Matching transistors - Measuring hfe

[CBS240]

Hi Shawn,

'snip' What set up did you use to match the Vgs of your devices.

A simple circuit I got from this forum once. I used an old 741.
http://www.diyaudio.com/forums/showthread.php?postid=1096777#post1096777

What devices were you using and how many pieces did you start and end with, I'm very curious.

http://www.fairchildsemi.com/ds/KS/KSK595H.pdf

I have not seen on this forum anyone using this part, so I'm kinda out on a limb here. Regardless of the tiny operating current, I like it and haven't found any real drawbacks in performance or sound. In fact the sound is excellent if used properly. The drawback is in the fact that there is a wide % variance in Vgs on these, at least much more than the j201's I measured. Went through 100 or so in the first measurement run. Measured typical Vgs@100uA of around -0.160V to -0.300V for the KSK595.
First I measure each device and sort them as similar groups of 10-20% or so. When I need a MP, I re-measure from a group to chose the closest two. Once you have them sorted into groups, it isn't so hard to find many matched pairs. In this circuit, there is about 5V Vds on each input transistor and on each cascoding pair. 5 pairs total...one could grow old looking through a pile of 100 jfets just to find matches. The input diff looks like a ladder ...one of my so called 'experiments' ...that actually seems to work.


I use these cheap storage boxes to keep them sorted. When I need more MP's, I now have a narrower margin to look through.

(The top row is some 5087 BJT's sorted by Hfe)

 

[darkfenriz]

darkfenriz, do you care much about hfe in pre/vas stages? Or does it all depend on the topology? Sounds like a huge factor?

Well, honestly I do my best making a design as Hfe-independent as possible, but that's probably not the answer you expect.
I tend to measure output devices hfe to check if they are genuine.
As far as diff and vas stages are concerned hfe of more than 100 is OK for me, same for drivers and predrivers of output buffer.

 

[TomWaits]

Hi Shawn,
A simple circuit I got from this forum once. I used an old 741.

Thanks for sharing that twice. Slowly sinking in...I downloaded it this time.

Well, honestly I do my best making a design as Hfe-independent as possible, but that's probably not the answer you expect.

Good answer.

I tend to measure output devices hfe to check if they are genuine. As far as diff and vas stages are concerned hfe of more than 100 is OK for me, same for drivers and predrivers of output buffer.

Thanks,

Shawn.

 

[AndrewT]

Hi Cbs,
your Vgs measurement set-up shows the DMM connected to the FET gate.
Is there a chance of causing oscillation doing this?

Since there is no current through the 100r could the DMM be connected to pin 6 instead? Thus avoiding risk of oscillation.

 

[CBS240]

Hi Andrew T

Not sure, but you did point out a mistake in drawing. The DMM was measured across the 1K resistor on pin 6. Didn't notice that

 

[Mark A. Gulbrandsen]

Shawn,

I think getting yourself a good used curve tracer might be the better route if you're really serious about this. You have measured HFE... but at what operating current? The curve tracer will measure all that at what ever given operating current you want up to device destruction . They can be had cheaply on E-Bay. I paid just 30.00 for mine!

Mark

 

[TomWaits]

Curve Tracer

Shawn,

I think getting yourself a good used curve tracer might be the better route if you're really serious about this. You have measured HFE... but at what operating current? The curve tracer will measure all that at what ever given operating current you want up to device destruction . They can be had cheaply on E-Bay. I paid just 30.00 for mine!

Mark

Mark, I'm looking higher in the tree and I do want a better result than some $5 DMM. This for the price may beat the pants off anything else right now in my budget.

Thanks,

Shawn.

 

[Mark A. Gulbrandsen]

Now yer cookin! I may try one of those out myself. Ita alot lighter than the old tektronix is.

Mark

 

[TomWaits]

Now yer cookin! I may try one of those out myself. Ita alot lighter than the old tektronix is.

Mark

I picked up on it in Zero Cool's Curve Tracer Thread. Seems there may be downloadable software upgrades as it evolves.

Screen capture/print and share + easy for achiving. You can't beat it unless you have mucho $$$ to spend. I'm sold... I got to have one!

Shawn.

 

[dutch diy]

hfe is not 'a constant' factor

One thing which is adressed in 'curve tracing' in this topic is that hfe is not constant in relation to Ic.

