that you have some place to go when it gets loud. In your original question you wanted to raise the level of signals with frequencies near 50Hz. If you simply raised their levels you’d reach the loudest possible signal quickly and cause clipping. In the digital part of your system to make things around 50Hz louder you need to instead make everything else softer. Then you need to make up for that in analog somewhere, probably in your preamp by turning the analog volume up by the same amount you lowered the digital.

If you don’t try to make any particular range of frequencies louder with an EQ you don’t need to mess with all of this…

Thank you Ted,

I now completely understand what you mean.