>Where can you buy C3600/A1406 Sanyos?
>Can you get these devices in B'lore? Is there a risk of fakes with the Sanyos?
I had them on my master list, but my regular vendor didn't appear to have them. There's a possibility of counterfeits with anything except rock-bottom priced commodity semiconductors - e.g. Philips BC550C/560C, BD139/140, ...
The Fairchild KSC3953 with the same vendor seems to be authentic - moulding is clean, wets well when soldering, beta is consistent and sounds good in a VAS. I can get you a 100 of these for US 30c each + shipping, if you can manage with 120v Vceo, hfe clustered around 60.
I can also source certain low-valued silver mica (10, 12, 22, 27, 47, 54, 470 pF - 300V) & styroflex caps (22, 68, 220 pF and some nF+ values, 100V) in any mix, small quantities for a flat rate of US 50c each for the SM, and 20c each for the styroflex. These are NOS - the leads are oxidized and will have to be scraped clean before soldering, but they're otherwise perfect and measure to within 2-5% of their specified values.
>Can you get these devices in B'lore? Is there a risk of fakes with the Sanyos?
I had them on my master list, but my regular vendor didn't appear to have them. There's a possibility of counterfeits with anything except rock-bottom priced commodity semiconductors - e.g. Philips BC550C/560C, BD139/140, ...
The Fairchild KSC3953 with the same vendor seems to be authentic - moulding is clean, wets well when soldering, beta is consistent and sounds good in a VAS. I can get you a 100 of these for US 30c each + shipping, if you can manage with 120v Vceo, hfe clustered around 60.
I can also source certain low-valued silver mica (10, 12, 22, 27, 47, 54, 470 pF - 300V) & styroflex caps (22, 68, 220 pF and some nF+ values, 100V) in any mix, small quantities for a flat rate of US 50c each for the SM, and 20c each for the styroflex. These are NOS - the leads are oxidized and will have to be scraped clean before soldering, but they're otherwise perfect and measure to within 2-5% of their specified values.
Not sure about the brand(s) on the Silver Mica - some have a red-brown coating, thick stiff leads, and have only the value and voltage printed on them (10-54pF), and a few others (470 pF, some smaller values) have a darker brown coating, thinner leads, and are labeled SAN, with value, tolerance and voltage (?). The red-brown ones are more prevalent, and could be Saha.
FET??
Interesting thread... How about FET's in the VAS? Personally i've just ripped out the BJT's and the nessesary sprinkling of resistors in the LTP of my new ClassA amp and replaced them with FET's. I can't tell the difference, sonically they sound fine to me and according to the scope i have slightly less noise in the background ( about 2mV background compared to approx 4mV with BJT's ) Don't know if this lower noise floor is due to eliminating the resistors by using fet's, or if it is the active devices themselves. The fet's in question i have used are ZVN / ZVP4424A's. High voltage and very fast.
LaterZ
Mad.P
Interesting thread... How about FET's in the VAS? Personally i've just ripped out the BJT's and the nessesary sprinkling of resistors in the LTP of my new ClassA amp and replaced them with FET's. I can't tell the difference, sonically they sound fine to me and according to the scope i have slightly less noise in the background ( about 2mV background compared to approx 4mV with BJT's ) Don't know if this lower noise floor is due to eliminating the resistors by using fet's, or if it is the active devices themselves. The fet's in question i have used are ZVN / ZVP4424A's. High voltage and very fast.
LaterZ
Mad.P
I've used Hitachi transistors with good results, but honestly, I have
not done a lot of comparing, so they may not be the best...
2SA1084 (PNP)
2SC2547 (NPN)
There are 3 voltage variations. I can't remember which the above are,
but I think there is a 120v version. These are very low noise transistors,
which is why I picked them.
not done a lot of comparing, so they may not be the best...
2SA1084 (PNP)
2SC2547 (NPN)
There are 3 voltage variations. I can't remember which the above are,
but I think there is a 120v version. These are very low noise transistors,
which is why I picked them.
Good VAS Transistor Pair
I have had good luck with the Panasonic 2SA1535A/2SC3944A complementary pair. Both are rated at 180V with an Ft of 200 MHz. The application in which they are used requires +/- 75V rails, so a 180V rating is required.
When used in a cascode configuration, the 200 MHz Ft is also sufficient for my needs. I run them at an idle current of approx 15 mA, so at least one device in each cascode pair requires an external heatsink.
I have had good luck with the Panasonic 2SA1535A/2SC3944A complementary pair. Both are rated at 180V with an Ft of 200 MHz. The application in which they are used requires +/- 75V rails, so a 180V rating is required.
When used in a cascode configuration, the 200 MHz Ft is also sufficient for my needs. I run them at an idle current of approx 15 mA, so at least one device in each cascode pair requires an external heatsink.
> Panasonic 2SA1535A/2SC3944A complementary pair
They have a Cob of 30..50pF, which is too high for a VAS if you intend to use a larger Cdom for compensation. But it's probably a nice driver transistor in a TO-220 package, with ample margins on Ic to drive a lot of parallel output transistor bases in darlington EF configuration.
They have a Cob of 30..50pF, which is too high for a VAS if you intend to use a larger Cdom for compensation. But it's probably a nice driver transistor in a TO-220 package, with ample margins on Ic to drive a lot of parallel output transistor bases in darlington EF configuration.
Does this say anything about the requirements for a VAS EF buffer or for the LTP input stage?linuxguru said:
2sa1535/c3944 makes an excellent driver pair but not a VAS.
It might be OK as the cascode for the VAS. Anyone with a thought on this?
Although slower, the 2sb649/d669 still seems to better it in most respects.
> Does this say anything about the requirements for a VAS EF buffer or for the LTP input stage?
Hmm - fT isn't that critical for a driver/pre-driver - in fact, if it's too fast it might be prone to instability that has to be tamed with larger base-stoppers or external Miller capacitors. Anything between 30 - 150 MHz should be fine.
Since the Hitachi c2547/a1084 was earlier identified as low-noise, it might be suitable for an LTP also.
> 2sa1535/c3944 makes an excellent driver pair but not a VAS.
> It might be OK as the cascode for the VAS. Anyone with a thought on this?
It's OK as the lower transistor, but then just about any transistor, including any commodity low-voltage small-signal transistor is fine there. For the upper transistor, Cob is still critical - <5 pF seems to be indicated, maybe <2 pF if lead-compensated a la AKSA - otherwise Cob variation with Vce will contribute to non-linear charging during voltage swings.
>Although slower, the 2sb649/d669 still seems to better it in most respects.
The b649/d669 are actually faster, with an fT = 140 MHz and lower Cob at ~15pF . However, Vceo at 120V is inferior (the d669a at 160V is almost comparable).
Hmm - fT isn't that critical for a driver/pre-driver - in fact, if it's too fast it might be prone to instability that has to be tamed with larger base-stoppers or external Miller capacitors. Anything between 30 - 150 MHz should be fine.
Since the Hitachi c2547/a1084 was earlier identified as low-noise, it might be suitable for an LTP also.
> 2sa1535/c3944 makes an excellent driver pair but not a VAS.
> It might be OK as the cascode for the VAS. Anyone with a thought on this?
It's OK as the lower transistor, but then just about any transistor, including any commodity low-voltage small-signal transistor is fine there. For the upper transistor, Cob is still critical - <5 pF seems to be indicated, maybe <2 pF if lead-compensated a la AKSA - otherwise Cob variation with Vce will contribute to non-linear charging during voltage swings.
>Although slower, the 2sb649/d669 still seems to better it in most respects.
The b649/d669 are actually faster, with an fT = 140 MHz and lower Cob at ~15pF . However, Vceo at 120V is inferior (the d669a at 160V is almost comparable).
Hi,
the 160Vce0 649a/669a has superior SOA cf the 180Vce0 1535/3944, but the 1535 is considerably faster at any of the lower Ic values.
But generally, higher SOAR, the low Cob, higher hFE, flatter gain curve (except at high IC) show the 649 in a good light, surprisingly, since it's in the To126 package.
Is there a cheap but improved equivalent to replace this obsolete device?
the 160Vce0 649a/669a has superior SOA cf the 180Vce0 1535/3944, but the 1535 is considerably faster at any of the lower Ic values.
But generally, higher SOAR, the low Cob, higher hFE, flatter gain curve (except at high IC) show the 649 in a good light, surprisingly, since it's in the To126 package.
Is there a cheap but improved equivalent to replace this obsolete device?
Not sure about price, availability or obsolescence - but here are some alternatives for drivers/pre-drivers (not all of which have any/some/all specs superior to the b649/d669):
2sa1930/2sc5171 (Toshiba, plastic TO-220)
2sa1837/2sc4793 (Toshiba, plastic TO-220)
2sa1011/2sc2344 (Sanyo, TO-220AB)
2sa968b/2sc2238b (Toshiba/CDIL, TO-220AB)
2sa985a/2sc2275a (NEC, TO-220)
2sa1110..12/2sc2590..92 (Panasonic, TO-126B/TO-220AB)
2sa965/2sc2235 (Toshiba, TO-92L, very flat hFE vs. Ic up to 0.2A)
2sa1606/2sc4159 (Sanyo, TO-220ML)
2sa1220a/2sc2690a (NEC, TO-126B)
2sa1930/2sc5171 (Toshiba, plastic TO-220)
2sa1837/2sc4793 (Toshiba, plastic TO-220)
2sa1011/2sc2344 (Sanyo, TO-220AB)
2sa968b/2sc2238b (Toshiba/CDIL, TO-220AB)
2sa985a/2sc2275a (NEC, TO-220)
2sa1110..12/2sc2590..92 (Panasonic, TO-126B/TO-220AB)
2sa965/2sc2235 (Toshiba, TO-92L, very flat hFE vs. Ic up to 0.2A)
2sa1606/2sc4159 (Sanyo, TO-220ML)
2sa1220a/2sc2690a (NEC, TO-126B)
>Can you say more about the effects or problems of a transistor with a too high Cob in a VAS ?
The Cob of the VAS transistor has to be charged by the current flowing into it, either from a CCS or bootstrapped collector load. This current is small - 5 to 15 mA is the norm here to keep the collector dissipation low. The larger the Cob, the lower the slew rate of the VAS - hence, we'd like to minimize it.
The second effect is that the Cob is a non-linear function of Vce - hence, the voltage at the collector of the VAS will not increase linearly when a constant current source charges the (non-linear) Cob. The way around this is to use a linear capacitor (say a styroflex or silver mica) in parallel with a Cob, and much larger than the Cob to reduce the effect of the non-linear Cob. This is also known as the Miller capacitor or Cdom, used for lag compensation. But if the Cob is too large, Cdom will have to be large, limiting the slew rate again.
>And why it need to be <2 pf Cob if we use lead-compensated a la AKSA ?
The capacitance used for lead compensation tends to be relatively small - say 1-10 pF, and one terminal sometimes goes to the collector of the VAS. It helps to keep all other capacitances from the collector of the VAS relatively low - especially the non-linear Cob, to allow the linearity of the lead compensation cap to dominate.
The Cob of the VAS transistor has to be charged by the current flowing into it, either from a CCS or bootstrapped collector load. This current is small - 5 to 15 mA is the norm here to keep the collector dissipation low. The larger the Cob, the lower the slew rate of the VAS - hence, we'd like to minimize it.
The second effect is that the Cob is a non-linear function of Vce - hence, the voltage at the collector of the VAS will not increase linearly when a constant current source charges the (non-linear) Cob. The way around this is to use a linear capacitor (say a styroflex or silver mica) in parallel with a Cob, and much larger than the Cob to reduce the effect of the non-linear Cob. This is also known as the Miller capacitor or Cdom, used for lag compensation. But if the Cob is too large, Cdom will have to be large, limiting the slew rate again.
>And why it need to be <2 pf Cob if we use lead-compensated a la AKSA ?
The capacitance used for lead compensation tends to be relatively small - say 1-10 pF, and one terminal sometimes goes to the collector of the VAS. It helps to keep all other capacitances from the collector of the VAS relatively low - especially the non-linear Cob, to allow the linearity of the lead compensation cap to dominate.
No, it needs a test and measurement setup to correctly bias the transistor and measure the waveform at the collector load in response to a specific input waveform (usually a pulse train) to be able to measure Cob - it's a non-linear function of Vce, so that further complicates the measurement.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.