DIY Audio Analyzer with AK5397/AK5394A and AK4490

Member
Joined 2009
Paid Member
Thanks.

I found the problem with the noisy channel. It has been fixed now and the picture below shows a loopback from generator output to the left input.

The generator output level was -9dBFS and -19dBFS. Input set to no gain/attenuation.

The right channel still doesn't work correctly.
 

Attachments

  • 1kHz loopback_256k FFT_average over10_-9dBFS out.png
    1kHz loopback_256k FFT_average over10_-9dBFS out.png
    38.5 KB · Views: 3,091
  • 1kHz loopback_256k FFT_average over10_-19dBFS out.png
    1kHz loopback_256k FFT_average over10_-19dBFS out.png
    37.5 KB · Views: 2,482
Member
Joined 2009
Paid Member
This is the noise floor. The A-weighted noise level is measured at -126.3 dBFS.
By the way: the PCB has not been put into a box yet, so there is no shielding at the moment.
 

Attachments

  • Noise floor_256k FFT_average over 100_220ohm across inputs.png
    Noise floor_256k FFT_average over 100_220ohm across inputs.png
    35 KB · Views: 2,328
Member
Joined 2009
Paid Member
OK, got the right channel working also. I had deliberately left out the capacitors on the ADC right channel input to make it easier to experiment with different values if needed. I had forgotten all about until a moment ago :eek:

Now they are in place and the results are similar to the results for the left channel, see below.

So the basic functionality now seems to be OK. I still need to adjust it and check the performance further.
 

Attachments

  • 1kHz loopback_256k FFT_right channel_average over10_-9dBFS out.png
    1kHz loopback_256k FFT_right channel_average over10_-9dBFS out.png
    38.5 KB · Views: 2,286
  • 1kHz loopback_256k FFT_right channel_average over10_-19dBFS out.png
    1kHz loopback_256k FFT_right channel_average over10_-19dBFS out.png
    37.5 KB · Views: 2,225
I finished mounting the PCB yesterday. I have attached a picture of the assembled PCB.

JensH,

Nice work ! Did you use a reflow oven or did you solder them by hand ? Also did you use a stencil ?
Also could you tell how long did it take you to
1. place the components and solder paste (if any)
2. to solder them if manually soldered
I'm asking these questions just to get an idea of the effort needed for such a board, should I ever attempt something like that.
 
Member
Joined 2009
Paid Member
I soldered the components by hand. No stencil. It probably took me something like 20-25 hours to do so. Excluding the time it took to find the components in different sample kits etc. I used parts from these sample kits for all the uncritical parts. The critical parts, e.g. 0.1% resistors, I ordered from Farnell and Mouser.

I have a Mantis Vision microscope, which helps a lot in the assembly process. Without it, I would have no chance to assemble it by hand. Maybe I could have done it 20 years ago, but not today.

It is a big task to solder it by hand and I have no plans to assemble a second one by hand! But for a first prototype, to verify the performance, it was OK.
 
Member
Joined 2009
Paid Member
Two of the attached screen dumps show the noise floor at 192 kHz sampling.

The measurements were done at the 0dB input setting. This means a direct input, no attenuation and no additional gain, max input around 3Vrms.
There is a small noise peak at probably 48kHz on both channels. The right channel is slightly noisier than the left channel. There are also a couple of noise peaks between 80 and 90 kHz on the right channel.
But there doesn't seem to be any discrete noise frequencies apart from these. And there is still no shield/box around the PCB.

I have also measured the frequency response at 192 kHz sampling.
The attached figure shows the loop back response. It does drop a bit at high frequencies. This is primarily due to the generator. So I need to fine tune the filters on the outputs.
Measured separately the input side drops by 0.1 dB at 80 kHz and 0.3 dB at 90 kHz.
 

Attachments

  • Noise floor_left channel_relay 0dB_1M FFT_average over 100_220ohm across inputs.png
    Noise floor_left channel_relay 0dB_1M FFT_average over 100_220ohm across inputs.png
    34.7 KB · Views: 983
  • Noise floor_right channel_relay 0dB_1M FFT_average over 100_220ohm across inputs.png
    Noise floor_right channel_relay 0dB_1M FFT_average over 100_220ohm across inputs.png
    35.5 KB · Views: 844
  • Frequency response_-11dBFS out.png
    Frequency response_-11dBFS out.png
    27.8 KB · Views: 954
Member
Joined 2009
Paid Member
@ Mogens,

I have a number of options for the external generator:

1. A Panasonic VP-7722A. I have used this for some initial distortion tests.
2. A home made 1 kHz low distortion oscillator. I haven't used it yet for this design. Distortion is below -150 dB.
3. A generator in my Agilent oscilloscope. It goes to 5 MHz, so I have used this for the input frequency response measurement.
 
Member
Joined 2009
Paid Member
I have modified the post-DAC filter to improve the frequency response at the top. I also changed the gain a little, so that I now have a maximum level of 20Vrms balanced and 10Vrms single ended.

The attachment below shows the loop-back frequency response of the modified channel.
 

Attachments

  • Loopback frequency response left channel 150811.png
    Loopback frequency response left channel 150811.png
    55.4 KB · Views: 1,952
Dear JensH,

Very nice work and impressive results!
May I ask you if you had done test at higher sampling rates?
Like a loopback test for best distortion, like at -10/-11dB at 96K sampling, the usual ~1kHz test frequency?
I'm asking because I don't understand something in my own setup, with AKM4490 into an EMU card's 5394ADC. I see a distortion rise more than I would have expected.

Ciao, George
 
Member
Joined 2009
Paid Member
@ Joseph K,

Yes, I have done some testing at 192 kHz sample rate. The loop-back frequency response and noise floor measurements are shown above.

I will try the measurement you describe when I get home and report it here. So far I have mainly tested at 48 and 192 kHz sample rates.