How know bjt transistors are fast w/o testing?

How do you know the VAS or driver transistors are fast without buying & testing?
People are always speifying BD139-140 (tgm8) or 2SA1837 2SC4793 (apex AX6).
I'm in the USA. newark has BD139-149 but they are from fairchild. The datasheet guarentees no Ft or output capacitance. Neither does the Phillips datasheet I downloaded. If they just put new paint on a TIP41 how would I know? Aside from the package. I'm already using TIP41C, they are about $.26 instead of $37 for BD139-140 but TIP41C Ft is 3 mhz which is not 30 or 70. TIP41 has more SOA current. Which is important when I do stupid stuff like yesterday leaving apex6 prototype on the lightbulb for 30 minutes when I thought I had plugged in the soldering iron.
Digikey has 2SA1837 2SC4793 for $.85 but they are from digital technology, not toshiba. When you download the datasheet from digikey, you get the toshiba datasheet, not the digital technology one. 70 mhz Ft sounds good but trying to buy real toshiba stuff puts you into the e-bay/counterfeit market.
Peavey designed the PV-1.3k with 2sa968 2sc 2230 which are 100 mhz Ft and c0b of 30 pf, but they did a service bulletin to change to MJE15032/33 at 30 mhz Ft because maybe they could get real ones from on semi instead of imitations specified with a silk screen and paint?
 

Mark Johnson

Member
Paid Member
2011-05-27 3:27 pm
Silicon Valley
Maybe it's worth asking yourself, what's the simplest circuit you can build which correctly answers the question:
  • Here are two NPN transistors, X and Y. Which one has the larger fT?
You could then have an fT tournament between 3 BD139s, 3 2SC4793s, 3 2SC2230s, 3 2SC3601s, 3 KSC2690As, and 3 MJE340s. Find the highest fT among the diyAudio preferred candidates of TO-126 / TO-225 medium power packaged transistors. Use three examples of each type to avoid categorization mistakes due to random weaklings.
 
Last edited:
The question as originally put was about vas transistor selection. I think that the appended "fast" didn't help a lot, as it is a bit of a red herring.

A good vas transistor does have a number of attributes, and I a sure there are numerous threads on this topic and people willing to throw their 2 bits worth in.

Low Cre, that capacitance that adds to the miller feedback capacitance is a BIG deal with a vas transistor.

I note you quote bd139/140. These are venerable transistors that fall right in between power devices like the mje340 and more typical devices. To cite another Philips type, the bf469 /470 ticks the typical vas boxes.

Things that I would expect to see are
- low capacitance.
- fairly low current and power, which relates to the low capacitance
- you need a decent voltage rating, as this device sees the full rail to rail voltage, and in this regard the bd139 is limited.

That said pretty much any transistor will "work". Just look for behaviour coming out of clipping on negative excursions (mje340 is rubbish) and the devices can't a orange. Should be a few pf
 
But you have picked that one specification out of the list as if that was the salient factor. Peavey designed the amp with a 2SC2230? OK, that is a TO92 (large version) with a 160v rating, 100ma current rating, and 800mw dissipation. They replaced that part with the MJE15032? That is a TO220 part, with ratings of 250v, 8 AMps, and 50 watts dissipation.

I am somehow left with the impression that frequency was the last thing on their minds. They changed to a much heftier part for reliability.
 
The question as originally put was about vas transistor selection. I think that the appended "fast" didn't help a lot, as it is a bit of a red herring.
A good vas transistor does have a number of attributes, and I a sure there are numerous threads on this topic and people willing to throw their 2 bits worth in.
Low Cre, that capacitance that adds to the miller feedback capacitance is a BIG deal with a vas transistor.
I note you quote bd139/140. These are venerable transistors that fall right in between power devices like the mje340 and more typical devices. To cite another Philips type, the bf469 /470 ticks the typical vas boxes.
Things that I would expect to see are
- low capacitance.
- fairly low current and power, which relates to the low capacitance
- you need a decent voltage rating, as this device sees the full rail to rail voltage, and in this regard the bd139 is limited.
That said pretty much any transistor will "work". Just look for behaviour coming out of clipping on negative excursions (mje340 is rubbish) and the devices can't a orange. Should be a few pf
Thanks for that response.
Interesting, the mje340 is popular among the speaker capacitor amps, like recent design Apex AX6. No Ft or Cob spec. Speaker capacitor=primitive design, IMHO, all except for part about it won't burn your $600 speaker if one solder joint lets go.
The Bd139/140 is used in the Bigun TGM8 and the JohnE Basic 50. Pity the speed or capacitances are completely unspecified.
The "coming out of clipping part from negative excursion" scenario is why I haven't heard any problems with these obsolete RCA TO5 VAS in my dynaco ST120. Neither the slow old SK something RCA transistor I replaced it with This amp, with 90 VAC power trasnformer for 60 w/ch, at <2 vac input NEVER clips. The power is heat sink limited, not voltage excursion limited.At least since I bought it and put the 20 gain NTE60 output transistors in. In 1966 with 5 gain 2N3055 derivitive output transistors the best available, that "rediculous" rail voltage was needed to get 21 VAC out into 8 ohms.
Sorry, Enzo, over 65 year old eyes error, the PV-1.3k had 2SC2238B/2SA968B TO220 predrivers with 100 mhz Ft. Scrap plastic now, along with 141 other parts.
If I get this AX6 point to point copy working without oscillating, I'll compare the 3 mhz TIP41C I've got in as VAS against BD140 I have in the parts box. On tinkly bells and top octave piano, I imagine, to stress the high frequecies. My ears go to 14 khz, as do the SP2.XT speakers.
 
Last edited:
Use three examples of each type to avoid categorization mistakes due to random weaklings.
Sometimes it also requires using different date codes as well. I've come across hundreds of similarly bad parts from the same date code in one shipment..at least the manu and distribution were consistent..

I've also come across transistors where exactly 25% of the devices were "different", unable to slew collector voltage low to max under constant current operation without blowing up, a melt under an emitter finger being the problem.
Turned out to be a mask problem at the wafer level.

For power devices, I've seen entire lots of product with bad die attach or bad wirebonding.

John
 
For power devices, I've seen entire lots of product with bad die attach or bad wirebonding.
John
Farnell US sold me some NJW21194G with Pd of 200 mw instead of 200 W. That wasn't a misprint, when I put 85 volts on them at idle current they shorted across. They were $1.94 instead of $4 for real MJLxxxx TO247 parts. These were genuine On semi Korea rejects, I suppose. Farnell doesn't sell counterfeits, and they did warn me in the selector table. Stupid me I bought 30 of them before I installed 20 in the PV-1.3K. I don't know what they are good for. If you paint dots on them I suppose you could make miniature dominos of them.
 

mt490

Member
2010-06-09 2:32 pm
I am currently using 2SA2205 and 2SC6099 as my driver transistors in my latest amplifier. Certainly has guaranteed fast included (too fast?), although capacitance is nothing special.

Previously I was using A1708/C4488 - some similarities to the BD139/140 but much more thoroughly specified.
 
For VAS, Cob and Early Voltage is important. Please read Samuel Groner comment on Douglas Self's book.
I use 2SA1381 and 2SC3503 for VAS if I want high quality amplifier.

For output transistor, if you want high slew rate amplifier, you should use high fT BJT or it will blown like Thimios or Still4given on slewmaster thread.

Example of my modification of AX16(AX14) who built by Still4given: http://www.diyaudio.com/forums/solid-state/164093-100w-ultimate-fidelity-amplifier-572.html#post4547111 You can compare the specification or listening test impression with original design.
 
The 2SA1381 and 2SC3503 datasheets from Fairchld have actual Cob and Ft numbers, 3pf and 150 mhz. Mouser shows as obsolete, digikey shows as non-stock, allied elec has no listing and newark(farnell) is substituting On semi MJE340-350 which have no Cob or Ft spec and googlyone has already pointed out are "rubbish".
I'm happy you can buy real toshiba and sanyo transistors where you are . The only vendors of those brands I have seen here, mcmelectronics doesn't label the part number with vendor in catalog or packaging. Digikey downloads datasheets from toshiba but tells you the part is made by someone else. Other vendors I view as likely counterfeiters with a good paint gun.
The 100 W ultimate fidelity amplifier quoted above is a split supply "Speaker Toaster" and there is NO chance I'm going to connect transistors directly to my $600 each speakers. I had a bad solder joint short a driver transistor across last week, collapsing Vcc to 4.6 v through the light blub. My point, introducing a speaker capacitor in the design leaves one in the dark ages in amp design, 1970 at latest. Even mike Slavonic used a MJE340 in his speaker capacitor design AX6 which has no cob spec and is "rubbish" in googlyone's post.
bf469/470 also come up no record at newark, digikey, and mouser.
To googlyone's point, BD139/140 with Vceo of 80 are pretty light duty considering I'm installing AX6 in a amp with 70 VAC transformer, or 90 v peak. However, I did actually manage to buy those from newark. Am building a voltage regulator board to 72 v today, using 6 1.3 watt 12 v zeners stacked and five parallel TIP142 as regulators for maybe 20 amp peaks on cannon shots in 1812 overture
 
Last edited:

Mark Johnson

Member
Paid Member
2011-05-27 3:27 pm
Silicon Valley
Rochester sells Fairchild 2SC3503F ("F" = highest beta bin) for $0.20 and the 2SA1381F for $0.12. (link)

I bought fifty of each back in 2014 when prices were higher. I paid $0.41 and $0.25 respectively. The invoice is copied below, see line items 6 and 7.

If you wish, I would be glad to sell you any amount up to 20 pieces of each, although I prefer that you support the new-old-stock ecosystem by purchasing from Rochester instead. To encourage you to source from Rochester, my price is $2.50 per transistor, either type, plus $5.00 per order for the shipping envelope and for my time spent at the post office waiting in line, plus $10.00 per order "convenience fee" if you prefer to use PayPal. If you pay by check or money order or cash, disregard the PayPal fee. Send me a PM with your order.

_


493196d1436713918-john-curls-blowtorch-preamplifier-part-ii-an_order_conf_email.png
 
Okay, thanks Mark Johnson.
I checked with octopart/rocelec.com (rochester) and they have a minimum line charge of $100. They are worse than hamilton/avnet. So I don't think I can use 910 each of 2SA1381/2SC3503 in my life time, as much as I am hoarding leaded parts for my furture projects.
I found 1 each Sanyo 2SA1352 for $.64 and 5 each Toshiba 2SC3423 for $1.38 at mcmelectronics. Last ones, guys. I needed a DTV converter from mcm anyway, so five channels of Apex AX6 is taken care of. We'll see if it sounds better than TIP41C in there now, with Cob under 3 pf. mcm is farnell, so even if the bag label describes only the part number and quantity as they do, I'm not out much money. Mcm are probably not counterfeit.