I am sure it can and will be solved. If A-P can do it so can we. 
Here is what I did tonight ---- I found that you can set the osc with the SW on the XP and then remove the USB cord and the freq stays where it was set. Now no noise/ground issues from the XP. Still same problem.
I also moved the WIN8 A-P computer USB over to the XP computer and ran the A-P software from the XP. Same data. So, it is now less likely from the USB(s).
I also added 25 Ohms in series with supply with 10,000 mfd to ground after the 25 Ohms for extra filtering. no change. Noise changes around here and there but I am focused on the line freq harmonics.
I played around with power cords, isolation transformers and power grounds. No removal of those pesky line harmonics; no change in thier amplitudes.
Tomorrow is another day. I'll try batteries for +/- 15 v ... using 12v batteries (13.8v, n.l.). The current draw is fairly high -- 200mA per side as I recall. But just to see if the line harmonics diminish... if it is the HP power supply. Then on to extending the osc far from every thing else.
-RM
Here is what I did tonight ---- I found that you can set the osc with the SW on the XP and then remove the USB cord and the freq stays where it was set. Now no noise/ground issues from the XP. Still same problem.
I also moved the WIN8 A-P computer USB over to the XP computer and ran the A-P software from the XP. Same data. So, it is now less likely from the USB(s).
I also added 25 Ohms in series with supply with 10,000 mfd to ground after the 25 Ohms for extra filtering. no change. Noise changes around here and there but I am focused on the line freq harmonics.
I played around with power cords, isolation transformers and power grounds. No removal of those pesky line harmonics; no change in thier amplitudes.
Tomorrow is another day. I'll try batteries for +/- 15 v ... using 12v batteries (13.8v, n.l.). The current draw is fairly high -- 200mA per side as I recall. But just to see if the line harmonics diminish... if it is the HP power supply. Then on to extending the osc far from every thing else.
-RM
Last edited:
If the dominant noise source is the SYS-2722 analyzer, and your preamp has lower noise than this, yes. But why do you want to measure distortion of so weak signals?
Samuel
There are many things that it could be helpful.
Testing passive parts of all sorts would be an example. Good vs bad contacts. There are some very large, expensive equipment that measures resistors at a few spot freqs for 3H to -160, IIRC. They are big and clumsy to use and we cant see what happens at different voltage and freqs. And, of course, amplifiers, ADC/DAC, contacts, metals, magnetic field affects, anything of interest, etc.
It will make SOTA amplifier testing easier, also. Combined with a best ADC and notch filter and you have a freq flexible worthy test system that doesnt cost $30,000 dollars. Those ADC used with a notch filter can be reasonably accurate to -110-120. So if the source is better (-130 or more), we can forget about it. Now we need a good variable freq notch filter to compliment the ultra-low distortion variable freq generator. Near SOTA would be within financial reach of a lot more people.
Then there is the challenge and beauty of designing and learning from the experience and use.
Lots of good things to do with it.
THx-RNMarsh
Testing passive parts of all sorts would be an example. Good vs bad contacts. There are some very large, expensive equipment that measures resistors at a few spot freqs for 3H to -160, IIRC. They are big and clumsy to use and we cant see what happens at different voltage and freqs. And, of course, amplifiers, ADC/DAC, contacts, metals, magnetic field affects, anything of interest, etc.
It will make SOTA amplifier testing easier, also. Combined with a best ADC and notch filter and you have a freq flexible worthy test system that doesnt cost $30,000 dollars. Those ADC used with a notch filter can be reasonably accurate to -110-120. So if the source is better (-130 or more), we can forget about it. Now we need a good variable freq notch filter to compliment the ultra-low distortion variable freq generator. Near SOTA would be within financial reach of a lot more people.
Then there is the challenge and beauty of designing and learning from the experience and use.
Lots of good things to do with it.
THx-RNMarsh
Last edited:
I have trouble doing stuff in my head, but isn't -135dB below 3V about 500nV?
Also, if it were 178nV, then I think 1780nA, or 1.78uA through 0.1 ohm would come out to 178nV.
Still quite small.
Cheers,
Bob
Dick,
We both look at the same thing and see different results! I use log scale on my FFTs and you use linear, but for what it is worth the odd order AC line harmonics are dominant in your results. That to me is the signature of transformer distortion. Even order harmonics seem to indicate rectification noise. In my playing increasing the filter capacitance works when you see even order noise and for odd order shielding works better.
Now where the noise is coming from is a different issue. Trying to improve shielding is actually quite difficult. I find even moving a .062" thick piece of steel between the source and the receiver has little effect. Now distance would help if the source of the noise were small. Often it isn't (for example the currents in the power wiring are a very large source size) so moving things about may have no noticeable effect. When the noise source is small and close i expect to see more high order harmonics than you show.
This is where a Faraday cage is supposed to be the most benefit, but not having tried one on odd order distortion I have no useful contribution for that.
What I have started doing is setting the lower limit of the FFT above the harmonics to get a better reading. As noise is constant bandwidth the loss of say 600 out of 20,000 hertz bandwidth ain't too bad.
ES
Last edited:
As the oscillator is tuned closer to the supply noise the noise in the narrow band will increase.
Since the oscillator gain is not flat the noise can't be white. We have a compound effect from this. If the oscillator is tuned exactly on the noise the noise is now partly running the oscillator. The ALC must compensate for this, something it has no control over. It can get quite ugly.
