I am only interested in measurements from serious sources.
As it was a comparison between the two different caps, we only noticed the similar construction (both were radially leaded foils encapsulated in epoxy filled plastics housings of not to different dimensions) and checked for the actual capacitance (both were spec'ed with +- 10%) with a measurement bridge (0.2% accuracy at 10kHz), confirming that the difference in capacitance was below 0.5%.
Source impedance was well below of 100 Ohms while the input impedance was at 47kOhms.
Are you joking!?
The notation of physical units is well defined (at least the SI ones, and the ones derived from that; I am not so familiar with other systems). But we were not discussing units at all!?
Are YOU joking? It was you who brought up that content-free rant.
Unit symbols are nothing that the author can define; there are two language
versions of ISO/EN/Ö 80000 of equal rank.
It's silly to blame the WIMA author for falling back to U in a single case.
U is applicable law here and what you learn in 5th grade. V is at most tolerated.
And WIMA _is_ here. ( 90 min. driving from my place btw.)
If you run a business here and sell 27" screens, then be prepared for
an invasion of lawyers sent by the competition. Legally allowed is
only metric. We had a wave of that just 3 or 4 years ago.
You'd better contributed some on-topic facts.
/Gerhard
Correct me if I'm wrong , but 2,2 uF/47k is about -6dB at 10 Hz , not much music is going that low , but then again if you're playing around with (software) synthesizers , you don't want the input to weaken those frequencies to speakers or HP's that already fall off below 30-35 Hz.
Hmmm....... I thought the -3 db point of an RC filter is equal to the inverse of 2πRC........1/(2πRC). For 47KΩ and 2.2µF, that's 1.54 Hz. That's PLENTY low enough for ANY audio circuit.
I have used nothing else since I saw that circuit in the 70's. I haven't compared it with the usual circuit. It pairs nicely with matched mosfets or jfets used in the front end diff stage, and with proper fet matching, I never have to use a servo.
I have a stash of fairly compact 2u2/63V polycarbonates that work well in that position.
Slight correction as it is -3dB at roughly 10 Hz with parts meeting the exact values and of course variying according to the tolerances of the parts. (If possible, matching to <1% is recommended).
As there is an accompanying phase shift, group delay might approach the audible region, but more importantly is what you've mentionened in the second part above; combination of high pass filters in a chain leads to altered frequency response overall(using the -3dB point as reference) and accumulation of phase shift (group delay).
Given that polyester is slightly more nonlinear as a dielectric than polypropylene it is not surprising that some careful tests show this, whether listening or measurements.Jakob2 said:
However, he said
and gave a table of measured results, presumably the tests he referred to. I asked how those tests showed that MKP were the best, given that distortion was below the measurement threshold for all caps in that test.RickTH said:
My question was not about MKP being better, but how the tests he referred to showed that they were better. The figures from Elektor seemed to me not to be sufficiently sensitive to detect a difference.
2.2uF and 47k gives an LF rolloff at 1.54Hz. To a first approxmation, varying phase shift extends from about x10 below to x10 above the corner frequency, so 0.154Hz to 15.4Hz. Unlikely to be audible unless you have unusual speakers, and an unusual room, combined with certain tastes in music.
Shame on me; I should have noticed that.....
It depends on the chain, means the number of high passes already involved and their realisations, we noticed the difference in direct comparisons, but at that time couldn't check the contribution of the other gear. Room and loudspeakers were capable to allow for 20Hz and below reproduction.
The "elektor tables" showed what to expect wrt losses, styroflex type first, mkp types as "runner up", mkt/mks behind.
Last edited:
Still hard to say, as I am only aware of two or three papers covering the low frequency region.
But the recommendation to aim for lower group delay in the bass region goes back to Fincham in the 80s and was imo corrobated by newer experiments testing the phase equalization of loudspeaker above and below 100 Hz.
The usual reference (Blauer/Laws from 1976) examined mainly the region above 500 Hz, but extrapolation down seems to corrobate Finchams recommendation as well.
Assuming the usual spread of human abilities, it seems to be wise to account for it.
Thanks, I wondered if you knew of anything more recent. Rod Elliott says: "For what it's worth, accepted wisdom indicates that group delay should not exceed 2 complete cycles of the waveform at any frequency significantly below 500Hz". When talking about subwoofers, Earl Geddes commented that 4 or 5 cycles are required to recognise the frequency.
Newer are Choisel/Martin from 2008, Mäkivirta et al. and Lisiki et al. both from 2018; Choisel/Martin exploring more interchannel differences wrt phase, Mäkivirta et al. simulating digital phase equalization without real listening tests while Lisiki et al. examined the audibility of group delay characteristics of loudspeakers, but all above 63 Hz (Choisel/Martin) resp. above 100 Hz Lisiki et al.
Its best practice to aim for the 3dB point around 1Hz or so, for two reasons:
1) Several bits of equipment each with 3dB roll-offs around 10Hz or so will combine to make a very noticable bass drop off extending well above 20Hz. Several bits of equipment all with 3dB at 1Hz or so will not show appreaciable drop off in the audio band.
2) Electrolytic capacitors create some distortion, but this strongly depends on signal voltage across the capacitor - having a 1Hz roll off means that only 5% of the signal voltage is across the capacitor at 20Hz, only 1% at 100Hz, so the impact of any distortion is greatly diluted.
A graph of THD v. frequency will often show an upturn at LF if over-small electrolytic coupling capacitors are used.
1) Several bits of equipment each with 3dB roll-offs around 10Hz or so will combine to make a very noticable bass drop off extending well above 20Hz. Several bits of equipment all with 3dB at 1Hz or so will not show appreaciable drop off in the audio band.
2) Electrolytic capacitors create some distortion, but this strongly depends on signal voltage across the capacitor - having a 1Hz roll off means that only 5% of the signal voltage is across the capacitor at 20Hz, only 1% at 100Hz, so the impact of any distortion is greatly diluted.
A graph of THD v. frequency will often show an upturn at LF if over-small electrolytic coupling capacitors are used.
3dB roll-off at 10 Hz is 1.25dB at 20 Hz
What is the roll-off of your speakers at 20 Hz ?
The thread is not about electrolytic capacitors.
What are measured distortions of film capacitors ?
The point is lots of places in the audio path may have low-end roll-offs - so each roll-off needs to be conservative so the combined effect isn't deliterious.
The principle applies to film caps as well - if you are using a film cap in a LPF at 1Hz or so it can be polyester without ill-effects since only a tiny fraction of the signal can be distorted by it.
The behaviour of polyester v polypropylene is well covered in this fairly famous set of articles:
Cyril Bateman's Capacitor Sound articles | Linear Audio NL
The principle applies to film caps as well - if you are using a film cap in a LPF at 1Hz or so it can be polyester without ill-effects since only a tiny fraction of the signal can be distorted by it.
The behaviour of polyester v polypropylene is well covered in this fairly famous set of articles:
Cyril Bateman's Capacitor Sound articles | Linear Audio NL
If you need a REALLY small audio capacitor for amp input then Nichicon Es is your best option: Nichicon Muse ES bipolar caps measured: <-120dB THD, <-140dB IMD