ARTA Measurement of Power Amp

[JensH]

@ 1audio,

The ARTA plot seems to contain a large DC component. Is that included when calculating the noise and could that explain the 1.64%?

 

[1audio]

I was using 1.7.1. I upgraded to 1.8. I found the "dc" component and removed it. The AK5394A needs a small DC offset to remove some distortion things that happen at zero. It also has an internal high pass filter. I turned that on and the DC went away. But the problem was still present.

It seems ARTA is finding a single very narrow tone at 94998 Hz. Praxis doesn't even look that high. I'll check with AKM about it today. Its generated internally to the ADC.

I would be helpful to set the measurement range in ARTA to preclude ultrasonic stuff. Meanwhile, having found this I'll work on getting rid of it.

 

[JensH]

It looks like the 1 kHz mirrored around half the sample rate.

I am very interested in your findings on the AK5394A, since I am close to finishing a PCB layout with this and the AK4399.

In the analyzer SW I use, I can determine the noise bandwidth included in the calculation. I am using AudioTester V3.0d.

 

[1audio]

I am embarking on something similar with those parts. We should cooperate as much as possible. I'm working on a replacement card for the Juli@'s analog section. I'm happy to share what I have done.

I'll try audiotester. See how it compares.

 

[1audio]

Found the problem. I had the demo board for the AK5394A misconfigured. Now that all the settings are right I'm getting better results.

 

[skidave]

Demian,

Thanks for taking time to grab some screen shots and results. Everyone else, thanks for adding information as well.

Looking at the plots from the internal generator the 5W plot suggests everything is -95 dB or lower. The noise is very low really. When the level is increased other spurs appear and that may be normal as power supply harmonics for example.

--Yes, this amp reaches high distortion at about 120 Watts.

I do not trust ARTA's distortion calculation since there is not enough data on how its done. I tried it with a source good for -110 dB THD+N and it gave me -85. I think the window around the fundamental may be too narrow for an unsynchronized analog source.

--Agreed and I have not found a way to high pass filter like real devices (30KHz or 80KHz)

I'm using ARTA 1.7.1 and have not updated yet.

Also, when I use an external generator, it is from my HP8903B.

Here is a screen shot of loop back on my work laptop with some additional grounding with the best results:

An externally hosted image should be here but it was not working when we last tested it.

A few things with my loop back; 1) The high frequency grass is the same on different laptops, but a bit better with additional grounding. 2) This EMU 0204 does not do well with 192KHz sampling, so I use 96KHz. 3) This EMU 0204 also tends to 'skitz' out every once in a while and I think it is an internal issue. If I am in loop back and watching my averages...after about 20 seconds it stops the sine wave but continues to record. Sometimes the sine wave comes back and sometimes it does not. Does this with REW and Audio Tester.

I think at this point, I have what I have and this is the best I am going to get with 'my' hardware.

Dave

 

[skidave]

Now that all the settings are right I'm getting better results.

Wow! I'll take one when ready!

Dave

 

[1audio]

I have had simialr experiences with the EMU stuff. I think its driver issues. With the EMU 1212 installed in a system getting 4 hours of operation on anything before the system would lock up solid was unlikely. I remove the hardware and software, put back the ESI card and no issues at all. I have an 0404 and a tracker pre. Both are flaky but better than the 1212. I suspect setting up a Linux system with the hardware would be far more reliable (many eyeballs on the drivers to get them right). For now I have better places to use my time (like the PCB layout for the project above).

 

[skidave]

I made a bunch of changes to grounding, divider network and some Windows changes. I have much better results and EMU stability is solid.

Here are screen shots of 10 Watts & 95 Watts. For internal, I used the ARTA sine generator. For external, I used my HP 8903B as the source.

The power amp I am using definitely could use a power supply improvement from 1) design and 2) age of caps. But the focus here is to see what is in the spectrum and I'm finally very happy with the results.🙂

An externally hosted image should be here but it was not working when we last tested it.

External Sine @ 10 Watts

An externally hosted image should be here but it was not working when we last tested it.

Internal Sine @ 10 Watts

An externally hosted image should be here but it was not working when we last tested it.

External Sine @ 95 Watts

An externally hosted image should be here but it was not working when we last tested it.

Internal Sine @ 95 Watts

It's still a little grassy, but I can live with that. My THD+N measurements from ARTA are equal with the THD+N measurements from my 8903. Also, I calculated THD with the various harmonics by hand (well using my Excel spreadsheet) and they calculate correctly to the ARTA values (not that I was expecting different).

DAve

 

[feds27]

AK5394A Eval Board Settings

Found the problem. I had the demo board for the AK5394A misconfigured. Now that all the settings are right I'm getting better results.

I may be playing around with a AK5394A in the upcoming week. Are your settings shared somewhere? Also, are you connecting via spdif to a PC soundcard or are you using a I2S to USB type of XMOS asynchronous input like the USBStreamer? Thanks

 

[1audio]

Are you looking at the chip or the demo board? I had to do some significant tweaks to the demo board to get the best performance from it.

I may be able to document those later this week. They involve using a better external clock. The gate oscillator on the board has sidebands from the other digital stuff going on. I swapped the opamps to LME 49710's to get lower distortion. There are some power supply tweaks etc that I'll need to check that seemed to make a difference.

I have the EMU 1212 and 1616 working well now. Big learning curve plus fresh OS install to get stability. I upgraded the 1212 with the LME49720's and improved its performance. The 1616m is next on the upgrade effort but its much more involved to upgrade.

 

[feds27]

Thanks Demian. I'm using the demo board.

