FCCID: RAXWG4005G
P75233 AWGR54 WG4005G-LF-AK 141400520017J REV:01
U1: Atheros AR2317-AC1A with stick on or solder on heat sink
U8: IC+ IP175C LF 5 Port 10/100 Ethernet Integrated Switch with integrated linear regulator driver
Q4: UTC 2SB1412 B1412 PNP HIGH VOLTAGE SWITCHING TRANSISTOR C259:3.3v -> EC180:1.88v->L79->EC179 regulated by the IP175C
U10: Anpec APM9435 P-Channel Enhancement Mode MOSFET
U12: Matrix Microtech MT34063M Universal DC to DC Converter C289:12v -> C291,C169:3.3v
D11: SK24 SMD 3.3v freewheel diode
U11: ISSI IS42S16400B-7T 64 MBit SDRAM 1Mb x 16 x 4
iESR=in circuit ESR
ESR=out of circuit ESR
WV=Working voltage of a functional unit
C289: Taicon VT 100uf 25v 6.5x12 iESR=0.75 ESR=0.85 WV=12v transformer voltage
EC179: Elcon 47uf 25v EYA 5x11 iESR=0.03 ESR=0.12 WV=1.88v
EC180: Elcon 47uf 25v EYA 5x11 iESR=0.03 ESR=0.12 WV=1.88v
EC182: Elcon 47uf 25v EYA 5x11 iESR=0.03 ESR=0.12 WV=3.3v
C291: Elcon 470uf 10v 6.5x12 iESR=0.05 ESR=0.55 WV=3.3v parallel with C169
C169: Taicon 100uf 25v 6.5x12 VT iESR=0.05 ESR=0.85 WV=3.3v parallel with C291
6 units, all unreliable. All fail, some run for days, some fail within minutes. The wireless and web configuration may work for a short time after the router dies and cuts off the Internet. A power cycle and they are working again. No obvious bad caps. C169 and C291 are in parallel and are very different. Maybe the circuit was designed for two 470uf 10v caps and the smaller cap got substituted by mistake or to save a penny. The Taicon datasheet does not list ESR for C169. 1 ohm ESR is appropriate for a 100uf low ESR capacitor. The Elcon EYA data sheet shows 0.13 ohms impedance for a 470uf 10v. The series of C291 is not labeled but if it were an EYA, 0.55 ohms would be 5 times the proper value.
With the router becoming unreliable when the Elcon 470uf reaches 0.55 ohms the perfectly good 1 ohm Taicon 100uf is completely useless in position C169. In C169 I put a Panasonic or Rubycon 820uf 6.3v which have an ESR of 0.02. At 1/5th the ESR of the best original, my cap takes the entire load so C291 is no longer important. EC179-182 have way better ESR than anything I can supply. The ESR of C289 does not need to be very low.
One cap done!
It appears that many low cost networking products have the same flaw and the same fix works for most of them. The engineers design a reliable circuit. Bean counters cut costs until the design is marginal but still works for most of the warranty period. Semiconductors aren't easy to cut because they don't substitute well. Capacitors can be cut right to the edge of reliability. The capacitors slowly degrade which stresses the marginal circuitry to the breaking point. Soon the product is unreliable but can be coaxed into functioning. Eventually the product fails completely. The capacitors look and test good so the problem is assumed to be in unfixable circuitry and the product is discarded.
Caps more reliable, larger, and with better ESR than the originals in the right places makes the circuit no longer marginal allowing the product to be as reliable and long lasting as the engineers originally intended.
So far all indications are that this fix is perfect. All test runs have passed and the deployments I've done are working flawlessly. Fortunately I ran across another unmodified unit so I can provide scope pics.
3191-Top: 12v at C289
3191-Bottom: 3.3v at C291
Cap testing at C289 shows that the 12v waveform can be smoothed out but it does not improve the 3.3v waveform. Cap testing at C291 shows the wave can be smoothed out substantially. Unfortunately a 1500uf cap test on C291 surprised the MT34063 and burned it and the IP175C and AR2317 and IS42S16400B out. APM9435 was still good. Fortunately I can get scope pics from another already serviced unit.
C169: Rubycon MBZ 470uf 16v
3192-Top: 12v at C289
3192-Bottom: 3.3v at C169 with Rubycon MBZ
3193-Top: 3.3v at C169 with Rubycon MBZ
3193-Bottom: 1.8v at EC179
Reducing the ripple from 40mv to 10mv on the 3.3v rail would be why this fix works so well. A better cap could reduce it even more.
P75233 AWGR54 WG4005G-LF-AK 141400520017J REV:01
U1: Atheros AR2317-AC1A with stick on or solder on heat sink
U8: IC+ IP175C LF 5 Port 10/100 Ethernet Integrated Switch with integrated linear regulator driver
Q4: UTC 2SB1412 B1412 PNP HIGH VOLTAGE SWITCHING TRANSISTOR C259:3.3v -> EC180:1.88v->L79->EC179 regulated by the IP175C
U10: Anpec APM9435 P-Channel Enhancement Mode MOSFET
U12: Matrix Microtech MT34063M Universal DC to DC Converter C289:12v -> C291,C169:3.3v
D11: SK24 SMD 3.3v freewheel diode
U11: ISSI IS42S16400B-7T 64 MBit SDRAM 1Mb x 16 x 4
iESR=in circuit ESR
ESR=out of circuit ESR
WV=Working voltage of a functional unit
C289: Taicon VT 100uf 25v 6.5x12 iESR=0.75 ESR=0.85 WV=12v transformer voltage
EC179: Elcon 47uf 25v EYA 5x11 iESR=0.03 ESR=0.12 WV=1.88v
EC180: Elcon 47uf 25v EYA 5x11 iESR=0.03 ESR=0.12 WV=1.88v
EC182: Elcon 47uf 25v EYA 5x11 iESR=0.03 ESR=0.12 WV=3.3v
C291: Elcon 470uf 10v 6.5x12 iESR=0.05 ESR=0.55 WV=3.3v parallel with C169
C169: Taicon 100uf 25v 6.5x12 VT iESR=0.05 ESR=0.85 WV=3.3v parallel with C291
6 units, all unreliable. All fail, some run for days, some fail within minutes. The wireless and web configuration may work for a short time after the router dies and cuts off the Internet. A power cycle and they are working again. No obvious bad caps. C169 and C291 are in parallel and are very different. Maybe the circuit was designed for two 470uf 10v caps and the smaller cap got substituted by mistake or to save a penny. The Taicon datasheet does not list ESR for C169. 1 ohm ESR is appropriate for a 100uf low ESR capacitor. The Elcon EYA data sheet shows 0.13 ohms impedance for a 470uf 10v. The series of C291 is not labeled but if it were an EYA, 0.55 ohms would be 5 times the proper value.
With the router becoming unreliable when the Elcon 470uf reaches 0.55 ohms the perfectly good 1 ohm Taicon 100uf is completely useless in position C169. In C169 I put a Panasonic or Rubycon 820uf 6.3v which have an ESR of 0.02. At 1/5th the ESR of the best original, my cap takes the entire load so C291 is no longer important. EC179-182 have way better ESR than anything I can supply. The ESR of C289 does not need to be very low.
One cap done!
It appears that many low cost networking products have the same flaw and the same fix works for most of them. The engineers design a reliable circuit. Bean counters cut costs until the design is marginal but still works for most of the warranty period. Semiconductors aren't easy to cut because they don't substitute well. Capacitors can be cut right to the edge of reliability. The capacitors slowly degrade which stresses the marginal circuitry to the breaking point. Soon the product is unreliable but can be coaxed into functioning. Eventually the product fails completely. The capacitors look and test good so the problem is assumed to be in unfixable circuitry and the product is discarded.
Caps more reliable, larger, and with better ESR than the originals in the right places makes the circuit no longer marginal allowing the product to be as reliable and long lasting as the engineers originally intended.
So far all indications are that this fix is perfect. All test runs have passed and the deployments I've done are working flawlessly. Fortunately I ran across another unmodified unit so I can provide scope pics.
3191-Top: 12v at C289
3191-Bottom: 3.3v at C291
Cap testing at C289 shows that the 12v waveform can be smoothed out but it does not improve the 3.3v waveform. Cap testing at C291 shows the wave can be smoothed out substantially. Unfortunately a 1500uf cap test on C291 surprised the MT34063 and burned it and the IP175C and AR2317 and IS42S16400B out. APM9435 was still good. Fortunately I can get scope pics from another already serviced unit.
C169: Rubycon MBZ 470uf 16v
3192-Top: 12v at C289
3192-Bottom: 3.3v at C169 with Rubycon MBZ
3193-Top: 3.3v at C169 with Rubycon MBZ
3193-Bottom: 1.8v at EC179
Reducing the ripple from 40mv to 10mv on the 3.3v rail would be why this fix works so well. A better cap could reduce it even more.
Comment