New DDDAC build with high pitched noise / hiss
I just finished another DDDAC build (details below). It works ok, except from a high-pitched noise / hiss at the audio output. It’s not very loud, but loud enough to be disturbing during playback of silent music. I believe the noise is somehow related to the I2S connection between the RPI/Kali (see tests below). It would be great if anybody could help me fixing this… Any ideas or suggestions where to start debugging?
The system is as follows:
- Blue DDDAC main board (got this second hand from a member here, looks to be in good shape, and works well with a WaveIO board)
- 4 DAC boards (the new version)
- Output is balanced from DDDAC into Sowter output transformers, and into RCA output connectors (isolated from chassis)
- Digital signal source is an RPI with Kali I2S reclocker. The Kali I2S is connected to the I2S input of the DDDAC mainbaord. All I2S data lines have their own GND connection running between the Kali and the DDDAC.
- Software running on RPI is Volumio. It gets all it’s music data via WLAN (no wired data connections!)
- Power supplies are the “Doede designs”. There’s a 12 VDC supply for the DDDAC, and a 5 VDC supply for the RPI/Kali.
The GND scheme is as follows:
- The “negative” supply of the 12 VDC PSU forms the “audio GND”. It is connected to the chassis safety earth using a ground loop breaker (Rod Elliot / ESP type).
- The “rings” of the RCA audio outputs (Sowter secondary) are connected to the audio GND.
- The RPI/Kali are grounded via the GND wires of the I2S connection to the DDDAC. The 5 VDC supply is NOT connected to GND.
Things I have already tested:
- Disconnect output of Sowter secondary from DDDAC GND —> this causes substantial hum on the speakers; I can't tell if the high-pitched noise changes, because the hum is too loud.
- Removed I2S connection between RPI/Kali and DDDAC mainboard —> THE NOISE IS GONE!
- Audio resampling turned on/off in Volumio —> does not affect the noise at all
- Start / stop music playback on RPI —> there is more noise with music playing. I don’t think the additional noise is in the recording (I am quite familiar with some of these recordings, and they are not noisy).
I just finished another DDDAC build (details below). It works ok, except from a high-pitched noise / hiss at the audio output. It’s not very loud, but loud enough to be disturbing during playback of silent music. I believe the noise is somehow related to the I2S connection between the RPI/Kali (see tests below). It would be great if anybody could help me fixing this… Any ideas or suggestions where to start debugging?
The system is as follows:
- Blue DDDAC main board (got this second hand from a member here, looks to be in good shape, and works well with a WaveIO board)
- 4 DAC boards (the new version)
- Output is balanced from DDDAC into Sowter output transformers, and into RCA output connectors (isolated from chassis)
- Digital signal source is an RPI with Kali I2S reclocker. The Kali I2S is connected to the I2S input of the DDDAC mainbaord. All I2S data lines have their own GND connection running between the Kali and the DDDAC.
- Software running on RPI is Volumio. It gets all it’s music data via WLAN (no wired data connections!)
- Power supplies are the “Doede designs”. There’s a 12 VDC supply for the DDDAC, and a 5 VDC supply for the RPI/Kali.
The GND scheme is as follows:
- The “negative” supply of the 12 VDC PSU forms the “audio GND”. It is connected to the chassis safety earth using a ground loop breaker (Rod Elliot / ESP type).
- The “rings” of the RCA audio outputs (Sowter secondary) are connected to the audio GND.
- The RPI/Kali are grounded via the GND wires of the I2S connection to the DDDAC. The 5 VDC supply is NOT connected to GND.
Things I have already tested:
- Disconnect output of Sowter secondary from DDDAC GND —> this causes substantial hum on the speakers; I can't tell if the high-pitched noise changes, because the hum is too loud.
- Removed I2S connection between RPI/Kali and DDDAC mainboard —> THE NOISE IS GONE!
- Audio resampling turned on/off in Volumio —> does not affect the noise at all
- Start / stop music playback on RPI —> there is more noise with music playing. I don’t think the additional noise is in the recording (I am quite familiar with some of these recordings, and they are not noisy).
