DIY Analog-to-Digital Converter project.Audio measurements tool - Page 6

abraxalito · 31st March 2010, 12:56 AM

Quote:

Originally Posted by ikoflexer

Before seeing this discussion I was thinking of building the attached.

Lest you wield the iron in vain, run a monte-carlo simulation to see how the depth of those notches holds up when the component values don't have 23bit-precise matching. Also 100k resistors at the opamp inputs won't help the distortion performance

iko · 31st March 2010, 01:12 AM

Quote:

Originally Posted by abraxalito

Lest you wield the iron in vain, run a monte-carlo simulation to see how the depth of those notches holds up when the component values don't have 23bit-precise matching. Also 100k resistors at the opamp inputs won't help the distortion performance

Nah, you saved me from wasting my time, thanks so much. Could you please point me to an opamp based notch filter schematic which is not sensitive to well matched RC values? The 100k nasties I can fix.

abraxalito · 31st March 2010, 01:39 AM

Quote:

Originally Posted by ikoflexer

Could you please point me to an opamp based notch filter schematic which is not sensitive to well matched RC values?

If your aim is to get better distortion measurements, you could do a lot worse than Bob Cordell's distortion analyser (full schematics and explanations available at his website). In order to get better than, say a 60dB notch, the filter's centre frequency needs to be auto-tuned within a control loop. Bob's design does this very elegantly, you can just lift his auto-tuned notch filter without building the whole project.

Cordell Audio: Papers: Build a THD Analyzer

If you just want a precise notch circuit which is insensitive to component values, I don't have a clue

jackinnj · 31st March 2010, 12:31 PM

A good notch filter is also sensitive to layout, not just component values!

Texas Instruments has a filter utility which is free to download "FilterPro" -- it will allow you to see the effects of using "exact" values, and resistors from E12 to E192, caps E6 to E24.

I'll say this 'til I'm blue in the face -- the HP3581 has one of the best filter systems you're ever going to see for audio -- I just picked up another HP3581 for $49 a few weeks ago. It's RMS detector is better than any unit you'll find in a hand-held DVM. Of course, unless you're Hercules you're not going to hand-hold a '3581!

iko · 31st March 2010, 01:10 PM

Quote:

Originally Posted by jackinnj

A good notch filter is also sensitive to layout, not just component values!

Texas Instruments has a filter utility which is free to download "FilterPro" -- it will allow you to see the effects of using "exact" values, and resistors from E12 to E192, caps E6 to E24.

Yep, that's the utility I used.

Quote:

I'll say this 'til I'm blue in the face -- the HP3581 has one of the best filter systems you're ever going to see for audio -- I just picked up another HP3581 for $49 a few weeks ago. It's RMS detector is better than any unit you'll find in a hand-held DVM. Of course, unless you're Hercules you're not going to hand-hold a '3581!

What do you mean, option filters? I do have an HP 3581a.

jackinnj · 31st March 2010, 05:31 PM

HP3581A has bandwidth filters of 300,100,30,10 and 3 Hz -- and they have skirts like a crystal filter should. You have to use a slow sweep with the lowest bw xtal filter. The HP3581 can be "automated" if you have a sweep generator and data-acquisition card.

iko · 31st March 2010, 06:02 PM

Oh, but not a notch filter to null the fundamental. Perhaps you're saying use the band filter to isolate and measure individually the 2nd, then 3rd, etc. harmonics? But I could do the same with the hp 3585a, simply choose the start and stop frequency around the harmonic number(s) that I want to measure. Then the dynamic range would apply only to those harmonics. Or am I delusional here?

Frex · 2nd April 2010, 04:40 AM

Hello,

After some advice of Ghg, i have build a twin-T notch filter at 10kHz, to try to measure the THD of the ultra-pure oscillator from the AN67 of Linear Tech.
For that, i use a small shielded box with 0.1% metal resistors and 1nF 1% low drift mica capacitors.After the passive notch filter, i add an OPA134
low THD OPAMP as follower. Is it powered with clean linear +/-12DC supply.
I used Spectraplus for plot the filter response in frequency domain.
As you can see in this first file : 10k_NotchFilter.pdf , the notch filter allow approximately -68dB attenuation at 10kHz.
This performance is very good for non trimmed filter.

Then, i make a measurement of oscillator spectrum with this notch filter inserted between the oscillator and the ADC.
For this purpose, i use my old ADC (AKM AK5394 IC).

You can see the resulting spectrum here : THDwithNotch.pdf
The THD (not THD+N!) measured by Spectraplus is 0.035% !
Only the second and third harmonic are visible, at levels of -155 and -160dBFS (here 0dBFS is 7.07Vrms or 20Vpp).
In green on the same plot, you can see also the notch filter response.

To calculate the real THD, you must add some gain for the harmonics, corresponding to the notch response.
The notch filter attenuate H2 to about 9dB, H3 about 5dB and H4 about 3dB.
The input level at of the signal at notch filter input is -14dBFS, so attenuation at 10kHz is about 68dB.
After correction of the level of harmonics, and correction of the level of fundamental, the recalculated THD is about 0.000039%.
It is corresponding to -168dBFS !

To be sure that the ADC doesn't introduce THD at this input level, i have also build a -70dB passive attenuator.
I have inserted this attenuator between oscillator and ADC, and was made same spectrum measurement (With this attenuator,
the ADC input level become same as notch filter, but all harmonics are attenuated by 70dB).
The resulting spectrum is THDwith70dBAtt_lin.pdf show clearly no harmonics contain.
This allow to say that previous THD measurement with notch filter is valid.

I really think that it will be extremely difficult to build better oscillator than that !

Frex.

