Hi, thanks for the suggestions.
I tried to copy all the settings from CyberPit, but I forgot to set the sampling Frequency to 96kHz. Once I did this, the sine wave had 1kHz frequency as expected.
I tried to copy all the settings from CyberPit, but I forgot to set the sampling Frequency to 96kHz. Once I did this, the sine wave had 1kHz frequency as expected.
Alright, here are some first results ( sorry for the long post 😆)
My measurement setup:
Octavia board sitting without a case on the desk, raspberry pi 3B running. I connected the differential output from the header to my Komplete Audio 1 interface (MIC input). The interface has an THD-N of 0.001% and a noise floor of about -150db ( a bit higher at lower frequencies). On the desk are also monitors and a docking station. They might be responsible for some peaks in the measurement as well. I powered the board with a 12V MeanWell power supply. Don't have the lower voltage supply yet.
I measured the default ocatvia configuration with several input amplitudes:
Very high input amplitudes (-3dbfs - 0dbfs): Here the THD-N is a bit worse compared to the optimum. On one side I get close to the maximum of my audio interface (komplete audio 1) and on the other hand has the DAC not the best THD-N at 0dbfs.
Input amplitued of -4dbfs to -8 dbfs: In this section I measured the best performance. This also makes sense when looking into the DACs datahsset. This graph shows a similair optimal range.
Here is the best measurement at -7dbfs I measured:
With THD-N of -91.8dB we are already very close to the DACs performance (-93dB at -1dbfs and from the plot above a bit better at -7dbfs, so maybe 95-96dB). The last few dBs could are probably a combination of lower noise and lower harmonics. Also this peak at 150Hz (probably caused by a device on my desk) makes the THD-N a bit worse. So I think the performance is not bad at all give the limits of the DAC.
Additionally, I tried a different LC lowpass at the output of the DAC with 470 Ohms and 2.2nF. I also used a film capacitor. This however did not seem to change the results. I migh do some more measurements in the following days.
I also tried 10k resistors instead of 4.7k resistors at the opamps. My Idea was to have higher load resitance. The harmonics did not change but the noise incereased, which was to be expected. So this is not useful either. I'll maybe try with 1k resistors as well. This could lower the noise a bit more and hopefully not increase the harmonics. Lets see...
Tomorrow I can remove some muting transistors and an opamp at work to do some measurements without them.
One thing I also noticed, when I power up the board the harmonics not so high and they will increase over time to a steady point shown in the graphs above. Maybe this is dues to heat? I have to see whether the measurements change when I point a fan at the board.
This is it for now. I'll keep you updated.
And if you have any ideas what I can further test, feel free comment 😉
Ludwig
My measurement setup:
Octavia board sitting without a case on the desk, raspberry pi 3B running. I connected the differential output from the header to my Komplete Audio 1 interface (MIC input). The interface has an THD-N of 0.001% and a noise floor of about -150db ( a bit higher at lower frequencies). On the desk are also monitors and a docking station. They might be responsible for some peaks in the measurement as well. I powered the board with a 12V MeanWell power supply. Don't have the lower voltage supply yet.
I measured the default ocatvia configuration with several input amplitudes:
Very high input amplitudes (-3dbfs - 0dbfs): Here the THD-N is a bit worse compared to the optimum. On one side I get close to the maximum of my audio interface (komplete audio 1) and on the other hand has the DAC not the best THD-N at 0dbfs.
Input amplitued of -4dbfs to -8 dbfs: In this section I measured the best performance. This also makes sense when looking into the DACs datahsset. This graph shows a similair optimal range.
Here is the best measurement at -7dbfs I measured:
With THD-N of -91.8dB we are already very close to the DACs performance (-93dB at -1dbfs and from the plot above a bit better at -7dbfs, so maybe 95-96dB). The last few dBs could are probably a combination of lower noise and lower harmonics. Also this peak at 150Hz (probably caused by a device on my desk) makes the THD-N a bit worse. So I think the performance is not bad at all give the limits of the DAC.
Additionally, I tried a different LC lowpass at the output of the DAC with 470 Ohms and 2.2nF. I also used a film capacitor. This however did not seem to change the results. I migh do some more measurements in the following days.
I also tried 10k resistors instead of 4.7k resistors at the opamps. My Idea was to have higher load resitance. The harmonics did not change but the noise incereased, which was to be expected. So this is not useful either. I'll maybe try with 1k resistors as well. This could lower the noise a bit more and hopefully not increase the harmonics. Lets see...
Tomorrow I can remove some muting transistors and an opamp at work to do some measurements without them.
