Bipolar transistor grading

Gepetto

Administrator
Staff member
Joined
May 15, 2011
Messages
13,473
Location
Sterling, MA
Tagline
Old 'Arn Enthusiast
#1
For device matching, usually you are most concerned about base emitter voltage matching.

What usually works well is a fixture with the base grounded, the collector attached to a positive DC source and the emitter connected through an appropriate value resistor to a negative DC source.
DMM connections are positive lead to ground and negative lead to emitter of device under test (DUT). Keep DUT wattage low during the test or it will take a long time for the reading to stablize. Keep your lead length short and ensure good connections.

All power sources should be well regulated sources or your readings will drift all over the place.

The positive supply voltage is less critical for absolute setting than the negative supply. It is just ensuring that the transistor is not in saturation.

For NPN small signal transistors my set up is:
Positive voltage source = 4VDC
Negative voltage source = -14VDC
Emitter resistor = 68,000 ohms 1/4W
This produces a Collector-Emitter current of ~200uA

For TO-3 NPN power transistors my set up is:
Positive voltage source = 3VDC
Negative voltage source = -14.3VDC
Emitter resistor = 100 ohms 10W. Mount this resistor away from the DUT as it will be dissipating ~1.5W and you do not want its heating effects screwing up your readings.
This produces a Collector-Emitter current of ~137mA and a DUT dissipation of ~1/2W

Do all your matching as close in time as possible to get the same ambient temperature in the room you are testing in.

It will take a while for the device to temperature stabilize and for you to get a stable reading. Be patient.

I have a long chart that I constructed of quadrille paper for sorting the devices, 8.5x11" sheets taped end to end vertically.
The lowest marking is 0.625V and then there is a major marking every five squares at 0.630, 0.635, 0.640 and so on up to 0.690. This will give you a millivolt resolution chart.

I lay this sheet on my working bench and lay the graded devices next to the millivolt gradation line corresponding to how the device graded out. Don't let your cat in the room or he/she will knock all your hard work off your chart :)

Your body heat temporarily affects the DUT as all bipolar semiconductors have a negative tempco on the base emitter junction. Wear gloves to minimize this or use needle nose pliers to install the DUT in the tester socket.

Hope this helps.
 

Gepetto

Administrator
Staff member
Joined
May 15, 2011
Messages
13,473
Location
Sterling, MA
Tagline
Old 'Arn Enthusiast
#2
Wow. No cat huh?? that could be a problem. So you are looking for voltage matches at a regulated current through the BE jct? (Lee's question)
 

Gepetto

Administrator
Staff member
Joined
May 15, 2011
Messages
13,473
Location
Sterling, MA
Tagline
Old 'Arn Enthusiast
#3
Yes that is the idea. Remember the output drivers all have their bases connected with a fat bus wire and their emitters connected to a fat bus wire with an intervening emitter power resistor. So this Vbe matching is the critical matching parameter.

For the front end, long tailed pair, the theory is very similar but small signal in nature. The emitters of these are directly tied together and connect to R6. Thus the amp offset is directly affected by the Vbe of Q1 and Q2.

The matching method may be different for other non-Phase Linear amp applications but this is the best method for use within PL amps.
 

Gepetto

Administrator
Staff member
Joined
May 15, 2011
Messages
13,473
Location
Sterling, MA
Tagline
Old 'Arn Enthusiast
#4
The gains will be less significant for this application than the Vbe matching.

Here is an article that does a decent job explaining the method that is appropo for the Phase Linear application.

http://home.comcast.net/~ijfritz/projec ... mat001.pdf

And definitely keep your cat out of the room when you do this. She thought those little transistors were fun to play with and shuffled all my earlier sorting.
 

orange

Veteran and General Yakker
Joined
Jul 6, 2010
Messages
17,704
Tagline
Broken beyond repair but highly affable
#6
So is it true you can tell a bipolar transistor by what it's doped with?
 

mlucitt

Veteran and General Yakker
Joined
Jun 24, 2011
Messages
3,339
Location
Jacksonville, FL
#8
I had the opportunity to talk with John Hillig at Musical Concepts http://www.musicaldesign.com/
They work on Hafler gear and produce some nice (and expensive) mods for the Hafler and Dynaco amps. I was asking John about Kevin's P500 and if there were any common failure modes. He said there really are not any.

I described the symptons: One channel OK, the other channel distorted, clipping light comes on with minimun input, top of waveform flattened (clipped). He said it sounded like bad outputs. He said to use a "Super Cricket" transistor/FET tester, nothing else will really work on the MOSFETs. He also said to check the resistance on the bias resistors; he said if they are 470 Ohms they should read close to that (Hafler used 5%). If they read 460 or maybe 450 Ohms, then the MOSFET connected to that resistor is bad. I guess they fail "shorted" when they go. He also said to check the amplifier bias, pull the rail fuses and connect the DMM across the fuse clips to read the current flow.

I wondered how that check would translate to the Phase Linear amps? (required PL content) Lee, wanna try it? It would be interesting to find a value that could be adjusted and compared to the bias reading described in the Service Manual for the PL 400 and the PL700.

Mark
 

laatsch55

Administrator,
Staff member
Joined
Jan 14, 2011
Messages
74,124
Location
Gillette, Wyo.
Tagline
Halfbiass...Electron Herder and Backass Woof
#9
Toi write a bias SETTING procedure using current draw off the rails instead of measuring resistance accross R38??
 

mlucitt

Veteran and General Yakker
Joined
Jun 24, 2011
Messages
3,339
Location
Jacksonville, FL
#10
Maybe it is more of a load balancing than a "bias" like we are used to in the PL world. The Hafler world is a little more distorted and slightly more out of phase.

From the Hafler P500 Operation Manual:
"P2 adjusts the voltage to bias the amplifier for class AB operation and therefore determines the quiescent (idle) current in the driver and output stages."

There is no reference in the manual telling where to make the measurement.

Mark
 

mlucitt

Veteran and General Yakker
Joined
Jun 24, 2011
Messages
3,339
Location
Jacksonville, FL
#12
Well, we are talking about MOSFETs but there are common points for the gates and drains. I need to study the schematics a bit more. The real problem is the "block" type of heatsink, like you have on the Frankenamp, makes it difficult to get at the measurement points when the amp is powered up.
The PL amps are great to work on if you remove the front panel and tip the main board forward.
 

Gepetto

Administrator
Staff member
Joined
May 15, 2011
Messages
13,473
Location
Sterling, MA
Tagline
Old 'Arn Enthusiast
#13
MOSFETs have a postitive Tempco D-S so they naturally share well in a paralleled configuration. MOSFET D-S channel ON resistance increases with increasing temperature. This paralleling is done all the time in high power switching, SMPS converters and inverter applications. This is opposite the behavior of bipolar transistors.

A frequent problem of paralleled MOSFETs is parasitic high frequency oscillation that occurs when the gates of paralleled devices are directly connected to each other and driven in common. It is good practice to insert a small series gate resistor to each device (~22 ohms) to prevent this type of oscillation. Matching type and date code of MOSFETs still remains good practice however. Date code matching, however, does not always guarantee that the die in two different devices within the same date code came from the same wafer.
 

laatsch55

Administrator,
Staff member
Joined
Jan 14, 2011
Messages
74,124
Location
Gillette, Wyo.
Tagline
Halfbiass...Electron Herder and Backass Woof
#14
Were you an EE in a previous life also?? We have yet to ask a question of anything electronic in nature that you couldn't aswer exhaustively. A lifetime (or two) of learning obviously.
 

laatsch55

Administrator,
Staff member
Joined
Jan 14, 2011
Messages
74,124
Location
Gillette, Wyo.
Tagline
Halfbiass...Electron Herder and Backass Woof
#15
A positive temperature co-efficient does not allow for uncontrolled thermal runaway, correct? So what are the drawbacks to mosfets??
 

Gepetto

Administrator
Staff member
Joined
May 15, 2011
Messages
13,473
Location
Sterling, MA
Tagline
Old 'Arn Enthusiast
#16
No drawbacks Lee. Just different.

And yes, the positive tempco behavior prevents thermal runaway in paralleled configurations.
 

nakdoc

Chief Journeyman
Joined
May 11, 2011
Messages
619
Location
Nashville, TN Music City
Tagline
highly biased
#17
The description for sorting based on Vbe is measuring the wrong parameters. Matching bipolars in audio applications is done to choose transistors with similar response to a fixed base current, response being collector current. Typically the transistor under test is placed in a jig exactly like the one described, but rather than measure Vbe, Ic is measured. This value is the current gain of the transistor. For early PL outputs, current gain would vary between 25 and 60, so the matching process might select a set of outputs with a gain of 40. All of these transistors would "share the load" nearly equally. However, transistor characteristics are highly dependent on Vc and Ic, and as described, temperature, to the point that a graphical display of transistor gains tells the best story.

I recommend an old book 'Transistors for Audiofrequency" by G. Fontaine, published by Hayden Book Co. to anyone who wants to teach themselves everything about transistors
 

wattsabundant

Chief Journeyman
Joined
Apr 26, 2011
Messages
842
Location
Central Ohio
#18
I'm not aware of any requirement by P/L for matching of outputs. The Crown DC150's and 300's did match outputs but those are the only ones I've seen. It's my understanding that the emitter resistors force current sharing. All of the P/L manuals have a procedure for verification of load sharing which measures the voltage drop across the emitter resistors.

I check outputs before I install them on a curve tracer and with a test set I built that uses a constant current source on the base of outputs and measures the collector current through a shunt. Although I've never had a failure right out of the box for On Semis and the like, one of the very few NTE replacements I bought was shorted. For TO92's I use the hfe test in a DVM.
 

oldphaser

Chief Journeyman
Joined
Sep 8, 2012
Messages
845
#19
The gains will be less significant for this application than the Vbe matching.

Here is an article that does a decent job explaining the method that is appropo for the Phase Linear application.

http://home.comcast.net/~ijfritz/projec ... mat001.pdf

And definitely keep your cat out of the room when you do this. She thought those little transistors were fun to play with and shuffled all my earlier sorting.
The link above no longer works.

I found a new link: http://ijfritz.byethost4.com/MiscProj/transmat001.pdf?i=1

and an archived copy at: https://web.archive.org/web/20121213171727/http://home.comcast.net/~ijfritz/projects/transmat001.pdf
 
Last edited:

WOPL Sniffer

Veteran and General Yakker
Joined
Aug 10, 2015
Messages
11,223
Location
Minnie-Soda
Tagline
Screw it
#20
Anybody recommend a couple BOOKS to have on hand for transistor lookup/substitution? I'd like to have a couple RCA books on hand which is useful for checking the parameters of transistors, or testing them on my new tester. Yeah, I could probably download a couple and print them by why kill more trees when there re a million out there looking for new owners.
 
Top