I've read and searched but haven't managed to find a satisfactory answer to this most basic of basic questions:
You have a solid state amplifier circuit designed, you know the current and voltage and power dissipation of the transistor .. this narrows the field down to a certain package, say TO-92 or TO-126, and if you have unusually high voltages this will limit the choice somewhat, but if you working with low voltages as is often the case you are looking at close to a dozen valid options. How do you go about narrowing down the choice from there?
The biggest differentiator that I can see is Hfe, often a given type will have 3 variants, low-med-high current gain. This is probably a really obvious point but I can't figure out why you'd want to use anything other than the high current gain types.
Then there is noise figures. Then there is "general purpose" vs. "audio" transistors, which here again I'm not fully appreciating what it is that makes up the difference.
Should I just use standard BC557B/547Bs (if they match the basic circuit requirements re. power etc.) or should I be really concerned about optimizing?
You have a solid state amplifier circuit designed, you know the current and voltage and power dissipation of the transistor .. this narrows the field down to a certain package, say TO-92 or TO-126, and if you have unusually high voltages this will limit the choice somewhat, but if you working with low voltages as is often the case you are looking at close to a dozen valid options. How do you go about narrowing down the choice from there?
The biggest differentiator that I can see is Hfe, often a given type will have 3 variants, low-med-high current gain. This is probably a really obvious point but I can't figure out why you'd want to use anything other than the high current gain types.
Then there is noise figures. Then there is "general purpose" vs. "audio" transistors, which here again I'm not fully appreciating what it is that makes up the difference.
Should I just use standard BC557B/547Bs (if they match the basic circuit requirements re. power etc.) or should I be really concerned about optimizing?
Every transistor has a mix of properties in varying ratio that can affect optimum performance in their role and hence the overall performance of the amplifier. Generally we focus on the noise, linearity, bandwidth & power handling qualities in a particular duty. These vary in every discrete parts design but if the price and performance of the product can be no different by using just a few strategically designed high performance types, these tend to be used almost universally.
Small-signal types like the obsolete 2SC2240/A970 and KSC1845/A992 come to mind as excellent general purpose types for high quality, class AB amplifiers. Note that voltage ratings preclude the use of many BCXXX types and the choices for new stocks of all TO92 transistors is shrinking every year, so unless you use old stock or cheap copies, the choices are actually not as wide as we may think. YMMV
To have more than an overview of transistor design and parts selection requires serious study of electronics and that's what electronics engineers train for. Textbooks like "The Art of Electronics" are standard basic texts for this but there are probably equivalents in your own language. https://www.google.com.au/url?sa=t&...=CWPRpDNFly8k6ougbZWw3w&bvm=bv.85970519,d.dGY
There are on-line basic courses in electronics in most languages but you may come to the realization that specific skills and knowledge are necessary for individual design analysis and part specification. That is not something that can be taught in a few forum posts so you will need to study in depth or just copy what the other guys do by using simulation programs to perform the calculations and eliminate the learning curve.
LT Spice and TI Tina are respectively, examples of advanced and basic free sim. programs that can help you understand the requirements and prove your designs and parts selections without buying anything. Try TINA V9 from the TI website, if this fast-track type of learning appeals more to you. 😉
Hi Ian, thanks for your reply.
Horowitz and Hill has had a place on my bookshelf for many years, and I'm reasonably well-versed in LTSpice. Actually the reason I posted this thread is because I found I could substitute many different transistor models into my circuit in LTSpice with little obvious effect.
If I could draw you on specifics, what it is about them that makes the 2SC2240/A970 better than the BC547/557?
For reference, the circuit I'm mulling over is the headphone buffer shown below.
Horowitz and Hill has had a place on my bookshelf for many years, and I'm reasonably well-versed in LTSpice. Actually the reason I posted this thread is because I found I could substitute many different transistor models into my circuit in LTSpice with little obvious effect.
If I could draw you on specifics, what it is about them that makes the 2SC2240/A970 better than the BC547/557?
For reference, the circuit I'm mulling over is the headphone buffer shown below.
Attachments
Your circuit has a requirement for some low noise, high gain types at the input stage and higher current types at the ouput. BC547 is normally considered to be a general purpose amplifier and switch - not a low noise, low level amplifier. You would probably benefit from using BC549 or BC550 at the input stage instead.
