... Silent here ...
Well, let me put something forward. I have just listened to the latest version of my dac, which uses John's latest DEM circuit (one reclocker line, and a 33pf cap to the other pin). The dac is driven by a squeezebox via i2s.
I just noticed that when I reduce the volume on the squeezebox (digitally), the noise levels increase. I have not yet tested with a test tone, so don't have any measured noise figures yet, but the noise becomes noticable with headphones when I do this. At full output, everything sounds great...
Two questions:
- How much noise can one reasonably expect from the TDA? Should it ever be noticable with -xx db levels? My understanding of DEM reclocking is that optimum performance is reached when: there is a lock with the internal oscillator AND dem reclocking is in synch with BCK (due to transients impact)
- Any objections on the 33pf? I would have thought it's either a lock or no lock, and I thought I would hear a no lock more clear than I do now.... so is there a lock or no lock??
Well, let me put something forward. I have just listened to the latest version of my dac, which uses John's latest DEM circuit (one reclocker line, and a 33pf cap to the other pin). The dac is driven by a squeezebox via i2s.
I just noticed that when I reduce the volume on the squeezebox (digitally), the noise levels increase. I have not yet tested with a test tone, so don't have any measured noise figures yet, but the noise becomes noticable with headphones when I do this. At full output, everything sounds great...
Two questions:
- How much noise can one reasonably expect from the TDA? Should it ever be noticable with -xx db levels? My understanding of DEM reclocking is that optimum performance is reached when: there is a lock with the internal oscillator AND dem reclocking is in synch with BCK (due to transients impact)
- Any objections on the 33pf? I would have thought it's either a lock or no lock, and I thought I would hear a no lock more clear than I do now.... so is there a lock or no lock??
digital volume has these days, surpassed any other type of volume control, some sub-par software implementations are still a bit dodgy, but most are now at the point that if you use anything else with a digital source, you are adding noise, not the other way around. try building an analogue control with 40-64bits noise floor and perfect channel matching. where are you controlling the digital volume studiostevus?
shouldnt be that then, it would be at least 32bit and most likely higher. excuse my ignorance here as i'm not really familiar with the concept. but DEM reclocking, how and where is it derived? could it be that when you shift bits in the i2s stream for volume, that you are causing some kind of interlacing/moire type overlayed noise in the bitstream?
Last edited:
Building a 40bit digital volume control would be fairly pointless as nobody's got anything near a 40bit DAC, and the DAC is ultimately the noise limit for such.
I agree for DACs with >120dB DNR an analog volume control is probably not required, but with a 16bit DAC it most certainly would be. I doubt anyone really wants to be only listening to the bottom 8 bits of their DAC when the volume's set to -48dB.
Last edited:
ermm you do realize thats the point yeah?
using a 40+bit vol control doesnt mean it tries to output 40+bits, it means the volume control is calculated at 40bits floating point. puremusic for example calculates the volume and filters at 64bit floating point, the dac then uses what it is sent.
Last edited:
it does, the volume control DOES have 40-64bit noise floor, exactly what I said.. the data is processed at that width and then outputs whatever you want up to the limit of your dac.
this is exactly the same as the analogue control, the analogue control is fed the analogue signal and adds a certain amount of noise or error, we are comparing the volume controls here in my statement are we not? we are not comparing the data, of course it cant make up for lack of data, but how does that have any baring at all in the comparison of a volume control?
but hey I suggest you argue with someone else... maybe they'll be more fun, more fun than a ghost anyway
Last edited:
Then you've misunderstood. If you process at 40bits then send to a real world DAC (less than 40bits) you get quantization distortion. Show me an analog volume control which generates quantization distortion at the LSB level of the DAC that's being used?
Are you or are you not saying that digital volume control beats analog in practice? If you are then you're mistaken in the context of this thread which is about a 16bit DAC.
If though you are saying that software multiplication can be done to greater precision than analog hardware, I would agree. But that's not the context here which is about real noise coming from a real DAC.
No this is good enough fun
Yay, connect your dac to your soundcard, run -40db FS sine and check the FFT.
Take 0dB FS and run it again.
Put some constant signal with offset (or slooowly varying sine at 0.1Hz and check it's output with oscilloscope. Maybe youv'e got some DEM cell damaged in the LSBs. Maybe the DEM capacitor broke (been there for MSBs, got crackling noises).
And do you know where do you take the DEM clk from, which rate are you working at?
I2S is usually at 64Fs bclk. You need to divide it down to 4Fs-8Fs for the DEM to work properly at 44.1kHz Fs.
Take out the DEM clock and throw a 470-680pF cap in place, do you still hear these noises?
Digital volume with 16bit DACs? That's a bad thing altogether. You'd want to use the volume pot in the amplifier chassis, and keep the DAC's output as high as possible to overcome the induced noise on the wiring.
Maybe your's amp oscillates at DAC's glitches - running NOS without proper filtering?
Take 0dB FS and run it again.
Put some constant signal with offset (or slooowly varying sine at 0.1Hz and check it's output with oscilloscope. Maybe youv'e got some DEM cell damaged in the LSBs. Maybe the DEM capacitor broke (been there for MSBs, got crackling noises).
And do you know where do you take the DEM clk from, which rate are you working at?
I2S is usually at 64Fs bclk. You need to divide it down to 4Fs-8Fs for the DEM to work properly at 44.1kHz Fs.
Take out the DEM clock and throw a 470-680pF cap in place, do you still hear these noises?
Digital volume with 16bit DACs? That's a bad thing altogether. You'd want to use the volume pot in the amplifier chassis, and keep the DAC's output as high as possible to overcome the induced noise on the wiring.
Maybe your's amp oscillates at DAC's glitches - running NOS without proper filtering?
Last edited:
Thanks for the suggestions!
I guess I would be looking at a comparison between the noise at 0db and -40db?
fDEM = fBCK = 2.8 Mhz
Will try this on the weekend. My output stage measures not very great though, so this will impact results (high THD), also using NOS (so more THD+N).
I guess I would be looking at a comparison between the noise at 0db and -40db?
Can you explain this one?
fDEM = fBCK = 2.8 Mhz
I am using NOS indeed. The headphone amp is an aikido... I don't expect this to be the issue, given the noise only is present at low levels.
This sounds to me like you are using the single Jfet buffer output stage? Using headphones I wouldn't be surprised if you are hearing distortion, you probably have over %5 THD including the tube aikido with lower level signals. That "single jfet buffer" I/V stage isn't ideal for headphones especially multiplied by the Aikidos distortion. It was obviously design specifically for the ecd ciclotron amp, don't even try to go without a coupling cap with other amplifers.
Also digital volume control is just about the antithesis of the NOS philosophy. You would need a modern DAC if you need digital volume control.
Last edited:
Thanks for your comments. Actually i am not using the jfet stage, but another discrete stage (pedja rogic diamond stage).
What strikes me is that nobody of you considered the dem circuit and my implementation of it (33pf @ 2.8mhz) as the possible source....
Reclocking affects jitter - doesn't it ? Jitter affects signals regardless of level, doesn't it ? The latter is because jitter is phase modulation - isn't it ?
nowhere,
what they mean by "DEM reclocking" is not the reclocking itself, but forcing TDA's "dynamic element matching" circuitry to run with external clock. So the DEM averaging will always contain at least 4 steps for each sample (herein it's rate should be 4Fs or greater, there are just 4 positions of current dividers).
The DEM isn't supposed to run this way, and it's native frequency is around 300-700kHz, self oscillating and running free.
What studiostevus did is connected the BCLK directly to DEM, and forced it to run on 2.8MHz which is a bit more than 300-700kHz, hence his strong suspectism.
what they mean by "DEM reclocking" is not the reclocking itself, but forcing TDA's "dynamic element matching" circuitry to run with external clock. So the DEM averaging will always contain at least 4 steps for each sample (herein it's rate should be 4Fs or greater, there are just 4 positions of current dividers).
The DEM isn't supposed to run this way, and it's native frequency is around 300-700kHz, self oscillating and running free.
What studiostevus did is connected the BCLK directly to DEM, and forced it to run on 2.8MHz which is a bit more than 300-700kHz, hence his strong suspectism.