OK, so the right channel signal gets lost at U10. That also explains the huge offset at the output of U28/U30. You could power off everything and have a very good look at U10 to see if everything is soldered as it should.
Also with power off, you could do a couple of resistance measurements to check connectivity. If the pins of your multimeter leads are too thick, which they probably are, you can use needles: put the needle at the SMD pin and press the multimeter leads to the needles. Of course you will have the usual contact problems that you always have with resistance measurements: a large resistance can either mean no contact or that you haven't pressed hard enough.
Does pin 1 of U9 make contact with pin 1 of U10?
Does pin 5 of U10 make contact with R32?
Are pins 6, 7 and 8 of U10 all connected to the +5 V?
Does pin 4 of U10 make contact with ground?
Does pin 9 of U6 make contact with pin 2 of U10?
Also with power off, you could do a couple of resistance measurements to check connectivity. If the pins of your multimeter leads are too thick, which they probably are, you can use needles: put the needle at the SMD pin and press the multimeter leads to the needles. Of course you will have the usual contact problems that you always have with resistance measurements: a large resistance can either mean no contact or that you haven't pressed hard enough.
Does pin 1 of U9 make contact with pin 1 of U10?
Does pin 5 of U10 make contact with R32?
Are pins 6, 7 and 8 of U10 all connected to the +5 V?
Does pin 4 of U10 make contact with ground?
Does pin 9 of U6 make contact with pin 2 of U10?
I can't say because I was testing just for audio out not if pos and neg are present, I suppose it didn't change,
Left channel was alive from the beginning but only the pos node was functioning.
The good news is that PPY's reclock is working, of course it must me responsible for the half speed behavior I realize.
Periklis
It's a pity that there are no loose connections. Do you see or measure any (unintended) shorts between nearby pins? To check that U10 is really a flip-flop, is the part marking on U10 the same as on U9?
(As an aside, I once spent many days trying to figure out why a switched-mode power supply didn't work until a colleague pointed out that the Schottky diode was actually a tantalum capacitor.)
(As an aside, I once spent many days trying to figure out why a switched-mode power supply didn't work until a colleague pointed out that the Schottky diode was actually a tantalum capacitor.)
That's possible. It is also possible that you damaged it by electrostatic discharge or that there was some latent defect in this flip-flop that didn't show up during production test.
In my experience, it happens far more often that a pin either didn't get soldered to the PCB correctly or got shorted by minuscule solder bridges, though. If your measurements show that it is definitely not that, then you will have to replace the flip-flop. That's doable with a hot air soldering station, but very difficult with a normal soldering iron.
Not hard to remove 8-pin leaded SMD devices with a normal solder iron. So long as the leads are fully exposed and not underneath the device, that is.
I use a wide chisel tip to heat all four leads on one side at once, then lift up that side of the chip while the solder is still molten. The leads on the other side of the chip bend a little to allow it. Then unsolder the other four pins and lift the chip out. Wick pads to complete old solder removal. Quick and easy.
I use a wide chisel tip to heat all four leads on one side at once, then lift up that side of the chip while the solder is still molten. The leads on the other side of the chip bend a little to allow it. Then unsolder the other four pins and lift the chip out. Wick pads to complete old solder removal. Quick and easy.
pgour, did you try to resolder the pins of U10?
I was in the same boat as you till I finally went to test the voltages on the smd parts. As there was no signal after U6 I searched around a bit and found that L1 seemed to be soldered alright, but did not pass any voltage.
A dab of solder added to both pads and everything was fine.
Maybe you have the same problem and the IC still intact?
---
I get (good) sound now, but have at least one major problem:
I get a high pitched noise when the DAC gets muted. The noise is permanent.
At restarting playing again, there's a high pitched beep and then everything is silent again and music plays.
I was in the same boat as you till I finally went to test the voltages on the smd parts. As there was no signal after U6 I searched around a bit and found that L1 seemed to be soldered alright, but did not pass any voltage.
A dab of solder added to both pads and everything was fine.
Maybe you have the same problem and the IC still intact?
---
I get (good) sound now, but have at least one major problem:
I get a high pitched noise when the DAC gets muted. The noise is permanent.
At restarting playing again, there's a high pitched beep and then everything is silent again and music plays.
Does the high-pitched noise sound like the beep, but some 40 dB weaker?
As I was told that a relay would be too slow, I used J109 JFETs as mute switches in the raw DSD valve DAC. Their on resistance is not zero: typically 6 ohm, maximum 12 ohm at 0 V gate-source voltage. That limits the mute attenuation to about 40 dB.
Does increasing R40 to 2.2 Mohm eliminate the beep or make it shorter? R40 and C16 together set a delay in the release of the mute.
As I was told that a relay would be too slow, I used J109 JFETs as mute switches in the raw DSD valve DAC. Their on resistance is not zero: typically 6 ohm, maximum 12 ohm at 0 V gate-source voltage. That limits the mute attenuation to about 40 dB.
Does increasing R40 to 2.2 Mohm eliminate the beep or make it shorter? R40 and C16 together set a delay in the release of the mute.
The permanent noise in the muted state is more fizzly, dither-noisy, the beep is more like a proper "sound".
I have no way of modding this part right now. Also, the mute delay as such is not the problem, as there is no problem when the DAC is not muted.
Is there maybe a way to prevent muting? Like permanently setting the mute signal low? What could be a downside of doing so?
Yes, more than once.
I also did point to point connectivity test through the DSDR signal path.
I am taking a break until the replacement part arrives and if it does not cure the problem I'll reheat all the parts on the path again, adding some fresh soldering paste.
Thanks
---
This behavior is well known on NoDacs but Marcel's ValveDac has some serious precautions. Check DSDON between tracks and on tracks. Do the same for MUTE.
How about the beep? Does the sound get some 40 dB weaker when you make the mute input high?
Thanks for this comment, it points in a different direction than I was thinking about!
diyAchim, is the fizzy noise there when playing silence (like in a track with only zeros)? Does it depend on what sigma-delta algorithm you use or on the settings of the timing trimming potmeters RV3 and RV4? Does it get better at lower DSD rates?
Just checking this out after the Christmas break - belated season's greetings to you all BTW and let's hope 2021 turns out to be a better year.
Anyway, I think Marcel just beat me to it regarding the fizzy noise - I was going to ask whether you had tried playing silence (all 0s) and adjusting the trim pots. It will also be interesting to know how you're sourcing your DSD signal, i.e. what hardware/software is upstream of the valve DAC.
It's good to see a couple of other builds getting close to fruition but I' wondering if I should be feeling lucky that my build worked straight off and just required a bit of tweaking to null the noise,
Anyway, I think Marcel just beat me to it regarding the fizzy noise - I was going to ask whether you had tried playing silence (all 0s) and adjusting the trim pots. It will also be interesting to know how you're sourcing your DSD signal, i.e. what hardware/software is upstream of the valve DAC.
It's good to see a couple of other builds getting close to fruition but I' wondering if I should be feeling lucky that my build worked straight off and just required a bit of tweaking to null the noise,
Just one observation - the dac structure is almost like a cascode. I found this point about improving cascode PSRR for consideration: Improving the Cascode's PSRR (page 3)
Thanks, but with its negative supply, the valve DAC doesn't have the PSRR problem that a normal cascode amplifier has. In fact that's one of the reasons for choosing a negative supply.
Normally the ripple on the positive supply ends up on the output with practically no attenuation, but when the positive supply is ground, there is no ripple. You can't apply this trick with a negative supply to a normal cascode amplifier, because otherwise you get an even bigger PSRR problem at the input.
If for whatever reason (to combine it with another valve circuit, for example) you have to redesign the circuit for a positive rather than negative supply, then you could use the trick you linked to. That would boil down to superimposing the correct amount of supply ripple on the +168 V and +163 V (previously -132 V and -137 V) supplies for the flip-flops that drive the upper valves.
Mind you, these supplies need local decoupling to the supply and ground planes to give the data signal currents a good return path, otherwise the noise floor goes up substantially. You then probably experimentally have to find a compromise value for the capacitors, large enough to conduct the switching spikes yet small enough not to suppress or phase shift the deliberately introduced fraction of the supply ripple. You can also choose the ratio of the decoupling caps to the ground and supply planes such that the capacitive voltage divider that they form also puts roughly the right amount of ripple on the +168 V and +163 V.
Normally the ripple on the positive supply ends up on the output with practically no attenuation, but when the positive supply is ground, there is no ripple. You can't apply this trick with a negative supply to a normal cascode amplifier, because otherwise you get an even bigger PSRR problem at the input.
If for whatever reason (to combine it with another valve circuit, for example) you have to redesign the circuit for a positive rather than negative supply, then you could use the trick you linked to. That would boil down to superimposing the correct amount of supply ripple on the +168 V and +163 V (previously -132 V and -137 V) supplies for the flip-flops that drive the upper valves.
Mind you, these supplies need local decoupling to the supply and ground planes to give the data signal currents a good return path, otherwise the noise floor goes up substantially. You then probably experimentally have to find a compromise value for the capacitors, large enough to conduct the switching spikes yet small enough not to suppress or phase shift the deliberately introduced fraction of the supply ripple. You can also choose the ratio of the decoupling caps to the ground and supply planes such that the capacitive voltage divider that they form also puts roughly the right amount of ripple on the +168 V and +163 V.