What happened to HP and Tektronix semiconductor divisions?

[Wolfgang]

Well, if you like TO-3 devices, there are still some on eBay and in ham radio shops. What you can still get is:

- 2N3866
- 2N4427
- 2N5109

PNP ones are much harder to find. I was able to get a little box of the following:

- 2N5160 (PNP complement to 2N3866/2N5109)
- 2N5161 (PNP complement to 2N3375). I was really lucky

what is still no problem is

- 2N2219A
- 2N2905A

what I also managed to get are "propeller" case transistors like the BRF96 (NPN) and BFT95 (PNP).

[David Hess]

Well, if you like TO-39 devices, there are still some on eBay and in ham radio shops. What you can still get is:

- 2N3866
- 2N4427
- 2N5109

PNP ones are much harder to find. I was able to get a little box of the following:

- 2N5160 (PNP complement to 2N3866/2N5109)
- 2N5161 (PNP complement to 2N3375). I was really lucky

The 2N3866, 2N5109, and 2N5160 are good examples which show up in a lot of old application notes for medium power high frequency output stages.  What I want to know is what replaced them?  There is sure nothing in TO-126/TO-225 or TO-220 packages and the surface mount alternatives are lower power and rely on the printed circuit board for heat sinking.  What I have seen some designs do is use multiple lower power transistors in parallel for each one.

I forget the part numbers but the TO-39 output transistors in my function generators are another example.

[Wolfgang]

I am not exactly an SMD freak, so I go along with the TO-39 parts as long as I can get them.

An SMD part that could be interesting would be the BFG135 from NXP.

[David Hess]

I am not exactly an SMD freak, so I go along with the TO-39 parts as long as I can get them.

An SMD part that could be interesting would be the BFG135 from NXP.

NXP has some great parts but they lack the voltage and power ratings.  Also, availability has been hit and miss.  The last time I looked, the BFG31/BFG97 and BFQ149/BFQ19 (1W 15V 100mA) were the most interesting to me but now all but the BFQ19 are discontinued.

The BFG135 page says, "Archived content is no longer updated and is made available for historical reference only."

[Wolfgang]

Agreed, but it is still widely available and cheap. Do you need your components for mass-producable industrial design or just on a hobby scale ?
I am a radio amateur myself (DL1DWG), and as long I have enough parts in supply for what I need (and maybe some colleagues) I have no bias against vintage parts.

[David Hess]

Agreed, but it is still widely available and cheap. Do you need your components for mass-producable industrial design or just on a hobby scale ?
I am a radio amateur myself (DL1DWG), and as long I have enough parts in supply for what I need (and maybe some colleagues) I have no bias against vintage parts.

I was considering them for a hobby level production run of 10s to 100s of units.  Those projects will be put on hold until an alternate source is found.

[Wolfgang]

Do you have a schematic of what you want to make ?

[David Hess]

Do you have a schematic of what you want to make ?

Among other things, it is just replicating the NIST reference level pulse generator output stage and similar stuff which requires PNP RF transistors.

[Wolfgang]

I googled this but did not find something relevant. Are they similar to the broadband amps of Jim Williams ("High Speed Amplifier Techniques") ?
Do you have a link to find the docs ?

[james_s]

I feel the same feeling of sadness/nostalgia. Perhaps because these brands evoked a sense of quality or familiarity that is unmatched by the newer names - I know that Philips, RCA, GE, Motorola and a few others do that to me.

Brands used to be relevant. A company designed and produced their own products, which carried their brand. Employees at the better companies took pride in their job and cared about more than quarterly gains. Over time there has been a gradual shift toward brands being nothing more than a name that gets marketing types excited over creating a brand image, which is often bought/sold and slapped onto various stuff manufactured by the lowest bidder. Today I consider brands to be virtually irrelevant, most of them are just a name for marketing purposes. Most of the brands I recognize have no connection with the companies founded with them.

[G0HZU]

An SMD part that could be interesting would be the BFG135 from NXP

This is a nice part and it is a shame it is obsolete. However, a few years ago I bought a mini reel of them by mistake from Farnell. I've still got it somewhere (still vacuum sealed). Luckily the price was fairly cheap as they were going obsolete and Farnell were selling them cheap in large qty. I think the mini reel has 150-200 genuine BFG135 parts in it and I couldn't be bothered to send it back to Farnell as it was too much hassle.

[David Hess]

I googled this but did not find something relevant. Are they similar to the broadband amps of Jim Williams ("High Speed Amplifier Techniques") ?
Do you have a link to find the docs ?

No, these are mostly cascode switching designs although I was also considering the transistors for class-A and class-AB output stages.  The NIST reference level pulse generator design is very similar to other reference level pulse generator designs like the Tektronix PG506 and 7000 series standardizers.  Jim Williams published his own variation as well shown below which I forgot about.  The NIST (NBS) documents are too large to post here but you should be able to find them as NBS technical note 1067.

[schmitt trigger]

When I worked for General Instruments, we had our own semi division.

GI Semi had produced the first video game on chip IC, as well as some sound generators. They produced some low horsepower microcontrollers.
I remember that fact because for our top of the line cable box we had to employ a Zilog device.

Eventually corporate became tired of the Semi division and spun it off. Which became Microchip.

[Wolfgang]

When I look at the circuit diagram (Fig. 3-6) of your proposed NBS technical note 1067 I can only see garden variety 2N3906 and 2N4250 PNP transistors in it.
These are widely available even today, in TO-92 or even in SMD. The only exotic part is the BFR90, which can still be bought new from german ham radio stores.

I cant see why it could be a problem to build this in 2018. Did I miss something ?

[David Hess]

When I look at the circuit diagram (Fig. 3-6) of your proposed NBS technical note 1067 I can only see garden variety 2N3906 and 2N4250 PNP transistors in it.
These are widely available even today, in TO-92 or even in SMD. The only exotic part is the BFR90, which can still be bought new from german ham radio stores.

I cant see why it could be a problem to build this in 2018. Did I miss something ?

I do not want to duplicate the NBS/NIST design and its performance; I already have something better.  I want to build a faster one and see just how fast an edge it can generate.  And the polarities have to be swapped to reverse the pulse edge so an NPN output version and a PNP output version are required for both edges.

[xaxaxa]

The 2N3866, 2N5109, and 2N5160 are good examples which show up in a lot of old application notes for medium power high frequency output stages.  What I want to know is what replaced them?  There is sure nothing in TO-126/TO-225 or TO-220 packages and the surface mount alternatives are lower power and rely on the printed circuit board for heat sinking.  What I have seen some designs do is use multiple lower power transistors in parallel for each one.

I forget the part numbers but the TO-39 output transistors in my function generators are another example.

If all you want is a medium power RF amplifier in the VHF band, to-39 bipolars (or bipolars in general) are absolutely not what you want to use these days; any cheap LDMOS transistor made for the purpose will be much more efficient; e.g. RQA0009, or any cellular base station transistor.

[duak]

Wavetek used a lot of 2N3866A and 2N5160 transistors in their function and pulse generator outputs.  These were from Motorola's RF line and were optimized for VHF.  I believe they had multiple ballasted emitters.  I was able to get a few of each last summer from Digikey.  I think Central still makes them.  I have a bunch of 2N4427s from a project that developed a 33 channel laser diode array driver.  Something like a 2N2219 was too slow.

HP used those or 2N5943 and 2N5583 which have an even higher FT.

Does anyone know if Maxtek is still around?  We had them package a number of our ASICs in a hybrid for us in the early 2000's.  I believe the fellow that ran it (Jeff?) also designed/worked on the 453 scope in the 60's.  When he toured me around the facility, I said that when I was younger I had thought that Tek would be a company I'd like to work for.

xaxaxa: is there a P-channel complement to the RQA0009?  If so, then an output stage for a 50 MHz function gen might be an application.

Cheers,

[dzseki]

HP used those or 2N5943 and 2N5583 which have an even higher FT.

There were MRF variants of these: MRF5943 and MRF5583, packed in SO-8, interestingly their ft data seem to be the opposite of the 2N counter part, ie. 1.3GHz on NPN, and 2.1GHz on the PNP part.
My all time favorite mid power RF transistor is the 2SC3616 with its ruler flat gain-current curve, sadly it is without a PNP complement...

[David Hess]

The 2N3866, 2N5109, and 2N5160 are good examples which show up in a lot of old application notes for medium power high frequency output stages.  What I want to know is what replaced them?  There is sure nothing in TO-126/TO-225 or TO-220 packages and the surface mount alternatives are lower power and rely on the printed circuit board for heat sinking.  What I have seen some designs do is use multiple lower power transistors in parallel for each one.

I forget the part numbers but the TO-39 output transistors in my function generators are another example.

If all you want is a medium power RF amplifier in the VHF band, to-39 bipolars (or bipolars in general) are absolutely not what you want to use these days; any cheap LDMOS transistor made for the purpose will be much more efficient; e.g. RQA0009, or any cellular base station transistor.

These were transistors intended for medium power RF amplifier applications but used in linear and sometimes switching circuits instead where a MOSFET part would usually be at a significant performance disadvantage not even counting the lack of complementary pairs.

Wavetek used a lot of 2N3866A and 2N5160 transistors in their function and pulse generator outputs.  These were from Motorola's RF line and were optimized for VHF.  I believe they had multiple ballasted emitters.  I was able to get a few of each last summer from Digikey.  I think Central still makes them.  I have a bunch of 2N4427s from a project that developed a 33 channel laser diode array driver.  Something like a 2N2219 was too slow.

Central makes a lot of these obsolete parts but at a premium price.

At one point Tektronix had a special version of the 2N3866 made for them with the TO-39 pinout changed to swap the base and collector for common-base cascode applications.

[daqq]

IBM did an interesting thing with its semiconductor manufacturing some years ago:
https://arstechnica.com/information-technology/2014/10/struggling-ibm-pays-1-5-billion-to-dump-its-chipmaking-business/

Yeah, it's sad... and I don't see how new companies can even hope to enter the semiconductor business, with the initial costs being that extreme.

[Richard Crowley]

Struggling IBM pays-1.5 billion to dump its chipmaking business.

[rfeecs]

Yeah, it's sad... and I don't see how new companies can even hope to enter the semiconductor business, with the initial costs being that extreme.

Rock's Law:
https://en.wikipedia.org/wiki/Moore%27s_second_law

Rock's law or Moore's second law, named for Arthur Rock or Gordon Moore, says that the cost of a semiconductor chip fabrication plant doubles every four years.[1] As of 2015, the price had already reached about 14 billion US dollars.

[james_s]

I wonder if that will ever stop, Moore's Law already is not holding up so well anymore as I recall. At some point we will reach the physical limits of the materials. Already computers have reached the point where they are not needing to get more powerful, a 10+ year old PC is still usable unless you're gaming or video editing.