Complementary Questions

[Ian Finch]

Hi folks
A FAQ about complementary transistors is: Are are they really complementary or just close enough for some tasks? Mullti-sourcing of small-signal types makes it difficult to compare datasheets since they can vary in data format and depth. I'm sure there are experienced professionals and amateurs who have a clearer understanding of what is fact and not in silicon matters.

The economical NPN series BC546,7,8,9,50 and PNP BC556,7,8.9, 60 have long been cited as complementary. eg BC546 + BC556 or BC549 +BC559.
Are they really? When Philips, Siemens and only a few other Europeans produced them they certainly were not considered so. One recommended audio complement for BC549 in a Wien oscillator circuit was actually 2N or PN4250!

Are the later BC550/560 indeed considered complementary? Reading ON Semi data sheets might lead you to agree, since some graphs appear to be just copies! Not all manufacturer specs are actually like that and I wish to stay with this series as it is...(ahem)..was..widely available.

Do you have any thoughts or an opinion on whether to trust these TO92 types as compllementary for use in audio preamps or buffers for example?

 

[lineup]

Those you mention from BC-series are as complementary as small signal transistors usually are.
Term COMPLEMENTARY does not mean they are same in HFE or VBE
but they are similar in many ways.
So, they are good to use together.

To make one NPN and one PNP with perfect match is close to impossible.
What I know it has never been done and I doubt it ever will be.

If you use BC546/556 or BC550/560 it is a fair match.
Gain, hFE and VBE, BE-voltage will be different.
How much different is from case to case.
Depends on the production.
Sometimes is a bit better match, other times less.

 

[Ian Finch]

Thanks for the reply, Linie. I consider it a compliment. I suspected that the newer types you show in bold type were considered better complements than some of the earlier types.

I plan to make a low power amp with basic mirror image input stage, like Rotel, Electrocompaniet and many others have done. Without CCS or Current Mirrors, I only guess that match may be a little more important.

 

[east electronics]

Ian's coment and post make absolute sense from many aspects

----now days we dont know any more what exactly we buy ....
got my shelf 200 pcs of BC549 made by philips all of them same batch but hfe varied from 180 to 450 ... makes no sense

got myshelf 100pcs BC557 from siemens ( ?? ) had no batch number onthem only the number 341 ....99% of them had hfe of 341 !!!!!!probably military specs that fell of in the market by accidednt

---- now days it seems that market is full with many diferent parts and especially small things need to be extensivelly tested before used in high quality applications

obviously even in the most simple audio application that uses a diferential LTP or any other similar tranistor application transitor matching is mantatory and will result to better performance

even in the most simple application the P3A using carefully matched transisitors increased tightness of the bass made it more punchy and the all amlifier behaved by far better ( of course in my version of the P3A the LTP stages features beyond matched LTP pair a thermal junction . )

a living example of the importance of complementary is the TITAN amplifier from the ELEKTOR circuits that anybody that builded found out that was a machine impossible to stabilize and make operate properly unless used tranistors that had the specific data and performance of the simulation and then againo those needed to be matched with the oposite pairs on the other side ...

regards sakis

( hello Ian !!!! how are you ??? any news ???)

 

[Ian Finch]

Hi Sakis
Those BC557 will take some time to use. I would think twice about that supplier next time. Still, you now know the tester is accurate, eh?

There are other parameters like Noise Figure, Ft, Cob etc. that are a problem. Just when you think you have a pair, some have PNP with 4 times the NPN figure or vice versa. Greg Erskine's recent transistor listing on his website seems like an impossible task when you think about High-end demands.

O/T..... Yes, apologies for slow progress, my friend. I will email.

 

[CBS240]

IMHO, if 'high end' is the goal, why not fork out a few extra bucks for some matched pairs, like THAT 3XX, MAT, or ALD matched mosfet devices?
Then you don't have to worry about it.

The THAT 3XX transistors have very good NF and Ft, Hfe is typically 80 - 100. But you don't need much current gain for a line input stage anyway. They make great mirrors, no degeneration required.

 

[knutn]

I don't know who said the following: "PNP and NPN transistors are as complementary as men and women".

It's true!!

 

[lineup]

Haha.
That one was good, knutn

 

[Ian Finch]

If I remember correctly, Doug Self credited NAIM designer Julian Verekher with that phrase. It was used to explain his choice of quasi-complementary OP stages. Somehow, it seems to give a different meaning in that context. I won't go any further with that!

Hi CBS, thanks for the suggestion. I should not have referred to Hi-end really. Guys building it would probably not even consider BC series parts.

 

[Greg Erskine]

I don't know who said the following: "PNP and NPN transistors are as complementary as men and women".

It's true!!

Hi knutn,

Is that a reference to the different majority carriers, electrons and holes?

 

[CBS240]

....Hi-end really...... Guys building it would probably not even consider BC series parts.

I suppose it depends on what the intended circuit function of the device is. Heck I'd even use an old 3904/3906 type for some functions where Hfe, Ft, Cob, or Vceo isn't a factor. But for these types of input stages, I would go with 'manufacture-ally matched'. Current & voltage symmetry just comes out better. Thanks to the modern age there are some really high class signal transistors out there for reasonable price.

 

[sajti]

----now days we dont know any more what exactly we buy ....
got my shelf 200 pcs of BC549 made by philips all of them same batch but hfe varied from 180 to 450 ... makes no sense

got myshelf 100pcs BC557 from siemens ( ?? ) had no batch number onthem only the number 341 ....99% of them had hfe of 341 !!!!!!probably military specs that fell of in the market by accidednt

I had same experience with the Onsemi NJW0281/0302. The 140 pcs of NPN has the Hfe between 84 and 122, the 140 pcs of PNP has the gain 98 and 120, but 90% of them between 99 and 106!!!!.
I measured them with 100mA current.

