Harsh or not is not only depend on driver transistor. mje15034/35 can sound good but not the best if designed properly.
I used 2SC5171/2SA1943 or 2SC4793/2SA1837.
Definitely. But i think 2sa1837 isn't suitable for two pair of mjl93/94, +/-55volt supply & 4ohm load.
I think 2 pair is capable. Even for 4 pairs with PSU 63V still capable but it in the edge, you can not give them high idle current or you can run them in class B.
Are you sure? If so i'll try to get some, cheap & quite available[KEC]. Btw i heard that mje15034/35 sounds great with high idle current.
Of course. I designed an amplifier using those driver for 4 pairs output transistor with 63V PSU. It built by many in my country.
If you use high idle current like 40mA or more, you can parallel them like ASTX's design. I think it still better than mje15034/35.
Then i think i did a big mistake buying mje34/35 🙁 2sa1837 & it's complementary very much available here with very very affordable price.
Hi nirupambhowmick,
I wouldn't stress too much, MJE150034/35 driving two pairs of MJL21193/94 in an EF2 will be absolutely fine. Did you decide on a VAS BJT in the end?
Cheers
Paul
I wouldn't stress too much, MJE150034/35 driving two pairs of MJL21193/94 in an EF2 will be absolutely fine. Did you decide on a VAS BJT in the end?
Cheers
Paul
Hehehehe...
🙂
How about both? I mean ksa1381 & ZTX558 with complimentary. I'll look for both type. Regards.
🙂How about both? I mean ksa1381 & ZTX558 with complimentary. I'll look for both type. Regards.
Last edited:
2.80 a pair isn’t a big deal when you need 2 pairs. My personal preference in EF2 would be the C4793 BUT only the original Toshiba. I don’t trust the supply chain on them anymore since Toshiba discontinued them, so I can no longer recommend their use. You pay 20 cents, it’s worth 20 cents (if that). Like I said, I’ve got a big stash of originals that came from Digikey when they still sold the C3281(!) - they find their way into everything, even non audio. So what’s next? The MJEs. The full-pack TO-126’s aren’t physically big enough to remove the heat at load (even idling, 3 watts is steep even on a heat sink). Full-pack TO-220’s we’re a great compromise - you could get a 23 pF Cob and still pull 7 or 8 watts of heat out of it.
For EF3, the *only* option is the MJE, IMO (for a handful of pairs). All the reasons to use something smaller vanish, since you would have something smaller driving it. 200 MHz fT in that position is quite possible - it’s NOT going to eat into phase margin.
For the VAS it comes down to whether you can stand having non-hFE-matched C3503/A1381, or start looking at other options. Mine are different gain grades - I chose not to worry about that when things started disappearing. There are other ways around that problem.
Don't underestimate KEC, a well known Korean semiconductor manufacturer. Although they're not as good as Toshiba but you CAN trust them. Here you can get some parts at an unbelievable price! I mostly like to buy from local retailers because of that. Sometimes the price can be 1/10 compared to mouser, Farnell or RS. I mean products like 5532, tl072/4, tda7294/93, lm3886, lm1875, three pin voltage regs like 7xxx, lm317/37, logic ic's, Philips BJTs(NOS), 2n3055, 2sc5200/a1943, ST BD139/40 & many more. New & 100% authentic! On the other hand when it's all about DIY i don't think too much about $$... I just want the right component for my work but same time i don't like to waste money for snake oil.
Btw i ordered 5pairs of mje15034/35 & i believe they're not waste. Rod recommends those for his P-3A amp, +/-35v supply. I also heard praise for mje34/35 using that amp. But Toshiba is simply phenomenal in their work.
🙂 🙂 🙂
My first choice is ZTX558/458 because of their smaller package & excellent spec, i like them
Anyway do you recommend them for VAS?Mje15034/35 as driver & MJL93/94 as power trannys & double emitter follower configuration(EF2). Forget my ignorance, still learning so everyones view & ideas VERY important for me. 🙄 Regards.The issue with using then for VAS is the high dissipation. I’d run the VAS at 20 mA, and thats a stretch even for a ZTX transistor clamped to a piece of metal. 10 mA is safer, but you may run out of gas. I’ve been known to parallel two in the VAS position to get around dissipation problems. But then you have to ask yourself if you can get a single higher power transistor with the same combined Cob.
