Low-distortion Audio-range Oscillator - Page 9 - diyAudio
Go Back   Home > Forums > Design & Build > Equipment & Tools

Equipment & Tools From test equipment to hand tools

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
Reply
 
Thread Tools Search this Thread
Old 9th February 2012, 06:06 PM   #81
diyAudio Member
 
thaumaturge's Avatar
 
Join Date: Aug 2011
Location: Utah
Quote:
Originally Posted by dirkwright View Post
Anyone have a source for really low tolerance capacitors? I can barely find any +/-1%. I looked into getting an LCR bridge to match my own, but heck those things are killer expensive now. I can't believe I used to have a GR 1650 bridge that was perfectly fine but sells for something like $800 now!?!?! Even that bridge was only good for +/-1%. I seriously doubt that the capacitance measuring feature on my DMM is worth a crap.
Old Tek Scope timebase plug-ins. Have typicallly 1% or better mylar timing caps. The 5000 series ought to go for a song nowadays.
Doc
__________________
Ne timeas a facie mulierum ea ignorare
  Reply With Quote
Old 9th February 2012, 10:31 PM   #82
coluke is offline coluke  Italy
diyAudio Member
 
Join Date: Mar 2009
Just to clarify my previous post - I've attached couple of spectra from two of my low THD generator prototypes: the first one is based on a SVF ring of NE5534s, a T&H/S&H peak detector and a low cost analog multiplier (AD633) as level controlling element; the second one is a SVF ring too, but with venerable LM318s in place of NE5534s, a 4-phase peak detector and a couple of paralleled AD633 as level controller. Both generators are equipped with integrating ALC loop, and employs 5-10% red WIMA pp boxes (FKP) as timing capacitors. As you can see, both generator outperform the simple APF generator shown in the first post; they have lower THD, lower THD+N, and far better sideband performances. LM318 are definitely not known as ultra low thd opamps, still they are capable of remarkable performances - THD figures are probably entirely due to the EMU0204 used to capture the spectra: both generators are in 0.0001% (-120dBc) range or better - my notch says something near 0.00003% @1kHz, 20Vpp out.

The simple APF generator is very far from these figures, so building it with cost-no object parts doesn't make too much sense, IMHO.

Ciao,

L.
Attached Images
File Type: png SVFGEN01.png (23.5 KB, 598 views)
File Type: png SVFGEN03.png (36.4 KB, 574 views)
  Reply With Quote
Old 9th February 2012, 10:39 PM   #83
diyAudio Member
 
Join Date: Nov 2008
coluke: the graphs look great, so show us the schematics! E
  Reply With Quote
Old 9th February 2012, 10:43 PM   #84
diyAudio Member
 
dirkwright's Avatar
 
Join Date: May 2011
Location: Virginia
I'm confused. Is the one I want to build an SVF type or something else? I don't even understand the abbreviation "APF" (all pass filter?).
  Reply With Quote
Old 9th February 2012, 10:51 PM   #85
mlloyd1 is offline mlloyd1  United States
diyAudio Member
 
mlloyd1's Avatar
 
Join Date: Feb 2001
Location: illinois
if you want low THD, it's tough to go wrong with a SVF (State Variable Filter) based oscillator.
there are comments elsewhere on the board regarding cordell's oscillator from his distortion analyzer from years ago. it's still a good one.

mlloyd1
  Reply With Quote
Old 10th February 2012, 07:09 AM   #86
richiem is offline richiem  United States
diyAudio Member
 
richiem's Avatar
 
Join Date: Sep 2010
Location: Grapeview, WA
Hi all -- just dialed in to the thread, saw some references to my web pages, and got caught up.

I have yet to find a variable frequency oscillator better than the HP 239/339 (these are slightly different in their ALC circuits but have very similar performance), and these are about as simple as good oscillators get, with individual distortion products for 1kHz fundamental below the 0.0001% level.

It is impossible to evaluate distortion of an oscillator whose THD is lower than the THD of your sound card (or THD analyzer). The HP 331 series, for example, has a floor of about 0.01% in the mid-band; the HP339 has a floor of about 0.001%. The EMU 0202 or 0204 line inputs have a residual of about 0.0002% or so at 1VRMS at 1kHz. So if your oscillator is better than these, you won't know it.

My solution is to use an active notch filter that reduces the level of the fundamental without attenuating the harmonics. 60dB attenuation of the oscillator fundamental seems to be enough to increase the dynamic range of my EMU 0204 enough to allow measuring harmonics down in the 0.00003% and lower, with the noise floor as the limit. I generally tune for 80dB fundamental suppression, which is easy with the filter I built.

As Davada mentioned, I'm away from home for an extended period, so have made no progress on the IG-18 conversion to the HP 239 circuit. This is potentially the easiest way to get to very low levels of THD -- lower than 0.001% THD+noise.

I agree that it is likely that a SV (bi-quad) design is overall the best, IF the ALC issues can be solved or improved. I don't know how good the Jim Williams Wien bridge design can be -- I think that ALC issues are it's limit, if really good opamps are used, such as the LT1468. BTW, don't try to put a level control ahead of the LT1010 -- it's output drives the ALC circuit, putting it an an overall loop, so another output amp driven from the 1010 input should be used instead for a level controlled stage.

See my web page at Home -- article index and click the link to Active Twin-T Notch Filter to read about the details. Please note that the circuit drawing does not show the bypass switch for setting the input reference level -- that's a single-pole switch that connects to the input jack and to the U1 amp input. Closed, the switch takes the filter out of the circuit; open, the filter is in the circuit and can be tuned.
  Reply With Quote
Old 10th February 2012, 09:31 AM   #87
PChi is offline PChi  United Kingdom
diyAudio Member
 
Join Date: Jan 2012
Location: Staines, UK
Default More Variable Frequency Oscillator thoughts

Hello,

My opinion is that I agree with Ciao that the way to go is a Sample and Hold or peak detector which theoretically eliminates modulation of signal by the amplitude control. The idea has been well proven.
I read years ago that the Radford Audio Oscillators used this technique.
The Hewlett Packard HP239A uses a peak detector.
The Black Star Audio Oscillator used something similar.

Another opinion of mine is that Ciao is correct that an Analog Multiplier is the most linear gain control element. Cyril Bateman used an SSM2018P. I have tried out an Analog Devices AD534 because I had one. I have a Texas Instruments MPY634 which I plan to use on the PCB that I am designing.

A couple of other thoughts are that I suspect that the low distortion of the HP339 relies on the well matched switched components to keep the gain variations low hence the voltage across the gain controlling FET low hence distortion low. The Black Star used a well matched dual gang potentiometer. Also I have always had problems stabilizing variable frequency integrator based, state variable, designs in the past so have chosen a phase shift arrangement.

I agree that Robert Cordell's Oscillator in the THD Analyzer is good but it is limited to switched frequencies. I have found that a variable frequency Audio Oscillator useful for testing where a loudspeaker low frequency cut off point is, checking loudspeaker drive units and crossover frequencies so low distortion isn't really required.

I think that the various Jim Williams designs are good but the ones that I have seen are fixed frequency.
  Reply With Quote
Old 10th February 2012, 12:21 PM   #88
diyAudio Member
 
dirkwright's Avatar
 
Join Date: May 2011
Location: Virginia
I can't afford an HP 339, and the HP 239 seems to very rare. Hence, I want to build something myself. I wish I understood this stuff better.
  Reply With Quote
Old 10th February 2012, 06:19 PM   #89
richiem is offline richiem  United States
diyAudio Member
 
richiem's Avatar
 
Join Date: Sep 2010
Location: Grapeview, WA
@PChi -- when we're playing in the 1 part per million arena, I've learned to not take anything for granted. No ALC scheme involving rectification (or frequency multiplication) is totally without some distortion contribution. Heavy filtering leads ultimately to stability issues or unacceptably long settling times. Of course, smart designers will find clever solutions. Fluke resorted to quadrature rectification in their high-precision calibrators (5720) which allowed better filtering for a given loop time constant. As you note, the HP 239/339 use a half-wave peak detector and an integrator with adjustable Tau to provide drive to the JFET ALC element. And as Bob Cordell noted, the AC signalsat the ALC JFET's gate and drain need to be *very* small for lowest distortion.

For anyone who wants low THD from an oscillator (0.001% or less), but is not trying to milk mice, then the HP 239 circuit is a winner all around. Although it has switched frequency selection, it also has a vernier to allow tuning between the steps. It has a low parts count, and is easy to recreate, as I hope to show later this year, with IG-18 #3 (see my site).

I look forward to seeing some results for the design you all are working on here, and I offer to make some detailed measurements for anyone who wants to send me their unit, once I get back to my bench.
  Reply With Quote
Old 10th February 2012, 07:02 PM   #90
diyAudio Member
 
Join Date: Sep 2006
Quote:
Originally Posted by richiem View Post
Hi all -- just dialed in to the thread, saw some references to my web pages, and got caught up.

I have yet to find a variable frequency oscillator better than the HP 239/339 (these are slightly different in their ALC circuits but have very similar performance), and these are about as simple as good oscillators get, with individual distortion products for 1kHz fundamental below the 0.0001% level.

It is impossible to evaluate distortion of an oscillator whose THD is lower than the THD of your sound card (or THD analyzer). The HP 331 series, for example, has a floor of about 0.01% in the mid-band; the HP339 has a floor of about 0.001%. The EMU 0202 or 0204 line inputs have a residual of about 0.0002% or so at 1VRMS at 1kHz. So if your oscillator is better than these, you won't know it.

