Noise & Distortion
Judging from your rather laborius way of reaching the nether abyss of noise & distortion measurements, it seems to me that my investment in a R&S UPV years ago was a smart move.
Having all harmonics presented to you in a bargarphlike manner, live, as you adjust relevant parameters, really make it easier to tolerate bad weather&temper.
I have had both Boonton 1120 and Panasonic VP7722. The Boonton is a nice toy, but the Panasonic is a real gem, harmonics down to the fifth is available at the touch of a button. I ultimately sold it to Per Abrahamsen, earlier of EC fame, were I believe it still does a good job. Nowadays they can be bought sh for as low as US$ 1K3.
Keep up the good work.
Regards
RMalmin
Judging from your rather laborius way of reaching the nether abyss of noise & distortion measurements, it seems to me that my investment in a R&S UPV years ago was a smart move.
Having all harmonics presented to you in a bargarphlike manner, live, as you adjust relevant parameters, really make it easier to tolerate bad weather&temper.
I have had both Boonton 1120 and Panasonic VP7722. The Boonton is a nice toy, but the Panasonic is a real gem, harmonics down to the fifth is available at the touch of a button. I ultimately sold it to Per Abrahamsen, earlier of EC fame, were I believe it still does a good job. Nowadays they can be bought sh for as low as US$ 1K3.
Keep up the good work.
Regards
RMalmin
Guided tour (although in French) through the AP SYS-2722 (ignore the wrong model number in the title): 18 - Audio Precision System Two Cascade SYS-2522A - YouTube
I think at ~30 min he misinterprets the notch filter circuits as auxiliary generators (but my French hasn't improved the last 10 years, and it has not been good back then
).
Samuel
I think at ~30 min he misinterprets the notch filter circuits as auxiliary generators (but my French hasn't improved the last 10 years, and it has not been good back then
).Samuel
Give me a few days I'll come up with something.
BTW: I'm going to go in a different direction than Brisbois.
At the moment I'm mostly occupied with with a Class D design.
Phase modulation, THD in the low ppm range, approximately constant switching F and very fast recovery from clipping.
Attaced Class D stuff.
Stein
A goal: To have as low of THD +N as possible. That is with lowest noise as well.
Some hypo numbers would be individual harmonics <-140db re 1v rms into >1K -10K and noise as low as possible (?).... about equal to 1K Ohm or less.
And, do it at 3-4 freqs.
There are many osc circuits floating around and several commercial units.... many Greened circuits. But nothing that is low and quiet and multiple freqs. Nor for the non-corporate budget DIY'er.
The oscillator situation reminds me of all the iterations of amp toplogies being done when only a few are really SOTA. Distilled down to the few best topologies, emplimentation and finally the one or two best of the best. The High-End version of an oscillator. This osc should hold everyone for at least the next generation and then some. Bragging rights go to All contributors.
Of course it has to be buildable by others and measured/verified. SIM only wont get it .... no gold star on your forehead.
Thx-RNMarsh
Some hypo numbers would be individual harmonics <-140db re 1v rms into >1K -10K and noise as low as possible (?).... about equal to 1K Ohm or less.
And, do it at 3-4 freqs.
There are many osc circuits floating around and several commercial units.... many Greened circuits. But nothing that is low and quiet and multiple freqs. Nor for the non-corporate budget DIY'er.
The oscillator situation reminds me of all the iterations of amp toplogies being done when only a few are really SOTA. Distilled down to the few best topologies, emplimentation and finally the one or two best of the best. The High-End version of an oscillator. This osc should hold everyone for at least the next generation and then some. Bragging rights go to All contributors.
Of course it has to be buildable by others and measured/verified. SIM only wont get it .... no gold star on your forehead.
Thx-RNMarsh
Last edited:
low distortion sources are always helpful. For inspiration... what did it take to design and then measure this Sony product..... A rare FM super tuner (called the Direct Comparator model) from the same period as many audio oscillators we now use; Check out the thd, noise etc. and those 120dB numbers all over the place:
And, this was 25-30 years ago ! Are we as good today? Who were these unsung hero's? Do we have some hero's here for our most difficult projects?
Will we add R.Moore, R.Cordell, DiMitri and others to those already on board?
Thx-RNMarsh
And, this was 25-30 years ago ! Are we as good today? Who were these unsung hero's? Do we have some hero's here for our most difficult projects?
Will we add R.Moore, R.Cordell, DiMitri and others to those already on board?
Thx-RNMarsh
Last edited:
There is a lot of interesting stuff that goes on when the natural frequency of the oscillator is slightly different than the injection frequency. It is easier to bring a low natural frequency up than it is to lower it.
When you model the injection signal be sure it is not a perfect sine wave, else you may get false low readings. I use that transversal filter out of cmos that you don't like. It is only -60 without much cleanup.
Then if you have a batch of locked oscillators you can parallel them and get less noise.
The problem with lower impedance tuning elements is the capacitors add vibration induced noise.
I find looking at the control signals with an FFT starting around .01 Hz is quite interesting.
A quadrature squared level detector didn't work for me in the mid 70's when it was intended for one of those military funded applications.
I like the idea of using sample and holds in a chain to form a transversal filter. That should allow fast settling and error compensation at the same time.
Another issue is that there are two basic uses for a test oscillator. One is to sweep frequencies and see what happens. For that use I don't like the normal band ranges of 2-20, 20-200 etc. I want 20 and 20K in the middle of the sweep, as that is often the area of interest. That to me means overlapping ranges. 2-20, 5-50, 20-200, 50-500 etc.
The second basic use is as a low distortion source which is what most of the discussion has been about. If you are looking at THD the oscillator is critical. That is why once things look too good I use IM. It is easy to use two tones and have the difference a nice filter distance away. The harmonics of the test oscillators also mostly can be safely picked out and ignored, although there are some combinations that under some conditions may annoy you.
Currently I have my crew playing with microphone preamps. All measure flat from 5 to 22 khz. at a few mv input. The input transformer is the usual limit. The ones with servos drop like a rock around 3 hz. Yet some of the test listeners describe the differences in terms of tonality.
The worst unit has IM distortion products at -80 or so at 2 mv input.
I can certainly produce test curves that all look the same even though there are differences everyone seems to discern and agree on. (The FET class a preamp is "warmer.")
ES
Last edited:
That makes no sense at all. What is a false low reading? Injected THD will appear at the output, instead I would inject a perfect sine wave and get a real low reading.
By coincidence one of the guys at work was experimenting with heat sinks. He found a solid block of polished copper 3x3x.5 in. worked better than a carefully machined, anodized, finned aluminum heat sink. I actually can't figure that one without seeing what he is actually doing.
Last edited:
Having actually done the injection lock thing its a combination of levels and waveform that affect things. The setup I used (into a KH4400) confirms that he waveform has some effect but only to the extent that a square wave will lock at a lower level than a side wave (.1V vs 1V) and its harmonics get through, attenuated about 60+ dB. A 1V sine wave (coupled inside with a high value resistor) will lock the system pretty reliably. Oscillators tend to want to lock, a real problem with measuring high precision oscillators since they will lock to the reference you are using to check them. I'll post the details later (when I open it up to see what I did).
I would use a sound card to inject a sine wave through a big resistor, seems like a no brainer to me everything off of the same clock.
The Direct Comparator
Looks as if Silicon Labs has some more work to do on the Si4770 to do better.
Found this on the net, interesting!!
"From a 1983 US ES catalog:
"[...] Quartz synthesis tuners were inherently more noisy than the best analog models. So Sony produced one of the most highly acclaimed tuner circuits ever invented. The Direct Comparator.
The Direct Comparator improves audible reception because it functions above the audible frequency range. By making the required PLL comparison directly at higher frequency, the D.C. in tuners like the ST-S555ES cuts through the blanket of noise so prominent in FM reception. Result? A spectacular S/N ratio of 92dB in mono and 86dB in stereo."
Something like a HP 11715 test set. Used these to test 8901A/B modulation analyzers and hi-perf. FM tuners. A retirement project is to try to build one of these or find a cheap one.
Looks as if Silicon Labs has some more work to do on the Si4770 to do better.
Found this on the net, interesting!!
"From a 1983 US ES catalog:
"[...] Quartz synthesis tuners were inherently more noisy than the best analog models. So Sony produced one of the most highly acclaimed tuner circuits ever invented. The Direct Comparator.
The Direct Comparator improves audible reception because it functions above the audible frequency range. By making the required PLL comparison directly at higher frequency, the D.C. in tuners like the ST-S555ES cuts through the blanket of noise so prominent in FM reception. Result? A spectacular S/N ratio of 92dB in mono and 86dB in stereo."
Looks like I can pull the oscillator 10% in frequency without issue. It's like a free lunch, the distortion is still only determined by the AGC element and output amplitude like before. You can see the lock-in overcoming the natural oscillation frequency after you turn it on. Upon settling the output is a perfect sine wave at the lock-in frequency, maybe there are some noise sideband issues that don't show up here? This requires more research.
10% is enough to overcome normal component tolerances.
10% is enough to overcome normal component tolerances.
Couldnt this procedure be used with Virtins FFT program to lock a very low thd oscillator to an integer frequency of the FFT so as to use a rectangular window and remove spectral leakage and picket fencing?
Virtins Technology: Turn a PC into Virtual Instrument - Products
FYI -- View attachment 342057
So there you are. Clear now? Good.
[Physicsforums.com/archive/index.php/t-131758.html]
Dick, the comparison was aluminum with fins, anodization, and all the other good stuff. the copper was just a lump. I keep pointing out there is nothing to learn unless ALL the factors are covered.
yes. You are 100% correct. That physics explaination was about 2 lumps of metal.
I would suspect from description that copper is sooo much more efficient that much larger surface radiating area of alum with fins et al is needed to equal copper with same bulk (volume).
If cost is an issue - use alum. If space is at a premium ... maybe copper is better (eg Beryllium copper ?).
-RM
I would suspect from description that copper is sooo much more efficient that much larger surface radiating area of alum with fins et al is needed to equal copper with same bulk (volume).
If cost is an issue - use alum. If space is at a premium ... maybe copper is better (eg Beryllium copper ?).
-RM
Last edited: