£3,648 for a volume knob (i.e. why not DAC into Power Amp)?

perizoqui posted:

Richard Dane posted:
Oh, please don't.  The kit was not designed to be used this way.  As has been explained the NDX2 is a source only and should not be used directly into a power amp.

Well, I may not for long. I'll try it, and compare it to a Classé preamp in the middle, and compare it to a Naim preamp in the middle if I can get a loaner. Of course, I'm hearing that a 252 is the requirement to match the NDX 2 and 300, so it'd be a $19,000 add-on if I were to buy new. More than I paid for NDX and 300 put together!

I once ran a Yamaha CD player with volume control directly into a power amp.  It didn’t sound good, but I was also losing bits, which I didn’t realize at the time. 

I also used a PS Audio preamp at one time which could be switched between passive and active modes.  Active sounded much better.  Just my experience, YMMV, etc. 

perizoqui posted:

Richard Dane posted:
Oh, please don't.  The kit was not designed to be used this way.  As has been explained the NDX2 is a source only and should not be used directly into a power amp.

Well, I may not for long. I'll try it, and compare it to a Classé preamp in the middle, and compare it to a Naim preamp in the middle if I can get a loaner. Of course, I'm hearing that a 252 is the requirement to match the NDX 2 and 300, so it'd be a $19,000 add-on if I were to buy new. More than I paid for NDX and 300 put together!

Well, you could trade the 300 in for the Supernait 2 integrated amp, and that would achieve exactly what you want.

perizoqui posted:

Richard Dane posted:
Oh, please don't.  The kit was not designed to be used this way.  As has been explained the NDX2 is a source only and should not be used directly into a power amp.

Well, I may not for long. I'll try it, and compare it to a Classé preamp in the middle, and compare it to a Naim preamp in the middle if I can get a loaner. Of course, I'm hearing that a 252 is the requirement to match the NDX 2 and 300, so it'd be a $19,000 add-on if I were to buy new. More than I paid for NDX and 300 put together!

Assumimg the NDX2 filters out any spurous ultrasonics, and provided that the volume control Retains the last value on going in and out of standby (or otherwise drops to zero but I that seems unlikely) - the first could be asked of Naim support, and the second of anyone with one - even the dealer - I can’t see how the combination would cause harm.  Personally I would have no hesitation trying myself, without time limit - however I cannot advise that you do it, Why not ask Naim support if there is any risk of harm (if it was me I’d say simply I dont have pream at the moment, any harm running direct even if not optimal?

I have used a chord TT straight into a 250DR, and get result was very good. 

I tried my Dave into my nap 250dr but didn’t like it that much. It sounds better with the 282 in between 

perizoqui posted:

Simon-in-Suffolk posted:

Although Roon allows for gain management and volume control if you enable those functions, as it support AirPlay after all, unless I missed it I saw no area where you could increase the sample word size... or add dither so as to reduce digital artefacts in using digital volume control.

I believe Roon does both these things. It first converts 16 (or higher) bit words to 64 bit floating point, then applies volume control, before converting back to the native resolution of the DAC (32 bits for the NDX). So for data streaming between 16 and 24 bit resolutions (everything I own), it's demonstrably lossless.

[@mention:70131329546614416], unfortunately, due to the maths of signal processing, that isn't actually lossless and will also add artefacts (but whether or not you can hear them is another question, which depends on the combination of equipment and the particular listener involved).

Huge, you understand amp design and I don't. What does time alignment bring to the party when a proper pre-amp is used? Based on reading that is one of the biggest differences next to filtering I believe.

Huge posted:

[@mention:70131329546614416], unfortunately, due to the maths of signal processing, that isn't actually lossless and will also add artefacts (but whether or not you can hear them is another question, which depends on the combination of equipment and the particular listener involved).

My understanding is that the null test (subtracting the digital word after conversion and back from the original) determines whether it's lossless (by yielding a result of zero). Roon is demonstrably so (at least the folks over there say they've shown it to be. So while I agree that the math will show that it could be lossy, the data show that in Roon's DSP volume it isn't lossy, for reasonable levels of attenuation.

perizoqui posted:

Huge posted:

[@mention:70131329546614416], unfortunately, due to the maths of signal processing, that isn't actually lossless and will also add artefacts (but whether or not you can hear them is another question, which depends on the combination of equipment and the particular listener involved).

My understanding is that the null test (subtracting the digital word after conversion and back from the original) determines whether it's lossless (by yielding a result of zero). Roon is demonstrably so (at least the folks over there say they've shown it to be. So while I agree that the math will show that it could be lossy, the data show that in Roon's DSP volume it isn't lossy, for reasonable levels of attenuation.

To implement a volume control in the digital domain, and output in an integer format must be lossy unless the conversion is always done in 6dB steps, and the attenuation is less than the increase in resolution from the input format to the output format - this lossy nature of the processing is due to round-off errors.  This applies even if the conversion is done in FP (which introduces it's own approximations); furthermore these round-off errors are correlated to the signal, and therefore dither should be added after the processing, so taking it even further from a lossless process.

Ardbeg10y posted:

Huge, you understand amp design and I don't. What does time alignment bring to the party when a proper pre-amp is used? Based on reading that is one of the biggest differences next to filtering I believe.

I'm not sure offhand, I'd have to look at the exact circuit, including all the other circuitry in the preamp (and analyse it all) to work out what it's aligning and why.

In musical signal terms, time alignment is important as the human ear-brain combination is very sensitive to time information - so much so that we can break the Fourier limit in our time discrimination.

Huge posted:

To implement a volume control in the digital domain, and output in an integer format must be lossy unless the conversion is always done in 6dB steps, and the attenuation is less than the increase in resolution from the input format to the output format - this lossy nature of the processing is due to round-off errors.  This applies even if the conversion is done in FP (which introduces it's own approximations); furthermore these round-off errors are correlated to the signal, and therefore dither should be added after the processing, so taking it even further from a lossless process.

In theory yes. In practice the degree of loss depends on the resolution involved 64 bit floating point has errors so small that they round off to the correct bit every time for reasonable degrees of attenuation. So even though the floating point operations have the potential to introduce error (as you describe), in practice they don't (as they've shown). They've got the data over on the Roon forums I believe, to show that this is true. 

perizoqui posted:

Huge posted:

To implement a volume control in the digital domain, and output in an integer format must be lossy unless the conversion is always done in 6dB steps, and the attenuation is less than the increase in resolution from the input format to the output format - this lossy nature of the processing is due to round-off errors.  This applies even if the conversion is done in FP (which introduces it's own approximations); furthermore these round-off errors are correlated to the signal, and therefore dither should be added after the processing, so taking it even further from a lossless process.

In theory yes. In practice the degree of loss depends on the resolution involved 64 bit floating point has errors so small that they round off to the correct bit every time for reasonable degrees of attenuation. So even though the floating point operations have the potential to introduce error (as you describe), in practice they don't (as they've shown). They've got the data over on the Roon forums I believe, to show that this is true. 

You still need to look at the round trip, rather than just each step in isolation.

perizoqui posted:

Mike-B posted:

NDX DIN & RCA outputs are 30? maximum & the required minimum load impedance is 10k?

Thank you sir! I couldn't find that anywhere. 

.......   In the Manual (pdf)