 

[bbp]

Are you looking at the chip or the demo board? I had to do some significant tweaks to the demo board to get the best performance from it.

I may be able to document those later this week. They involve using a better external clock. The gate oscillator on the board has sidebands from the other digital stuff going on. I swapped the opamps to LME 49710's to get lower distortion. There are some power supply tweaks etc that I'll need to check that seemed to make a difference.

I have the EMU 1212 and 1616 working well now. Big learning curve plus fresh OS install to get stability. I upgraded the 1212 with the LME49720's and improved its performance. The 1616m is next on the upgrade effort but its much more involved to upgrade.

Hi 1audio,

How do you think about AD7960? A 5M 18bits SAR ADC with ~100dB SNR and ~120dB THD.

I got AP ATS-2 in hand for noise measurement, but ATS-2 has a limit -120dB THD for 1kHz signal. When the frequency rises, its THD increase at the same time. For a HI-END ES9018 DAC design, ATS-2 is not good enough when I testing high freuqncy signals (range ~7kHz to ~100kHz).

I think AD7960 ADC can help me to enhance the measurement for high frequency THD. So, I've order its demo kit from ADI, maybe it will arrive in 3 weeks later.

Best Regards
Paul

 

[1audio]

I just saw this: LTC2378-20 - 20-Bit, 1Msps, Low Power SAR ADC with 0.5ppm INL - Linear Technology which is really interesting. Its nearly $50 but does seem to have the performance you would need. However for ESS DAC you need to use analog preprocessing. Its one of the few DACs that approach the performance of the best ADC's.

I wish I could figure out why it doesn't seem to sound as good as it measures.

 

[bbp]

I just saw this: LTC2378-20 - 20-Bit, 1Msps, Low Power SAR ADC with 0.5ppm INL - Linear Technology which is really interesting. Its nearly $50 but does seem to have the performance you would need. However for ESS DAC you need to use analog preprocessing. Its one of the few DACs that approach the performance of the best ADC's.

I wish I could figure out why it doesn't seem to sound as good as it measures.

I‘ve also notice that this LT2378-20 from linear tech.
But, per its DATASHEET tells, the THD is not so impresive at high frequency range. That's one of the reason that I didn't choose it. The other reason is that the analog bandwidth seems only 500kHz. same as AP2, but not so enough for me.

Do you have tried MultiInstrument for FFT analyse? I have no experience with ARTA, but MI seems excellent to my experience.
MI has the 'calibration' which can deliver correct absolute voltage from the monitor window, this function is very useful when the absolute noise/voltage is need.

For why so many people say ES9018 does not sound good, I think I've already find the root cause.
²»Ð¡ÐÄÏûÃðES9018µÄÉíÉϵÄÒ»Ö»³ô³æÁË£¬M´óÏÀÀ´¿´¿´¹þ - Powered by Discuz!

My DAC presents unique -30dBFS to -45dBFS range performance than any other manufactures', even the 'TOP of HI-END'.🙂
You can refer to the measurement below:

 

[1audio]

I tried to read through the linked thread using Google translate. I think you have identified a "bump" in the distortion vs. levle curve that is very interesting. My guess is that they have a distortion cancelling scheme in the ESS chip that is doing something to get the really good numbers but is not so happy with complex signals.

I think you said you found a workaround for the distortion "bump". If so, that's pretty impressive.

I'll try multi-instrument. Normally I use Praxis since its really flexible but its too expensive for most DIY'ers. The APx is nice but a lot of money and you have demonstrated its easy to run into its limits. The AP 2722 is way too much money and seems to be at end of life.

I have one of these: http://www.diyaudio.com/forums/equi...requency-analog-digital-signal-generator.html which is proving invaluable for DAC development.

 

[31697B]

My guess is that they have a distortion cancelling scheme in the ESS chip that is doing something to get the really good numbers but is not so happy with complex signals.

The AP 2722 is way too much money and seems to be at end of life.

I have one of these: http://www.diyaudio.com/forums/equi...requency-analog-digital-signal-generator.html which is proving invaluable for DAC development.

___________________________________

My feeelings as well re ESS chip.

AP2722 - Unless it got used and abused a lot... it is like a TEK or HP instrument and will go for decades.

I will get one of those ADC too. Looks like a great tool... and maybe useful/adaptable for music recording, too??


THx-RNMarsh

[still in the middle of no where - Nepal]

 

[skidave]

I started this thread almost a year ago and I thought I would post some recent images during testing of a project I finished. I have come a long way with my limited hobby time---here and there. The images below are from a tube rectified push-pull KT88 power amp.

Square Wave:


Input Shorted:


Both Channels @ 4 Watts:


1% Distortion:


1% Distortion on my HP 5380 Spectrum Analyzer (sweep starts at about 900Hz and is 1000Hz / division):


A few quick notes...I did not grab any screen shots while using my notch filter. Also, when using ARTA, I use -5dB as the peak of the sine wave. If I adjust the sound card input for 0dB with the sine wave, the level indicator on the sound card peaks and it distorts the signal. If anyone has any suggestions...please let me know. I'm fine with using -5dB; if calculate manually, I just compensate the -5dB in the calculation.

Dave

 

[SyncTronX]

SkiDave,

I see you have a 3508A. I'm jealous.
Not just for the Analyzer, but for the KNOBS.

Can actually read them. ; )

cheers,

 

[skidave]

I see you have a 3508A. I'm jealous.
Not just for the Analyzer, but for the KNOBS.

Amazing find via Craigslist relatively local to me. Yes, the knobs are intact, but need delicate fingers when turning.