HN3C51F and HN3A51F NPN and PNP matched pairs from Toshiba. So far they seem to be unobtainium, but worth pursuing.
150V VCbo, high and linear Hfe, and a whopping 1 db noise figure! JFET territory for sure...
Here are links to the datasheets:
http://www.bestweb.net/~jgedde/HN3C51.PDF
http://www.bestweb.net/~jgedde/HN3A51.PDF
150V VCbo, high and linear Hfe, and a whopping 1 db noise figure! JFET territory for sure...
Here are links to the datasheets:
http://www.bestweb.net/~jgedde/HN3C51.PDF
http://www.bestweb.net/~jgedde/HN3A51.PDF
darkfenriz said:
... and 2SC2240 - 2SA970 are just great low noise transistors
and even though rather low power is recommended
they can take 120 Volt Collector-Emitter !!
Which makes them suitable, without cascoding,
as higher power amplifier input differential dual pair.
Compared to other common TO92, like BC550 - BC560 'only' 45 Volt C-E
Lineup Audio Lab
... uses both BC550 and 2SC2240 in circuits, frequently
... and results are very rarely disappointing
For the ones I found low power is the word of the day also.
Note that the HN3*** beats the noise figure of the 2sa970 by 2db. Better yet, they are matched pairs (3%) matching. This means it is possible (a good idea?) to not use emitter degeneration resistors.
Since they are also Hfe matched, it means reductions in THD without having to go through the trouble of matching discrete transistors by hand.
I'm designing an amplifier that will use these (if I can find any). I was curious about opinions from those you have reviewed the datasheets.
Note that the HN3*** beats the noise figure of the 2sa970 by 2db. Better yet, they are matched pairs (3%) matching. This means it is possible (a good idea?) to not use emitter degeneration resistors.
Since they are also Hfe matched, it means reductions in THD without having to go through the trouble of matching discrete transistors by hand.
I'm designing an amplifier that will use these (if I can find any). I was curious about opinions from those you have reviewed the datasheets.
i did have a look at datasheet
and test diagrams
i must say it all look very nice and good
I would say, even though Toshiba have used 1mA for to get good data
I think they work the absolute best between like: 0.5 - 1.0 mA
It is a bit depending on the resistance at the input of your circuits.
Source resistance will a have some little effect due to some little bias current noise in resistors.
But also, of course one factor is what collector load you need to drive with input pair.
Sometimes we need more than 1mA to get good drive of collector loading.
and test diagrams
i must say it all look very nice and good
I would say, even though Toshiba have used 1mA for to get good data
I think they work the absolute best between like: 0.5 - 1.0 mA
It is a bit depending on the resistance at the input of your circuits.
Source resistance will a have some little effect due to some little bias current noise in resistors.
But also, of course one factor is what collector load you need to drive with input pair.
Sometimes we need more than 1mA to get good drive of collector loading.
If the voltage is not so important BC847BS is quite OK as a pair, very cheap also, less than a 1/4 dollar.
http://sjostromaudio.com/hifi_pics/hifi_100pr/qsxm2_sot363_detail.jpg
http://sjostromaudio.com/hifi_pics/hifi_100pr/DIL6toSOT363.jpg
http://sjostromaudio.com/hifi_pics/hifi_100pr/qsxm2_sot363_detail.jpg
http://sjostromaudio.com/hifi_pics/hifi_100pr/DIL6toSOT363.jpg
2SC2240 and 2SA970 - RIP. Replacements and dual versionss
This line of parts are in my estimation the lowest-noise BJTs now available (lowest r_bb') that have official low-noise specs**, now that Rohm's 2SD786 and 2SB737 have disappeared.
Sadly the 2SC2240 and 2SA970 (npn, pnp) super-low-noise trannies in TO-92 packages no longer appear on Toshiba's Bipolar Transistors Product List, and have been designated NRND by Toshiba (Not Recommended for New Designs). [link] This means the factory is presumably still taking (large) orders, but they are destined for the dust-bin of history. FYI, last chance, B&D still has some inventory.
But, not to fear, Toshiba offers what appears to be the same (or better) die in SOT-23 packages, the 2SC3324 and 2SA1312. These are stocked by Mouser, and cost only 15 cents qty 100. Their datasheets are fairly complete, including two NF contour plots. The 1kHz plot shows the 3dB contour dropping 35 ohms at 1mA and below 20 ohms at 10mA. That corresponds to voltage noise levels e_n of 1.4, 0.75 and 0.56nV at 0.1, 1 and 10mA. Using the 100Hz plots, we can calculate that the 1/f corner frequencies are about 14Hz, 35Hz and 170Hz, not too bad.
The hn3c51F and hn3a51F dual transistors, original subject of this thread, also appear on Toshiba's official Bipolar Transistors Product List. They come in easy-to-solder SC-74 (6-pin SOT-23) packages, and apparently use the same dies as the 2sc3324 and 2sa1312. Their datasheets have the same two impressive noise-figure contour plots. But even though several years have gone past since this thread started, no distributors have picked up these parts. If we want to use them, we'll have to make a large group purchase. A standard reel is 3000 parts.
