Which topology for a low-pass filter would give the least thd/harmonics from the opamp used, on the output? Something with a cap on last stage from output to neg input. I presume??
THx-RNMarsh
THx-RNMarsh
I maintain an almost militant pro-activeness towards unsubscribing from any marketing list that isn't extremely relevant to what I'm doing (read: almost nothing escapes my wrath).
And why are we re-covering old grounds wrt capacitor materials? It's the same lines trotted out again and again. In most cases for the capacitor materials in question, used for AC coupling, it requires a contrived situation to drive them into substantial distortion. Worth sweating the last nth when you're working on test equipment? Yes. Worth sizing up voltage-wise to keep things polite in the audio regime? Sure. Hand wringing past that? No.
High K dielectrics, albeit not super linear, are nice in PSU bypass to maintain low Z into MHz, not in coupling. Size accordingly to their voltage deration. That way we can build circuits and used universally hated IC's that have near perfect integrator GBW out into the 10's - 100's of MHz and the attendant gigantic amounts of NFB throughout audio bandwidth. 😉
They are space savers in SMPS design, and as you point out, have excellent HF ESR in chip form. I am building a Li charger prototype and it would not be possible given the size constraints without them. It is just important to remember the Cv characteristic and size accordingly. I try to fit the highest voltage rating/case size I can given the size constraint to minimize the de-rating of C when it is critical.
On a related subject, building SMD protos has become one of my least favorite jobs...of all the things of youth I miss, good vision is one of the most missed. I get pretty tired taking glasses on and off, looking through the stereo microscope to even see the 0603 chips...I tried a USB microscope, but that was worse...I want bionic eyes like the million dollar man...where's my Excedrin?
Howie
I don't the cause, but possibly some manufacturer's NP0/C0G formulations differ -
Bruce Hofer was surprised when I mentioned my measurement in one of his Analog Master Class seminars
for dominant pole compensation with >30 dB loop gain at 20 kHz it would seem to be a non problem
but 2-pole wins again - much less audio band contribution of compensation cap's nonlinearities
Bruce Hofer was surprised when I mentioned my measurement in one of his Analog Master Class seminars
for dominant pole compensation with >30 dB loop gain at 20 kHz it would seem to be a non problem
but 2-pole wins again - much less audio band contribution of compensation cap's nonlinearities
Sam's first conclusion was they don't distort. After his oscillator adventures I am aware the two of you did some more accurate measurements, but still using a reference capacitor. No I have not read the article. I have been using my techniques to select parts.
The beauty of a reciprocity test is you can continue to improve accuracy.
I have cleanly measured the decrease in distortion of Mylar capacitors before and after desiccation. I can measure the difference between polypropylene capacitors rated 50 volts or 100 volts DC. Higher voltage capacitors have less distortion with all else the same.
C0G capacitors can be quite good but there are better.
The second issue is of course sensitivity to vibration. Mounting position on some capacitors can change results.
Last edited:
DPH, I reintroduced the subject of cap non-linearity, because several contributing members of this website have ignored its consequences and happened to mention my name in the process. One individual actually regularly contributes to THIS thread as well, so I thought it should be reviewed. A single high Q ceramic cap might have more inherent distortion than the rest of a following IC circuit. Is this rational?
Last edited:
Not really but they are so low at line levels who cares? Destroyer X put me onto a very interesting guy, huge body of experience, spends an inordinate amount of time critically listening and collecting LP's, and thinks 15W SET's and horns are basicly the SOTA. BTW I still think you are using the term reciprocity improperly.
This really happened today. I was walking past the TV wall at COSTCO and simply noticed one screen going from one mode to another, it was an on/off Dolby ATMOS demo when I went and looked. Turn it off is all I said! I'm not sure what I'm going to do if my TV's die. I guess stock up like the guys that stashed Toshiba FET's away.
This also brings back an important experience in my life, that changed my job prospects and directly led me to be a better audio designer.
Back in 1970, I was hired by a company called Alembic to design a solid state PA mixing board for the Grateful Dead. I didn't have to work alone, and I was largely in charge of the active parts of the board. I did not have to put it together or select the passive parts.
Since this board would be IC based, I chose a really good one (also expensive) $5, about an engineer's hourly wage each. It was the Harris Assoc (formerly Radiation Inc) HA909 or HA911 (I don't remember if there was any difference between the two numbers. I had earlier used this IC as a common mode servo at Ampex Research for a balanced bridge power amplifier, and it had several advantages. It was about 5 times faster than the uA741, quieter, able to drive 50ma, and could operate at +/- 24V. I really thought that I had found a winner, even if up subjectively against open loop single tube/channel mixers (Ampex) then used by the GD for years.
This is where things fell short:
My associate (boss) innocently bought a number of 2.2uF CERAMIC CAPS for coupling. These were the ugliest caps of that kind that I had ever worked with, but what the heck, they were small, relatively cheap to buy, and available. Later, I tried prettier 2.2uf ceramic caps and they were about the same quality inside.
