Progress
Tix88 boards arrived on Friday. Saturday evening was a good day to test the simple reflow station. Power is more than sufficient for this type of parts. Works like a charm. Like magic when all capacitors and resistors start floating and aligning themselves. So glad having this for soldering the FPGA connectors!
For those resistors directly on the clock and tube sections I used MELF and Susumu parts and went to hand soldering with traditional solder. With smd’s this size, no unsolvable problem, as long as you follow the tips on YouTube and use extra flux when needed... and have very thin tipped tweezers to handle the parts.
Tix88 boards arrived on Friday. Saturday evening was a good day to test the simple reflow station. Power is more than sufficient for this type of parts. Works like a charm. Like magic when all capacitors and resistors start floating and aligning themselves. So glad having this for soldering the FPGA connectors!
For those resistors directly on the clock and tube sections I used MELF and Susumu parts and went to hand soldering with traditional solder. With smd’s this size, no unsolvable problem, as long as you follow the tips on YouTube and use extra flux when needed... and have very thin tipped tweezers to handle the parts.
Attachments
The other side...
Is not finished yet.... but Sunday evening was productive. With parts shortage you have to take what is available. Decided mostly for dale RN, where available 65 and E, a really tight fit! Where not available 60s and 55s D or C. Also managed to fit 1 AB for gridstoppers and avoided the string of 2 and 5 by using parts with higher rating and sufficient voltage drop capability, like mills MRA 12 with 9k1 instead of 5x1k82. Polishing all leads with fine grade steel wool not only removes some patina, makes them shiny as new and soldering very easy (highly recommended !) but also made the leads of the mills fit, tightly - but without forcing into the pad hole. Before it simply did not enter!
Mouser also delivered the LX75 as indicated in September. I have not yet checked if one can program this alone without computer guru knowledge.
Hope to make good progress next weekend solder the remaining parts and wind some coils. Will keep you posted.
Alex
Is not finished yet.... but Sunday evening was productive. With parts shortage you have to take what is available. Decided mostly for dale RN, where available 65 and E, a really tight fit! Where not available 60s and 55s D or C. Also managed to fit 1 AB for gridstoppers and avoided the string of 2 and 5 by using parts with higher rating and sufficient voltage drop capability, like mills MRA 12 with 9k1 instead of 5x1k82. Polishing all leads with fine grade steel wool not only removes some patina, makes them shiny as new and soldering very easy (highly recommended !) but also made the leads of the mills fit, tightly - but without forcing into the pad hole. Before it simply did not enter!
Mouser also delivered the LX75 as indicated in September. I have not yet checked if one can program this alone without computer guru knowledge.
Hope to make good progress next weekend solder the remaining parts and wind some coils. Will keep you posted.
Alex
Attachments
Marcel, can I just confirm that the section in post #1 about output levels assumes a single-ended arrangement. If the downstream amplifier operated in differential mode the signal level would be 2.48V RMS?
I ask because I am contemplating having a go at the SusyLu amplifier
SuSyLu Where Are You?
I ask because I am contemplating having a go at the SusyLu amplifier
SuSyLu Where Are You?
Last edited:
No. When you have an output transformer with negligible losses, you get 1.24 V across the load no matter whether it's single-ended or differential. With transformer losses you get a bit less, typically something in the 1 V to 1.2 V range, again independent of whether the load is single-ended or differential.
Without output transformer, you get about 1.24 V into a differential load and about 0.62 V and somewhat increased noise into a single-ended load.
Without output transformer, you get about 1.24 V into a differential load and about 0.62 V and somewhat increased noise into a single-ended load.
Hi Anthony,
Do you mean just a better crystal or a completely redesigned oscillator and buffer?
A better crystal could reduce close-in phase noise. There are people on this forum who are very much into reducing close-in phase noise, see The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator . Unfortunately the thread starter recently got banned.
What I'm about to write now is bound to be contested: I see the theoretical advantage of less close-in phase noise, but I can't imagine it has any practical advantage. Less close-in phase noise means less unintended frequency modulation of the reproduced signal, but even with a dirt cheap crystal and very basic oscillator, the undesired frequency modulation is negligible compared to the undesired frequency modulation (a.k.a. wow and flutter) from the very best electromechanical analogue reproduction equipment like tape recorders and turntables.
Completely redesigning the oscillator and buffer such that the E88CCs see steeper clock signals is likely to reduce the noise floor and possibly noise modulation. I made several improvements in August 2016 and always measured lower noise afterwards. Don't ask me how to improve it further, because if I knew I would have put it in the circuit already.
Marcel
Do you mean just a better crystal or a completely redesigned oscillator and buffer?
A better crystal could reduce close-in phase noise. There are people on this forum who are very much into reducing close-in phase noise, see The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator . Unfortunately the thread starter recently got banned.
What I'm about to write now is bound to be contested: I see the theoretical advantage of less close-in phase noise, but I can't imagine it has any practical advantage. Less close-in phase noise means less unintended frequency modulation of the reproduced signal, but even with a dirt cheap crystal and very basic oscillator, the undesired frequency modulation is negligible compared to the undesired frequency modulation (a.k.a. wow and flutter) from the very best electromechanical analogue reproduction equipment like tape recorders and turntables.
Completely redesigning the oscillator and buffer such that the E88CCs see steeper clock signals is likely to reduce the noise floor and possibly noise modulation. I made several improvements in August 2016 and always measured lower noise afterwards. Don't ask me how to improve it further, because if I knew I would have put it in the circuit already.
Marcel
I just mean a better clock, not a circuit redesign.
Have previously tried to decode TWTMC terminology and would need to spend quite some time figuring out what is needed for 27MHz suitable for the ValveDac...or even if it could still be got.
I get the low phase noise hesitancy, but I would also like to give it a go so any direction I can be pointed in is appreciated. I looked at something simple like the CCHD-957 but it looks like 27Mhz is not available.
Have previously tried to decode TWTMC terminology and would need to spend quite some time figuring out what is needed for 27MHz suitable for the ValveDac...or even if it could still be got.
I get the low phase noise hesitancy, but I would also like to give it a go so any direction I can be pointed in is appreciated. I looked at something simple like the CCHD-957 but it looks like 27Mhz is not available.
27MHz is an unusual frequency for an ultra-low phase noise audio clock. IIUC that was by design in order to help avoid possible intermodulation products from various clock harmonics.
Also, not clear how much an ultra-low phase noise clock would affect this particular dac design. Some dacs seem to be more sensitive than others. Regarding theoretical musing verses listening reports from multiple listeners using various different dacs, my experience is that some people are very good at noticing very tiny differences in SQ that other people don't notice. Maybe the people who don't notice such things don't hear them because they are expecting to hear only music, not artifacts, nor tiny imperfections.
Regardless, you would probably have to have custom crystals manufactured to build an ultra-low phase noise 27MHz audio clock. It would likely be an expensive proposition for one person since there is usually a minimum order quantity for custom crystals.
For one person a custom designed low-phase noise PLL clock might be more practical, but it would probably take an EE background to design one. Don't know if that's something you would want to take on.
Also, not clear how much an ultra-low phase noise clock would affect this particular dac design. Some dacs seem to be more sensitive than others. Regarding theoretical musing verses listening reports from multiple listeners using various different dacs, my experience is that some people are very good at noticing very tiny differences in SQ that other people don't notice. Maybe the people who don't notice such things don't hear them because they are expecting to hear only music, not artifacts, nor tiny imperfections.
Regardless, you would probably have to have custom crystals manufactured to build an ultra-low phase noise 27MHz audio clock. It would likely be an expensive proposition for one person since there is usually a minimum order quantity for custom crystals.
For one person a custom designed low-phase noise PLL clock might be more practical, but it would probably take an EE background to design one. Don't know if that's something you would want to take on.
For what it's worth, I tried powering the clock with an ultra low noise PSU separately and didn't notice a change. Went back to letting the board power it via the 5V reg (I'm using a low noise IC based one).
The biggest jump in performance came from using nicer tubes. Going from some JAN Phillips to nicer matched tubes significantly lowered the noise floor and changed the sound much in a way a tube amp is affected by tube changes.
The biggest jump in performance came from using nicer tubes. Going from some JAN Phillips to nicer matched tubes significantly lowered the noise floor and changed the sound much in a way a tube amp is affected by tube changes.
Does anyone happen to know if off-the-shelf 27 MHz (or close to 27 MHz, like 27.125 MHz, CB channel 14) third-overtone crystals are still sold anywhere? They were quite common 30 years ago, but now most 27 MHz crystals are fundamental mode.
A thing one could try to reduce close-in phase noise without drastic redesign is to use a third overtone crystal. Hence the question. You then also need to add an LC network somewhere to suppress the fundamental mode, I haven't thought about how to do that yet.
A thing one could try to reduce close-in phase noise without drastic redesign is to use a third overtone crystal. Hence the question. You then also need to add an LC network somewhere to suppress the fundamental mode, I haven't thought about how to do that yet.
Last edited:
There is this fellow...
AUDIOPHONICS TCXO Jitter Clock 27MHz 0.5ppm U/I-Sabre compatible - Audiophonics
...and then expensive fellow like the New Class D Neutron Star...both at 27Mhz.
AUDIOPHONICS TCXO Jitter Clock 27MHz 0.5ppm U/I-Sabre compatible - Audiophonics
...and then expensive fellow like the New Class D Neutron Star...both at 27Mhz.
No guarantee that it will work or that it will improve anything at all, but what you could try is replacing the crystal with the series connection of:
A third-overtone crystal, such as the one syn08 found or this one Blocked
An RC parallel network, consisting of 100 ohm in parallel with 150 pF.
At least the pole-zero extraction program I have indicates that it should work, and using a third-overtone crystal should reduce close-in phase noise. Far-off phase noise may well get worse due to the higher motional resistance.
A third-overtone crystal, such as the one syn08 found or this one Blocked
An RC parallel network, consisting of 100 ohm in parallel with 150 pF.
At least the pole-zero extraction program I have indicates that it should work, and using a third-overtone crystal should reduce close-in phase noise. Far-off phase noise may well get worse due to the higher motional resistance.
The original valve DAC is designed to work with a quartz crystal rather than a quartz crystal oscillator. If you want it to work with an external crystal oscillator, it would make sense to kick out the ECC81 valve that's originally used as the oscillator and to kick out the two EF80 clock buffers, and add two EXOR gates to make the clock signal differential. That is, you essentially get the clock path of the DSD-only variant. The original valve DAC PCB isn't designed for that, though.