Low-distortion Audio-range Oscillator

[davada]

Originally Posted by RNMarsh

The 339A is beginning to see the effects of the THD of the opamps used in itself and also the parts and even contacts of the switches. All issues which an audio amp builder would also come to address at these levels of harmonics. The performance will be limited by the parts we use. I am installing sockets everywhere to try different opamps in different places.... will be much easier until the parts are settled as to which are best ones and stable. Thx-RNMarsh

I have the 1468 working fine in the input stage and an LME49710 in the buffer stage. This brought the noise way down on the FFT from the output of the notch filter. We have a max gain of 30dB in the input stage and need a GBP of at least 25MHz. At 30dB gain the LT1468 is 25MHz.

I may have to deal with some RF noise from the oscillator. When the range switch is set to off, which puts a 600 ohm load on the analyzer input, the meter goes to near zero with a -80dBV range. It is garbage getting in. We can filter the output of the the oscillator no problem.

Cheers,

[davada]

Here is the result for replacing the input amplifier with a LT1468 and the buffer with an LME49710. All the hash seen in prior plot is near gone and what remains is power supply harmonics and oscillator distortion.

All the caps found parallel the resistors in the input amplifier gain range stage have to be removed. The LT1468 doesn't like this. I'll work on filtering the input but for this I have to
do a frequency and level sweep.

[Damon Hill]

How much further do we have to go before hitting quantum noise?

I'm not surprised seeing how formerly passive parts are becoming measurable contributors to residual distortion and noise, to say nothing of ambient RFI and increasingly critical layout of parts and grounding to achieve better results at this low level. The CFLs in my work room are becoming an annoyance--the lamps directly radiate stuff around 40KHz and above.

I'm trying to build a copy of the Cordell SV oscillator, using good metal film resistors and even a few bulk metal foil resistors (I have about 90 2K0000 RNC90Ys if anyone's interested--I'll give away free samples). I'd use more foil resistors and 25 ppm or better metal films if I could afford them. And a Nichicon 220 uF bipolar cap for output coupling.

I'm kind of hoping to demonstrate qualification of passive parts using ultra low distortion measurement, a poor man's Radiometer CLT-1 or the Danbridge equivalent:

Danbridge

[1audio]

I think you will find that getting the CLT-1 distortion level will be really difficult. The CLT-1 has a low distortion oscillator and a power amp set up to drive a complex filter network to get everything above the second harmonic to -160 dB under load. And then a narrow band filter to see how much HD3 is still present.

After all of that most good parts are at the residual of the instrument. Pots are not. Cheap parts maybe not. The bulk foil parts and wirewound resistors are at the residual as are good metal film resistors. LDR's are not even near the residual but you can see the nonlinearity of those with a conventional analyzer. Looking at the esr of an electrolytic cap may be interesting. I should try that. However at 50 milliohms we will be below the range of one of these things. I should get an assortment of film caps from Richard Marsh to look at and see if there are significant variations to learn from.

[davada]

Originally Posted by Damon Hill

How much further do we have to go before hitting quantum noise?

I'm not surprised seeing how formerly passive parts are becoming measurable contributors to residual distortion and noise, to say nothing of ambient RFI and increasingly critical layout of parts and grounding to achieve better results at this low level. The CFLs in my work room are becoming an annoyance--the lamps directly radiate stuff around 40KHz and above.

I'm trying to build a copy of the Cordell SV oscillator, using good metal film resistors and even a few bulk metal foil resistors (I have about 90 2K0000 RNC90Ys if anyone's interested--I'll give away free samples). I'd use more foil resistors and 25 ppm or better metal films if I could afford them. And a Nichicon 220 uF bipolar cap for output coupling.

I'm kind of hoping to demonstrate qualification of passive parts using ultra low distortion measurement, a poor man's Radiometer CLT-1 or the Danbridge equivalent:

Danbridge

Hi Damon,

It's a particularly bad time of year for RFI where I am. For some reason the intensity RF propagation is greater in the very cold and dry. Where I am it was -31C, -24F last night.

A while back I was setting up a comparator for a zero crossing detector and found a 14.25MHz ringing that just wouldn't go away. I thought it was my circuit. It turned out the Cisco cable modem sitting beside me is emitting a burst signal at that frequency. I measured 30mVpp on the oscope holding the probe about 18" from the modem. In RF terms that's about 29.5 dBmV. If a cable TV system had a leak that strong on an aeronautical frequency they would be given 24 hours to fix it or risk loosing those channels.

In contrast to when the 339a and others where designed consider the difference in the amount of RFI from then to today.

I had to add a large amount of hysteresis to my zero crossing detector to tolerate the environment and that changed my trigger point. Can't win.

Cheers,

[Damon Hill]

Originally Posted by 1audio

I think you will find that getting the CLT-1 distortion level will be really difficult. The CLT-1 has a low distortion oscillator and a power amp set up to drive a complex filter network to get everything above the second harmonic to -160 dB under load. And then a narrow band filter to see how much HD3 is still present.

After all of that most good parts are at the residual of the instrument. Pots are not. Cheap parts maybe not. The bulk foil parts and wirewound resistors are at the residual as are good metal film resistors. LDR's are not even near the residual but you can see the nonlinearity of those with a conventional analyzer. Looking at the esr of an electrolytic cap may be interesting. I should try that. However at 50 milliohms we will be below the range of one of these things. I should get an assortment of film caps from Richard Marsh to look at and see if there are significant variations to learn from.

I don't doubt it will be a challenge, but fun trying.

Reading the application notes on the CLT-10/20, I see that very low resistor Tc corresponds with lowest distortion, as should low voltage coefficient. Good wirewound resistors were virtually unmeasurable, as I'm sure bulk foil resistors are. Apparently it can measure down to -170db or so--impressive!
That must be >some< power amplifier used to drive the test components at 10 KHz.

Gives me a bit of perspective, and perhaps a means for testing my collection of less-premium components just to be sure they're still good. Defects in end caps and other quality issues apparently stand out. I would imagine old carbon track potentiometers are real stinkers.

Going to test my limited collection (may have to rob other projects) of opamps on a simple Wein bridge breadboard before I power up the oscillator. With a AD797, bulk foil resistors and polystyrene caps, I hit my Tek AA501A's residual of .001%, but most of them are 5534s or slightly better. I'm a little surprised the original SV design did so well with the available opamps of the time.

[jackinnj]

Originally Posted by Damon Hill

And a Nichicon 220 uF bipolar cap for output coupling.

Try the Elna Silk caps as well (2, back to back)-- I ran some distortion tests and these did best.

With respect to the CFL's -- "Abandon hope all ye who enter here..."

[RNMarsh]

[QUOTE=1audio;3277457
After all of that most good parts are at the residual of the instrument. Pots are not. Cheap parts maybe not. The bulk foil parts and wirewound resistors are at the residual as are good metal film resistors. LDR's are not even near the residual but you can see the nonlinearity of those with a conventional analyzer. Looking at the esr of an electrolytic cap may be interesting. I should try that. However at 50 milliohms we will be below the range of one of these things. I should get an assortment of film caps from Richard Marsh to look at and see if there are significant variations to learn from.[/QUOTE]

Check with Brian Elliot.. I think he still lives in Stanford area. he did origianl reserach into distortion limits/mechanisms for IBM... then when he retired from IBM, went back to Stanford and worked for HP for awhile (PHD EE Stanford). He did capacitor tests using a bridge setup and found some interesting things... one I think was that on some film caps the thd/harmonics went up with DC bias on them. The MultiCap was best brand he measured. Dont have to duplicate tests if someone else has already done them, is my thinking. Thx-RNMarsh

[RNMarsh]

Originally Posted by davada

Here is the result for replacing the input amplifier with a LT1468 and the buffer with an LME49710. All the hash seen in prior plot is near gone and what remains is power supply harmonics and oscillator distortion.

All the caps found parallel the resistors in the input amplifier gain range stage have to be removed. The LT1468 doesn't like this. I'll work on filtering the input but for this I have to
do a frequency and level sweep.

Now, what does the meter on the 339A indicate? Is it off the zero position? Dropped in several op-amps today (LT1468's). I think I may need another FFT to verify this thing is even turned on and working !! Thx-RNMarsh

[davada]

Originally Posted by RNMarsh

Now, what does the meter on the 339A indicate? Is it off the zero position? Dropped in several op-amps today (LT1468's). I think I may need another FFT to verify this thing is even turned on and working !! Thx-RNMarsh

I take it your results are good?

Cheers,