At low voltages, there is no reason to use improved high voltage types like 2SC2240 but there are others, such as 2SC1815/A1015, that are equally good or perhaps better here. The main concern at the input stage is lowest noise/high Hfe
Rather than use multiple output transistors, you could also use a single driver pair, like BD139/40, BC327/337, BC639/40 etc. or even 2SD669/B649 for their better overload capability.
The problem we all face with simulation is the inaccuracy of models and access to good ones. If you find all BC transistors give the same simulated result, it's probably because they all use the same generic model. That is, you need to look elsewhere for good models. For power amplifiers, Bob Cordell's site offers quite a few models that he has reworked to the best accuracy he can. How you add these Spice models to your simulator's library and distinguish them from the generic models will depend on the software but you can see on the many threads here, how they have been downloaded and added to LT Spice simulations here. Just download an appropriate ASC. file and see or follow the LTSpice WIKI.
As an example of how different the transistors really are, look at the datasheets for original BC549 V BC547: Philips - datasheet pdf
Philips - datasheet pdf
Noise figure is 2-10 versus 4 maximum and this significant, even for these standard, low gain types with no suffix.
Last edited:
I think that is a very good question. I have been asking very similar question after about 3 months reading and studying SS audio amplifier designs. Other than the basic parameters like Vce, power dissipation, beta(at different current), fT, input and output capacitance, I really don't get why people choose one transistor over the other. I honestly do not see why everyone say BC560 is better fit than KSA992 for front end use. I know I can draw load line on the collector curve to look for linearity. BUT we all know to use emitter degeneration not only to take the transistor linearity out of the picture, degeneration on the IPS help to increase slew rate of the VAS. So why one transistor is better than the other if the basic parameters mentioned above are similar?
Back to this circuit, correct me if I am wrong. The circuit has gain of unity. Q1 and Q2 are EF driving a composite emitter follower. Total gain is unity at best. Why noise comes into the picture?
Matter of fact, why noise of the IPS input transistor even that important? You only have gain of say 20. I can understand noise can be an issue if it is the preamp or a phono preamp. Why noise is important for power amp?
From looking at the circuit, it seems to me that it would be better serve if there is an adjustment on the voltage driving the power stage. It's like the Vbe multiplier in conventional VAS stage to adjust the bias current to optimize cross over distortion. Again, this is just my inexperience observation.
Back to this circuit, correct me if I am wrong. The circuit has gain of unity. Q1 and Q2 are EF driving a composite emitter follower. Total gain is unity at best. Why noise comes into the picture?
Matter of fact, why noise of the IPS input transistor even that important? You only have gain of say 20. I can understand noise can be an issue if it is the preamp or a phono preamp. Why noise is important for power amp?
From looking at the circuit, it seems to me that it would be better serve if there is an adjustment on the voltage driving the power stage. It's like the Vbe multiplier in conventional VAS stage to adjust the bias current to optimize cross over distortion. Again, this is just my inexperience observation.
Last edited:
Thanks again Ian. You've helped me a lot.
The circuit is part of a headphone amplifier. It will be driven directly by an OPA134 (until the day comes when I can design a discrete voltage stage that beats it). In principle the op amp will determine the output noise ... I expect, at any rate, but since I'm not completely sure and since it's no sweat to use BC549/550 it seems prudent to. The thing is since its a unity gain buffer, I wonder if just replacing the input stage transistors would be useful, or whether I'd need to replace the full set.
I was going to use BD135/136 for the output transistors, the parallel version with TO-92 is an alternate. Both options are on the table at present.
My own concern with simulations with LTSpice is not so much the accuracy of the model, but that it shows you only what you care to look at. I mean, you get the distortion figures, and you try to minimize them, and before you know it all you end up doing is optimizing for lowest distortion ... things like thermal stability, PSRR, input impedance and other practical concerns are forgotten about. Still, as a virtual bench - to put things together and see how the different circuit elements work - it's a wonderful tool.
The circuit is part of a headphone amplifier. It will be driven directly by an OPA134 (until the day comes when I can design a discrete voltage stage that beats it). In principle the op amp will determine the output noise ... I expect, at any rate, but since I'm not completely sure and since it's no sweat to use BC549/550 it seems prudent to. The thing is since its a unity gain buffer, I wonder if just replacing the input stage transistors would be useful, or whether I'd need to replace the full set.
