Hi,
I observe non-flat frequency response in SRC (SRC4192).
I admit a tiny unevenness on 18-20 kHz, but I have a slope down starting from <1 kHz that ends with – 5-6 db at 18-20 kHz (freq2.png)
I would deeply appreciate somebody could suggest me an idea for how to explain that and what to try to fix that.
The stand A (expected flat frequency response freq1.png)
(connected sequentially)
1 A PC based sweep generator
2 USB sound card (analog out)
3 Passive mode I2S ADC
4 DSP (all filters are off)
5 DAC
6 Level Meter
The stand B (non-flat frequency response freq2.png)
1 A PC based sweep generator (the same)
2 Tried 2 options
2.1 USB to SPDIF (tried also USB to COAX) with SPDIF\COAX receiver (I2S output finally)
2.2 Or USB to I2S (without a receiver obviously)
3 SRC4192 based sample rate converter (passive IN\passive OUT)
4 DSP (all filters are off, the same)
5 DAC (the same)
6 Level Meter (the same)
Tried to verify 2.1 (USB to SPDIF/COAX) frequency response separately it's flat as expected
Tried to change input sample rate (from 44.1 to 98 kHz) and input signal bitness (16 and 24) without any difference and still have significant signal degradation starting from < 1kHz
SRC4192 board is pretty straightforward, just default connection layout with the both In and Out in passive mode
Would be happy if somebody could find an explanation of the above and/or suggest an idea for what to try to make FR flat.
Thank you in advance,
Serge
I observe non-flat frequency response in SRC (SRC4192).
I admit a tiny unevenness on 18-20 kHz, but I have a slope down starting from <1 kHz that ends with – 5-6 db at 18-20 kHz (freq2.png)
I would deeply appreciate somebody could suggest me an idea for how to explain that and what to try to fix that.
The stand A (expected flat frequency response freq1.png)
(connected sequentially)
1 A PC based sweep generator
2 USB sound card (analog out)
3 Passive mode I2S ADC
4 DSP (all filters are off)
5 DAC
6 Level Meter
The stand B (non-flat frequency response freq2.png)
1 A PC based sweep generator (the same)
2 Tried 2 options
2.1 USB to SPDIF (tried also USB to COAX) with SPDIF\COAX receiver (I2S output finally)
2.2 Or USB to I2S (without a receiver obviously)
3 SRC4192 based sample rate converter (passive IN\passive OUT)
4 DSP (all filters are off, the same)
5 DAC (the same)
6 Level Meter (the same)
Tried to verify 2.1 (USB to SPDIF/COAX) frequency response separately it's flat as expected
Tried to change input sample rate (from 44.1 to 98 kHz) and input signal bitness (16 and 24) without any difference and still have significant signal degradation starting from < 1kHz
SRC4192 board is pretty straightforward, just default connection layout with the both In and Out in passive mode
Would be happy if somebody could find an explanation of the above and/or suggest an idea for what to try to make FR flat.
Thank you in advance,
Serge
Attachments
Is the de-emphasis turned on by mistake?
Edit: browsing through the datasheet, I don't see any de-emphasis function. The SRC4392 has de-emphasis so I assumed the SRC4192 to also have it, but apparently it hasn't (or I haven't read the datasheet well enough).
Edit: browsing through the datasheet, I don't see any de-emphasis function. The SRC4392 has de-emphasis so I assumed the SRC4192 to also have it, but apparently it hasn't (or I haven't read the datasheet well enough).
Last edited:
>Is the de-emphasis turned on by mistake?
There is nothing de-emphasis related in SRC4192 :-(
>Have you tried raising the output sample rate?
No chance, SRC output is passive and clocked by DSP (48 kHz) only.
The SRC board itself is pretty primitive SRC4192+ 24.576kHz Crystal Oscillator + 5-3.3v power converter + a couple of caps in the power line.
I found that the soldered crystal oscillator should be powered by 5v (by the specifications), but actually by 3.3v. The clock signal in oscilloscope is stable though.
Could you admit some minor (invisible evidently) irregularity in the reference clock could cause that.
There is nothing de-emphasis related in SRC4192 :-(
>Have you tried raising the output sample rate?
No chance, SRC output is passive and clocked by DSP (48 kHz) only.
The SRC board itself is pretty primitive SRC4192+ 24.576kHz Crystal Oscillator + 5-3.3v power converter + a couple of caps in the power line.
I found that the soldered crystal oscillator should be powered by 5v (by the specifications), but actually by 3.3v. The clock signal in oscilloscope is stable though.
Could you admit some minor (invisible evidently) irregularity in the reference clock could cause that.
You could try and feed it 48kHz input and enable the bypass mode (BYPAS input (pin 9) high) for SRC4192 to see if it's the one actually causing this.
Bypass won’t work unfortunately, I would need active output mode for that.
In my case (say 48.0001) passive input can’t by bypassed to (say 47,9999) passive output.
Thanks for the idea though
In my case (say 48.0001) passive input can’t by bypassed to (say 47,9999) passive output.
Thanks for the idea though
SergeG,
Would you be willing to post a schematic and register settings for devices? Also, sometimes helps to see hi-res pics of the setup.
Might I also ask if you have listened to the sound of music using the setup? Does it sound to you like simple frequency response roll-off?
Would you be willing to post a schematic and register settings for devices? Also, sometimes helps to see hi-res pics of the setup.
Might I also ask if you have listened to the sound of music using the setup? Does it sound to you like simple frequency response roll-off?
Hi Mark,
I have no a scheme.
It’s
I2S frequency up 192K/24BIT board, SRC4192 frequency up board, DAC upgrade board|Connectors| - AliExpress
So the construction is primitive.
The only change I made is pin 26 (MODE0 of SRC4192) reconnected from 0 to 1, to make the output passive.
SRC4192 is not SRC4193 so has no registers at all. All the configuration is made via pins
>Does it sound to you like simple frequency response roll-off?
Looks like that. I didn’t listen it carefully so far, but the sound is generally “normal”.
Anyway all the FFT analyzes are decent enough (excepting FR). It’s just a part of
Digital (coax/spdif) input for DBX Driverack 260
I have no a scheme.
It’s
I2S frequency up 192K/24BIT board, SRC4192 frequency up board, DAC upgrade board|Connectors| - AliExpress
So the construction is primitive.
The only change I made is pin 26 (MODE0 of SRC4192) reconnected from 0 to 1, to make the output passive.
SRC4192 is not SRC4193 so has no registers at all. All the configuration is made via pins
>Does it sound to you like simple frequency response roll-off?
Looks like that. I didn’t listen it carefully so far, but the sound is generally “normal”.
Anyway all the FFT analyzes are decent enough (excepting FR). It’s just a part of
Digital (coax/spdif) input for DBX Driverack 260
When SRC4192 is set for I2C slave mode for both input and output, seems like there have to be 2 internal PLLs they use to sync with the two sets of external I2S clocks. IME, ASRCs tend to sound their worst when a good crystal reference clock is not being used at RCKI, and a PLL is used instead. IME power quality matters too for the PLLs to be as stable as possible. Not clear to me how those things could alter frequency response though unless maybe something is very wrong with the input signals or power. In that case there might be something obviously wrong with the sound.
Other than that, could be the AK4192 chip is a fake or damaged. Sometimes cheap Chinese boards are fitted with used parts to keep the cost down.
Last thing I would say is that I usually sketch out a schematic of ebay boards so I can understand what they have done with the design. Not hard to do with two layer boards.
Other than that, could be the AK4192 chip is a fake or damaged. Sometimes cheap Chinese boards are fitted with used parts to keep the cost down.
Last thing I would say is that I usually sketch out a schematic of ebay boards so I can understand what they have done with the design. Not hard to do with two layer boards.
Last edited:
>When SRC4192 is set for I2C
I2S actually.
> seems like there have to be 2 internal PLLs
Maybe, but the SRC4192 datasheet doesn’t notice that :-(
>Not clear to me how those things could alter frequency response though
Yes, that’s the main puzzle.
>Last thing I would say is that I usually sketch out a schematic
Yes, I have it in my mind, there is nothing extraordinary in the schematics, just default connections as in the datasheet recommendations
I2S actually.
> seems like there have to be 2 internal PLLs
Maybe, but the SRC4192 datasheet doesn’t notice that :-(
>Not clear to me how those things could alter frequency response though
Yes, that’s the main puzzle.
>Last thing I would say is that I usually sketch out a schematic
Yes, I have it in my mind, there is nothing extraordinary in the schematics, just default connections as in the datasheet recommendations
The problem is rather explained than solved.
SRC is fine and not involved.
I connected a logical analyzer to SRC output (DSP input) and (then) to DSP output, converted I2S to samples (to WAV file).
As it turned out FR on SRC output is pretty flat, but on DSP output the FR looks like processed by a de-emphasis.
Finally I was able to find an analog pre-emphasis made by the (native) analog input with a simple RC filter.
So from analog input FR is flat finally, but from digital input is not
SRC is fine and not involved.
I connected a logical analyzer to SRC output (DSP input) and (then) to DSP output, converted I2S to samples (to WAV file).
As it turned out FR on SRC output is pretty flat, but on DSP output the FR looks like processed by a de-emphasis.
Finally I was able to find an analog pre-emphasis made by the (native) analog input with a simple RC filter.
So from analog input FR is flat finally, but from digital input is not