Low-distortion Audio-range Oscillator

Jump to Latest
S

simon7000

Member
Joined 2008
vicnic said:
What type of capacitors you have tested in this way? Metallized film, film/foil or other? What kind of dielectric?
Click to expand...

Metalized polyester mostly, also film and foil Wimas. I was comparing the change from break in to that of using a silica gel drying chamber.

I find storing capacitors in a metal paint can full of silica gel packets does lower their distortion.

Six months in a drying chamber seemed to be about the same as a week of break in.

My take on this is to put silica gel packets inside my gear that uses film capacitors. Things like a phono preamp, a preamp with tone controls but not my audio power amplifiers that use very large film capacitors on the voltage rails. (Might just try it though!)
 
1

1audio

Member
Joined 2004
Paid Member
Polystyrene is very hygroscopic and should not be washed in a normal PCB line. Also temp sensitive with a max temp that is pretty low. Moisture can affect caps that are not hermetic in many ways. hermetic caps and resistors are an expensive PITA but necessary in some applications.

The network will need to be tuned to be useful. Its easier to tune the oscillator if provided than the network. Victors oscillators drift a little on warmup and over time. All parts do to some extent. A precision voltage source or resistor will need 3-6 months of "break-in" before it can really be calibrated or certified. Fully characterized resistors cost much more than even audiophile resistors, same for caps.
 
jan.didden

jan.didden

AX tech editor
Joined 2002
linearaudio.net
Paid Member
Years ago I attended an AES presentation of the effect of film cap constriction and expansion with voltage. The effect was measured using a laser displacement measuring system. Result: max they saw at very large voltages was a dimensional change of less than molecular size.

Two conclusions:

1 - to measure is to know;

2 - no use losing sleep over this.

Jan
 
diyralf

diyralf

Member
Joined 2016
Amazing NE5532

I'm just doing a few measurements with notch filters. For this I use a 40dB amplifier with LM4562 with these values:


LM4562

H2: -93dBFS,

H3: -102 dBFS

noisefloor: -133dBFS


I have then used a NE5532 and come to the following values:

H2: -99dBFS,

H3: -115 dBFS

noisefloor: -130dBFS.



Again, a proof of the excellent quality of the NE5532.
 
M

Monte McGuire

Member
Joined 2013
What's your analyzer's loopback distortion reading?

This is done by connecting the generator to the analyzer with nothing between them, to establish a baseline distortion reading, to which all of your DUT readings will be added (in a vector sense, which means that phase can cause the observed harmonic levels to be higher or lower than the analyzer's residual, depending on the DUT harmonic's phase relative to the analyzer's residual harmonic phase).

I ask because either your circuit is torturing those amplifiers with a high noise gain, or your analyzer residual us very high. These numbers are not the best that either chip can do by quite some margin. Relative rankings are also not possible when the analyzer has a high residual relative to the actual DUT distortion, because the DUT and analyzer harmonics don't add linearly. The analyzer has to be about 20dB better than the DUT to be able to accept the raw numbers without correction.
 
M

Monte McGuire

Member
Joined 2013
Jan is right - dBFS is meaningless. You can be off easily by 10-20dB just by being casual with assumptions, and that's a lot.

Even if 0dBFS is +20dBV and you have a fundamental level of 0dBV, giving you something around -120dBc harmonic levels, the numbers are far worse than the chips themselves when run in a low gain circuit. These numbers are more in line with the residual of a quality ADC and DAC - you're looking at your test set. This is why you need to determine your test set's residual, so you don't play games with yourself and make relative decisions based on measurement fallacies.

Also, the noise floor is not the level of the "noise floor" bin level that you see in an FFT, because the wideband noise is divided among all of the FFT bins. How many bins is that? 4096? 4 million? This affects that number greatly - the wideband noise level will be reduced by roughly 3dB * log2(FFT length). With a 1M point FFT, that means your bin levels will be 60dB less than the actual broadband noise level. With a 4096 point FFT, the bin levels will be only 36dB less than the broadband levels. And, what's the sample rate? Are you measuring over a 20kHz passband or a 96kHz passband? Without specifying the FFT length and the sampling rate, even a calibrated dBc level for a noise floor is meaningless.
 
V

Vovk Z

Member
Joined 2011
Last edited:
s3tup

s3tup

Member
Joined 2004
s3t.it/audio/
Noise floor should be expressed either in total RMS power in the measured bandwidh, or normalized to 1Hz BW (dBV/rtHz). And still, there is ENBW of window function to be taken into account.
I wonder whether relative noise floor level to the carrier could have any meaning for measurement/display. (expressed as peak Vrms value of carrier versus Vrms/rtHz of noise floor).
By the way, the displayed pixel is a peak value of set of corresponding bins, and the "visual noise level" depends on resolution of the screen as well as displayed frequency range, and is quite non-linear due to log frequency scale... It's funny how a clear picture is soo misleading))


Separate measurement of PSD (power spectral density) should me more or less sufficient to express the noise density. Preferrably either without window function or corrected for it's ENBW (scaling for noise per AP docs)
 
Last edited:
V

vicnic

Member
Joined 2004
Some years ago I made opamps measurement in this RIAA-78 stage schematic (40dB gain at 1kHz)-
https://content6-foto.inbox.lv/albums/v/viccc/PhonoSC/PSsch.jpg
All spectrum measurements was done at 5V RMS 1kHz output signal level.
OPA627-
https://content6-foto.inbox.lv/albums/v/viccc/PhonoSC/OPA627-5V.jpg
NE5532-
https://content6-foto.inbox.lv/albums/v/viccc/PhonoSC/NE5532-5V.jpg
OPA2134-
https://content6-foto.inbox.lv/albums/v/viccc/PhonoSC/OPA2134-5V.jpg
OPA637-
https://content6-foto.inbox.lv/albums/v/viccc/PhonoSC/OPA637-5V.jpg
LM4562-
https://content6-foto.inbox.lv/albums/v/viccc/PhonoSC/LM4562-5v.jpg
All results are in good corellation with the official data sheets.
The NE5532 is good opamp, but not the best.
 
Last edited:
V

vicnic

Member
Joined 2004
EUVL said:
Since you are cascoding, 2SK3557 is a touch lower noise and higher transconductance.
:)
Patrick
Click to expand...
When I designed this schematic, I have searched, what I can really get. Later I got BF862, but not tried yet them in this place.
Anyway the open loop gain of this module is very high: around 120...130dB at 1kHz (LME49720 has around 95dB at 1kHz) , but the noise is low enough.
 
Last edited:
You must log in or register to reply here.
Top Bottom