With a multimeter only a small Ic is used to measure Ib and thus hfe can be calculated.

When matching small signal devices there is not a huge miss.

When matching output devices the intended Ic should be used.

I did match 100 MJE15003 and found there was a + and - 2 % spread between measurements with Ic of .7 A and an Ic of .9 A. The average Hfe for .7A was 65.2 and for an Ic of .9 it was 64.7.

The hfe from the datasheet for this device was hfe min 25 and hfe max 150.

 

[john curl]

Folks, I am coming in late, but Hfe is important to match, if you can. However, you MUST measure Hfe at or near Iq (where you are going to operate the device). All else is useful, but not competely accurate.

 

[lineup]

Folks, I am coming in late, but Hfe is important to match, if you can. However, you MUST measure Hfe at or near Iq (where you are going to operate the device). All else is useful, but not competely accurate.

I too, can confirm John Curl advice.

Even if
2 transistors have PERFECT MATCH at some current level
both
- hFE and - Vbe
...... this does not have to say they match at another current level

I did long ago some tests of a number of BC550.
I measure the exact hFE at like 5 different currents levels.
In fact
those exemplars with almost same data at 1 current
did not have same data at other currents.
They could have lower hFE or higher hFE at another point of bias.
Even at currents Not Much different from 1 current.


lineup

 

[TomWaits]

Folks, I am coming in late, but Hfe is important to match, if you can. However, you MUST measure Hfe at or near Iq (where you are going to operate the device). All else is useful, but not competely accurate.

John,

hfe is application specific? What are your general thoughts of this device in general. It seems worthy to me but I would like to hear your concerns. After reading about hfe for days it "seems" to me that a curve tracer may provide a "fuller" result. This device seems like it would do what I want with small signal and driver transistors.

Shawn.

 

[lineup]

it "seems" to me that a curve tracer may provide a "fuller" result

A Curve Tracer tester for transistors
is as perfect tool we can get.
So. You are very right, TomWaits.

lineup is not yet happy owner of Transistor Curve tracer
.............. which would have saved him from hours of manual measurings

 

[lineup]

hfe is application specific? What are your general thoughts of this device in general. It seems worthy to me but I would like to hear your concerns. After reading about hfe for days it "seems" to me that a curve tracer may provide a "fuller" result. This device seems like it would do what I want with small signal and driver transistors.

Shawn.

I this topic it seems like they are ready to organize
A GROUP BUY of Curve Tracer for diyAudio members!

Curve Tracer group buy

We should set up a group buy of the Circuit Ed tracer.
There seems to be enough talk about it to get a group together.
I noticed there was mention of it in another thread as well.
Zc

I'm in!

I am in too!

Regards
lineup

 

[gpapag]

Hi all

Monitoring audio signals on oscilloscope, I observed that there is a lot of audio waveforms which are highly asymmetric. For example:

Human voice - male in particular - in plane speech and in recitative

Complex music passages, where many instruments play different scores at the same time .

Plunked instruments - piano, classic guitar ect - on low register chords (although on single notes produce very symmetric waveforms).

Bass String instruments - viola, cello, contra bass ect - depending on the bowing style.

In all these examples, the frequency spectrum does not extend high, nor the attack time is very short.

There are other musical instruments - brass wind instruments, “x”phones (‘x” being glass, metal, some woods), triangles – which also produce asymmetric waveforms.
They possess a short attack time and power on the higher end of audio spectrum.

On the first group (the low frequency content and long attack time), the duration of asymmetry of their waveform is long.

On the second group (the high frequency content and short attack time), the duration of asymmetry of their waveform is short.

I hypothesize that due to the asymmetry of all these signals, a pre or power amplifier featuring a complementary (N+P) symmetric topology may show an audible degradation just because (*) of the subject discussed in this thread (dynamic mismatch of “complementary” N+P pairs).

(*) I mean excluding the effects of PSU dynamic behaviour.

What is your opinion?

Regards
George

 

[roender]

Hi all,
I want to share with you some hfe measurements I made on newest and oldest ONsemi power devices.
I was very surprised to see how well NJL0302/0281 power ThermalTrack BJT's are matched.
I think that this is the future for amplifiers with BJT's
Mihai

---------------------------------------------------------------
nb-PNPtype, V(15ohm) - V(10kohm), Ic - Ib, hfe(Ic/Ib), Vbe
---------------------------------------------------------------
1-mje15035, 1.070V - 3.39V, 71.30mA - 0.339mA, hfe-208, Vbe-595mV
---------------------------------------------------------------
1-mjw0302a, 1.036V - 6.08V, 69.07mA - 0.608mA, hfe-113.6, Vbe-572mV
2-mjw0302a, 1.036V - 6.06V, 69.07mA - 0.606mA, hfe-113.9, Vbe-573mV
-------------------------------------------------------------------
1-njl1302d, 1.044V - 5.64V, 69.60mA - 0.564mA, hfe-123.4, Vbe-568mV
2-njl1302d, 1.040V - 5.85V, 69.33mA - 0.585mA, hfe-118.5, Vbe-569mV
-------------------------------------------------------------------
1-njl0302d, 1.046V - 5.50V, 69.73mA - 0.550mA, hfe-126.8, Vbe-581mV
2-njl0302d, 1.050V - 5.36V, 70.00mA - 0.536mA, hfe-130.6, Vbe-580mV
3-njl0302d, 1.050V - 5.33V, 70.00mA - 0.533mA, hfe-131.3, Vbe-580mV
4-njl0302d, 1.048V - 5.44V, 69.86mA - 0.544mA, hfe-128.4, Vbe-581mV
5-njl0302d, 1.047V - 5.50V, 69.80mA - 0.550mA, hfe-126.9, Vbe-581mV

---------------------------------------------------------------
nb-PNPtype, V(15ohm) - V(10kohm), Ic - Ib, hfe(Ic/Ib), Vbe
---------------------------------------------------------------
1-mjw0281a, 1.043V - 5.88V, 69.53mA - 0.588mA, hfe-118.3, Vbe-572mV
2-mjw0281a, 1.048V - 5.44V, 69.87mA - 0.544mA, hfe-128.4, Vbe-572mV
-------------------------------------------------------------------
1-njl3281d, 1.037V - 6.17V, 69.13mA - 0.617mA, hfe-112.1, Vbe-569mV
2-njl3281d, 1.057V - 5.03V, 70.47mA - 0.503mA, hfe-139.9, Vbe-571mV
-------------------------------------------------------------------
1-njl0281d, 1.052V - 5.37V, 70.13mA - 0.537mA, hfe-130.6, Vbe-582mV
2-njl0281d, 1.051V - 5.50V, 70.07mA - 0.550mA, hfe-127.4, Vbe-582mV
3-njl0281d, 1.039V - 6.23V, 69.27mA - 0.623mA, hfe-111.2, Vbe-581mA
4-njl0281d, 1.049V - 5.66V, 69.93mA - 0.566mA, hfe-123.6, Vbe-581mV
5-njl0281d, 1.049V - 5.65V, 69.93mA - 0.565mA, hfe-123.8, Vbe-581mV
-------------------------------------------------------------------
1-mje15034, 1.070V - 3.65V, 71.30mA - 0.365mA, hfe-195.3, Vbe-595mV
2-mje15034, 1.070V - 3.66V, 71.30mA - 0.366mA, hfe-194.8
3-mje15034, 1.070V - 3.63V, 71.30mA - 0.363mA. hfe-196.4
4-mje15034, 1.070V - 3.71V, 71.30mA - 0.371mA, hfe-192.2
5-mje15034, 1.070V - 3.72V, 71.30mA - 0.372mA, hfe-191.6
6-mje15034, 1.061V - 4.64V, 70.73mA - 0.464mA, hfe-152.4

 

[roender]

This is the tester used

 

[wds]

I too, can confirm John Curl advice.

Even if 2 transistors have PERFECT MATCH at some current level both
- hFE and - Vbe
...... this does not have to say they match at another current level

This is an example of 2- 2N3906 tests using the CT100 USB Curve Tracer. One by Fairchild and the other is Motorola. Dynamic testing produces revealing characteristics:

2N3906 PNP:

An externally hosted image should be here but it was not working when we last tested it.

 

[sletol]

Interesting curves, but I wonder if the result does not look worse than if expressed, say, in numbers. (As far as I can see the relative deviation is less than 5 %).