Low-distortion Audio-range Oscillator

Member
Joined 2004
Paid Member
There are two limits on distortion on these oscillators today: noise and AGC. Noise is best addresses through impedance of the tuning network. However too low and the network takes power to drive with attendant distortion but 600 Ohms at even 7 Volts it a real option with today's opamps.

The AGC seems to be the real limit on distortion. AP had a patent on a digital multiplying DAC for AGC. The patent has expired. I think it would be possible with a fairly simple processor to steer the gain with a digital pot like either the Maxim or the JRC and have the AGC issue become history. It may be possible to use a very simple logic with some of these parts that can be toggled up and down with single inputs. E.G. DS1802 Dual Audio Taper Potentiometer with Pushbutton Control - Overview
 
There are two limits on distortion on these oscillators today: noise and AGC. Noise is best addresses through impedance of the tuning network. However too low and the network takes power to drive with attendant distortion but 600 Ohms at even 7 Volts it a real option with today's opamps.

The AGC seems to be the real limit on distortion. AP had a patent on a digital multiplying DAC for AGC. The patent has expired. I think it would be possible with a fairly simple processor to steer the gain with a digital pot like either the Maxim or the JRC and have the AGC issue become history. It may be possible to use a very simple logic with some of these parts that can be toggled up and down with single inputs. E.G. DS1802 Dual Audio Taper Potentiometer with Pushbutton Control - Overview

I'm planning on trying this with an Mdac as a multiplier.

I also plan to try a high res Dac and sine look up table to provide a stimulus. Capture the osc peak and adjust the amplitude of the stimulus accordingly. The osc filter operates as a resonator - tuning fork. It also takes care of the frequency pulling at the same time.
I'll need some DSP for this. Too many of these are in a BGA. Any recommendations.

A low distortion buffer on the output of an op amp takes care of the loading issue.

Lots to play with here.


Cheers,
 
Disabled Account
Joined 2012
My conclusions are:

I pick up 6dB on AGC topology changes and at least another 6db from oscillator opamp changes. Another few dB's from polar cap changes to bipolar; Say about 15+dB's for simple mods. But you have to start with an Ulra-low thd oscillator to begin with. Gets down to four zeros and a one with care and tweeking a few values. To go better in noise and distortion with either LT1468 or AD797, the circuit Z values have to be lowered with possible add-on output stage or buffer. After that level of change, the composites would have to be used to advance performance.

Thx-RNMarsh
 
Last edited:
Member
Joined 2004
Paid Member
I'm planning on trying this with an Mdac as a multiplier.

I also plan to try a high res Dac and sine look up table to provide a stimulus. Capture the osc peak and adjust the amplitude of the stimulus accordingly. The osc filter operates as a resonator - tuning fork. It also takes care of the frequency pulling at the same time.
I'll need some DSP for this. Too many of these are in a BGA. Any recommendations.

A low distortion buffer on the output of an op amp takes care of the loading issue.

Lots to play with here.


Cheers,

Another solution would be to use a DDS to injection lock the oscillator (similar to what you were thinking, just takes one part). The DDS is not perfect but pretty good: http://www.analog.com/static/imported-files/data_sheets/AD9837.PDF look at page 8. You would need probably 100-500 mV to lock the oscillator. There are others but they are over $5.
 
Disabled Account
Joined 2012
four zeros and a 6 -

While you are working on designs for stablizing the amplitude on the new SVO, I tried another opamp for the oscillator; if compounding for higher openloop and greater feedback can be done to reduce thd, then a higher open loop gain single opamp would improve thd as well. That coupled with lower noise has worked on what are already ultra-low THD generators. The last one was the AD797.

This time I tried the LT1037 opamp for is extreamly high gains (+130dB) and rejection of CMode and is almost as quiet as AD797. The results are as I expected, even lower THD+N and the 2H and 3H are now at -138 and 136db respectfully. Noise floor is even lower, which is interesting. Now approaching what I guessed was -160dB for the AD725D... this opamp combo is at -155dB. THD+N = .00006% at 7v p-p/1KHz/100KHz BW. The newest A-P is between .00003 and .00005%.

[This is from the KH-4402B. Only a slight touch-up of the AVC pot was needed.]

LT1037 KH4402B.jpg


Darn near my goal of -140dB THD+N :)

Thx-RNMarsh
 
Last edited:
AX tech editor
Joined 2002
Paid Member
Don't recall this, and thought I've unscored all of them. What's the number? 4,631,522 is about the MDAC used in the System One for frequency tuning.

Samuel

Yes that's the one they used, it wa a custom chip with an additional on-chip feedback resistor for less drift.
I don't think they ever used one for AGC.

jan
 
Member
Joined 2004
Paid Member
See Hou, The new design of AGC circuit for the sinusoidal oscillator with wide oscillation frequency range, Instrumentation and Measurement, IEEE Transactions on, Oct. 2004. It's not a smart implementation though, it uses an up-down counter approach which results in very slow settling times.

Samuel

I was first thinking of the Maxim part with up and down buttons controlled from a comparator. The steps take 100 mS so it would not be fast but its real simple. If you have a micro Maxim has some resistors with nS response times. There are FETs used as switches inside but their linearity may drop out if its scaled correctly. The DS1802 lists its distortion as .002% at 1V and - 6 dB. If its used to tweak a gain from -1 to +1 dB that may be enough. It would take 7 seconds from one end to the other using the internal clock. It also has zero crossing detection etc. which could slow settling at lower frequencies.

Its a very low power device so you could provide a battery to enable it to maintain settings while the power is off.

Unfortunately the log taper may be a real limitation but Maxim has many other alternatives to look at.