I was going to use BD135/136 for the output transistors, the parallel version with TO-92 is an alternate. Both options are on the table at present.
My own concern with simulations with LTSpice is not so much the accuracy of the model, but that it shows you only what you care to look at. I mean, you get the distortion figures, and you try to minimize them, and before you know it all you end up doing is optimizing for lowest distortion ... things like thermal stability, PSRR, input impedance and other practical concerns are forgotten about. Still, as a virtual bench - to put things together and see how the different circuit elements work - it's a wonderful tool.
Another parameter not mentioned so far. Early Voltage.
http://en.wikipedia.org/wiki/Early_effect
The Early voltage is rarely stated.
But it can be inferred from the Ic vs Vce plot, if shown in the datasheet.
Early voltage tends to havea loose proportionality to Vce0.
Looking at the Ic vs Vce plot (see Fig2) you see the graph of Ic rising steeply from 0Vce to a few Vce. Then it starts turning the corner and becomes a gently rising slope of increasing Ic as Vce goes past 5V all the way up to it's maximum Vce.
The slope of that top part of the Ic plot is directly proportional to the Early voltage (EV).
The flatter the slope, the higher the EV.
Delta Ic/delta Vce is that slope. If projected back till it eventually meets the 0mA axis, where it crosses the 0mA axis is the EV.
EV effectively determines how much the current Ic changes as Vce changes.
A high EV results in a low change in Ic, whereas a low EV results in a high change in Ic.
EV varies a lot between device types. EV is critical to good operation of the VAS transistor.
It also affects operation as Vce nears saturation, eg. where the output devices have a low Vce while trying to pass current to the load.
http://en.wikipedia.org/wiki/Early_effect
The Early voltage is rarely stated.
But it can be inferred from the Ic vs Vce plot, if shown in the datasheet.
Early voltage tends to havea loose proportionality to Vce0.
Looking at the Ic vs Vce plot (see Fig2) you see the graph of Ic rising steeply from 0Vce to a few Vce. Then it starts turning the corner and becomes a gently rising slope of increasing Ic as Vce goes past 5V all the way up to it's maximum Vce.
The slope of that top part of the Ic plot is directly proportional to the Early voltage (EV).
The flatter the slope, the higher the EV.
Delta Ic/delta Vce is that slope. If projected back till it eventually meets the 0mA axis, where it crosses the 0mA axis is the EV.
EV effectively determines how much the current Ic changes as Vce changes.
A high EV results in a low change in Ic, whereas a low EV results in a high change in Ic.
EV varies a lot between device types. EV is critical to good operation of the VAS transistor.
It also affects operation as Vce nears saturation, eg. where the output devices have a low Vce while trying to pass current to the load.
Not one, not two, but three different pinouts for TO-92 package. Ugh. Though the nxp/fairchild parts are simply 180 deg rotated, so as long as you know, you can just flip the parts.
For low voltage audio applications at 0.5~5 mA typical emitter currents, I cannot see any particular advantage between, say, 2SC1815, BC327, or BC547 .. though in principle the low noise variants BC550 might be desirable, its hard to compare the noise data between manufacturers since it is very dependent on measurement conditions.
Since my circuit is running on 12 V rails, the BC549/559 pair seems the best choice, unless I went with the Toshiba 2SA1015 as being of better audio pedigree.
I'd have to settle this before the getting the boards fabbed, unfortunately, given the different pinout. Then there's the question of availability of the 2SA1015-GR part (its non-stocked at Mouser ... and though its pretty easy to track down in Japan I'm not sure how it is in Europe etc.)
For low voltage audio applications at 0.5~5 mA typical emitter currents, I cannot see any particular advantage between, say, 2SC1815, BC327, or BC547 .. though in principle the low noise variants BC550 might be desirable, its hard to compare the noise data between manufacturers since it is very dependent on measurement conditions.
Since my circuit is running on 12 V rails, the BC549/559 pair seems the best choice, unless I went with the Toshiba 2SA1015 as being of better audio pedigree.
I'd have to settle this before the getting the boards fabbed, unfortunately, given the different pinout. Then there's the question of availability of the 2SA1015-GR part (its non-stocked at Mouser ... and though its pretty easy to track down in Japan I'm not sure how it is in Europe etc.)