My solution is to use an active notch filter that reduces the level of the fundamental without attenuating the harmonics. 60dB attenuation of the oscillator fundamental seems to be enough to increase the dynamic range of my EMU 0204 enough to allow measuring harmonics down in the 0.00003% and lower, with the noise floor as the limit. I generally tune for 80dB fundamental suppression, which is easy with the filter I built.

As Davada mentioned, I'm away from home for an extended period, so have made no progress on the IG-18 conversion to the HP 239 circuit. This is potentially the easiest way to get to very low levels of THD -- lower than 0.001% THD+noise.

I agree that it is likely that a SV (bi-quad) design is overall the best, IF the ALC issues can be solved or improved. I don't know how good the Jim Williams Wien bridge design can be -- I think that ALC issues are it's limit, if really good opamps are used, such as the LT1468. BTW, don't try to put a level control ahead of the LT1010 -- it's output drives the ALC circuit, putting it an an overall loop, so another output amp driven from the 1010 input should be used instead for a level controlled stage.

See my web page at Home -- article index and click the link to Active Twin-T Notch Filter to read about the details. Please note that the circuit drawing does not show the bypass switch for setting the input reference level -- that's a single-pole switch that connects to the input jack and to the U1 amp input. Closed, the switch takes the filter out of the circuit; open, the filter is in the circuit and can be tuned.
A couple of thoughts/observations here.

First, as with any THD measurements, the oscillator THD performance may very well not be as good at higher frequencies like 20 kHz as at 1 kHz. Unfortunately, most sound cards do not capture the spectra of many of the harmonics of 20kHz because of their limited sample rate.

I have an HP339 and have measured its oscillator THD with my own THD analyzer (which has a floor of less than 0.001% at 20 kHz, lower at 1 kHz), and have not really been impressed by the THD performance of the 339, so I'm surprized at your comments about it being really good (if I read you right).

The issue of the AGC element's distortion is always an interesting one. The tradeoff between using JFETs (which I used) and analog multipliers is always an interesting one. Keep in mind it comes down to a signal to noise ratio versus distortion issue. The analog multiplier is often capable of less distortion at a given operating level, but it often generates more noise than a JFET-based agc circuit. In my oscillator, I am able to operate the JFET at a very low signal level because the noise of the circuit can be kept low. Also bear in mind that the necessary amount of control correction range comes in. If you need a large correction range, you will have to operate the agc element at a higher level or suffer more noise from it. Designing an SVF osillator carefully so that it does not need a large agc correction range can be very helpful in reducing the distortion and/or noise introduced by the agc element and/or the amplitude detector.

Finally, when using a JFET agc element, it is important to use the feedback distortion correction circuit where approximately 50% of the drain signal is fed back to the gate, as used in my THD analyzer oscillator. For best results, the exact percentage of feedback for this should be trimmed for lowest distortion (assuming you can measure it!). This trim I did not do in my original design, but have incorporated it since. Note also that you can get a picture of the distortion contribution from the agc detector and agc element by looking at the output of the agc correction element, which will have a higher level of distortion than at the oscillator output because only a fraction of this signal is fed into the SVF for correction.

It also matters what kind of JFET you are using for the agc element. The one I chose for my oscillator was not by accident. It had a high threshold voltage and a fairly low Rds_on. Many ordinary JFETs will not perform as well.

It is also always interesting to look at which function is creating the most distortion: the agc element or the (rectifier) control circuit that generates the agc control signal. Looking at the harmonic structure of the THD can sometimes help sort this out. It is not always correct to just assume up front that the amplitude detector is the dominant source of distortion in the oscillator, especially if its filter circuit is switched in accordance with the frequency range being used, as the one in my THD analyzer is. This also relates to oscillator settling time, which is more often an issue at lower frequencies. BTW, for this reason, one should always characterize the distortion of the oscillator at the lowest frequency as well. Some HP oscillators actually had a low-distortion switch that traded distortion and settling time.

Cheers,
Bob
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Radford Low Distortion Oscillator Series 2 audiomik Equipment & Tools 21 19th February 2014 11:46 AM
ultra-low distortion audio oscillator geekysuavo Analog Line Level 16 26th March 2013 04:04 PM
Low distortion oscillator? rjm Equipment & Tools 30 4th May 2011 11:45 PM
Can we improve this low distortion sine oscillator ? gaetan8888 Solid State 22 29th March 2009 01:30 PM
Simple, low distortion 1kHz oscillator jackinnj Solid State 4 6th October 2003 04:58 AM


New To Site? Need Help?

All times are GMT. The time now is 11:12 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2