BJTs for VAS Stages

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
In the old days, I used 2N3440 & 2N5415 transistors. Obviously, these are not a particularly contemporary choice.

I haven't seen a lot of discussion on transistor choices for anything but the output and driver transistors.

I like the availabilty (at least in the US) of the ON Semi parts and I can be confident that they are "real"

What are some favorite complementary pairs?
 
A fast and low cob part is most ideal. I personally love to use either the
800mw 2sc2705/a1145 parts or the 5w 2sc3423/a1360 for the vas 1.8pf for the npn/2.5pf for the pnp. These are 200mhz 50ma devices with an average Y-beta of about 200.. I see most designs using 1amp parts but these are emtremely unneccesary not to mention have an average collector capacitance anywhere from 10-30pf..


Colin
 
danville said:
In the old days, I used 2N3440 & 2N5415 transistors. Obviously, these are not a particularly contemporary choice.

I haven't seen a lot of discussion on transistor choices for anything but the output and driver transistors.

I like the availabilty (at least in the US) of the ON Semi parts and I can be confident that they are "real"

What are some favorite complementary pairs?



but did you do any searching?

like for instance "vas" + "search in tiltle"?

maybe you would have picked up on:

http://www.diyaudio.com/forums/showthread.php?s=&threadid=66234

(admitedly I cheated, rather than look thorough all 82 threads I also know to look for AKSA as poster)

(which also brings up the apparent issue that "search in title" currently seems to pick up on more than just Thread Titles - should it do this?)
 
Yeh, I searched VAS but I missed the earlier thread. Too many irrelevant hits.....

I've looked at many of the suggestions and they look good but most are thru hole parts. I plan to build my VAS stage with SMT parts. This really restricts the recommendations and most of the parts don't seem to have SMT equivalents.

I haven't seen a selector guide from either Sanyo or Toshiba. This might make it a little easier to search.
 
A fast and low cob part is most ideal. I personally love to use either the
800mw 2sc2705/a1145 parts or the 5w 2sc3423/a1360 for the vas 1.8pf for the npn/2.5pf for the pnp. These are 200mhz 50ma devices with an average Y-beta of about 200.. I see most designs using 1amp parts but these are emtremely unneccesary not to mention have an average collector capacitance anywhere from 10-30pf..

Colin


I have looked at a lot of transistors since last night and the Toshibas that Colin mentioned would be excellent for my purposes. I haven't found anything better on paper than the 2SA1145/2sc2705 pair.

The only catch I see is that there doesn't seem to be SMT equivalents. I have looked at Sanyo, Zetex, Toshiba, ON-Semi & Fairchild so far for similar performance. This seems a bit odd since high volume today is in SMT.

Any new suggestions?
 
Here's a picture of a doubled 2SC2705 in the Vas stage of an expensive buggar from the 90s.
A single 2SC3955(+2SA1540) only costs 10 cents more.
 

Attachments

  • doubled 2sc2705 vas.jpg
    doubled 2sc2705 vas.jpg
    55.2 KB · Views: 1,469
If you prefer lower power VAS as I do, these are from Fairchild, which can be ordered from Mouser here in the US.

http://www.fairchildsemi.com/ds/KS/KSA1174.pdf
http://www.fairchildsemi.com/ds/KS/KSC2784.pdf


http://www.fairchildsemi.com/ds/KS/KSA992.pdf
http://www.fairchildsemi.com/ds/KS/KSC1845.pdf

Now if you like to cascode your VAS as with a common base, were Vce is of non-issue, the 5089 has Hfe, I measured almost 600!
http://www.fairchildsemi.com/ds/KS/KST5088.pdf

Or the good old BC850 / BC860

I'm sure there are those who find these lacking in power for this purpose, but I like to use a smaller VAS and utilize current gain of the output stage.:warped: Lower Ic and higher Hfe in the VAS stage, means less current change in front end, depending on design of course.
:)

:2c:
 
I want to thank everyone we has replied so far. I thought this would be an easier process:

Here is an update of some of the issues I have looked at. These may apply to many other designs.

My VAS uses cascode circuits and also wilson mirrors. The wilson mirrors are configured to have some gain (about 2.5). This means several things:

1. The only transistors that have a large Vce requirement are one of the transistors (the non-matched one) of each wilson mirror.
The wilson mirrors drive the bias voltage source as typical in most designs. This means that the high voltage transistors have the highest current requirement as well. Maybe a TO-92L might be OK but I would prefer an SOT-89 or larger.

2. The remaining transistors have relaxed requirements. None of the transistors will need high Vce (60-80V would be OK). I think that an SOT-23 might be OK even though they can dissipate about 300mW. The Ft of the input transistors are also less critical since the cascode configuration helps in this regard.

3. The Toshibas parts are less desirable in my case since I would probably have to buy a whole reel if SMT. Fairchild & ONSemi are good choices for Americans since Mouser and DigiKey carry these lines.

On a different note, I am considering the NJL3281/1302 for outputs since the internal diode would be convenient for bias tracking. I used to clamp the bias transistor to a TO-3 case in the past. This is much better than mounting on a heatsink, but not as good as being on the same die as the transistor.
 
I have finally decided on a set of VAS transistors

I was looking for SMT packages that are easy to procur (at least in the US)

Here are the results:

BC850C - BC860C (Fairchild). These are SOT23 versions of the popular BC550/560. They can't handle the same current however. I am using these for 2/3 of a wilson mirror and the bias circuit (assuming I don't use a BC560 glued to the output Q). I think I will use the embedded diode in a NJL1302.

BCP56-16/BCP53-16 (ON Semi). These are not as fast as some of the TO-92 suggestions but handle more power. They are in SOT-223 packages. I have cascode circuits which allow the 80V Vce to be adequate.

KSA1381/KSC3503 (Fairchild). I gave up finding an SMT part that was good enough. These are 200-300V parts that have an Ft of 150MHz. You can buy F versions with high beta. They come in a TO-126 package. The only SMT part I found was the MJD340/MJD350 (ON-Semi). These are SMT variants on the MJE340/350. The high frequency characteristics seem to be a secret with these parts, which suggests they are not too special. The beta was not real high either.

I am thinking of building a VAS module that I can use with a variety of output stages. My VAS design should be good for half bridge amplifiers as large as 250 - 300 Watts (75 - 80V rails)
 
danville said:
I am considering the NJL3281/1302 for outputs since the internal diode would be convenient for bias tracking. I used to clamp the bias transistor to a TO-3 case in the past. This is much better than mounting on a heatsink, but not as good as being on the same die as the transistor.

danville,

What Thermaltrak bias circuit will you use? The OnSemi application note shows a simple string of several of the diodes, while Leach suggests a Vbe multiplier transistor on the main PCB plus thermal diodes from the heatsink in its C-B bias leg. Leach feels the RC in this leg provides more stability than using long wires to the Vbe multiplier transistor.

I've been debating using ThermalTrak with the Leach circuit, but the specs on the Sanken 1215/2921 specs look attractive for my low voltage application, unfortunately... much more expensive.
 

Attachments

  • vbemult.gif
    vbemult.gif
    2.3 KB · Views: 973
I've never done it with an external diode. My thought was to add a single diode in series with the emitter. The BE junction of the transistor will be fairly constant if it is not near the heatsink. The variation will then be caused by the diode.

I use a two transistor circuit. The second Q just lowers the impedance.

I can tell you that you can control bias quite well when the tracking transistor is mounted directly on an output Q (not on the heatsink). The diode should be better.

I don't like the whole string of diodes method. This doesn't give you flexibility in setting the bias point in a fairly arbitrary way.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.