Sajti

 

[Ian Finch]

I know that you are right there for serious projects, CBS. I'm just a tightwad when it comes to Semis. Just consider the power and quality now possible from a mere $5 pair of plastic OP transistors!

I did have a lot of success with Supermatch LM394H years ago. I also used a few MAT 02,3,4 in low level mic./MC preamps but the price/availability see-saw was really annoying. Today, long distance buying of these rarer components is a breeze for sure. If vendors were more considerate about shipping options/charges, I would buy a lot more "manufacturally-matched" and specialised parts.

If our eyes, hands and PCB skills are up to it, there are very good and cheap SMT array options too. However, adapting older and high VCE requirement circuits in amps, such as we love to do here on DIY-A, is still gonna be difficult for pairs and complements.

It's probably O/T but considering linearity again; I wonder why there is wide preference for mirror image/complementary input topology? If complementary matches are never so good as same type NPN or PNP, then the imbalance must increase input stage distortion significantly. Randy Slone's liking for it was based on PA type issues, so no help there.

 

[ostripper]

I had same experience with the Onsemi NJW0281/0302. The 140 pcs of NPN has the Hfe between 84 and 122, the 140 pcs of PNP has the gain 98 and 120, but 90% of them between 99 and 106!!!!.
I measured them with 100mA current.

Sajti

I use these lovelies almost exclusively.. at $1.40 each what a deal!! most are between 90-115 P and N channel. I have no problem finding 4 pairs or 2 pairs that are within 2 Hfe. I just bought 25 pair, all the same date code.

On-semi is VERY consistent with both the models and the actual devices.

OS

 

[CBS240]

I know that you are right there for serious projects, CBS. I'm just a tightwad when it comes to Semis. Just consider the power and quality now possible from a mere $5 pair of plastic OP transistors!

If our eyes, hands and PCB skills are up to it, there are very good and cheap SMT array options too. However, adapting older and high VCE requirement circuits in amps, such as we love to do here on DIY-A, is still gonna be difficult for pairs and complements.

It's probably O/T but considering linearity again; I wonder why there is wide preference for mirror image/complementary input topology? If complementary matches are never so good as same type NPN or PNP, then the imbalance must increase input stage distortion significantly. Randy Slone's liking for it was based on PA type issues, so no help there.

I'm a tightwad too...my output stage consists of two pair of TO-220 fets that cost a total of $4.50. 22S of Gm and Pd over 40W at 125C...TO-220 As for linearity, it seems to be pretty good being driven by HEC network.

As for PCB layout and component choice, I'm a big fan of using as little current as possible and using acordingly sized components. Arrays are better because it's less components to solder together, even if it is SOT-563. Now they have SOT-963. I have not gotten the courage to try them yet, they seem to be expensive compared to SOT-563 part and the tiny space saved is not worth it. I do use some SOT-923 transistors in this circuit, such as these and these nice linear gain (~300 @ 1mA) transistors from Toshiba.

But then there are a lot of 'odd' (for around here) things done in this amplifier. It is a balanced amp though, and I think that the symetry of drive Z and CGBW has something to do with the drive toward symetry. This ver prototype amp is designed to have very low distortion and particularly common mode distortions and would be nearly impossible without matched arrays.

 

[sajti]

I use these lovelies almost exclusively.. at $1.40 each what a deal!! most are between 90-115 P and N channel. I have no problem finding 4 pairs or 2 pairs that are within 2 Hfe. I just bought 25 pair, all the same date code.

On-semi is VERY consistent with both the models and the actual devices.

OS

I have lot of NPN-s with hfe 84-95, and no PNP pairs. So finally I made about 70 well matched pairs (less than 1%!). For more pairs I recommend to order more NPN-s. But for triple output the NPN/PNP match is not very important.
I like them too, and I use many of them, because they are cheap, have very low internal capacitances.

Sajti

 

[Ian Finch]

Oh, whoa....Nice bit of PCB work there. Has this been posted with schematic etc. before, CBS, or is that not the plan? Thanks for posting the spec sheets too. All I need now is new close up specs to check it out and maybe think about some more SMT myself!

 

[CBS240]

If your going to go SMT, you might as well go with SMD resistors, and caps too. The SMD metal film resistors are nice in terms of noise and the small COG pf caps are great for compensation duty, and they're cheap. I'm going to use some 0201 size resistors in the next model. For resistors that have uW to a few mW of power, why not? I do know that if you intend to have to hand solder 0201's, you should leave a bit longer solder mask pads in the PCB and never bury any pads within large copper fills.

Oh, and that amp is the latest version of my modular bridge amp but I have not posted all the schematics, just a couple of sections. There's a thread somewhere I will have to look it up.

But now I have to get ready drink:)to go to a football game (not soccer)

 

[Ian Finch]

At the OP stage; have you guys tried NJW21193/4 as triple complements - ie triple EF or close? It's not straightforward comparing Hfe at 8W with the 3W spec of the Toshiba clones 3281/1302 and 5200/1493, but I assume the 21193/4 are lower with lower Ft too, which might be safer in that application.

OK, price is higher and availability is not so good (do any major distributors hold stock?). Stability is a greater issue though. Os, you work with plenty of triples - any thoughts?

 

[sajti]

I used 21193/94. They are really rugged devices. If You want to build PA amplifier, than You have to use them. Unfortunately the Hfe is worse than the 3281/1302 family. I measured hfe=30 on some NPNs, with Ic=100mA, and there was no PNP less than 80...

Sajti