I find it odd that disconnecting the Sowter output connections from DDDac ground would produce hum. I am assuming that by outputs you mean one side of the output windings of the Sowter is being disconnected from DDDac ground. You should not need to connect the output of the transformer to DDDac ground at all. The transformer should work just fine and indeed better if the secondary is not connected to the same ground as the primary.
I’m not real familiar with the Kali but it seems to me that the - side of the Kali +5v supply may need to be connected to the dddac supply ground.
I would try connecting the Kali supply and dddac supply grounds together and ungrounding the Sowters on the output side. One of the major advantages of the transformer is that it can perform galvanic isolation between the DAC and the amplifier. You loose this when you wire the secondary side back to the DDDac ground.
It sounds like some sort of hf tone is being produced by a process of some kind that is going on in the Kali or the RPI. If the Kaali outputs are not totally isolated from the rest of the RPI/Kali then some current may end up going through the I2S cable and the DDDac board to Power Supply ground, causing issues. By joining the kali and DDDac grounds together you would give that current a shorter path back to ground.
Once again I am not familiar with the details of the implementation and am offering up ideas for you to consider. Good luck finding the problem.
I agree. I didn't look into this very much yet. I somehow believe the hum might be related to the fact that both the DDDAC and the following stage (amplifier) have a ground loop breaker.
The Kali is grounded to the DDDAC GND via the GND wires of the I2S connection. But you're right, these might not provide a good low-impedance conncetion to GND. However, adding a direct low-impedance connection of the RPI/Kali GND to the DDDAC GND would result in a ground loop (two connections to GND!). I might just try and see what happens with the additional GND connection.
Also, I believe you're right by saying that "some sort of hf tone is being produced", because I found that if I set the RPI to upsample the audio stream to something higher than the standard 44.1 / 48 kHz data rate, the hiss goes away. Is it possible the I2S signal interferes with the analog signal of the DDDAC? This might become inaudible if the I2S frequency becomes high enough (by upsampling).
Ok, enough food to go and test a few things -- thanks!
I believe the HF noise you are hearing with 44.1 / 48 kHz is quantization noise. This is a normal thing for DDDAC caused by the NOS design!?!
see:
#post3910900
you can try to install some caps like described in this thread.
I have never heard any sampling noise with DDDAC - ever. I have also backed this up with measurements. RPi is not a good source of a digital sound and it simply should not be used as such.
You can find some measurements for that on the dddac page:
DDDAC 1794 NOS DAC - Non Oversampling DAC with PCM1794 - no digital filter - modular design DIY DAC for high resolution audio 192/24 192kHz 24bit
I can hear this noise on my system, too. Upsampling the source material removes this noise. Maybe you have already an oversampling running in your system? Or your chain is filtering something?
I have a speaker with air motion transformer which is very sensitive and I have to put my ears near the speaker to hear this noise. Means, it is not loud and I don't care about this... But, the difference between raw 41 kHz and oversampling exists...
There is no filtering of any kind in my system... apart from the filtering done by interconnects (which is a very low capacitance - no shielding design) and the filtering that happens in the amplifier. I also do not use output transformers. Additional, purposeful filtering would defy the pure NOS approach of DDDAC design.
I am not saying that there is no increase in quantization noise at 44.1 kHz sampling rate. When I tried 20Hz signal (at -110dB) at 44 kHz sampling, sent to the WaveI/O and then to DDDAC, I did not see the increase in quantization noise as large as Doede. I am not sure what type of power supply regulation Doedede used... In my case, I had a very low noise regulator on Wave I/O +5V DC supply, a very low noise regulator on +12V DC supply for the DAC decks (4 of them) and a separate, very low noise regulator feed of +5V for the motherboard logic (and Wave I/O isolator IC). All S/PDIF motherboard components were disconnected from all power supply rails and then shorted to PCB ground. I am not sure if this was supposed to make any difference in measurements' results.
What I did say was that quantization noise at 44khz sampling rates should not be audible. I do not have speakers with air motion transformers and I do not press my ear close to the speakers; however, I do believe that it is possible to hear it with a very efficient speaker. You probably have a single-ended valve gear to drive those speakers. The very high input impedance of such a valve amplifier and/or low capacitance interconnects would also emphasize the "issue". But, during normal listening (and especially if Sowters are sued), there should not be any quantization noise present at the speakers.
When I used similar Jensen transformers in my F5 power amps I mounted them using small blocks of cherry. I used a forstner bit to make cups in the wood that were a bit larger than the transformer. I then put some clear silicone caulk in the cups and bedded the transformers in it. The blocks of cherry are bolted to the chassis. They have some vibration isolation and are held securely. The tape is also a good solution that I never thought of. I would worry about the heat in a class A power amp. This wouldn’t be an issue in a Dac. Doesn’t Cinemag ship mounting brackets with their transformers?
Transformers could even be suspended on a few rubber bands... if sound/vibration isolation is important.
The double-sided 3M tape comes in various thicknesses. So the thicker one could also dampen vibrations... again, if the sound isolation is important.
Velcro solution is a non-permanent one...
Cinemags I ordered from the manufacturer, did not come with mounting brackets.
I would assume the grounding of the Cinemags would need to be taken care of as well... bolting them to the chassis may or may not be an optimal solution...???
The double-sided 3M tape comes in various thicknesses. So the thicker one could also dampen vibrations... again, if the sound isolation is important.
Velcro solution is a non-permanent one...
Cinemags I ordered from the manufacturer, did not come with mounting brackets.
I would assume the grounding of the Cinemags would need to be taken care of as well... bolting them to the chassis may or may not be an optimal solution...???
I don’t seem to get any audible noise from my single board DDDac. I am not using the filter caps on the main board. I upsample most of the time to 176 or 352. I have listened quite a bit at 44.1 and at idle even with my ear next to the horn in my 95db Pi 3 speakers, I hear very little noise. What noise is there is wide band in nature and there are no tones that are audible. That said, I have never hooked up a HP amp and listened for noise floor changes when upsampling or not. I am using Pure Music 3.06 to upsample. I don’t think quantization noise or aliasing noise would present as a single tone unless you were reproducing a single tone at the time. If you have the SPDIF input available, it’s a good way to eliminate the computer and software as a source of noise. When I use a CD player as input to the spdif it still sounds very good but is a bit less well defined than the wave I/o USB input. I haven’t noticed extra noise with spdif either.
The brackets that Jensen ships with their input transformers are pretty flimsy thin metal angles. I just couldn’t use them on my F5V3 monoblocks so I used the wood blocks that I had. The wiring on the transformer includes one for the case so it is still grounded. I was going to order a pair of the cinemags to install in a balanced tube preamp I am building. Since it is an input transformer that’s really the best place to have it. If you mount them in the DAC you would need to keep the output cables shortish to avoid hf attenuation. I think they recommend 3 ft as the limit.
No, thats not my issue. There is no audible Neiße with the WaveIO. And there is no audible noise in my other DDDAC.
That sounds like a good way to do it. I don't want to use the tape since it makes it harder to move things around at a later stage ...
Ok, I tried a few things:
- Removed the the GND connections in the I2S connection between the Kali and the DDDAC except one (i.e., there is only one GND wire running from Kali-I2S-out to DDDAC-I2S-in) ---> no change
- Added a direct GND connection from RPI/Kali 5VDC power supply to DDDAC GND ---> no change
- Disconnected the Kali from the RPI, which left the RPI unpowered and disconnected from everything ---> no change
Since the noise/hiss is still there if the RPI is not powered up and disconnected from everything, the RPI cannot be the source of the problem.
Since the noise/hiss goes away if I remove the connection between the Kali and the DDDAC (the I2S lines), I think the source of the problem is on the Kali board.
Any thoughts?
I have tried 16 bit, 24 bit and 32 bit settings in the Volumio resampling options. Didn't change anything.