Re: Finding low-ESR 400V capacitors
I can't give anything like an evolutionary history of electrolytic capacitors, as I'm too young (I don't say that often!). I read some late 50s TV repair magazines, in which they were regarded as royal PITAs (Pains In The Ass), due to inconsistent quality and frequent-friers being semi-custom multi-section, multi-value parts. The latter aspect meant difficulty keeping suitable replacement parts on hand.
Fast-forward to my days at DeVry (early-mid 70s),and about the farthest profs went beyond, "Assume an ideal capacitor ..." was to mention that electrolytics and tantalums are polarized.
By the time I started at Boschert, my first job working with switching power supplies, in 1980, low ESR capacitors and "Computer Grade" (beer can sized screw terminal) capacitors had been on the market for quite a while. Switching power supplies need output capacitors the can handle high ripple (AC) current without overheating, thus, low ESR as it was then called. United Chemicon RX and RZ series, Nichicon PA and PB series, and Sprague 672D and 674D series were the best available at the time (I didn't see Marcon, Rubycon, or Panasonic at that time).
Since then "low impedance" became the common usage, and impedances of series have fallen steadily. Going back to the, "Assume an ideal capacitor ..." line, the impedance of a part at a given frequency is the vector sum of the part's capacitive reactance, inductive reactance, and resistance (reactance is for AC what resistance is for DC, but changes with frequency). In the case of a real-world capacitor, it's the vector some of the capacitive reactance, and the parasitic inductive reactance (ESL) and resistance (ESR) due to the wire leads, swaging the leads onto the foils, the foils, and the electrolyte (in an electrolytic capacitor, the dielectric is an aluminum oxide layer on the anode foil; the electrolyte is a very very low resistance conductor).
As frequency increases, inductive reactance increases, and capacitive reactance decreases. At 100KHZ, the capacitive reactance of a 2200uF capacitor is ~.72 milliohms. In the vector sum, capacitive and inductive reactance subtract. The inductive reactance at 100KHz is probably higher that than the capacitive reactance, so the impedance of the 2200uF capacitor is close to the vector sum of the ESR and the inductive reactance of the ESL.
Back in the late 70s and early 80, switch frequencies were in the 20KHz-30KHz. Since inductive reactance is proportional to the frequency, multiplying the frequency by 5 or 10 multiplied the inductive reactance by the same factor. So ESL was less relevant back 1980ish.
Some the others here can be more precise about the time-frame that I am, but somewhere around 2000 computer MB started including VRMs that stepped the 12V or 5V from the PSU down to the voltage needed by the processor or memory. These started as little plug-in boards, but soon were integrated into the MBs. Nowadays there are server MBs that just take 12V power, and have all the integrated DC-DC converters necessary to generate the various voltages needed by the MB.
I can't give anything like an evolutionary history of electrolytic capacitors, as I'm too young (I don't say that often!). I read some late 50s TV repair magazines, in which they were regarded as royal PITAs (Pains In The Ass), due to inconsistent quality and frequent-friers being semi-custom multi-section, multi-value parts. The latter aspect meant difficulty keeping suitable replacement parts on hand.
Fast-forward to my days at DeVry (early-mid 70s),and about the farthest profs went beyond, "Assume an ideal capacitor ..." was to mention that electrolytics and tantalums are polarized.
By the time I started at Boschert, my first job working with switching power supplies, in 1980, low ESR capacitors and "Computer Grade" (beer can sized screw terminal) capacitors had been on the market for quite a while. Switching power supplies need output capacitors the can handle high ripple (AC) current without overheating, thus, low ESR as it was then called. United Chemicon RX and RZ series, Nichicon PA and PB series, and Sprague 672D and 674D series were the best available at the time (I didn't see Marcon, Rubycon, or Panasonic at that time).
Since then "low impedance" became the common usage, and impedances of series have fallen steadily. Going back to the, "Assume an ideal capacitor ..." line, the impedance of a part at a given frequency is the vector sum of the part's capacitive reactance, inductive reactance, and resistance (reactance is for AC what resistance is for DC, but changes with frequency). In the case of a real-world capacitor, it's the vector some of the capacitive reactance, and the parasitic inductive reactance (ESL) and resistance (ESR) due to the wire leads, swaging the leads onto the foils, the foils, and the electrolyte (in an electrolytic capacitor, the dielectric is an aluminum oxide layer on the anode foil; the electrolyte is a very very low resistance conductor).
As frequency increases, inductive reactance increases, and capacitive reactance decreases. At 100KHZ, the capacitive reactance of a 2200uF capacitor is ~.72 milliohms. In the vector sum, capacitive and inductive reactance subtract. The inductive reactance at 100KHz is probably higher that than the capacitive reactance, so the impedance of the 2200uF capacitor is close to the vector sum of the ESR and the inductive reactance of the ESL.
Back in the late 70s and early 80, switch frequencies were in the 20KHz-30KHz. Since inductive reactance is proportional to the frequency, multiplying the frequency by 5 or 10 multiplied the inductive reactance by the same factor. So ESL was less relevant back 1980ish.
Some the others here can be more precise about the time-frame that I am, but somewhere around 2000 computer MB started including VRMs that stepped the 12V or 5V from the PSU down to the voltage needed by the processor or memory. These started as little plug-in boards, but soon were integrated into the MBs. Nowadays there are server MBs that just take 12V power, and have all the integrated DC-DC converters necessary to generate the various voltages needed by the MB.
Half of the computer problems is caused by bad contacts 
Comment