Maybe. ShibaSoku has about a dozen large pcb cards filled with circuitry plugged into a mother board. I can see down to -160dBv. Davada's generator goes into the mid -140dbv area and is variable freq.
I was looking at a simpler way for a potential ULD generator. Digital plus smoothing.
THx-RNMarsh
Last edited:
Here is the distortion of the source side of the RTX 6001 via the 725D. Its scaled so the levels are corrected. HD2 is -133, HD3 is -126.
While cancellation seems like a good trick an analog filter may be able to get there with less hassle. This is at about the best level (-15 dBFS). You have about +/- 3 dB before the performance is degraded. So visions of digital level etc. are probably not productive. Ideally a passive filter would be a great solution. J Curl has suggested just putting a good cap across the output will get 6 dB at F2 and 12 dB at F3. The CLT-1 uses a passive filter with inductors to get to -160+ (mine seems to get to as much as -170 dB) but those are not off the shelf inductors. They a big pot cores and it runs at 10 KHz only.
P.S. The ADC/Analyzer side of the RTX is lower distortion than the source, but I wanted to remove any questions about the measurements.
While cancellation seems like a good trick an analog filter may be able to get there with less hassle. This is at about the best level (-15 dBFS). You have about +/- 3 dB before the performance is degraded. So visions of digital level etc. are probably not productive. Ideally a passive filter would be a great solution. J Curl has suggested just putting a good cap across the output will get 6 dB at F2 and 12 dB at F3. The CLT-1 uses a passive filter with inductors to get to -160+ (mine seems to get to as much as -170 dB) but those are not off the shelf inductors. They a big pot cores and it runs at 10 KHz only.
P.S. The ADC/Analyzer side of the RTX is lower distortion than the source, but I wanted to remove any questions about the measurements.
Attachments
To put this into perspective there is probably no recording or reproducing chain with enough performance that this setup cannot evaluate its total performance. Certainly, no analog chain and I don't know of a digital chain at this level. So while better performance is a good thing it's not going to reveal much.
doesnt matter so much to me. Its interesting electronics and a challenge to do the best one can.
HP et al they do the best T&M equipment they can, also.
I might have been OK with the HP339A (-100dbv) except the HPA was at its' residual. So I/we tried to make it better just so i can see the true HPA distortion. But eventually had to go to other gear - various Panasonic models, Audio Prec and all sorts of gen brands and their mods. Finally with 725D and QA400/401 I got low enough to read what the HPA was doing. Now maybe the HPA doesnt need to be that good. But one should be able to measure what we design/build.
Now, I would like to see what the best of the best for an affordable DIY'ers can be -- do as well but simpler and smarter designs and lower cost.
Maybe.... a LP filter - fixed and variable. Sharp enough cutoff to significantly atten 2H. Passive and active. As I showed last year -- such a sharp cut-off filter will make a modest generator's performance seem great. What can it do to a great generator? I proposed opamp type circuitry with C in FB path to atten amp harmonics as well.
-RM
HP et al they do the best T&M equipment they can, also.
I might have been OK with the HP339A (-100dbv) except the HPA was at its' residual. So I/we tried to make it better just so i can see the true HPA distortion. But eventually had to go to other gear - various Panasonic models, Audio Prec and all sorts of gen brands and their mods. Finally with 725D and QA400/401 I got low enough to read what the HPA was doing. Now maybe the HPA doesnt need to be that good. But one should be able to measure what we design/build.
Now, I would like to see what the best of the best for an affordable DIY'ers can be -- do as well but simpler and smarter designs and lower cost.
Maybe.... a LP filter - fixed and variable. Sharp enough cutoff to significantly atten 2H. Passive and active. As I showed last year -- such a sharp cut-off filter will make a modest generator's performance seem great. What can it do to a great generator? I proposed opamp type circuitry with C in FB path to atten amp harmonics as well.
-RM
Last edited:
Hi Richard,
-Chris
Completely agree with you on this.
Which is my motivation and also most of the others people around here.
Again, right along the same path where we do the best we can. These efforts should be supported by the community as they benefit us all. We have already seen some amazing performance achieved by some of our members.
-Chris
That filter has some serious limitations, starting with the common mode non-linearity. (Been there, done that. . .) At these levels you need to use inverting amplifiers. Those bring noise issues. A state variable filter may work but its the same as a state variable oscillator with no AGC so not necessarily a high value option. The oscillator may well have less distortion.
I'm not sure as a filter that it shows up in the output (above -3db corner). Of course there are other variations. I dont care about the noise much (FFT).
is there a better LP filter topology for this app.?
If I was at my home lab, I'd bread-board it fast and test.
-RNM
Last edited:
As Demian writes, a state variable filter will probably be a better option.
I have done some simulations and it looks promising. I plan to make a test board to verify the simulations and check the distortion (which cannot really be simulated with the op-amp models available).
The big question is whether it should be tuneable over the audio band or a few discrete frequencies would be sufficient?
I have done some simulations and it looks promising. I plan to make a test board to verify the simulations and check the distortion (which cannot really be simulated with the op-amp models available).
The big question is whether it should be tuneable over the audio band or a few discrete frequencies would be sufficient?
test board is best way IMO also.
have you or others tried this or similar --
http://www.analog.com/media/en/technical-documentation/data-sheets/1562fa.pdf
UAF42 Universal Active Filter | TI.com
ETC
These are variable analog filters.....would it be too much to ask that such LP filter be sync'ed to osc freq ?
THx-RNMarsh
have you or others tried this or similar --
http://www.analog.com/media/en/technical-documentation/data-sheets/1562fa.pdf
UAF42 Universal Active Filter | TI.com
ETC
These are variable analog filters.....would it be too much to ask that such LP filter be sync'ed to osc freq ?
THx-RNMarsh
Last edited:
My gut would say I'd rather be confident in a couple frequencies than less confident across a wide spectrum.
At these levels though, we're obviously in empirical territory for the last n'th in distortion.
No, I haven't tried those.
As far as I can see, the performance of the op-amps are not as good as the regular high-performance types like OPA1612, LME49720, LME49990 and AD797 etc.
E.g. the op-amps in the UAF-42 have GBW of 4 Mhz and input noise of 10 nV/rtHz.
The LTC1562 seems to be rather noisy and with a limited supply voltage range.
They are probably OK for less stringent requirements, especially if the number of external components must be kept at a minimum.
These look to me to be an old National part that wasn't very good, noisy and too much distortion.
Why not and 8-Pole Active Low-Pass Filter Optimized for Precision, Low Noise, and High Gain.... Low noise 0.2uVp-p (0.1Hz to 10Hz) ain't bad. in CN-0127: LINK
There was also another App Note or Circuit Note, et.al. that I came across and it intrigued me. I saved it. But I'll be damned if I can find it again.
And, here is a good discussion on the "matter": LINK
If I know more about programming, we could use my account with the good folks that allow the use of their quantum computer to crank out some really unique stuff.
I had a liking to FORTRAN.
I see I can get more use outta that 725D. And it's fast too, thanks Demian, Richard, Davada.
So if one also want to measure the digital side of things, from the basics of SPDIF, AES, etc... then where do we head off to? How to interface the Digital to the analog equipment we have? Or do we have to buy a whole 'nuther set of test and measurement equipment?
Then I found the .35um CMOS process, with power consumption of the filters at 7.21nW, while the THD level was 4% for an input signal of 220pA. The RMS value of the input referred noise was 0.43pA. And it has a 0.5V supply.
There was also another App Note or Circuit Note, et.al. that I came across and it intrigued me. I saved it. But I'll be damned if I can find it again.
And, here is a good discussion on the "matter": LINK
If I know more about programming, we could use my account with the good folks that allow the use of their quantum computer to crank out some really unique stuff.
I had a liking to FORTRAN.
I see I can get more use outta that 725D.
And it's fast too, thanks Demian, Richard, Davada. So if one also want to measure the digital side of things, from the basics of SPDIF, AES, etc... then where do we head off to? How to interface the Digital to the analog equipment we have? Or do we have to buy a whole 'nuther set of test and measurement equipment?
Then I found the .35um CMOS process, with power consumption of the filters at 7.21nW, while the THD level was 4% for an input signal of 220pA. The RMS value of the input referred noise was 0.43pA. And it has a 0.5V supply.
You and I are on the same page..... a complete filter in one IC. You found a better one. maybe more out there. Thats what the "ETC" means.
I am asking ..... the distortion -- how much does it matter above the filter's cutoff?
and the noise..... in doing FFT not thd+n measurements. How much does the noise matter then? I think your refer might be just fine.
I'd try the all in one packed filters just to get answeres on those two. Then, moving forward... maybe seperate opamps... if really needed for this app.
But we can also do it with passives. large core size (L) or not depending on C size/value. Large value film caps of PP dielectric are not hard to get now. 100mfd are available for spkr xovers from Solen. I have boxes full of 10mfd from REL also to try. Just limiited in number of them to make. Whereas, opamp based would be more flexible.
Lots of ways to try LP filtering of higher orders to reduce source gen distortion to sota levels which will not be likely challenged for a very long time.
THx-RNMarsh
Last edited:
That's another possibility.
What is a clean uncluttered way way to actually try to do it?Or something like this? YouTube
-RNM
Last edited: