Low Noise Transistors: ZTX851 versus ZTX689B Noise Performance

[mediatechnology]

I used the MC preamp as a noise test circuit and temporarily installed a ZTX689 pair in one channel and a ZTX851 pair in the other for direct comparison.

Rsource = 10Ω; Gain 62 dB (+/-0.5 dB); Rg = 1Ω; Ic 5.5 mA nominal


A noise comparison between the ZTX689 and the ZTX851

Test circuit:



Moving Coil preamp thread: https://proaudiodesignforum.com/forum/p ... f=7&t=1179

To make sure I didn't have a gain error I swapped input pairs between channels and measured identical results.

There's an approximate 0.7 dB level error due to converter loading which makes the Ein of both slightly worse.
Still, the ZTX689B comes in around 600-700 pV√Hz.
Working backwards from noise the rbb' looks to be more like 8Ω rather than the 5.5Ω I measured earlier.
If needed, 4 paralleled devices would get the ZTX689 back to the <2Ω rbb of the ZTX851.

The difference in this circuit is about 3 dB.
Although the ZTX851 is quieter the ZTX689B has the advantage of higher minimum gain of 500 vs the ZTX851 minimum gain of 100.
With higher source impedances, such as those typical of a mic preamp, the ZTX689 may be a better choice where the difference in noise is far less than the 3 dB we see here.

[terkio]

Thanks for the news.
I understand the low noise of ZTX851 and ZTX689 comes from it's multiple transistors.
A very interesting technology, can I find more about it ?
How it is implemented. Other transistors of this kind. Are they duals of this kind, is it the same as in LM394 ?

[mediatechnology]

The LM394 did use multiple transistors.

About all I can find on Zetex technology is the "Matrix Geometry" used to provide low distributed base resistance to what appears to be a single transistor. Rather than many transistors we have many base contacts.


Zetex "Matrix Geometry" technology used in the E-line series.

The Zetex transistors' low Vcesat and their large area provide low noise performance as a side benefit.
It seems ironic that one of the applications shown by Zetex for the ZTX851 is the output driver for a car horn.
We have a transistor designed to generate large amounts of obnoxious noise turning out to be exceptionally quiet.

Zetex published a few application notes about using the E-Line series of transistors for low noise audio applications beginning in 1995.
ZTX DN-11 "Low Noise Audio Input Stage" https://proaudiodesignforum.com/images/ ... _Stage.pdf


Zetex DN-11 Low Noise Audio Stage.

It was a Zetex application note that caused me to look at the ZTX689B and ZTX789A.
https://proaudiodesignforum.com/forum/p ... =12&t=1166

There are several others in the series that may also have low noise characteristics.

See also:

Moving Coil Phono Head Amp, Zetex AN-23, March 1996: viewtopic.php?f=6&t=980
Zetex Bipolar Transistors: https://proaudiodesignforum.com/images/ ... atalog.pdf

[terkio]

I had seen this Matrix geometry but lost it and was unable to find it again.
Good, I have it now. Trouble is, I do not understand it well enough.
I do know the planar transistor, the manufacturing process, I do know the multiple emittors as found in TTL...But this matrix geometry is something else, confusing.
It is clear where the collector base junction is, same as usual BJTs.
It is not clear where the emittor(s?) base(s?) junctions are.
Not clear as well how metallization connects to emittors, bases and collector.
What in this geometry makes the transistor very low noise ? The Zetex explanation is too sketchy for me.

[mediatechnology]

What in this geometry makes the transistor very low noise ? The Zetex explanation is too sketchy for me.

I don't know but I'll take the results.

[JR.]

I had seen this Matrix geometry but lost it and was unable to find it again.
Good, I have it now. Trouble is, I do not understand it well enough.
I do know the planar transistor, the manufacturing process, I do know the multiple emittors as found in TTL...But this matrix geometry is something else, confusing.
It is clear where the collector base junction is, same as usual BJTs.
It is not clear where the emittor(s?) base(s?) junctions are.
Not clear as well how metallization connects to emittors, bases and collector.
What in this geometry makes the transistor very low noise ? The Zetex explanation is too sketchy for me.

It is a pretty popular technique to integrate multiple base-emitter junctions in parallel to optimize sundry characteristics.

I recall a Japanese RET "ring emitter transistor" that promised improved power transistor Ft performance. Then there is the classic LM394 Dual transistor that used a highly parallel base-emitter structure. It was touted for excellent performance in logging applications, but the highly parallel architecture also delivered low Rbb for good noise performance (good for it's time).

Similar highly parallel architectures have been used in power mosfets for similar reasons.

To make low Rbb in bipolar junctions you generally have to throw more silicon (area) at the design... multiple small junctions in parallel can be better behaved than single large ones...... or something like that.

JR

[terkio]

I think, i see the trick.
For low noise you need a large base emitter jonction. A large planar transistor seems the answer...then the base would be connected by a metallization ring around the large transistor..then the trouble is from the base thinness.
No trouble for outer base emitter junction areas that are close to the peripheral ring.
Big trouble for inner base emitter junction areas
that are far from the peripheral ring, far so very resistive and delayed along a long way through the thin base.
The trick is to punch through the emitter to reach the base at many places over the large transistor.
I don't see clearly how metallization is organized to connect at the base and the emitter, I presume there are interleaved combs to reach at many points over the large transistor area. I presume the collector is reached from the substrate side.

Anyway as said, what matters is the result.
Thanks to mediatechnology measurements: 5.5 Ohm rbb'.
This to compare with 20 Ohm rbb' of two paralleled dual LM394.
However LM394 stays leader as far as matching is concerned with 50μV Vbe matching, 2%Hfe matching.
A dream would be a dual with Matrix geometry.

[mediatechnology]

The trick is to punch through the emitter to reach the base at many places over the large transistor.

That's what I'm seeing in the drawing - lot's of distributed base contacts.

My rbb of 5.5Ω may be low, my last guestimate was 8Ω.
I've asked Hans at diyaudio to run his sim for the MC preamp adjusting the model's rbb until it got 2.7 dB noisier.

A dream would be a dual with Matrix geometry.

You might want to wade through that data book.

EDIT: I just went through the small bag of ZTX689B that I have and measured the gain of a few devices. The test current is about 200 µA; Vce about 2V.
The lowest gain was 450, 500 was about typical. One device had about 650.

[JR.]

I think, i see the trick.
For low noise you need a large base emitter jonction.

its all about the silicon substrate resistance.

back in the bad old days a low noise design trick was to use a medium power transistor for low noise applications. The more robust base-emitter for handling higher currents, helped lower noise thanks to reduced Rbb, but process technology back in those days was pretty rough so one would have to select out the transistors with too much process related noise (usually 1/F).

Often the rejects got used for other GP transistor tasks around the company.

JR

A large planar transistor seems the answer...then the base would be connected by a metallization ring around the large transistor..then the trouble is from the base thinness.
No trouble for outer base emitter junction areas that are close to the peripheral ring.
Big trouble for inner base emitter junction areas
that are far from the peripheral ring, far so very resistive and delayed along a long way through the thin base.
The trick is to punch through the emitter to reach the base at many places over the large transistor.
I don't see clearly how metallization is organized to connect at the base and the emitter, I presume there are interleaved combs to reach at many points over the large transistor area. I presume the collector is reached from the substrate side.

Anyway as said, what matters is the result.
Thanks to mediatechnology measurements: 5.5 Ohm rbb'.
This to compare with 20 Ohm rbb' of two paralleled dual LM394.
However LM394 stays leader as far as matching is concerned with 50μV Vbe matching, 2%Hfe matching.
A dream would be a dual with Matrix geometry.

[billshurv]

My rbb of 5.5Ω may be low, my last guestimate was 8Ω.
I've asked Hans at diyaudio to run his sim for the MC preamp adjusting the model's rbb until it got 2.7 dB noisier.

He's done the sims and gets 5.3R. close enough for Rock and Roll. I have to say, for any sane* cartridge if we didn't have the ZTX851 we'd be going gooey over the ZTX689B.

* I do own insane ones as well. And not just to annoy JR