I'll try.....
But don't you think 20ma is quite high compared to usual 5-10ma? could you elaborate a little about voltage amplifying stage for EF2 ??
But don't you think 20ma is quite high compared to usual 5-10ma? could you elaborate a little about voltage amplifying stage for EF2 ??
Last edited:
When I built a similar power amp using C2344/A1011 drivers and parallel 2N3773/2N6031 outputs it needed 20 mA in the VAS to be able to drive 2 ohms without premature clipping. It WAS a single ended VAS, and therefore unable to go into class B like some of my other amps can. Sounded better at the high current compared to my usual 8-10 mA with ‘normal’ loads, too. In EF3 I’ve gotten away with 3-4 mA in the VAS without early clipping, but slew rate suffers so you still want 8-10.
Not yet. I've to research more about VAS & suitable components. I can use mje340/50 but i'll try for something better.
When I doubled up on VAS transistors for dissipation reasons it WAS MJE340/350, with MPSA42/92 running in darlington. +/-127 volt power supplies.
Doubling up ZTX458/558 would work to get 20mA VAS at 10mA each. Another option, if you are more adventurous, would be to use a differential VAS. Two devices running at 10mA (amounting to a similar end result) with a mirror at the other end would give you a 20mA peak capability with a reasonably linear 10mA due to the differential operation. But, these designs are often more difficult to stabilise.
Even with doubling up, each device would be dissipating 500mW so using a small heatsink is still advisable, although within the device ratings. (And what ambient temperature you might have).
Are you considering something along the lines of a Blameless (Self's "standard" approach) or something else?
Looking at the data sheet comparison between the MJE34/35 and C4793 reveals some differences in the fT:Ic characteristics. The C4793 has a higher fT overall, so is faster, and clears 20MHz at only 5mA. The MJE needs 10mA and seems to roll off more quickly below that (although the C4793 datasheet cuts off at 5mA).
It may be that the MJE needs to be run at higher currents to keep the fT high, as a low fT may well impact the slew rate and possibly be a cause of harshness that has been reported above.
With that in mind, I suspect that MJE drivers would not be a problem if the current range is, say, in the region of at least 25mA to 500mA, which should be enough for two pairs of MJL23/4.
I'd recommend you consider a 3EF as well. One configuration I have is C2911/A1209 - C4793/A1837 and three MJL3281A/1302A. Works well, and if repeating that now I'd probably use C3503/A1381 -MJE15034/5(probably at 50mA bias) - same output. You might also be able to use ZTX458/558 pre-drivers, firmly bolted down on the same heatsink as the others, but can't guarantee that because there is no thermal resistance info to know how safe that would be. The datasheet value works out to be 175C/W junction to free air, so junction to case is likely to be around 20C/W, but may be higher. That may make temperature stabilisation of the quiescent current a little less stable than using a TO_126 or TO_220 device, which may well be 10C/W junction to case.
(this is not because of the dissipation in the device, but simply to keep the heatsink to junction temperature closely aligned)
Even with doubling up, each device would be dissipating 500mW so using a small heatsink is still advisable, although within the device ratings. (And what ambient temperature you might have).
Are you considering something along the lines of a Blameless (Self's "standard" approach) or something else?
Looking at the data sheet comparison between the MJE34/35 and C4793 reveals some differences in the fT:Ic characteristics. The C4793 has a higher fT overall, so is faster, and clears 20MHz at only 5mA. The MJE needs 10mA and seems to roll off more quickly below that (although the C4793 datasheet cuts off at 5mA).
It may be that the MJE needs to be run at higher currents to keep the fT high, as a low fT may well impact the slew rate and possibly be a cause of harshness that has been reported above.
With that in mind, I suspect that MJE drivers would not be a problem if the current range is, say, in the region of at least 25mA to 500mA, which should be enough for two pairs of MJL23/4.
