I'll surmise that the C0G cap is in a place where there's considerable voltage variation across it at audio frequencies (say in the RIAA network), rather than a coupling cap. In which case C0G is called for not just for lack of voltage coefficient but also temperature stability.
Exactamundo.
I would say more about the overall design but I'm already skating on thin ice w.r.t. NDA provisions. But yes, stability and constancy with voltage are quite important. It is also a 1% tolerance part, which allows a reasonable distribution of EQ errors without trims.
Here's my effort
Distortion at 370W into 8 Ohms at 20k:-
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 7.761e+01 1.000e+00 -1.60° 0.00°
2 4.000e+04 9.052e-05 1.166e-06 -22.11° -20.51°
3 6.000e+04 4.899e-05 6.312e-07 130.16° 131.76°
4 8.000e+04 5.098e-06 6.568e-08 -162.10° -160.50°
5 1.000e+05 2.387e-05 3.075e-07 58.85° 60.45°
6 1.200e+05 1.352e-05 1.742e-07 -151.01° -149.41°
7 1.400e+05 3.768e-05 4.855e-07 116.72° 118.32°
8 1.600e+05 2.056e-05 2.649e-07 -142.91° -141.31°
9 1.800e+05 4.202e-05 5.414e-07 118.14° 119.74°
Total Harmonic Distortion: 0.000158%
Distortion at 370W into 8 Ohms at 20k:-
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 7.761e+01 1.000e+00 -1.60° 0.00°
2 4.000e+04 9.052e-05 1.166e-06 -22.11° -20.51°
3 6.000e+04 4.899e-05 6.312e-07 130.16° 131.76°
4 8.000e+04 5.098e-06 6.568e-08 -162.10° -160.50°
5 1.000e+05 2.387e-05 3.075e-07 58.85° 60.45°
6 1.200e+05 1.352e-05 1.742e-07 -151.01° -149.41°
7 1.400e+05 3.768e-05 4.855e-07 116.72° 118.32°
8 1.600e+05 2.056e-05 2.649e-07 -142.91° -141.31°
9 1.800e+05 4.202e-05 5.414e-07 118.14° 119.74°
Total Harmonic Distortion: 0.000158%
Johanson Technologies used to make huge ones, both electrically and
mechanically. But it seems they filled a badly needed void, I could
not find them on their web site any more.
regards, Gerhard
Aksa would tell us that the "profile" is all wrong since we can see that some of the odd harmonics are higher than their immediately preceding even harmonic. But the overall level of all the harmonics is so low, I doubt that "profile" adherence can apply any longer.Here's my effort
Distortion at 370W into 8 Ohms at 20k:-
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 7.761e+01 1.000e+00 -1.60° 0.00°
2 4.000e+04 9.052e-05 1.166e-06 -22.11° -20.51°
3 6.000e+04 4.899e-05 6.312e-07 130.16° 131.76°
4 8.000e+04 5.098e-06 6.568e-08 -162.10° -160.50°
5 1.000e+05 2.387e-05 3.075e-07 58.85° 60.45°
6 1.200e+05 1.352e-05 1.742e-07 -151.01° -149.41°
7 1.400e+05 3.768e-05 4.855e-07 116.72° 118.32°
8 1.600e+05 2.056e-05 2.649e-07 -142.91° -141.31°
9 1.800e+05 4.202e-05 5.414e-07 118.14° 119.74°
Total Harmonic Distortion: 0.000158%
THD <1.6ppm @ 380W of output !
Superb result.
All my efforts currently are in the 30W (class A) and 100 and 200W output categories.
I doubt I'll build the 500 Watter (which the plot is from) anytime soon - more an exercise to look at what is possible.
Re harmonic structure . . . if they are buried in the noise floor, I believe they are of no consequence.
Last edited:
Not that we should stop trying , but low THD (beyond a certain point) has
so very little to do with usability or perceived SQ.
"Diminishing returns " below <30ppm. That figure only being a testament
to how well the base design performs.
Thermal stability , how "gracefully" the design fails - for the few that do ...
Safety factors and protection - I just had a Kapton insulator burn through ,
amp just blew the 10A fuse ... one output device replaced , all is well.
(ALL mica now - no more cheap plastic OP insulators !)
SQ - SO hard to tell between a 5 and 50ppm design , long term listening
has made me favor high PSRR/above average slew/symmetric designs.
I sold my 5ppm "blameless" in favor of a lowly 30ppm "leach" amp. It has a
noticeable (likable) balance between bass and HF detail with all genre's.
AND won't go up in smoke with a blown output device.
OS
I agree - balanced all the way. I don't know how Doug could force that blameless thing on the world when he had the opportunity to do a decent balanced design. Leach got it right back in the 70's all ready.
For me its like a Citroen CV. Although they seem to perform ok, they look all at odds with bits of stuff attached here and there and then skinny wheels on top of it.
(Having said that, I am driving a late model Citroen C4 MPV now - neat car! Bentley will have to wait . . . )
For me its like a Citroen CV. Although they seem to perform ok, they look all at odds with bits of stuff attached here and there and then skinny wheels on top of it.
(Having said that, I am driving a late model Citroen C4 MPV now - neat car! Bentley will have to wait . . .
)
All my efforts currently are in the 30W (class A) and 100 and 200W output categories.
I doubt I'll build the 500 Watter (which the plot is from) anytime soon - more an exercise to look at what is possible.
Re harmonic structure . . . if they are buried in the noise floor, I believe they are of no consequence.
Andrew, if I may, I'd suggest you do this.
Build a 100W/8 mominal amp, using four pairs of output devices from the 150-200W class (e.g. MJL, NJW, etc types, very linear, comfortably cheap). That will give you excellent current output capability. Then split its supplies into semicondustor regulated higher lines for the voltage gan stage to include the current predriver stage, at say +/-62V or so, and standard capacitor egulated current gain stages at say +/- 56V (rising at no load up to say +/-59V). This will provide you with impulse "music" power of about +2 dBW above the nominal 20 dBW, which is a nice margin, and with no gimmicks or artsy circuits. In effect, all the unclipped power you are likely to need under home conditions.
The added benefit is that the voltage gain stages can be cleaned up very well indeed, which should improve your PSRR right off the bat without bankrupting you. If you add some current mirrors to your IPS and VAS, you can probably take the PSRR up to stratospheric levels even by modifying your topology as you have already used so well before.
i thought the same when I saw the circuit. I can only guess that he wanted to show he could get low thd numbers from just about any old circuit topology.
Many such topologies do not give symmetric slew nor symmetric PSRR.
THx-RNMarsh
Exactly so. I only bring out the test gear when it's 'close', or when the due diligence legalities are milling about.
Just rebuilt an amp with premium parts throughout, for the exercise of it. Left a few original caps in the thing. It sounded huge, spacious, highly detailed, and just a tad loose and harsh/bright. I take a closer look, and a pair of ceramics are in each zobel, back to back. Replace with one clean poly and the whole thing normalizes.
No more wall to wall sound, but it was a leverage and multiplied lie..... and slowly built up to being grating and irritating. Due to the introduced distortions in that particular important spot, harmonic structure was rung like a resonant bell and highly exaggerated. Micro detail and space were huge, but in actuality....were a complete farcical lie.
See into the recording, hear things you've never heard before! Right.......
It clearly illustrated the difference between sophomore design and high end design, and the kind of ears and minds that go with either group. You tend to see this in the audio rag test results. Bad measurements.... but the reviewer classifies the thing as being amazingly detailed.
No. Just no.
Last edited:
The one I remember was a 270uF (!) C0G cap that was about the size of a small shoebox, I believe made for the International Space Station by AVX. It may still be in service. I think they billed 100k dollars or thereabouts, back when that was a lot of money.
I'm sure with sufficient volumes requested one could find a manufacturer willing to make higher-capacitance C0G parts. But the tariffs will be steep.
I see Mouser has a non-stocked 1.8uF 25V C0G in a big fat SM package for $7.64 in thousands. Novacap 2225E185K250PHT
Of course the ubiquity of surface mount has pushed us into ceramics to our potential detriment, but as well, assisted in the development of really good resistors. So it has been a mixed bag. C0G will never be volumetrically efficient, alas.
Now that I fixed the primary distortion mechanism I'm on to second-order ones, including the servo cap that elicited the evil cackle from infinia, and the infrasonic caps, basically glorified coupling caps, also victims of an ad hoc substitution from mediocre mylar to X7R (but at least high voltage).
- Status
- Not open for further replies.