The noise is not nearly at the level you are measuring it at. Change the tuning to 10kHz and the apparent supply noise will drop. I'm using 'supply noise' in a general sense meaning the total of all sources. If you do a sweep of spot frequencies and plot a graph of the change in supply noise you'll have better picture of the amplitude of the source. When I do a high frequency measurement of the SVO the supply noise is vanishing.
Since the oscillator gain is not flat the noise can't be white. We have a compound effect from this. If the oscillator is tuned exactly on the noise the noise is now partly running the oscillator. The ALC must compensate for this, something it has no control over. It can get quite ugly.
The noise is not nearly at the level you are measuring it at. Change the tuning to 10kHz and the apparent supply noise will drop. I'm using 'supply noise' in a general sense meaning the total of all sources. If you do a sweep of spot frequencies and plot a graph of the change in supply noise you'll have better picture of the amplitude of the source. When I do a high frequency measurement of the SVO the supply noise is vanishing.
If the culprit is a magnetic field would it be possible to fold the field away from the oscillator. If you have some speaker parts around you can try moving the speaker around the oscillator at different angles and see if that has any effect on the noise.
I forgot that I have an old tape head degausser around. I could have tried with that.
I forgot that I have an old tape head degausser around. I could have tried with that.
@davada, & RNMarsh -- An old TV degaussing ring is good.
The line-related noise is, in my opinion, magnetically induced. Very difficult to deal with. I am now almost never measuring below 1kHz. Take advantage of any high-pass filtering available in the analyzer.
I have rarely seen any piece of wide-band gear perform worse at 100Hz or lower than at 1kHz, with a few notable exception -- almost always any loss in performance is in fact a measurement error due to poor filtering, induction, statics, etc. In my IG-18 mods, even really crappy magnetic shielding of the power transformer helped a lot.
In the case of this oscillator, since it settles quickly, and as David has described, it is well shielded and decoupled, any LF differences are not oscillator circuitry related and are just coming in from outside. Annoying but not a deal-breaker.
If it can do -115dB THD+noise, pretty much ball game. At -120dB THD+N, I doubt better can be done or at any rate, measured.
The line-related noise is, in my opinion, magnetically induced. Very difficult to deal with. I am now almost never measuring below 1kHz. Take advantage of any high-pass filtering available in the analyzer.
I have rarely seen any piece of wide-band gear perform worse at 100Hz or lower than at 1kHz, with a few notable exception -- almost always any loss in performance is in fact a measurement error due to poor filtering, induction, statics, etc. In my IG-18 mods, even really crappy magnetic shielding of the power transformer helped a lot.
In the case of this oscillator, since it settles quickly, and as David has described, it is well shielded and decoupled, any LF differences are not oscillator circuitry related and are just coming in from outside. Annoying but not a deal-breaker.
If it can do -115dB THD+noise, pretty much ball game. At -120dB THD+N, I doubt better can be done or at any rate, measured.
I can't do it in my head eiter. I used three different web calculators to get there and needed to scale 3 orders of magnitude to be in range of the calculators. It quite possible I missed somewhere. However the point is that its a very small signal and easy to sneak in in the best of experimental setups.
Super Duper osc -
Here is what I am getting now. Where i left off was thinking it is the power supply or EMI.
I changed to battery power and the harmonics dropped.
But only 2-3dB. By-passing the batteries with large caps didnt do anything. It is interesting that if I put my finger on the battery B+ or B-, the 60Hz level increases. What does this suggest?
The best effect was to move the unit further away from most electrical... but probably not far enough anyway.
Below are three sources all set to 10KHz and using same coax cable.... Krone-Hite, A-P and David's. Guess which one is Davide's?
K-H
A-P
Greater shielding is warranted. Steel box, perhaps. If I find one to drop the whole thing inside, I'll see what it can do.
IT's A KEEPER ! 
THx-RNMarsh
Here is what I am getting now. Where i left off was thinking it is the power supply or EMI.
I changed to battery power and the harmonics dropped.
But only 2-3dB. By-passing the batteries with large caps didnt do anything. It is interesting that if I put my finger on the battery B+ or B-, the 60Hz level increases. What does this suggest?
The best effect was to move the unit further away from most electrical... but probably not far enough anyway.
Below are three sources all set to 10KHz and using same coax cable.... Krone-Hite, A-P and David's. Guess which one is Davide's?
K-H
A-P
Greater shielding is warranted. Steel box, perhaps. If I find one to drop the whole thing inside, I'll see what it can do.
IT's A KEEPER !
THx-RNMarsh
Last edited:
I thought you were talking about sources with -100 dBu/dBV *fundamental* level. I don't understand what's wrong with the SYS-2722 notch filter tuning if you're talking about sources with -100 dB *THD* level.
Samuel
Question: Would a steel chassis help in reducing picked up hum current?
Also, I've always wondered, would adding a DC magnetic field onto the case help in reducing magnetic pick up by saturating the magnetic capability of the case? Neodymium magnets are cheap.
OK I see.
I was referring to measuring the harmonics at levels close to the noise floor and.
The notch quality isnt as good at low freqs.
Causes errors in accuracy of the displayed harmonic levels.
Comment was related to the ' as if' tuning becomes less sensitive... in affect. Or the counter or both.
-RM
Last edited: