Low-distortion Audio-range Oscillator

[RNMarsh]

Originally Posted by 1audio

No schematic. Have not found a source. Perhaps we should buy a manual or two.

The FFT requirements are simple since it converts everything to 420 Hz or so, so even the tenth harmonic is only 4 KHz. You could even use an ipod-iphone spectrum analyzer app.

I'd buy one if I could find one (manual). BTW -- here is the patent number for the ShibaSoku distortion analyzer: USA ---4,417,310 Nov.22, 1983.

[jackinnj]

Dick: I have a 4400 and 4402. They have a procedure for adjusting the oscillator (I think Demian sent me the manual) -- to get to this level it's almost mandatory to work on a weekend morning, use as short a cable as possible, and let it warm up for a half-hour or so without changing the frequency. You can watch the THD tick down as it relaxes. If you muck up with the frequency and controls it will still oscillate below 0.001% One of these had a bad power supply regulator and was pretty inexpensive.

[1audio]

Dialing in a 4400 is easy with the Shibasoku. Simply adjust for minimum 2nd. Its almost cheating. They are very good, much better than the specs. I have not explored opamp tweaks. With the Shibasoku generators I just don't feel the need.

I have modified the KH4400 for injection locking. It helps when doing the really high res stuff since the generator has enough drift and phase noise to limit measurements. Here are measurements of my KH4400 using the AKM5394A (and 16M point fft) (locked to my Boonton)

I saw a cheap KH4400 on ebay a few days ago. I don't think there are customers for them at the usual $500+ price range.

[RNMarsh]

Originally Posted by 1audio

I have modified the KH4400 for injection locking. It helps when doing the really high res stuff since the generator has enough drift and phase noise to limit measurements. Here are measurements of my KH4400 using the AKM5394A (and 16M point fft) (locked to my Boonton)

.

From what i've read, -120 is about it for stock used oscillators. A few new ones are better and the old one's can get improved. Below that -120 is possible with a bit of work.... but it looks like you then get into parts selection -- contacts, caps, pots et al. perhaps even changing the C & R in the twin-T for lower noise values. Those parts have to be the best and a lot of trial and error to find them. Eventually, the noise of R's etc get in the way. Fortunately, for audio amps IMO under -100 (.001) under load is quit good and nothing to be ashamed of. Thx-RNMarsh

[RNMarsh]

Interesting situation here --- Two oscillators - the 339A and Victors -- show very low on ShibaSoku.... verified Victors oscillator distortion spec. BUT, the FFT from the new unit QA400 shows both at -100-ish (<1.0 v rms). Now what the heck? -RNM

[dimitri]

Richard, you posted 339A measurements in #1580. Can you please mention data for Victor's oscillator? Can you put a load for Victor's oscillator, say 600 Ohm?

[Telstar]

Originally Posted by RNMarsh

Interesting situation here --- Two oscillators - the 339A and Victors -- show very low on ShibaSoku.... verified Victors oscillator distortion spec. BUT, the FFT from the new unit QA400 shows both at -100-ish (<1.0 v rms). Now what the heck? -RNM

Weird, can somebody borrows you an AP?

[davada]

Originally Posted by RNMarsh

Interesting situation here --- Two oscillators - the 339A and Victors -- show very low on ShibaSoku.... verified Victors oscillator distortion spec. BUT, the FFT from the new unit QA400 shows both at -100-ish (<1.0 v rms). Now what the heck? -RNM

Hi Rick,

Are you using a notch filter to suppress fundamental? If not then you are measuring the qa400 distortion.

I had the same trouble measuring Victor's oscillator. Reads high on the QA400.

On the other hand is the Shibasoku distortion unit calibrated?
You can test this by injecting known levels at harmonics of and along with a fundamental.
Use two 600 ohm oscillators in series or parallel. Set fundamental to 1Vrms and vary the harmonic level at different frequencies.

You can do the same for the QA400. Remember you have to do the math for any THD readouts on either but individual harmonics should be fine.
The injected harmonic(s) must be well above oscillator distortion to be meaningful.

Be aware that all windowing of FFT produces ripple in the pass band to some degree. Flat top is the only one that doesn't but the noise is higher.
Read the help file.

[HpW]

Well,

Originally Posted by davada

H
Be aware that all windowing of FFT produces ripple in the pass band to some degree. Flat top is the only one that doesn't but the noise is higher.
Read the help file.

It really depends of the used FFT Window... there are low and higher order functions even on the Flat top...

What really surprises (on the given spectrum) the repeated inter-modulation, more or less, over the full measured bandwidth. Interesting would also to zoom on the fundamental to have to phase noise..

HpW

[1audio]

Originally Posted by RNMarsh

Interesting situation here --- Two oscillators - the 339A and Victors -- show very low on ShibaSoku.... verified Victors oscillator distortion spec. BUT, the FFT from the new unit QA400 shows both at -100-ish (<1.0 v rms). Now what the heck? -RNM

You are bumping into the Crystal ADC's limits.

I have attached a plot of the same generator as above (KH4400) measured with the QA400. You can see the difference in the distortion floor. Other key tricks, you must get the frequency to be precise or the bins will miss. That's why I added the injection lock to the KH4400. Make sure the internal generator matches the frequency since that is what their software uses to determine the harmonics.

The QA400 software is a good start and will get useful results immedately. However if something like HPW-works were to support the hardware I would find that very helpful (hint!).