In the way-back, I remember grappling with peaking in filter sections (a cascade of simple pole/complex-pole pairs). I learned that if the low Q sections were placed first in the signal path, higher Q sections were largely protected from overloading amplitudes.
Good question. Probably that will be the case. For this reason I'm using a passive band-pass filter (in order to purify the output signal from DiAna).
Cheers,
E.
I tried it out a few years back, I did not like it for a number of reasons, but it does the job. So if it is a matter of using it for maintenance of this design it would fine. I think a license is required, I am not going to buy one that is for sure. It is possible to use David's license, another unknown.
I could do a re-design, using what I am familar with rather than fighting with a ecad package that I know is sub par by my standards at least. I will only get involved if it is necessary and there is enough interest. I have other designs I could be doing that would generate me more $ to fund my development efforts, in which others show a greater interest in, than having to spending time on this one. Just being upfront.
So in order to getting the design resurrected, someone has to take ownership or it will stall. I can offer my help on the pcb side, someone with more experience should take on the software as that is a long learning curve for me.
RM has one so, he is pretty quiet, no pictures, just re-posting the schematic that we already have. I would like to know what drives the SPI interface?
If we plan on this resurrection, I think a unique thread just for this design should be started and not bury it in this thread any longer? Would be nice to grab all the posts in this thread and put them all together. Okay so who's in?
I have a Vicktors Mickevics oscillator to test out today, so oscillators are on my mind
Last edited:
Did you actually try it yourself? What distortion did you measure an octave above -3db cutoff freq?
and I dont mean a single opamp filter stage. Just the topology which uses C in nfb. The opamp harmonics are also being attenuated.
Passive also works, as I said. I tried both ways.
How much additional atten do you want to reduce 2H/3H? You dont need great amount of atten if you have a fairly clean source to begin with. I used 24db/oct filter. Thats all I needed to push harmonics into the noise.
THx-RNMarsh
Last edited:
and -- if you used a filter with OPA1656 opamps and its -130dbv thd, when further attenuated by its filtering above Fo will get you very very low.
You can also use a variable dual passive notch filter to attenuate 2H and 3H. Did that years ago.
Whats your experience?
THx-RNMarsh
Last edited:
As to the filter distortion, there is a JAES paper in which the author analyzes several active filter topologies, and gives formulae for calculating the distortion:
Bonello, O. J. Distortion in Positive- and Negative-Feedback Filters, J. Audio Eng. Soc., April 1984, Vol. 32, No. 4, pp. 239-245.
According to the paper, I thought that if I have an op-amp with a THD of -180dB at unity gain, and my filter has an excess distortion of 30dB, I'm still left with a net THD of -150dB.
Am I missing something in this estimation?
Regards,
Braca
Bonello, O. J. Distortion in Positive- and Negative-Feedback Filters, J. Audio Eng. Soc., April 1984, Vol. 32, No. 4, pp. 239-245.
According to the paper, I thought that if I have an op-amp with a THD of -180dB at unity gain, and my filter has an excess distortion of 30dB, I'm still left with a net THD of -150dB.
Am I missing something in this estimation?
Regards,
Braca
Why should I try that? Many smarter people than I have done it and documented it, at the AES, IEEE etc. I then use an advanced concept called 'reading'.
Jan
You've missed the point. Conversation is about using LPF to bring down THD level from external SG. If filter does not attenuate distortion, there is no sense to use it. And if there is SG with -180dB THD , its does not need any filtering at all. So all this about improving performance common low price SG to the level of expensive lab equipment, AFAIU.
Regarding Davida circuits, its a miracle to me how
(data sheet Analog Device AD5543) could show such outstanding performance
Last edited:
Well, read some more then. Not all filter topologies are equally as bad for this purpose.
Understand this also...
Bonello, O. J. Distortion in Positive- and Negative-Feedback Filters, J. Audio Eng. Soc., April 1984, Vol. 32, No. 4, pp. 239-245.
THx-RNMarsh
I will get around to taking some pictures maybe next week of Davids SG.
Last edited:
My inclination would be to use a state variable bandpass filter after the oscillator source. The SVF BPF would be tuned accurately to the same frequency as the oscillator source. If we are talking about a fixed-frequency source like Victor's, this can be especially practical and convenient. A Q of 10 or more would provide good differential gain between the fundamental and the harmonics. As I noted before, and Richard has also noted above, the SVF has the advantage of using integrators with the cap right around the op amp.
As compared to an oscillator source made with similar technology, the SVF BPF will have no added distortion from the AGC circuit or the AGC element because there is none. That filter will also tend to reduce the noise of the oscillator source.
As you might guess, I would mate the SVF BPF with a state variable oscillator of almost the exact same topology and tuning components, so matching of the center frequency to the oscillator frequency would be straightforward.
For a switch-based "variable" oscillator with maybe 10 steps per decade, tracking would still be very good. I would probably include an AGC trim for each decade to reduce the required amount of AGC authority, which reduces oscillator distortion from the AGC circuit. Even a switch-based oscillator in a 1-2-5 sequence over several decades would be of sufficient granularity for most measurement applications. This would be relatively easy to implement with a rotary switch or a relay arrangement.
Even slight mis-tracking between the high-Q BPF and the oscillator frequency in this arrangement will cause slight amplitude deviations/errors. For that reason, a second AGC loop can be placed around the BPF and oscillator. Importantly, this loop would NOT control an AGC element in the BPF, but would rather tweak slightly the AGC level reference in the oscillator.
It is easy to nudge the frequency of an SVF oscillator with a control circuit not unlike that of the AGC control circuit. That would permit locking the oscillator to the center of an FFT bin. If introducing the frequency control function into the oscillator increased its distortion a bit (unlikely by much, probably by no more than 3 dB), then the approach of following the oscillator with an SVF BPF becomes even more valuable.
Finally, if tweaking the oscillator frequency introduces enough frequency mis-tracking between the oscillator and the BPF to degrade amplitude stability, the above-mentioned global AGC loop will correct that without introducing more distortion.
Cheers,
Bob
The op-amp I have in mind is the Groner/Polak composite, I already tested in a second-order MFB filter:
Low-distortion Audio-range Oscillator
According to Bonello, the excess distortion is modelled as a multiplication factor to the op-amp distortion at unity gain, so they add linearly in dB.
I'm at the beginning with this, hope to present more measurements soon.
Thanks for the discussion!
Regards,
Braca
Low-distortion Audio-range Oscillator
According to Bonello, the excess distortion is modelled as a multiplication factor to the op-amp distortion at unity gain, so they add linearly in dB.
I'm at the beginning with this, hope to present more measurements soon.
Thanks for the discussion!
Regards,
Braca
Last edited:
In looking over those old threads I realized I never got an answer to my question. Samuel's plots using a notch look like they are using the op-amp feedback Q enhancement with something like the GP op-amp trick. Is that so?
I don't know. I believe that that -180dB plot is actually the result of a 60dB (IIRC) noise gain circuit adjusted for that 60dB.
Jan
Not talking about that, I mean the noise floor does not show the tails of the notch where usually you have to correct for 2nd and 3rd levels. The noise floor is flat almost immediately before and after 1kHz
Lets see, DAC:
https://statics.cirrus.com/pubs/proDatasheet/CS4398_F2.pdf
My understanding, there is no way someone could get better than -105dB out of this IC. Though, digital correction loop is implemented. Next, ADC
https://www.es.co.th/Schemetic/PDF/AK5394A.PDF
Since, software does correction based on integral sum of THD product dac AND adc, it would full you around show much lower THD level than it's in reality. Third element is required, ether ultra low THD oscillator to confirm adc performance, or external audio analyzer.