Attachments
Self's Small Signal Design Handbook is good guide to fairly recent discrete parts selections and design too. http://www.amazon.co.uk/s/ref=nb_sb_noss_1?url=search-alias=stripbooks&field-keywords=Douglas+Self
BC327 or 328 for example, can handle higher peak currents, making them more suited to output current demands. They have other good qualities like low Rbb that make them suitable for extreme low noise and low voltage amplifier applications that may not be relevant here. I have used them in small amplifiers for many years and recommend them as excellent, economical devices for a small output stage like an output buffer or headphone amp. up to ~ 0.5 watt. However, some people now seem to expect these to be able to drive loudspeakers too, so it's not always obvious what people mean by "headphone" amplifier, even if it is pure fantasy.
More generally, beware that TO92 parts are now going the way of the Dodo. Toshiba 2SC1815/A1015 are obsolete and On-semi has now pulled the pin on BC550/60 and other manufacturers will follow suit as designers switch more completely to SMD. DIY and hand assembled specialties just became that much harder. There is some NOS available and there are 1 or 2 good Chinese and Taiwanese copies of the Toshiba parts. The point is, relying on TO92 parts into the future as a manufacturer, may not be wise. YMMV
Last edited:
Well, if the worst comes to the worst I can re-do the board for TO-126 or something. For the moment TO-92 is fine. I'll take your advice and go with the BC327/337. Hundreds of thousands in stock at Mouser.
Question is why worry so much about the input transistors? For one, I have been doing simulations for days now, distortion of IPS and VAS is at least 20to 25dB LOWER than the OPS. I don't even know why we talk so much about IPS when it's almost irrelevant in the whole scheme of things. I simulate many different IPS/VAS configurations using only 2N2222 and 2N2904. I get so so much lower distortion than after I added the OPS.
Then even if you talk ONLY about IPS only, the design seems to be more important than the transistor itself.
Seems like all you want is transistor that can take the voltage, reasonable fT, not noisy. Who cares about the finer details as it does not matter!!!! Get yourself an OPS that has signal to harmonic ratio of over 110dB, then worry about IPS.
Then even if you talk ONLY about IPS only, the design seems to be more important than the transistor itself.
Seems like all you want is transistor that can take the voltage, reasonable fT, not noisy. Who cares about the finer details as it does not matter!!!! Get yourself an OPS that has signal to harmonic ratio of over 110dB, then worry about IPS.
Last edited:
Peak sells an entry level curve tracer for less than $150 (link). That might be one way to find out if you like some devices more than others.
Here is a post showing 3 NPN transistors' measured IV curves : a 2N3904, a BC550, and a BC337. I find that I prefer one of them quite a bit more than the other two. But maybe that's just my own idiosyncrasy.
Here is a post showing 3 NPN transistors' measured IV curves : a 2N3904, a BC550, and a BC337. I find that I prefer one of them quite a bit more than the other two. But maybe that's just my own idiosyncrasy.
This is interesting. I am planning to match transistors for front end LTP and much more important, matching power transistors.
Is it good enough to measure and match big power BJT like MJL3281 and MJL1302? What is the max current it can provide?
I have measured one of those at around 1nV/rtHz. It was driven from a low impedance so I didn't bother about current noise, and I didn't measure the 1/f corner either, but still pretty good for a part mostly used for switching...
They're really really slow, though.
You'll need to have someone read the sales page and user's guide for you, to dig out that fact. I myself don't know the answer. Eyeballing the photos on the website, it appears to be physically smaller than a traditional "mainframe" curve tracer that sits on a scope cart.
http://www.diyaudio.com/forums/solid-state/151253-diy-curve-tracer-pc-47.html#post4214625
This one is very cheap, and can trace up to 36V/ 2.5A, or a bit more if you mod it.
I want back a few pages, I could not see the unit. Can you give me the direct link to the product?
Thanks
I have one myself, it serves the purpose but has a learning curve to use it properly (the curves could look funny when the transistors oscillate), once you get used to it then it's a great tool considering the cost.
An example curve that I obtained with that tracer:
http://www.diyaudio.com/forums/solid-state/151253-diy-curve-tracer-pc-46.html#post4009990
- Status
- Not open for further replies.