Note:
I have send files (including Gerber,bom,schematics..) to DIYaudio user for building this oscillator.
I think that if many people are interested to build it it could be a good idea to make a group buy for the PCB.
The price of it can be very low if 10 or more is ordered.
Please let me your opinion..

grenert · 2nd April 2010, 06:34 AM

Quote:

Originally Posted by Frex

Note:
I have send files (including Gerber,bom,schematics..) to DIYaudio user for building this oscillator.
I think that if many people are interested to build it it could be a good idea to make a group buy for the PCB.
The price of it can be very low if 10 or more is ordered.
Please let me your opinion..

Thank you for the files. If you decide to make boards, I'd be happy to buy one (or two, if that will help you get to ten). It's not every day you get the opportunity to put together a 0.000039% distortion oscillator

The best I've been able to do is about 0.0025% with a simple circuit from Elektor.

iko · 2nd April 2010, 12:28 PM

I'd also be interested in the pcb. How difficult do you think it would be to make the oscillator 1kHz, if possible at all?

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
need an analog-to-digital converter schematic	prorms	Digital Source	0	28th March 2008 08:29 PM
Digital to Analog Converter DAC With Cable	mg16	Swap Meet	1	11th September 2006 06:27 AM
Analog to Digital converter	amirmk	Digital Source	9	26th November 2005 12:50 PM
Differences between Digital and Analog Audio	oscarcasalini	Digital Source	1	7th January 2004 11:32 PM
Digital to Analog converter	condor655	Digital Source	4	15th May 2003 10:02 PM

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving

31st March 2010, 12:31 PM	#54
jackinnj diyAudio Member Join Date: Apr 2002 Location: Llanddewi Brefi, NJ	A good notch filter is also sensitive to layout, not just component values! Texas Instruments has a filter utility which is free to download "FilterPro" -- it will allow you to see the effects of using "exact" values, and resistors from E12 to E192, caps E6 to E24. I'll say this 'til I'm blue in the face -- the HP3581 has one of the best filter systems you're ever going to see for audio -- I just picked up another HP3581 for $49 a few weeks ago. It's RMS detector is better than any unit you'll find in a hand-held DVM. Of course, unless you're Hercules you're not going to hand-hold a '3581!

31st March 2010, 05:31 PM	#56
jackinnj diyAudio Member Join Date: Apr 2002 Location: Llanddewi Brefi, NJ	HP3581A has bandwidth filters of 300,100,30,10 and 3 Hz -- and they have skirts like a crystal filter should. You have to use a slow sweep with the lowest bw xtal filter. The HP3581 can be "automated" if you have a sweep generator and data-acquisition card.

31st March 2010, 06:02 PM	#57
iko diyAudio Moderator Join Date: May 2008 Location: Toronto	Oh, but not a notch filter to null the fundamental. Perhaps you're saying use the band filter to isolate and measure individually the 2nd, then 3rd, etc. harmonics? But I could do the same with the hp 3585a, simply choose the start and stop frequency around the harmonic number(s) that I want to measure. Then the dynamic range would apply only to those harmonics. Or am I delusional here?

2nd April 2010, 04:40 AM	#58
Frex diyAudio Member Join Date: Nov 2009	Hello, After some advice of Ghg, i have build a twin-T notch filter at 10kHz, to try to measure the THD of the ultra-pure oscillator from the AN67 of Linear Tech. For that, i use a small shielded box with 0.1% metal resistors and 1nF 1% low drift mica capacitors.After the passive notch filter, i add an OPA134 low THD OPAMP as follower. Is it powered with clean linear +/-12DC supply. I used Spectraplus for plot the filter response in frequency domain. As you can see in this first file : 10k_NotchFilter.pdf , the notch filter allow approximately -68dB attenuation at 10kHz. This performance is very good for non trimmed filter. Then, i make a measurement of oscillator spectrum with this notch filter inserted between the oscillator and the ADC. For this purpose, i use my old ADC (AKM AK5394 IC). You can see the resulting spectrum here : THDwithNotch.pdf The THD (not THD+N!) measured by Spectraplus is 0.035% ! Only the second and third harmonic are visible, at levels of -155 and -160dBFS (here 0dBFS is 7.07Vrms or 20Vpp). In green on the same plot, you can see also the notch filter response. To calculate the real THD, you must add some gain for the harmonics, corresponding to the notch response. The notch filter attenuate H2 to about 9dB, H3 about 5dB and H4 about 3dB. The input level at of the signal at notch filter input is -14dBFS, so attenuation at 10kHz is about 68dB. After correction of the level of harmonics, and correction of the level of fundamental, the recalculated THD is about 0.000039%. It is corresponding to -168dBFS ! To be sure that the ADC doesn't introduce THD at this input level, i have also build a -70dB passive attenuator. I have inserted this attenuator between oscillator and ADC, and was made same spectrum measurement (With this attenuator, the ADC input level become same as notch filter, but all harmonics are attenuated by 70dB). The resulting spectrum is THDwith70dBAtt_lin.pdf show clearly no harmonics contain. This allow to say that previous THD measurement with notch filter is valid. I really think that it will be extremely difficult to build better oscillator than that ! Frex. Note: I have send files (including Gerber,bom,schematics..) to DIYaudio user for building this oscillator. I think that if many people are interested to build it it could be a good idea to make a group buy for the PCB. The price of it can be very low if 10 or more is ordered. Please let me your opinion..

2nd April 2010, 12:28 PM	#60
iko diyAudio Moderator Join Date: May 2008 Location: Toronto	I'd also be interested in the pcb. How difficult do you think it would be to make the oscillator 1kHz, if possible at all?