This line of parts are in my estimation the lowest-noise BJTs now available (lowest r_bb') that have official low-noise specs**, now that Rohm's 2SD786 and 2SB737 have disappeared.
Sadly the 2SC2240 and 2SA970 (npn, pnp) super-low-noise trannies in TO-92 packages no longer appear on Toshiba's Bipolar Transistors Product List, and have been designated NRND by Toshiba (Not Recommended for New Designs). [link] This means the factory is presumably still taking (large) orders, but they are destined for the dust-bin of history. FYI, last chance, B&D still has some inventory.
But, not to fear, Toshiba offers what appears to be the same (or better) die in SOT-23 packages, the 2SC3324 and 2SA1312. These are stocked by Mouser, and cost only 15 cents qty 100. Their datasheets are fairly complete, including two NF contour plots. The 1kHz plot shows the 3dB contour dropping 35 ohms at 1mA and below 20 ohms at 10mA. That corresponds to voltage noise levels e_n of 1.4, 0.75 and 0.56nV at 0.1, 1 and 10mA. Using the 100Hz plots, we can calculate that the 1/f corner frequencies are about 14Hz, 35Hz and 170Hz, not too bad.
The hn3c51F and hn3a51F dual transistors, original subject of this thread, also appear on Toshiba's official Bipolar Transistors Product List. They come in easy-to-solder SC-74 (6-pin SOT-23) packages, and apparently use the same dies as the 2sc3324 and 2sa1312. Their datasheets have the same two impressive noise-figure contour plots. But even though several years have gone past since this thread started, no distributors have picked up these parts. If we want to use them, we'll have to make a large group purchase. A standard reel is 3000 parts.
Attachments
Not sure where the claim comes from that the parts are matched to 3%. Most will probably be reasonably well matched because most units will come from adjacent spots on a silicon wafer (after all, it would take special effort to mix them up) but unless the manufacturer takes explicit care not to pair one unit from the end of a row with another from the start of the next row, occasional badly matched parts will turn up.
Toshiba makes no mention of taking any care at all with regards to this.
Anyhow. I tested 10 parts for hfe matching. Two were at 3%, the remaining eight were below 0.5%. Frustrating, because it means the majority are really good, but the nongood ones are still a substantial minority.
Toshiba makes no mention of taking any care at all with regards to this.
Anyhow. I tested 10 parts for hfe matching. Two were at 3%, the remaining eight were below 0.5%. Frustrating, because it means the majority are really good, but the nongood ones are still a substantial minority.
Which part types exactly did you test?
I'd like to say, with regard to hFE matching, it's an easy measurement for manufacturers and users to make (but perhaps relatively hard for manufacturers to meet).
I'm always disappointed when I see an hFE matching spec but no Vbe-matching spec. That's because hFE matching isn't a very useful parameter. hFE is not predictable, and well-done designs work around that parameter and depend instead on Vbe and Vbe matching. Vbe is the parameter that counts for differential-amplifier stages, etc. One can hope that well-matched hFE will imply well-matched Vbe, but one cannot count on it.
I wonder if you tried Vbe matches for your 10 parts?
I'd like to say, with regard to hFE matching, it's an easy measurement for manufacturers and users to make (but perhaps relatively hard for manufacturers to meet).
I'm always disappointed when I see an hFE matching spec but no Vbe-matching spec. That's because hFE matching isn't a very useful parameter. hFE is not predictable, and well-done designs work around that parameter and depend instead on Vbe and Vbe matching. Vbe is the parameter that counts for differential-amplifier stages, etc. One can hope that well-matched hFE will imply well-matched Vbe, but one cannot count on it.
I wonder if you tried Vbe matches for your 10 parts?
Yes, I agree.
Set up an LTP pair and apply a common Vbe to both devices in the pair.
Discover just how well matched the output currents are into the exactly equal Collector loads.
Once I have found potential pairs I always test them thermally coupled as LTP pair. That really does show up mismatching of supposed good pairs.
The 2sa1312 datasheet is impressive.
1dB noise figure for 50r @ 7mA & 1kHz
Jumps to 6dB ref same 50r @ 10Hz.
Maybe for lower frequencies, around 2 to 3mA may be a better bias.
Any thoughts?
Set up an LTP pair and apply a common Vbe to both devices in the pair.
Discover just how well matched the output currents are into the exactly equal Collector loads.
Once I have found potential pairs I always test them thermally coupled as LTP pair. That really does show up mismatching of supposed good pairs.
The 2sa1312 datasheet is impressive.
1dB noise figure for 50r @ 7mA & 1kHz
Jumps to 6dB ref same 50r @ 10Hz.
Maybe for lower frequencies, around 2 to 3mA may be a better bias.
Any thoughts?
Last edited:
Update:
Happily, Mouser has been carrying these Toshiba dual transistors, wow! They're priced at only 18 cents, qty 100. We've measured these parts. For the npn and pnp, they have r_bb' of 35 and 20 ohms, and e_n at 10mA of 0.75 and 0.58 nV/rt-Hz, resp. The Early voltage is 560 and 180 volts.
There are lower-noise transistors, lower e_n anyway, with r_bb' down to the under 2-ohm region, see Table 8.1 in AoE 3rd edition for full details, but they're not specified as low-noise transistors. They're low-saturation-voltage designs made by Zetex / Diodes, Inc. For example the ztx618 and ztx718 (fmmt618 and 718 for sot-23 packages) with r_bb' of 9.3 and 7.3 ohms, and the ztx851 and ztx951 with r_bb' of 1.67 and 1.24 ohms. We use the latter (multiple pnp transistors in parallel) to make a record-breaking low-noise amplifier, only 70 pV/rt-Hz. That's low enough to be a ribbon-mic pre-amp, without the usual step-up transformer!
Last edited:
Hey, that is sneaky! I came to diyaudio to post that AoE 3rd will finally be available, and I searched and found this post. You are a master of subtlety. I mean there's nothing really underhanded about your profile, but I bet a lot of people on this forum didn't even notice. They should find whoever is "winhill" and just give you that profile name out of respect.
Hah, nothing sneaky about it, it's a reference people will want to know about. Actually, Paul and I spent many months buying parts, taking measurements, filling out giant spreadsheets, and charting the results. All to come up with the kind of noise and performance data that I had badly missed not having for the last 25 years. I'm hoping it's the kind of data some in the diyAudio crowd will be glad to see. The mass of new material in chapter 3 (JFETs and MOSFETs), chapter 5 (precision amplifiers), chapter 8 (low-noise techniques) and chapter 13 (digital meets analog), totaling 422 pages for those four chapters alone, is like a special book on this topic.
As far as my ID, winhill2, is concerned, I'm probably the one who originally took out winhill as an I.D., but then forgot my password! You know how that happens: your computer remembers it automatically in the browser, for many years, but then it fails, taking your passwords with it.
As far as my ID, winhill2, is concerned, I'm probably the one who originally took out winhill as an I.D., but then forgot my password! You know how that happens: your computer remembers it automatically in the browser, for many years, but then it fails, taking your passwords with it.
Does low voltage-noise tend to be a general characteristic of low-saturation-voltage transistors? When looking for other things, I've stumbled across some Low-Vce(sat) transistors (from NXP, and ON Semi, as I recall) that look like they might be useful in audio, but the data sheets don't characterize them for audio applications.
Dale
I'd say the answer is yes, although we tested only a dozen or so of such candidate transistors. These were low-saturation-voltage transistors that also go to high currents, over an amp. This was four years ago; I should go back and carefully examine /study my giant spreadsheet, etc. (chapter 8 comprises over 2000 files and 1.1GB in my computer), and resurrect some memory cells.
Another type of transistor that had surprisingly-low r_bb', down there with the best of them, was small-die high-voltage video transistors, made by Sanyo. These are especially valuable, because they have low capacitance, whereas the high-current low-saturation types tend to have high capacitance.
Unfortunately, the sales market for 200-volt video transistors has really dried up, along with the CRT market, ;-), so they have been discontinued. Nonetheless I included two in our table just for your reference (they're marked with special footnotes). Naturally, I'm highly protective of my inventory of those parts. They are spectacularly-useful for implementing the "Regulated Cascode" RGC scheme in certain types of TIA amplifiers.
I'm not sure how this works in semiconductor device physics, because normally the secret to low r_bb' is large die area. Anyway, the existence of these Sanyo transistors proves that's not necessarily the case.
Another type of transistor that had surprisingly-low r_bb', down there with the best of them, was small-die high-voltage video transistors, made by Sanyo. These are especially valuable, because they have low capacitance, whereas the high-current low-saturation types tend to have high capacitance.
Unfortunately, the sales market for 200-volt video transistors has really dried up, along with the CRT market, ;-), so they have been discontinued. Nonetheless I included two in our table just for your reference (they're marked with special footnotes). Naturally, I'm highly protective of my inventory of those parts. They are spectacularly-useful for implementing the "Regulated Cascode" RGC scheme in certain types of TIA amplifiers.
I'm not sure how this works in semiconductor device physics, because normally the secret to low r_bb' is large die area. Anyway, the existence of these Sanyo transistors proves that's not necessarily the case.
I don't follow what is the best transistor. For power amp, I don't think noise is the most important thing as gain is low at this point. I plan to use this two for the IPS and VAS and is selling in Digikey. I choose this because it's beta over 100, 120V and low capacitance.
https://www.fairchildsemi.com/datasheets/KS/KSC1845.pdf
https://www.fairchildsemi.com/datasheets/KS/KSA992.pdf
https://www.fairchildsemi.com/datasheets/KS/KSC1845.pdf
https://www.fairchildsemi.com/datasheets/KS/KSA992.pdf
Producing these data was a great service and for me well worth the cost of the 3rd edition! Thanks!
It is a shame about the loss of a significant market for the Sanyo parts.
Brad Wood
---
Just to say that this info is surpassed by the extensive measurements Paul and I made on a pile of candidate low-noise transistors, reported in detail in AoE III, low-noise Chapter 9. Happily we found many much better available parts. But not dual parts, sorry.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.