Well, the PA board was built, tested, and put on the road with the GD. While on the road, the mixing engineer, attempting to improve the sound quality of the PA board, put it on batteries, but unfortunately reversed the power supply polarities and blew up the board, so the GD had to return to the Ampex tube mixers that they had in reserve. Then they returned it to us, saying that it sounded poorly. Within weeks, I was laid off, and left to question: WHAT HAPPENED?
I immediately thought it might be the use of heavy negative feedback in the IC's, after reading Otala's first IEEE paper on the subject. This is when I pushed myself to make my own discrete OP AMPs that lead to the complementary symmetry jfet input stage that I still use today. However, that was only PART of the problem. The BIGGEST problem was the use of the 2.2uF CERAMIC caps in the studio board, but they apparently passed my original test measurements, how? Because I had only a 60Hz based IM analyzer to test distortion with. The caps did not act up much with my test, so it passed by me.
Two years later, after developing the Levinson JC-2 level of circuits, first for myself, then the GD, and finally Levinson, I succeeded with the GD to make solid state electronics that they liked.
It is interesting that independently Mark Levinson used the SAME IC in his own mixer board, originally designed by Richard Burwen, and the GD got to try it. It sounded relatively OK, so how did he get away with it? Well, the BIGGEST problem was the 2.2uF Ceramics that we(Alembic) used, rather than the 2uF film caps that Mark Levinson used. [And you guys wonder why I constantly bring up things like this '-) ]
Later Mark compared my discrete designs to the IC module made by Burwen and changed over entirely to my discrete designs, based on subjective listening, so the IC was not entirely out of the picture, although I blamed it initially for just about everything.
So that's the story, you decide what to do.
Back in 1970, I was hired by a company called Alembic to design a solid state PA mixing board for the Grateful Dead. I didn't have to work alone, and I was largely in charge of the active parts of the board. I did not have to put it together or select the passive parts.
Since this board would be IC based, I chose a really good one (also expensive) $5, about an engineer's hourly wage each. It was the Harris Assoc (formerly Radiation Inc) HA909 or HA911 (I don't remember if there was any difference between the two numbers. I had earlier used this IC as a common mode servo at Ampex Research for a balanced bridge power amplifier, and it had several advantages. It was about 5 times faster than the uA741, quieter, able to drive 50ma, and could operate at +/- 24V. I really thought that I had found a winner, even if up subjectively against open loop single tube/channel mixers (Ampex) then used by the GD for years.
This is where things fell short:
My associate (boss) innocently bought a number of 2.2uF CERAMIC CAPS for coupling. These were the ugliest caps of that kind that I had ever worked with, but what the heck, they were small, relatively cheap to buy, and available. Later, I tried prettier 2.2uf ceramic caps and they were about the same quality inside.
Well, the PA board was built, tested, and put on the road with the GD. While on the road, the mixing engineer, attempting to improve the sound quality of the PA board, put it on batteries, but unfortunately reversed the power supply polarities and blew up the board, so the GD had to return to the Ampex tube mixers that they had in reserve. Then they returned it to us, saying that it sounded poorly. Within weeks, I was laid off, and left to question: WHAT HAPPENED?
I immediately thought it might be the use of heavy negative feedback in the IC's, after reading Otala's first IEEE paper on the subject. This is when I pushed myself to make my own discrete OP AMPs that lead to the complementary symmetry jfet input stage that I still use today. However, that was only PART of the problem. The BIGGEST problem was the use of the 2.2uF CERAMIC caps in the studio board, but they apparently passed my original test measurements, how? Because I had only a 60Hz based IM analyzer to test distortion with. The caps did not act up much with my test, so it passed by me.
Two years later, after developing the Levinson JC-2 level of circuits, first for myself, then the GD, and finally Levinson, I succeeded with the GD to make solid state electronics that they liked.
It is interesting that independently Mark Levinson used the SAME IC in his own mixer board, originally designed by Richard Burwen, and the GD got to try it. It sounded relatively OK, so how did he get away with it? Well, the BIGGEST problem was the 2.2uF Ceramics that we(Alembic) used, rather than the 2uF film caps that Mark Levinson used. [And you guys wonder why I constantly bring up things like this '-) ]
Later Mark compared my discrete designs to the IC module made by Burwen and changed over entirely to my discrete designs, based on subjective listening, so the IC was not entirely out of the picture, although I blamed it initially for just about everything.
So that's the story, you decide what to do.
I don't recall Dick designing any IC's, lots of bricks and hybrids maybe. Don't think anyone is around who could sort this out anymore. How much does it add to the body of knowledge now going over how uA741's and other op-amps designed on the original planar process with essentially no PNP's don't work for audio?
