Since the AK4490 DAC in your QA401 can manage -112 dB best case, I suspect all your measurements are lying to you.
If you are serious, just ship a prototype to Tom. Sand the markings off the parts if you must
.I thought we just discussed that the point of averaging was to solve that issue along with the noise floor? Why bother having anything display below -112db if it's a lie?
What I need right now is quality time in front of a good analyzer so I can finalize optimizations, a one off measurement would not help me much at this stage. A combination of the distortion magnifier, a quality oscillator, and the QA401 should get me where I need to go right?If you are serious, just ship a prototype to Tom. Sand the markings off the parts if you must
Depends what you are building, IMHO. You might look at some sets of measurements they do at Sterephile for different types of equipment. DACs need a different set than power amps, for example.
Read Cordell's article in Linear Audio #0. It'll answer that question for you.
Tom
Tom,
I'm sure you already know this but just in case -
If the noise 'grass' is on -158dBV that means nothing WRT the RMS noise measurement. With audio we generally use a 20kHz BW. A good example of this are measurements of something like 24bit ASRC. -144dB DR but the noise floor will be around -170dB on a typical FFT.
The only reason you are seeing the grass on -158 is the measurement bandwidth is very narrow = less noise energy.
I learned this very quickly using an old, VHQ analog spectrum analyser. You could keep reducing the sweep bandwidth and magically see the noise floor disappear. Smoke and mirrors!
T
They really need to start using a notch filter at the AP IP more often. As gear gets more linear, the AP test set is being challenged. From memory they did this with the Benchmark power amp.
Other frustrating habits are when they measure transformer coupled gear and only do a 50Hz measurement. The transformers LF distortion will dominate. I think JA might be on to this more recently.
T
No disagreement that its a useless number in the absence of a bandwidth.
'A text without a context becomes a pretext'.
Haha, very tempting! If only I could before Christmas
If that's a serious offer I might be able to take you up on it at some point soon though.
Per the spec the APx525 and an equivalent input noise of 1.3 uV or equivalent of 5000 Ohms. That translates into 8 nV/rtHz. The best instruments are around 5 nV/rtHz.
Using FFT's you can see noise and artifacts well below that limit with high resolution FFT's. Still oscillators capable of harmonics below -130 dB are limited. Victor's would be the go to solution and very cost effective. The QA401 is really very good. Not quite an RTX but still very low distortion and very useful for this stuff. It uses the same DAC and a wider band ADC that has a higher distortion floor. On both you need to keep the levels about -15dBFS for lowest distortion.
The THX stuff is a bit of an outlier since they have focused on ultra low distortion for some time. Those measurements showing a harmonic at -140 dB from a headphone amp are not done with commercial equipment. AP can tweak an APx525 for around -120 dB THD+N but that is a hybrid analog/digital instrument. I can get there with a Victor oscillator and a Shibasoku 725. And the combination can just hit -140 dB on harmonics typically.
When the system and the DUT have similar harmonic levels they can add or subtract with big errors. If they are out of phase the level goes to zero. In phase then it goes up 6 dB. This is why the system needs to be significantly better than the DUT.
I can get to -170 dB with the CLT-1 but its a single frequency device generating 10 KHz and measuring 30 KHz.
The distortion magnifier will get you further when checking a linear amplifying stage. If everything else is optimal you can get pretty good results. Hopefully it will show what you want to see.
I would be happy to put your device on the bench if you want. PM me. It would force me to clean up the benches.
Using FFT's you can see noise and artifacts well below that limit with high resolution FFT's. Still oscillators capable of harmonics below -130 dB are limited. Victor's would be the go to solution and very cost effective. The QA401 is really very good. Not quite an RTX but still very low distortion and very useful for this stuff. It uses the same DAC and a wider band ADC that has a higher distortion floor. On both you need to keep the levels about -15dBFS for lowest distortion.
The THX stuff is a bit of an outlier since they have focused on ultra low distortion for some time. Those measurements showing a harmonic at -140 dB from a headphone amp are not done with commercial equipment. AP can tweak an APx525 for around -120 dB THD+N but that is a hybrid analog/digital instrument. I can get there with a Victor oscillator and a Shibasoku 725. And the combination can just hit -140 dB on harmonics typically.
When the system and the DUT have similar harmonic levels they can add or subtract with big errors. If they are out of phase the level goes to zero. In phase then it goes up 6 dB. This is why the system needs to be significantly better than the DUT.
I can get to -170 dB with the CLT-1 but its a single frequency device generating 10 KHz and measuring 30 KHz.
The distortion magnifier will get you further when checking a linear amplifying stage. If everything else is optimal you can get pretty good results. Hopefully it will show what you want to see.
I would be happy to put your device on the bench if you want. PM me. It would force me to clean up the benches.
Demian, I'm in good hands I think. Not too shabby with the Victor and the Shibasoku 725D I have to say. And now, will you let me know how you modified that 725 of yours? Or did you sent it up north to davada and have him do it?
Cheers,
Last edited:
I modified my 725. Davida did the 725D. I believe I have his notes for the 725D somewhere. Its similar to what I did with a few key opamp upgrades. The two have really similar performance, the first gen just lacks the lowest step in the range switch. Still the same noise and distortion floor.
Hint- get a few real LME49990's and stash them. You will need maybe 3 max.
Hint- get a few real LME49990's and stash them. You will need maybe 3 max.
My PicoScope 4262 with the largest FFT of ~1M samples and a ±1V range shows a noise floor lower than -125dBV. Averaging will get it (very slowly) below -130dBV. I've used it to measure the difference in distortion between op-amps before, so I'm sure it is useable for that. The real noise is about 42µVrms in the ±1V range, which would be a -84dB noise floor relative to a 1Vpk sine wave. It only has a 5MHz bandwidth though. Getting a true -100dB noise floor would require >16bits of resolution, and I don't recall ever seeing such a 'scope.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.