I expected it would work for mid band distortion. For sorting good from defective opamps HF distortion isn't necessary.
Your noise gain circuit has a series cap. I checked and that isn't a part of the commonly used circuit. Is there a specific reason for that?
Your noise gain circuit has a series cap. I checked and that isn't a part of the commonly used circuit. Is there a specific reason for that?
If I end up with something satisfying (and I'm positive I will) you'll get Gerbers (or perhaps PCBs).
The large electrolytic? Not sure what you consider the commonly used circuit, but without you'll get a tremendous DC offset. 1 mV at the input makes 1 V at the output with 60 dB noise gain (and 10 V with 80 dB!)...
Samuel
Demian has some good stuff on his Blog; another source (with mods) to obtain an inexpensive very low distortion oscillator.
-RM
-RM
Hello,
Has anyone tried offset balancing the OP-AMP especially the 5534?
This might lower it's distortion.
Has anyone tried offset balancing the OP-AMP especially the 5534?
This might lower it's distortion.
It often does reduce distortion... at expense of increasing dc offset at the output [in direct-coupled designs]
-RNM
-RNM
I am especially interested in whether offset accumulates in a SVO circuit.
All 3 OP-AMPS seem to feed a loop
All 3 OP-AMPS seem to feed a loop
What level of offset? The 1468's always seem to be in the uV region when I've measured them. I wonder if there is a dc servo effect in the ring. The op amps are referenced to ground in an SVO. The multiplier has the potential of introducing offset into the ring.
With all the op amps direct coupled it's difficult to determine which is introducing the offset.
I suppose the oscillator could be stopped and measurement taken.
If the ring is opened the integrators slam against the rails.
It's interesting question.
Last edited:
The DC gain around the loop seems to be unity. The offsets interact but don't grow. Some of the SVO's I have worked with don't even have offset trimmers.
Another experiment is forcing the op amps to operate in class A by pulling the output to one of the rails with a current source.
An integrator's gain increases as the frequency approaches DC and is limited by the open loop gain of the amplifier. I think this would drive the dc offset out like a dc servo.
Could that have a small affect/improvement on the overall distortion out of the oscillator?
-RM
It's something Samuel suggested to me a while back with the 1468's. I was chasing sources of distortion in my SVO. Had a third H that just wouldn't go away. It didn't help for this particular circumstance and I haven't pursued the effect further but it worth looking into.
It may help with the distortion caused from loading of the stages. SVO's have a tremendous amount of loading from multiple sources. the Ring's summing and tuning, the multiplier inputs the AGC inputs and output loading. It's a large demand on the op amps. Buffering the outputs to the AGC and possibly to the multiplier help with this.
Other op amps like the 5534 might do better with class A loading. The 1468 hasn't done well with the op amp tricks. Paralleling or this. It needs more attention in an isolated circuit before trying to apply the tricks to an SVO.
The compensation used on the 5534 in the AP sys1 is interesting. The manual describes it as a two pole compensation.
Last edited:
This brings up another question.
What is the distortion level of the current source.
In operation we have several volts p-p and we are looking for distortion in the -140dB range or lower.
The KH4400 has an actively modulated current source on the key opamp in the oscillator. I figured it was to lower the distortion of the opamp. Jim Williams published a design using a buffer chip on the output of the oscillator to improve the distortion.
I just got a small improvement in the KH4400 replacing the key opamps with LME49990's. The 2nd improved a little and the noise more.
I just got a small improvement in the KH4400 replacing the key opamps with LME49990's. The 2nd improved a little and the noise more.
Nearly irrelevant, as it is driven by a low-Z node. Just make sure it doesn't saturate during normal operating condition (which might rule out JFET solutions).
A pure resistor will usually result in more distortion than an active current source because it causes higher output current variation.
Samuel
Is a group buy for a PCB available for this oscillator?
Is there a group buy for a PCB and parts for this ocillator? If not, are Gerbers
available? Let me know. Take care.
Is there a group buy for a PCB and parts for this ocillator? If not, are Gerbers
available? Let me know. Take care.
Demian --- do you know what the THD and THD+N are on the KH4400 now?
[BTW/FYI - the A-P 2722 is showing me .0003% THD+N (80KHz BW) or -110 dBV using its internal analog osc+analyzer.]
-Richard
Last edited:
What drive level are you testing at? With 80 KHz bandwidth you are mostly measuring noise, unless the harmonics are pretty strong.
I am getting about -112 dB with 80 KHz and 10V drive with the 590AR. The KH4400 (at approx. 7V) is about -109 but there is some significant low frequency noise I need to explore. Both of those get about 2-3 dB better with 30 KHz bandwidth. I don't see any significant harmonics above -135 dB past no. 3 with the 590AR and about 7 with the KH4400.
I am getting about -112 dB with 80 KHz and 10V drive with the 590AR. The KH4400 (at approx. 7V) is about -109 but there is some significant low frequency noise I need to explore. Both of those get about 2-3 dB better with 30 KHz bandwidth. I don't see any significant harmonics above -135 dB past no. 3 with the 590AR and about 7 with the KH4400.
There isn't an oscillator yet to have a design. There are a number of approaches depending on different aspects. It has been mostly teaching/learning about the frontiers of ultra low distortion oscillators.