Scott,
Yes Sam initially concluded it was low not zero distortion.
https://www.merriam-webster.com/dictionary/reciprocity
Yes Sam initially concluded it was low not zero distortion.
https://www.merriam-webster.com/dictionary/reciprocity
Yes mutual as in mechanical/electrical in a microphone, mechanical/electrical in a motor generator system. Explain to me your interpretation, the distortion of a capacitor I/V characteristic exists in isolation, if we could instrument to measure it we would. The bridge setup in both cases removes confounders, Samuel estimated -170dB in our case.
Last edited:
Scott, Dick Burwen USED the 911 IC, he didn't design it. He would select out from a number of 911 IC's and put them in a HYBRID MODULE with a few surrounding passive that he sold to Mark Levinson.
The Harris 911 had a true PNP internal transistor. Check it out, this IC was made with an alternate process, that is WHY it was so expensive at the time. It was made from radiation sensitive environments. Please check things out first. The HA911 was a pretty darn good part, sometimes. It also often suffered from Xover distortion, but if you selected them out, they were reasonably OK.
The Harris 911 had a true PNP internal transistor. Check it out, this IC was made with an alternate process, that is WHY it was so expensive at the time. It was made from radiation sensitive environments. Please check things out first. The HA911 was a pretty darn good part, sometimes. It also often suffered from Xover distortion, but if you selected them out, they were reasonably OK.
Who is using those grades of ceramic caps for anything but power supply decoupling/bypass in audio equipment? I haven't seen it except maybe in cheap phones or MP3 players.
For power supply decoupling, especially for digital ICs, you aren't going to beat the combination of low inductance (package size) and capacitance.
Scott,
My setup has always used a bridge. It is two series pairs in parallel vs a single on one side and a single feeds the other side's pairs.
Reciprocity is the exchange of "equals". You can move the parts around the circuit to see how that changes each measurement. When used in the single position that weights more than when one of four.
In practice I set up nine parts and try the rest of the batch in the open position. I then can use the best of the batch for the nine reference units.
As to accuracy I can match parts easily to .1% and then look 130 dB below that with an FFT based analyzer. Even lower when the test signal is synchronized to the FFT and lots of averaging is used. So -170 is quite good it is not hard to go even lower.
No magic involved.
Of course for resistors I used a trimmer as opposed to matçhing the parts under test.
Here is the filter, from the JFET, 2.2Vac, 22.5 (center frequency)
to the Bandpass filter. 15kHz - 30kHz . The JFET is a 2N5485.
I set its gain at 22, so with 100mV input, I get 2.200 V output
throughout 10Hz to 100k Hz.
Yes I swept the freqnency from 5kHz - 45kHz
it's either rising or flat through the full sweep
though the bandpass filter.
Ed are you saying I could/
should add that 8.2Meg Ohm
resistor at the bottom left of the board?
One end to the output connection.
(The end of the 5k3 series R
[connected to meter, scope, Spec Analyser,])
and the other end to of 8.2MEG R to ground.
The ground ties the 8.2, .001ufCap, and 100K
resistor to ground.
I think I did that once and it made no difference.
I can try again.
to the Bandpass filter. 15kHz - 30kHz . The JFET is a 2N5485.
I set its gain at 22, so with 100mV input, I get 2.200 V output
throughout 10Hz to 100k Hz.
Yes I swept the freqnency from 5kHz - 45kHz
it's either rising or flat through the full sweep
though the bandpass filter.
Ed are you saying I could/
should add that 8.2Meg Ohm
resistor at the bottom left of the board?
One end to the output connection.
(The end of the 5k3 series R
[connected to meter, scope, Spec Analyser,])
and the other end to of 8.2MEG R to ground.
The ground ties the 8.2, .001ufCap, and 100K
resistor to ground.
I think I did that once and it made no difference.
I can try again.
Attachments
Hey John,
Good experience with the Alembic etc. I've got one of their Basses
with and active P/U. I also have one of their first series of bass pre amps.
(An old Red Knob job made in The City, by the man himself.)
I wonder if "Window Pane Willie" was the guy
who hooked up the power to the board wrong....
I think there were some of those Burwen discreet op amps for sale
again recently...(like the last few years that is.) I was wondering
how the discreet types compare against the good grade of
Audio DIPs...or now SOICs.?
Cheers,
Good experience with the Alembic etc. I've got one of their Basses
with and active P/U. I also have one of their first series of bass pre amps.
(An old Red Knob job made in The City, by the man himself.)
I wonder if "Window Pane Willie" was the guy
who hooked up the power to the board wrong....
I think there were some of those Burwen discreet op amps for sale
again recently...(like the last few years that is.) I was wondering
how the discreet types compare against the good grade of
Audio DIPs...or now SOICs.?
Cheers,
Last edited:
The IC's are pretty much universally better (by measurements, at least, say what you will about audibility...). Have a look at Groner's opamp compendium for reference. For filters, I'd be looking at the TI OPA164x series. JFET input, so far kinder on network impedance.
- Status
- Not open for further replies.