I've debugged the mute function tonight. I had forgotten to include an extra long delay for the filter behind the bandgap to settle, it is fixed in the attached schematic. I've also changed R120 to mounted and R121 to not mounted, so the output gets muted faster when the MUTE input goes high. The clock now has a coaxial connector. Based on what I've measured so far, giving the clock its own connector is a better idea than sharing a header with the data signals. There is no up-to-date PCB design yet, I made these changes on my PCB by three-dimensional soldering.
I've measured a couple of square wave responses to check feedback loop stability, found no issues except the usual problems with oscilloscope ground leads, and updated the PCB design. I've also put in U.FL connectors for the data, if you need them on a header, you can mount R1 and R2. See the attachments, which are meant for 360 um thick prepreg, a few traces have to be made narrower for thinner prepreg.
Mind you, due to the 13 mm height of C4, C6 and C9, normal headers are just too short to connect the filter PCB to the DAC PCB (with the filter above the DAC). It just makes contact when you mount them a bit higher than intended. I hope connectors such as the Samtec DW-06-11-F-S-630 and DW-11-11-G-S-665, both available at Mouser, work better.
The NST45011's need not be mounted, but I forgot to mark them as NM.
Mind you, due to the 13 mm height of C4, C6 and C9, normal headers are just too short to connect the filter PCB to the DAC PCB (with the filter above the DAC). It just makes contact when you mount them a bit higher than intended. I hope connectors such as the Samtec DW-06-11-F-S-630 and DW-11-11-G-S-665, both available at Mouser, work better.
The NST45011's need not be mounted, but I forgot to mark them as NM.
I think I fixed the bugs and I hope I haven't introduced any new ones, so I think you have a good chance of getting a usable DSD DAC when you build the latest version. By design, the part of the mute circuit that monitors the supply only monitors the +15 V, so the other supplies should not lag too far behind the +15 V at start up, and not disappear before the +15 V at power down.
Except for the noise floor and the noise of the voltage reference, nothing is known about the performance yet. I will do some more performance measurements on my PCBs to see how well they work. I can then also check the supply currents, I only have estimates now.
Except for the noise floor and the noise of the voltage reference, nothing is known about the performance yet. I will do some more performance measurements on my PCBs to see how well they work. I can then also check the supply currents, I only have estimates now.
Measured at 100 Hz, 1 kHz and 10 kHz and 0 dBFS, the distortion is mainly third harmonic (I don't see any of the others), and the third harmonic lies somewhere between -90 dB and -95 dB, probably closer to -95 dB than to -90 dB. So that's roughly 0.002 %.
I again used a 27 Mbit/s bit rate and my PWM8 algorithm. 0 dBFS means between 25 % and 75 % ones in the sigma-delta modulate, the same as full-scale DSD.
I guess so, but what kind of circuit do you have in mind and what is the idea behind it?
Hi Marcel ... Thanks for your feedback ... Just out of curiosity: Is this what you would expect - or would you have expected the results to be different?
Also, I notice that you write that there essentially are no other distortion products than the 3H (and here I assume that your measurement setup is capable of "much" less than the - 95 dB level) ... Do you think there is a way to amend this 3H level - or is it a "given" for some reason?
Cheers,
Jesper
Also, I notice that you write that there essentially are no other distortion products than the 3H (and here I assume that your measurement setup is capable of "much" less than the - 95 dB level) ... Do you think there is a way to amend this 3H level - or is it a "given" for some reason?
Cheers,
Jesper
I had no expectation. The performance of a single-bit sigma-delta DAC is limited by hard-to-predict second-order effects, so I don't know of any way to calculate it in advance. You could simulate it if you had accurate models of all components and the PCB, which I hadn't, but even that is not straightforward.
My measurement set-up at home is more primitive than you think. Anything below -105 dB would have gone unnoticed.
I have a 1950's distortion meter, but it would be unusable because it makes no distinction between distortion and out-of-band noise and besides, it can't measure below 0.01 %. What I used instead is an attenuator, a field memory recorder and an audio editing program that can calculate DFTs. I recorded at different levels while keeping the DAC output level constant, so I could distinguish between the distortion of the field memory recorder (increases with recording level) and the DAC (stays constant).
My measurement set-up at home is more primitive than you think. Anything below -105 dB would have gone unnoticed.
I have a 1950's distortion meter, but it would be unusable because it makes no distinction between distortion and out-of-band noise and besides, it can't measure below 0.01 %. What I used instead is an attenuator, a field memory recorder and an audio editing program that can calculate DFTs. I recorded at different levels while keeping the DAC output level constant, so I could distinguish between the distortion of the field memory recorder (increases with recording level) and the DAC (stays constant).
Frequency response, measured single-ended at the positive left output using a HAMEG HM-1505 oscilloscope and a 1:1 probe: within +/- 0.1 dB from the response of the digital part between 2.5 Hz and 20 kHz, at about -1 dB at 1 Hz. The cut-off frequency will shift up somewhat with a lower impedance load.
Out-of-band noise, estimated as one sixth of the quasi-peak-peak noise measured with the scope while playing silence: 8 mV RMS differential in PWM8 mode, 4 mV RMS differential in PWM4 mode, 1 mV RMS differential in chaotic mode. I needed to put the data cable through a ferrite ring to get sensible readings, otherwise the third harmonic of the clock was stronger at the scope than the out-of-band noise.
Out-of-band noise, estimated as one sixth of the quasi-peak-peak noise measured with the scope while playing silence: 8 mV RMS differential in PWM8 mode, 4 mV RMS differential in PWM4 mode, 1 mV RMS differential in chaotic mode. I needed to put the data cable through a ferrite ring to get sensible readings, otherwise the third harmonic of the clock was stronger at the scope than the out-of-band noise.