I'd recommend you consider a 3EF as well. One configuration I have is C2911/A1209 - C4793/A1837 and three MJL3281A/1302A. Works well, and if repeating that now I'd probably use C3503/A1381 -MJE15034/5(probably at 50mA bias) - same output. You might also be able to use ZTX458/558 pre-drivers, firmly bolted down on the same heatsink as the others, but can't guarantee that because there is no thermal resistance info to know how safe that would be. The datasheet value works out to be 175C/W junction to free air, so junction to case is likely to be around 20C/W, but may be higher. That may make temperature stabilisation of the quiescent current a little less stable than using a TO_126 or TO_220 device, which may well be 10C/W junction to case.
(this is not because of the dissipation in the device, but simply to keep the heatsink to junction temperature closely aligned)
Last edited:
EF2 VAS Options
When using paralleled outputs in an EF2 design, the driver obviously works harder and so does the VAS. The requirement to run these stages at higher currents will also increase their respective power dissipation. Selecting a suitable VAS transistor can be quite tricky in this situation.
If you decide on a single VAS stage with CCS at the other end, then as previously mentioned you will need to run somewhere close to 20mA. This forces your hand somewhat as you need a high voltage medium power transistor for the VAS. This is not ideal however, many have done just that and produced great results. The KSC3503 is a reasonable option due to its low Cob value. You will need to heatsink the VAS in this situation.
Alternatively, you could use a Cascoded VAS at the expense of some headroom. In my opinion the benefits are well worth it, allowing you to choose a much better VAS transistor. This shifts the power dissipation away from the VAS to the Cascode BJT where it has less affect sonically. The biggest advantage of the Cascode is that it holds the collector of the VAS at a fixed voltage therefore eliminating the nonlinear affects caused by the VAS collector-base junction capacitance. Its a beautiful thing 🙂
Cheers
Paul
When using paralleled outputs in an EF2 design, the driver obviously works harder and so does the VAS. The requirement to run these stages at higher currents will also increase their respective power dissipation. Selecting a suitable VAS transistor can be quite tricky in this situation.
If you decide on a single VAS stage with CCS at the other end, then as previously mentioned you will need to run somewhere close to 20mA. This forces your hand somewhat as you need a high voltage medium power transistor for the VAS. This is not ideal however, many have done just that and produced great results. The KSC3503 is a reasonable option due to its low Cob value. You will need to heatsink the VAS in this situation.
Alternatively, you could use a Cascoded VAS at the expense of some headroom. In my opinion the benefits are well worth it, allowing you to choose a much better VAS transistor. This shifts the power dissipation away from the VAS to the Cascode BJT where it has less affect sonically. The biggest advantage of the Cascode is that it holds the collector of the VAS at a fixed voltage therefore eliminating the nonlinear affects caused by the VAS collector-base junction capacitance. Its a beautiful thing 🙂
Cheers
Paul
I agree. As I mentioned, I used a cascoded 2SC2238 for the VAS. That was in a differential configuration. If you run amplifiers with a response to DC one of the less obvious advantages of the cascode is that the active transistor has much less power dissipation, and therefore less drift on warming up than a single transistor stage. With a differential VAS and current mirror, drift is also reduced anyway if the mirror transistors are thermally coupled to each other. Preferably in the Wilson configuration which again keeps the high dissipation away from the mirrors.
If, and that was the purpose of my question about the type of design, the amplifier is to be AC coupled then drift is less problematic.
You can always offset the voltage headroom lost by using separate higher voltage power supply lines.
My recommendation for this is that it can be generated using voltage doublers and a simple stabiliser rather than adding a second transformer or having to modify a toroid or specifying a custom design. The current demand is quite low so does not need a massive heatsink on the stabilier devices.
An additional bonus for that is that clipping can also be made less harsh by clamping to the main power rails (as long as VAS currents are limited).
If, and that was the purpose of my question about the type of design, the amplifier is to be AC coupled then drift is less problematic.
You can always offset the voltage headroom lost by using separate higher voltage power supply lines.
My recommendation for this is that it can be generated using voltage doublers and a simple stabiliser rather than adding a second transformer or having to modify a toroid or specifying a custom design. The current demand is quite low so does not need a massive heatsink on the stabilier devices.
An additional bonus for that is that clipping can also be made less harsh by clamping to the main power rails (as long as VAS currents are limited).
Last edited: