Need low power high beta transistors - please drop me a few part numbers...

Jump to Latest
peufeu

peufeu

Member
Joined 2001
audio.peufeu.com
#1
This is for a DAC I/V, using transistors and cascodes in the jocko style. Since it is open loop the main source of error is transistor base current.

I have spent a few hours hunting transistors and have so far found the perfect one, except noone sells it.

So I'd be very grateful if you could share your finds. Here are the prerequisites :

- Vce max : 30V, maybe 20, but I don't need a high voltage part unless it has other advantages.
- Maximum current 100 mA
- Power dissipation 0.2 W (will fry SOT323 chips)
- PNP and NPN, complimentary part would be nice

Now before someone says "BC547C" :

- high hFe (>2-300)
- constant hFe between 5 and 30 mA with the curves to prove it
- LOW CAPACITANCE

See these graphs :

- BC547C gives up long before 10 mA
- BC317/337 would be good but it oscillates in Spice
- ZTX851 starts its linear region AFTER 10 mA
- MMBT489 is perfect but too low power dissipation and high capacitance

Thanks to anyone who can help !

An externally hosted image should be here but it was not working when we last tested it.
 
#2
BC557C. Easily obtainable. I just picked one at random and obtained this - hfe = 400 @ 10mA.
 

Attachments

  • bc557c 10ma.gif
    bc557c 10ma.gif
    13.2 KB · Views: 741
#3
That was a quick reply, but unless the datasheet curve is wrong, this transistor isn't suited for the application... look more closely at the hFe versus Ic curve that I posted ; it has a knee at about 10 mA.

Using BCP56 or BC317-40 which has a flatter curve yields about 4x linearity improvement, but 317 oscillates in the simulator. BCP56 might be suitable but it has large capacitance.

Can your gear trace hFe vs Ic (this was Ic vs Vce for various Ib, right ?)
 
#6
peufeu said:
That was a quick reply, but unless the datasheet curve is wrong, this transistor isn't suited for the application... look more closely at the hFe versus Ic curve that I posted ; it has a knee at about 10 mA.

Can your gear trace hFe vs Ic (this was Ic vs Vce for various Ib, right ?)
Click to expand...

No, I can't plot hfe against Ic, but I can measure hfe at spot currents. For the transistor I measured at Vce = 2.76V, hfe = 400 @ 10mA, falling to 363 @ 18.2mA. If you increase Vce, the fall becomes less noticeable due to Early effect, so when I increased Vce to 6V, hfe = 432 @ 13mA and 442 at 11mA. Yes, I saw the claimed fall in hfe in the curves you posted. Me, I believe what my own measurements say under my own conditions.
 
#9
If some transistor produces oscillation in the simulator, don't discard it, blame the model instead.

Also, simulation becomes completely inaccurate when trying to predict exact gains and capacitances of bipolar transistors. Models are not that good.

And last but not least important, given the same part number, gain magnitude and linearity and capacitances change from one manufacturer to another and from batch to batch, so a curve tracer is strongly recommended instead of simulations.
 
#10
I looked at the datasheets and 2SA1845 might have potential.

Eva : I know. Some models are quite suspicious, especially BC317/327/337. I will test them in circuit when I make the test boards and we'll see.

So far JFET simulations fail to bring improved linearity ; it's difficult to find the right PJFET, too.
 
#11
#15
Alright !

I'm considering ease of procurement (ie. I have a huge electronic parts shop 5 minutes from my place so if they stock it at the right price (ie. 0.2 euro) I'll be a lot happier than getting them from DigiKey !)

So :

2SC2240 - This one looks very good, and it has a complimentary, 2SA970. Pair costs 0.60 EUR. So far, these win !

2SC3467F - Would be very suitable for cascoding but needs 3-5 volts headroom which I'm not sure to have

2sb716/d756 - hFe gives up at too low Ic
2sa1085/c2547 - idem
2sa872/c1775 - idem
2sa1350 - datasheet doesn't specify curves
2sa1815/c4432 - hFe gives up at too low Ic
2sa1845 -good characteristics except 130 pF collector capacitance
2sc373, 2sc372 - can't find datasheets
Mpsa12, Mpsa13 - I should try darlingtons but there are no Spice models available.
 
#16
And remember all those data sheet curves are TYPICAL as characterised from samples.

They are a sort of design aim.

Unless there are enough datapoints established in the test spec (ie hfe at 0.01, 0.1, 1, 10, 100ma guaranteed) they should be treated with caution.
 
#18
should be treated with caution
Click to expand...

Yes ; but I need to pre-filter or else this project will never be done !

2SA1015-BL and 2SC1815-BL
Click to expand...

Very nice. These curves are so straight they look like they were hand drawn by the marketing department.

Thanks.

I need to find or build some gear to measure transistors. Problem is, you have to use short pulses or else you measure self heating instead of early effect, so it's a lot more complicated than it looks.
 
#19
heh, heh, that's a pretty funny quote. you're assuming the marketing types can even draw a straight line 🙂

looks like no complement for the highest hFE grade 2SC1815 though. is that a problem for your application?

peufeu said:

... Very nice. These curves are so straight they look like they were hand drawn by the marketing department ...
Click to expand...
 
#20
heh, heh, that's a pretty funny quote. you're assuming the marketing types can even draw a straight line 🙂
Click to expand...

Maybe I spoke too fast. If marketing was involved, the datasheets would come in powerpoint format.

Anyway, by using 2SC2240 and its complimentary instead of BC547/557 I get 10-12 dB improvement in distortion in the simulator. If we can trust the sims, this looks like a free lunch.

This is for a simple folded cascode IV without fancy tricks.
 
You must log in or register to reply here.
Top Bottom