One thing I also noticed, when I power up the board the harmonics not so high and they will increase over time to a steady point shown in the graphs above. Maybe this is dues to heat? I have to see whether the measurements change when I point a fan at the board.
This is it for now. I'll keep you updated.
And if you have any ideas what I can further test, feel free comment 😉
Ludwig
Firstly congrats on getting the board finished and working! These measurements look quite encouraging 😀
Do you think with a lower voltage supply the heat related harmonics may reduce?
Thanks for sharing your findings Ludwig!
I still have to confirm that the heat is the cause. But I will test again with a fan and a 7.5V source once it is here.
I am not sure. Everything in the audio path can cause distortion. So yes, the measurements can look better, but the noise can also be worse due to the nature of unbalanced connection. I will remove some muting transistors today after work and also change two outputs to unbalanced and do some measurements on this.
Hello again,
I have a few more results to share.
First of all, the muting transistors don't seem to have any negativ impact on the performance.
Then I measured an unbalanced output. At first I thought that the performance is much worse, but for the unbalanced output I have to use a different input on my audio interface, which has a different gain. Therefore the input level was much lower and THD-N worse. Then I reduced the gain with the balanced output to have a compareable input level in the measurement. The THD-N results are compareable, the harmonics look a bit different:
unabalanced - orange, balanced - green
The harmonics look a bit different (more on that later), the noise is a little bit lower with the balanced connection. Unfortunately, my audio interface limits my measuring possibilities regarding the unbalanced connection here, or maybe I can connect it to the balanced input somehow 🤔. I'll think about it...
Now I have one more interesting finding. So I took measurement not only from one balanced channel and I noticed that they look differently:
Here for example are two channels, both with the original components:
channel32 - green, channel 16 - brown
The harmonics on both channels look different. On channel 16 (brown) the even harmonics (2nd, 4th, 6th) are lower compared to channel 32(green). Here I would say this is due to non identical pcb layout of both channels. As far as I have read from a TI note, symmetry of traces and components affect the even harmonics, which could be the case here. But please tell me if I am wrong here. By no means I am an expert here.
I didn't find the time to inestigate the heat thing yet.
Ludwig
I have a few more results to share.
First of all, the muting transistors don't seem to have any negativ impact on the performance.
Then I measured an unbalanced output. At first I thought that the performance is much worse, but for the unbalanced output I have to use a different input on my audio interface, which has a different gain. Therefore the input level was much lower and THD-N worse. Then I reduced the gain with the balanced output to have a compareable input level in the measurement. The THD-N results are compareable, the harmonics look a bit different:
unabalanced - orange, balanced - green
The harmonics look a bit different (more on that later), the noise is a little bit lower with the balanced connection. Unfortunately, my audio interface limits my measuring possibilities regarding the unbalanced connection here, or maybe I can connect it to the balanced input somehow 🤔. I'll think about it...
Now I have one more interesting finding. So I took measurement not only from one balanced channel and I noticed that they look differently:
Here for example are two channels, both with the original components:
channel32 - green, channel 16 - brown
The harmonics on both channels look different. On channel 16 (brown) the even harmonics (2nd, 4th, 6th) are lower compared to channel 32(green). Here I would say this is due to non identical pcb layout of both channels. As far as I have read from a TI note, symmetry of traces and components affect the even harmonics, which could be the case here. But please tell me if I am wrong here. By no means I am an expert here.
I didn't find the time to inestigate the heat thing yet.
Ludwig
After thinking about it, my comparison of the unbalanced output might not be correct. Maybe I should switch from dbfs measurement to dbu. I'll test this again tomorrow...
Alright,
here is the update to the unbalanced conenction/comparison.
I was right with my little note yesterday, the comparison was not right. I need to have the same gain for both measurement, everything else makes no sense. And then I have to calibrate the levels of the two inputs on my audio interface. I found the option in REW to do so. Now I have a comparable set of graphs:
unbalanced - brown, balanced - turquoise
So harmonics are higher and the noise floor is higher on the unbalanced conenction, therefore THD-N as well. So unbalanced connection is not as good as balanced connection. But keep in mind that there are differences between the channels as discussed in my last big post. Some channels might look better or worse... Still, differential design helps to reduce noise and even harmonics so the result is quite plausible.
Ludwig
here is the update to the unbalanced conenction/comparison.
I was right with my little note yesterday, the comparison was not right. I need to have the same gain for both measurement, everything else makes no sense. And then I have to calibrate the levels of the two inputs on my audio interface. I found the option in REW to do so. Now I have a comparable set of graphs:
unbalanced - brown, balanced - turquoise
So harmonics are higher and the noise floor is higher on the unbalanced conenction, therefore THD-N as well. So unbalanced connection is not as good as balanced connection. But keep in mind that there are differences between the channels as discussed in my last big post. Some channels might look better or worse... Still, differential design helps to reduce noise and even harmonics so the result is quite plausible.
Ludwig
Member
Joined 2018
Hello Ludwing-San,
Thanks for your investigation and the detailed analysis considerations.
Here are closeup traces around CH16 and CH32. Electrical connections are the same but the signal traces are different due to the x4 DAC mode configuration capability.
If the DAC outputs the same quality signals, the difference will come from the OPA1632 unbalanced to balanced output driver circuit or PCB patterns. I know It's very difficult to detect this kind of difference because it's close to the limit of the USB Audio interface performance. You'd better put the metal plate on your bench desktop and connect it to the USB-Audio I/F Chasis electronically. It will decrease the surrounding equipment's effects. Hope you'll figure out the root causes of this phenomenon.
Best Regards,
CyberPit
Thanks for your investigation and the detailed analysis considerations.
Here are closeup traces around CH16 and CH32. Electrical connections are the same but the signal traces are different due to the x4 DAC mode configuration capability.
If the DAC outputs the same quality signals, the difference will come from the OPA1632 unbalanced to balanced output driver circuit or PCB patterns. I know It's very difficult to detect this kind of difference because it's close to the limit of the USB Audio interface performance. You'd better put the metal plate on your bench desktop and connect it to the USB-Audio I/F Chasis electronically. It will decrease the surrounding equipment's effects. Hope you'll figure out the root causes of this phenomenon.
Best Regards,
CyberPit
Hi CyberPit-San,
i will take a closer look at the PCB design the next days to analyse the differences between the channels. I also started creating a small simulation of the opamp circuit to better check if something can be improved there.
Besides that, my plan is to make a small test PCB with a dac and opamp (and maybe supply) to test a different PCB design for the output stage.
Thanks for the tip with the metal plate. However, my audio interface has no metal housing and no grounding screw...
BR
Ludwig
i will take a closer look at the PCB design the next days to analyse the differences between the channels. I also started creating a small simulation of the opamp circuit to better check if something can be improved there.
Besides that, my plan is to make a small test PCB with a dac and opamp (and maybe supply) to test a different PCB design for the output stage.
Thanks for the tip with the metal plate. However, my audio interface has no metal housing and no grounding screw...
BR
Ludwig
Member
Joined 2018
Hi Ludwig-San,
Well, the PCB space is too narrow to work around the circuit trials.
It's nice to start component tuning with simulation before the actual modification.
Maybe the last boss will be the PCB trace tuning. The actual audio performance will be different even if the same schematic circuits are.
CyberPit
Well, the PCB space is too narrow to work around the circuit trials.
It's nice to start component tuning with simulation before the actual modification.
Maybe the last boss will be the PCB trace tuning. The actual audio performance will be different even if the same schematic circuits are.
In that case, you can connect to the USB connector's frame ground or audio input ground. (choose the better one)
CyberPit
Hi CyberPit-San,
However, this week is already very much booked, so I guess until next week there won't be much progress.
Ludwig
At least for filter setup and the opamp circuitry. THD analysis is difficult as the PCB layout has such a big impact on it.
I am already heads deep into some books regarding EMI and PCB layout.
However, this week is already very much booked, so I guess until next week there won't be much progress.
I'll try, thanks.
Ludwig
I have not forgotten you guys, but I am moving house at the moment so my spare time is rather limited.
Ludwig
Ludwig
Member
Joined 2018
Hello Ludwig-San,
I've got something to tell you.
Point-A is PCM5102A output, and Point-B is Differential Driver Output. Resistors C are jumpers between Cold-End and RCA-Sleeves.
Point-A: PCM5102A DAC Output Distortion Spectrum.
Not so bad as of this point...
Point-B: Differential Driver Output Distortion Spectrum.
Distortion increased significantly.
Point-B: Resisrotors C Removed Non-Differential Driver Output Distortion Spectrum.
Distortion was not so increased.
According to those data shows, a significant large signal distortion was produced by OPA1632 driver circuit which is connected as a Cold-side termination. The last spectrum shows a heavy-loaded condition was released by removing R196/R197.
A differential output can reduce the hum-noise in many cases, but a large signal quality has some damage.
Perhaps we need re-think the differential output circuit connection.
CyberPit
I've got something to tell you.
Point-A is PCM5102A output, and Point-B is Differential Driver Output. Resistors C are jumpers between Cold-End and RCA-Sleeves.
Point-A: PCM5102A DAC Output Distortion Spectrum.
Not so bad as of this point...
Point-B: Differential Driver Output Distortion Spectrum.
Distortion increased significantly.
Point-B: Resisrotors C Removed Non-Differential Driver Output Distortion Spectrum.
Distortion was not so increased.
According to those data shows, a significant large signal distortion was produced by OPA1632 driver circuit which is connected as a Cold-side termination. The last spectrum shows a heavy-loaded condition was released by removing R196/R197.
A differential output can reduce the hum-noise in many cases, but a large signal quality has some damage.
Perhaps we need re-think the differential output circuit connection.
CyberPit
Last edited:
I'm cheering this on from the sidelines.
Great project! 👍
As a long term Najda DSP user (since the first boards shipped, back in 2013, I think it was!), I've found it excellent and multi channel DSP really is the the thing for multi channel audio.
I have 10 channel requirement, but get around that by analogue XOs on the tweeter channels.
If there are plans for a 10 out Octavia, I'd be very interested.
Low noise is very important using sensitive compression drivers on horns - you can hear a pin drop!
USB input and I2S over hdmi would be good, as Najda lacks these.
I have used the native I2S input to Najda a lot though.
One question I have (don't think it was answered), what are the 4 potentiometers in a row for?
Other questions;
Is there a volume remote?
Is the volume analogue or digital?
Can it do FIR? How many taps?
What sort of X/O / PEQ / time delay is available?
How is set up done? Via PC?
How many preset setups can be stored and accessed?
Possible to switch between them on the fly?
Sorry if I misinterpreted some things and the questions are not relevant.
Once again, great project.
Great project! 👍
As a long term Najda DSP user (since the first boards shipped, back in 2013, I think it was!), I've found it excellent and multi channel DSP really is the the thing for multi channel audio.
I have 10 channel requirement, but get around that by analogue XOs on the tweeter channels.
If there are plans for a 10 out Octavia, I'd be very interested.
Low noise is very important using sensitive compression drivers on horns - you can hear a pin drop!
USB input and I2S over hdmi would be good, as Najda lacks these.
I have used the native I2S input to Najda a lot though.
One question I have (don't think it was answered), what are the 4 potentiometers in a row for?
Other questions;
Is there a volume remote?
Is the volume analogue or digital?
Can it do FIR? How many taps?
What sort of X/O / PEQ / time delay is available?
How is set up done? Via PC?
How many preset setups can be stored and accessed?
Possible to switch between them on the fly?
Sorry if I misinterpreted some things and the questions are not relevant.
Once again, great project.
Hello CyberPit-San,
nice investigation you did there. This definitly proves that the performance can be improved. But I am also a bit confused, because in my balanced configuration R196 is already desoldered and R212 is desoldered and I still measured results close to your second graph.
I think this week I will find some time to work on this project again. Lets see what we can find out.
BR Ludwig
nice investigation you did there. This definitly proves that the performance can be improved. But I am also a bit confused, because in my balanced configuration R196 is already desoldered and R212 is desoldered and I still measured results close to your second graph.
I think this week I will find some time to work on this project again. Lets see what we can find out.
BR Ludwig
I am not so familiar with the term "cold-side termination". Can you eloberate this?
Hi Speedysteve,
here are some answers to your questions.
I am not sure about the delay. There is porbably a value somewhere in the datasheet.
In general many things are possible but they all require a little bit of tinkering , but we are at diyAudio afterall 😉.
Best Regards,
Ludwig
here are some answers to your questions.
I think so far the plan is to improve the performance.
The potentiometers are just analog inputs to the DSP chip. You can define yourself what they do when programming/configuring the DSP. Of the top of my head I can think of volume control, gain control for specific channels, treble, base, balance, and probably many more. You could also switch between filter presets.
not directly. However it is possible if a raspberry pi is connected. The board has an IR receiver and the raspberry pi could change the volumes in the DSP registers.
The volume control is digital in the DSP.
The Adau1452 datahseet says: "24,000 FIR filter taps per sample at the standard 48 kHz audio sampling rate."
I am not sure about the delay. There is porbably a value somewhere in the datasheet.
The DSP is programmed using SigmarStudio. This is a PC application. The configuration can be written using a USBi adapter or via network if a raspberry pi with the appropriate software is used. The configuration can also be written to a EEPROM. This way the DSP will get the config on startup automatically.
It is hard for me to estimate this. You can store as many as the memory allows and then you can switch between them using one of the potentiometers for example. This is however not a ready to use feature, you have to program this yourself in SigmaStudio.
In general many things are possible but they all require a little bit of tinkering , but we are at diyAudio afterall 😉.
Best Regards,
Ludwig