Hi,
In this post http://www.diyaudio.com/forums/equi...n-audio-range-oscillator-206.html#post3437562 I ran 96khz fft simultaneous with the latency check. MS Security Essentils turned on and got approx 12uS latency on a 5 year old XP install. No hick ups!
The secret is solid hardware, drivers and most importantly no bloatware.
Uninstall browser tool bars that may have services running in the background. If necessary make exceptions in Antivirus for your audio application. Preinstalled PCs typically have tons of bloatware installed. Uninstall everything not needed and preferably use chosen PC solely for measurements. Be cautious with 'free' applications such as PDF readers etc that typically installs bloat during updates. Spot the opt out check boxes during those updates and uncheck everything that is not needed (likely all).
In this post http://www.diyaudio.com/forums/equi...n-audio-range-oscillator-206.html#post3437562 I ran 96khz fft simultaneous with the latency check. MS Security Essentils turned on and got approx 12uS latency on a 5 year old XP install. No hick ups!
The secret is solid hardware, drivers and most importantly no bloatware.
Uninstall browser tool bars that may have services running in the background. If necessary make exceptions in Antivirus for your audio application. Preinstalled PCs typically have tons of bloatware installed. Uninstall everything not needed and preferably use chosen PC solely for measurements. Be cautious with 'free' applications such as PDF readers etc that typically installs bloat during updates. Spot the opt out check boxes during those updates and uncheck everything that is not needed (likely all).
b
I know you guys are after bigger fish, but many people are reading this thread and also hoping for a simple yet good enough oscillator like the one above. So this result may feed the hope of the less picky ones (like me).
With an AD8597, night light bulb (120V/7W), and 39nF/4K02 RC values I got this result on a 5V P-P output. No notch filter yet. Circuit not encased. Measured with an USB EMU-0404. Good enough for me.
Last night I took my C7 red lamp and grabbed 6 - .033uF film caps out of the same bag and 6 - 1260 Ohm resistors sequentially off the same strip and wired up Mr. Brisbois' oscillator and a mating notch. Using the AD797 here's a quick picture of the results. Out of the box the notch was -46dB and the amplitude was 5.3V rms (~15.7V p-p) at ~4k Hz. There are my annoying Aglient vibration sidebands (without care they contaminate all my vibration sensitive measurements). THD pretty good though, -135dB seconds, -126dB thirds, after fourths in the noise.
On a fanless brick supply the vibration should go way down. This could be a great op-amp rolling opportunity.And BTW this is just a chip and wire on ratshack board.
EDIT - Yup forgot the notch correction, still below -120dB I think, jeesh.
I know you guys are after bigger fish, but many people are reading this thread and also hoping for a simple yet good enough oscillator like the one above. So this result may feed the hope of the less picky ones (like me).
With an AD8597, night light bulb (120V/7W), and 39nF/4K02 RC values I got this result on a 5V P-P output. No notch filter yet. Circuit not encased. Measured with an USB EMU-0404. Good enough for me.
Attachments
There are quit a few ultra oscillators for a single frequency. EDN and other have published them over the years as well as some IC companies. Those are fine for many people, I would think. Victor on eBay is the absolute best value around.... cant beat his work/performance and price.
BTW-- -120dB for harmonics is pretty standard for a really good oscillator, I've found. BUT that is not THD+N. The numbers I get with some of the hardware in existance can be tuned for a single freq extreamly low.... -140 to -150dB individual harmonic range and is done for a wide range of freqs via mods and upgrades. -130dB thd+N can be easily done now... for a single freq. and limited freq range (100-10K) -- NOT C weighted... 100Khz filter BW only. This is really good sh*t.
Thx-RNMarsh
BTW-- -120dB for harmonics is pretty standard for a really good oscillator, I've found. BUT that is not THD+N. The numbers I get with some of the hardware in existance can be tuned for a single freq extreamly low.... -140 to -150dB individual harmonic range and is done for a wide range of freqs via mods and upgrades. -130dB thd+N can be easily done now... for a single freq. and limited freq range (100-10K) -- NOT C weighted... 100Khz filter BW only. This is really good sh*t.
Thx-RNMarsh
Last edited:
There are quit a few ultra oscillators for a single frequency. EDN and other have published them over the years as well as some IC companies. Those are fine for many people, I would think. Victor on eBay is the absolute best value around.... cant beat his work/performance and price.
BTW-- -120dB for harmonics is pretty standard for a really good oscillator, I've found. BUT that is not THD+N. The numbers I get with some of the hardware in existance can be tuned for a single freq extreamly low.... -140 to -150dB individual harmonic range and is done for a wide range of freqs via mods and upgrades. -130dB thd+N can be easily done now... for a single freq. and limited freq range (100-10K) -- NOT C weighted... 100Khz filter BW only. This is really good sh*t.
Thx-RNMarsh
Iko, i'm with you on the less than stratospheric performance, at least for now. At present I'm more ineterested in the learning by doing side of it. If I can get something fairly basic to work respectably well then I'll be happy, I'm always trying to get the inadequate do the impossible
I was intrigued by the phase shift oscillator at the very start of this thread when I first saw it in 2003. I built it and found it worked like everybody here said, OK distortion, no level stability, phase noise, etc. except in my case the distortion below 200Hz was horrible. Thanks to Brian 'Pingrs' for his PCB layout which I cribbed and modified http://www.diyaudio.com/forums/equi...ion-audio-range-oscillator-2.html#post2886307 .
The other week I built a 2nd board with an integrating AGC a la Bob Cordell, etc. This solves the level flatness problem and it settles nicely in few seconds but some of the LF distortion issue remains. The HF range I added will run upto 175KHz but suffers from too much loss preventing me from reducing the LDR control range, which would lower the distortion, I might have to abandon that. It's all built with cheap generic parts, NE5532's on the oscillator, TL072's in the AGC, prabably lousy layout and grounding, so Lots of work to do - it's a good learning exercise.
I know that ultimately this phase shift osc is a dead duck as far as real performance is concerned and I'll eventually replace it with an SVF which wouldn't be any more complex.
Like RNM I think getting the THD+N is important and this oscillator doesn't cut it, it always shows a wide noise skirt.
Well that's my oscillator journey so far. I'd just like to thank all the contributors to this thread - it's been quite the most fascinating thread I've followed on DIYAudio. Many thanks to you all.
Here's how it's doing so far, Asus Xonar DX PCI-E card (CS5361 ADC) the noise spurs at exactly 1,2,3..KHz are from the PC.
Attachments
Last edited:
I'm pretty sure the source of noise in mine is poor interconnects, no case, and the lab supply. I'm not sure, but I might make it manually variable using parts (the variable dual gang capacitor and range setting resistor network) from an old HP 200x wave generator that I have. Could use the case as well, build a low noise dual power supply, and it'll be almost too good for testing tube amps.
The old HP has a pretty high impedance tank circuit. You can get the snr better but you need to boost the signal a lot. That box had 30V normally and the low distortion option set it to about 8V I think. If you could run the oscillator at 100V the SNR would be much better, but the devices will be the limitation. maybe something with the Supertex depletion mode transistors would work?
The spreading at the base of the fundamental has several sources, first the windowing function or the bandpass of a swept spectrum analyzer, second the frequency stability of the components and finally the Q of the frequency tuning network. The power supply noise contributes as does the noise of the amplifiers driving the oscillator by pushing energy within the bandpass of the tuning network. Its not easy to point to a single source for the "phase noise". There is lots of literature on this because it becomes a real issue when building microwave communications systems. They all have multiplied oscillators for references and the noise is multiplied as the frequency is. A 1000X multiplication will increase the spread at the base 1000 times which can make it unusable if its not really good to start with.
The spreading at the base of the fundamental has several sources, first the windowing function or the bandpass of a swept spectrum analyzer, second the frequency stability of the components and finally the Q of the frequency tuning network. The power supply noise contributes as does the noise of the amplifiers driving the oscillator by pushing energy within the bandpass of the tuning network. Its not easy to point to a single source for the "phase noise". There is lots of literature on this because it becomes a real issue when building microwave communications systems. They all have multiplied oscillators for references and the noise is multiplied as the frequency is. A 1000X multiplication will increase the spread at the base 1000 times which can make it unusable if its not really good to start with.
And you lock yours to a reference as do I.
Now raising the selectivity or "Q" of what used to be called a bi-quad and is now know under the alias of a State Variable filter has another advantage.
The gain of what can be the input stage is raised to get more "Q." That makes that stage the one that most effects noise. So just by improving the noise performance of that stage you improve the whole thing. I use parallel low noise amplifiers to improve that.
Now power supply noise is a very long and involved subject. But ganged RC filters can help a lot. (We could also talk about shunt regulators, but everybody is doing that these days.) So maybe we should talk about the regulators that are neither shunt or series!
So it is not very hard to get a bit less base spreading.
Its all covered right here in the fine print. For a single freq within 100-10KHz You/me can fine tune a used commericial low thd oscillator to very low levels. However, the compromise level is the one the makers have to use as they are offering a wider range of freqs than just one. So, the tuning is a setting that gives the lowest thd number across the widest freq range..... which isnt the lowest single freq tune.
Now then.... since ultra low thd oscillators have been pretty much left behind in development and with the newest opamps, thd numbers - both single freq tune or wide range tuned - drop when they are used. Some more than others depending on the basic topology of design.
Right now the tuned up KH4402B and stock Shibasoku generators are the lowest across the freq range. I am still trying others. The osc from Victor are just as low.... but single freq. tuned and a very limited number of freqs available.
Lowering the noise has been a big help as well.
Thx-RNMarsh
Last edited:
I believe there is subtle distinction between a State Variable and a Bi-quad filter. Although they share the same basic circuit structure the SV has LP, BP and HP outputs and take it's input from a different section than the the Bi-quad. Another distinction is The Bi-quad has two LP outputs one inverted from the other.
These distinctions don't seem to be recognized by all and the two filters are often grouped together.
THD
here is the spec for the KH-4402B attached. As you can see it is already very low and upgrading it isnt too hard. In the midrange it is already below -120dB, stock.
[BTW - KH is calling their newest model 4402B as beingfrom below .0005% to below .0002%..... did they just tighten up the spec to be less conservative or did they do some updates/upgrades like I am have done?
Thx-RNMarsh
here is the spec for the KH-4402B attached. As you can see it is already very low and upgrading it isnt too hard. In the midrange it is already below -120dB, stock.
[BTW - KH is calling their newest model 4402B as beingfrom below .0005% to below .0002%..... did they just tighten up the spec to be less conservative or did they do some updates/upgrades like I am have done?
Thx-RNMarsh
Last edited:
Dick, I think Samuel is talking about the noise. At 100K BW and -130dB below 1V rms that's 1nV/rt-Hz. That means that if the resistive components of your frequency determining network are <50 Ohms there is still no known op-amp for your oscillator. Then of course there is no room for the distortion. There is no substitute for long averaged FFT's, this brute force approach is really just academic.
Last edited:
True. But the biggest issue for me is the lack of a few key frequencies, such as 50 or 100hz to do a LF fft up to at least the 50th harmonic, and 19k and 20k for CCIF IMD test (plus some synchronization!)
I had Victor make an 11.025 KHz and 12KHz oscillators for jitter testing. They will need locking to get to ideal sync but they are also fine for a CCIF IMD test. I don't think the lower frequencies would let something with issues slip by. That is a stalled project (I have too many) but I'll post details when its working.