One strange thing I noticed is that for this new pair of drivers I ordered my DMM shows huge discrepancy in Hfe. The PNP shows 230 while the NPN shows only 3. This is really strange. MJE15032 specs say the Hfe should be in the range 10-70 depending on Ic. Does anyone have an explanation to this?
Some times multimeter do not measure gain properly espcally with power transisstors.
To test for actuall gain of output transistors, its best to make your own test circuit, I can recomend you a all purpourse gain tester also with funcionality to test fake transitors and put it transistors on stress.
Have you doubble checked your circuit connects to be all correct? If not I recomend doing it.
To test for actuall gain of output transistors, its best to make your own test circuit, I can recomend you a all purpourse gain tester also with funcionality to test fake transitors and put it transistors on stress.
Have you doubble checked your circuit connects to be all correct? If not I recomend doing it.
Last edited:
Usually the gain of a transistor is given for a range of current and voltage applied. Those devices are meant to carry fairly high current. Your meter is testing at next to no current.
When you are trying to test your amplifier, pay attention to the temperature of the Zobel resistor on the output. If it's getting hot the amp is oscillating causing your outputs to pop.
Hi DIYnovice,
That works but isn't necessary. A normal control allows you to try different bias levels and is generally pretty reliable. Put across the E-B junction in an Vbe multiplier (normal bias circuit), the bias will drop to zero if that part opens up. So it fails safe. Use an extra resistor to limit the maximum bias adjustment range to protect the amplifier.
One thing I do like about that method (BGW-750C for example) is that it keeps idiots from wrecking the amplifier.
-Chris
That works but isn't necessary. A normal control allows you to try different bias levels and is generally pretty reliable. Put across the E-B junction in an Vbe multiplier (normal bias circuit), the bias will drop to zero if that part opens up. So it fails safe. Use an extra resistor to limit the maximum bias adjustment range to protect the amplifier.
One thing I do like about that method (BGW-750C for example) is that it keeps idiots from wrecking the amplifier.
-Chris
Get yourself this tester (dy294): DUOYI Digital Transistor Parameter Tester DY294 | eBay
Good enough for most bjt and diode tests.
cheers,
Good enough for most bjt and diode tests.
cheers,
I think most DMM hFE faciticity use a nearly fixed level of base current and monitor the collector current to arrive at an hFE value for that base current.
I suspect the (slightly variable) base current is around 1uA.
Hi Andrew,
Actually, you would have to examine the meter. Most use the diode function current source which will be close to 1 mA. You can test this by running a "diode test" on an accurate ammeter. In a calibration lab, we would typically measure the voltage drop across a 250 R standard resistor and calculate the current. The reason for that is that current is not something that is as accurately measured in meters compared to voltage. By the same token, resistance measurements are not as high a performance as we would like it to be.
I think a reason for poor accuracy in resistance measurements is the size of the resistor used to create the voltage drop (which is then measured and the result calculated) is a tiny think inside meters. Even meters like my HP 3456A, 3457A and 34401A. Never mind how small they are in hand held meters! A resistance standard is a huge resistor - larger than a can of pop / soda. The amount of self heating that takes place is almost non-existent with the standard resistor, while in bench meters it is a factor. With hand held meters, this resistance is physically small enough to experience heating.
Long story short, meters do typically use the 1 mA current for the diode test to set either the emitter or collector current. The base current is then measured and the ratio displayed.
Meters like the Heathkit transistor checkers have the operator set the emitter or collector current to exactly 1 mA. The base current is then read off the scale calibrated in beta. I know it doesn't sound like an accurate way to do this, but it is considering all the other factors. The base current is measured using a higher value resistor while keeping the heating factor to a minimum. Beta is also variable an not measured accurately. That's one reason I created a transistor matching jig. It was the only way to keep the temperatures between the parts the same, and then to balance the currents between them. Once in equilibrium, you can measure the voltage drops in the jig and obtain accurate readings. Accurate for that test current and temperature that is. The idea is really to match a pair of transistors, not to figure out the beta.
-Chris
Actually, you would have to examine the meter. Most use the diode function current source which will be close to 1 mA. You can test this by running a "diode test" on an accurate ammeter. In a calibration lab, we would typically measure the voltage drop across a 250 R standard resistor and calculate the current. The reason for that is that current is not something that is as accurately measured in meters compared to voltage. By the same token, resistance measurements are not as high a performance as we would like it to be.
I think a reason for poor accuracy in resistance measurements is the size of the resistor used to create the voltage drop (which is then measured and the result calculated) is a tiny think inside meters. Even meters like my HP 3456A, 3457A and 34401A. Never mind how small they are in hand held meters! A resistance standard is a huge resistor - larger than a can of pop / soda. The amount of self heating that takes place is almost non-existent with the standard resistor, while in bench meters it is a factor. With hand held meters, this resistance is physically small enough to experience heating.
Long story short, meters do typically use the 1 mA current for the diode test to set either the emitter or collector current. The base current is then measured and the ratio displayed.
Meters like the Heathkit transistor checkers have the operator set the emitter or collector current to exactly 1 mA. The base current is then read off the scale calibrated in beta. I know it doesn't sound like an accurate way to do this, but it is considering all the other factors. The base current is measured using a higher value resistor while keeping the heating factor to a minimum. Beta is also variable an not measured accurately. That's one reason I created a transistor matching jig. It was the only way to keep the temperatures between the parts the same, and then to balance the currents between them. Once in equilibrium, you can measure the voltage drops in the jig and obtain accurate readings. Accurate for that test current and temperature that is. The idea is really to match a pair of transistors, not to figure out the beta.
-Chris
Hi janusz,
For the reasons given in the post above, that instrument you linked to will never give you an accurate reading on transistor beta. Junction voltage drop is also variable with temperature. So it may help with quick checks if your meter doesn't include them (most do these days), but it can only help you with a pre-sort for matching transistors. That is only if you if you scrupulously stick to a method that prevents the parts from warming up. Believe me, it's a pain. The reason for this is to have a hope that all the transistors will be at the same temperature when you test them, which will only be true if they sit at room temperature for at least 1/2 hour. You must also avoid breathing on them. Take it from me, this is difficult to do! Of course, this means that you aren't touching any parts. Use some tweezers, non-metallic being the best as they don't conduct heat energy.
Best, Chris
For the reasons given in the post above, that instrument you linked to will never give you an accurate reading on transistor beta. Junction voltage drop is also variable with temperature. So it may help with quick checks if your meter doesn't include them (most do these days), but it can only help you with a pre-sort for matching transistors. That is only if you if you scrupulously stick to a method that prevents the parts from warming up. Believe me, it's a pain. The reason for this is to have a hope that all the transistors will be at the same temperature when you test them, which will only be true if they sit at room temperature for at least 1/2 hour. You must also avoid breathing on them. Take it from me, this is difficult to do! Of course, this means that you aren't touching any parts. Use some tweezers, non-metallic being the best as they don't conduct heat energy.
Best, Chris
This is true (especially for mosfets as one has to match their VGS at quite high currents as well) but for most purposes this tester is good enough as bjt power transistors do not have to be precisely matched (emitter resistors are used to stabilize the bias point - practically compensate for some variation in hfe) and it is cheap enough.
To reduce variation in measurement conditions one should place measured power transistor on a heatsink (I use 6mm aluminium or copper plates). I use three identical heatsinks when measuring a number of transistors at one go so hfe of each one is measured at a similar temperature. Good commercial amplifiers do not have individual power transistor matched at all - the same hfe range is good enough. Also worth remembering is that temperature is not identical at every transistor location along the heatsink.
For small signal transistors I use PEAK ATLAS DCA55. There is also more advanced DCA75 Pro but it costs about $150 - three times as much.
cheers,
To reduce variation in measurement conditions one should place measured power transistor on a heatsink (I use 6mm aluminium or copper plates). I use three identical heatsinks when measuring a number of transistors at one go so hfe of each one is measured at a similar temperature. Good commercial amplifiers do not have individual power transistor matched at all - the same hfe range is good enough. Also worth remembering is that temperature is not identical at every transistor location along the heatsink.
For small signal transistors I use PEAK ATLAS DCA55. There is also more advanced DCA75 Pro but it costs about $150 - three times as much.
cheers,
How about this guys. I was thinking of getting one. What are your thoughts on this meter. Peak DCA75 Atlas Advanced Semiconductor Analyser with Curve Tracing | eBay
Hi Stuart,
Reading curves will take more time than even my jig waiting for thermal equilibrium. I wouldn't call a curve tracer useless, but you do have to interpret the information. I have an old Heathkit IT-3121 that works. I have been thinking of installing an oscillator to run it faster so the entire family of curves can be seen. The limit is about 6 at the current sweep speed.
-Chris
Reading curves will take more time than even my jig waiting for thermal equilibrium. I wouldn't call a curve tracer useless, but you do have to interpret the information. I have an old Heathkit IT-3121 that works. I have been thinking of installing an oscillator to run it faster so the entire family of curves can be seen. The limit is about 6 at the current sweep speed.
-Chris
Hi janusz,
I can mount 4 BJT outputs or 4 Mosfet outputs at a time in the jig I made. Two different jigs of course. There I run the parts up to a set temperature (close, but not close enough really for units not in that group). I have loads which allow me to measure Collector / Drain current and also Base current of Gate voltage. The supply can be run at any arbitrary voltage, but normally they sit around 10 V. The operating current is easily varied so you can even run the outputs at the current and voltage levels you will see in the application.
If I had a quad ADC instrument, a family of curves could be generated easily for all four.
-Chris
I can mount 4 BJT outputs or 4 Mosfet outputs at a time in the jig I made. Two different jigs of course. There I run the parts up to a set temperature (close, but not close enough really for units not in that group). I have loads which allow me to measure Collector / Drain current and also Base current of Gate voltage. The supply can be run at any arbitrary voltage, but normally they sit around 10 V. The operating current is easily varied so you can even run the outputs at the current and voltage levels you will see in the application.
If I had a quad ADC instrument, a family of curves could be generated easily for all four.
-Chris
I have one! Well you get to look at some graphsics , only error is the 1% tol. but that is within range of tol avg., if you put a 50pcs tran. or jfet throught it some gaphifcs may satifacton of math. pair. per your choiche. . I like the calmness of grapfics at 1 % it will satify you , if you have a good dmm in the 5/6 trust the vbe of that take the graphs witch are better , your ......Good enough should be ,if not better not to be bitter , done . Enough.
That sounds great. Thanks for your time and efforts explaining all that. I had a look on ebay and I see one of the Heathkit IT-3121 there. So how do you see the results. The actual curves using this.... do you hook it up to your scope?
Any chance you could post some photos of your setup snd what the output curves look like.
Is there another product that you would recommend for this application?
Hi Stuart,
Yes, the IT-3121 is connected to an oscilloscope and there is a calibration series of steps. Then you're off to the races. Most oscilloscopes are 2% accurate at best, although the new digital ones would probably be better. The line rate wouldn't be a problem then either.
I used to use an IT-18 for matching transistors, but noticed that sometimes the matches were very good, and other times they were pretty bad. Eventually I figured out that transistors are even more sensitive to temperature than I thought. That's when I designed the jig (for signal transistors) that kept a pair of transistors at the same temperature. The same issues also hold when using a curve tracer. To match two parts, they must be tested at exactly the same temperature. By that, I mean they need to be in physical contact and isolated from the outside conditions on your bench.
Curve tracers do look cool though.
-Chris
Yes, the IT-3121 is connected to an oscilloscope and there is a calibration series of steps. Then you're off to the races. Most oscilloscopes are 2% accurate at best, although the new digital ones would probably be better. The line rate wouldn't be a problem then either.
I used to use an IT-18 for matching transistors, but noticed that sometimes the matches were very good, and other times they were pretty bad. Eventually I figured out that transistors are even more sensitive to temperature than I thought. That's when I designed the jig (for signal transistors) that kept a pair of transistors at the same temperature. The same issues also hold when using a curve tracer. To match two parts, they must be tested at exactly the same temperature. By that, I mean they need to be in physical contact and isolated from the outside conditions on your bench.
Curve tracers do look cool though.
-Chris
Hi everybody,
I'm waiting for a pair of boards from the store and in the meantime I'm doing part selections (based upon the offcial BOM) from my local resseller (TME in Europe).
Can you tell me if I'm not making wrong choices here?
Q1-Q2: anything else than mpsa18 and ss9014?
Q16-Q18: 2SC5200
Q19-Q21: 2SA1943
Q14: 2SC4793
Q15: 2SA1837
Q11-Q13: MJE340STU
Q10: MJE350
Q3,Q4,Q7,Q8: BC550CTA
Q5,Q6,Q9: BC560CTA
D3: 1N4744A-FAI
D4,D5,D8,D9: 1N5408RLG
R7: 3296W-1-201LF BOURNS
R30: 3296W-1-501LF BOURNS
R17: 3296X-1-102LF BOURNS
R37-R42: MPR5W-0R22 (non inductive and seems like the legs shall be long enough to fit the board)
R49,R50: 3W-10R MOR03SJ0100A19
Other res and capacitors are quite straightforward so I'm not listing them here.
I'll be using Diyaudio PSU boards with 4x 10000uf 100V each and maybe a couple of 300Va 45V+45V trannies (Toroidy). I'm going the pure dual mono road.
I'm waiting for a pair of boards from the store and in the meantime I'm doing part selections (based upon the offcial BOM) from my local resseller (TME in Europe).
Can you tell me if I'm not making wrong choices here?
Q1-Q2: anything else than mpsa18 and ss9014?
Q16-Q18: 2SC5200
Q19-Q21: 2SA1943
Q14: 2SC4793
Q15: 2SA1837
Q11-Q13: MJE340STU
Q10: MJE350
Q3,Q4,Q7,Q8: BC550CTA
Q5,Q6,Q9: BC560CTA
D3: 1N4744A-FAI
D4,D5,D8,D9: 1N5408RLG
R7: 3296W-1-201LF BOURNS
R30: 3296W-1-501LF BOURNS
R17: 3296X-1-102LF BOURNS
R37-R42: MPR5W-0R22 (non inductive and seems like the legs shall be long enough to fit the board)
R49,R50: 3W-10R MOR03SJ0100A19
Other res and capacitors are quite straightforward so I'm not listing them here.
I'll be using Diyaudio PSU boards with 4x 10000uf 100V each and maybe a couple of 300Va 45V+45V trannies (Toroidy). I'm going the pure dual mono road.
Last edited:
Alright, I replaced all driver and output stage transistors with brand new ones from Mouser to avoid any potential problems with using damaged, leaky or fake parts (at my first power up I mistakenly mixed up PNP and NPN transistors and also used drivers from eBay), so that issue is out of the way.
Now, when I power the amp and measure voltage across temporary 10 Ohm fuse replacements I see around 1V which gradually goes up to 1.6V and then down to 1.3V as VBE gets warmed up. Is this normal? The Build Guide says anything more than 0.5V at idle means I need to check my work....
I am using +/- 67V PSU and the longest I have powered the amp so far is a few minutes. I am afraid to keep it up longer so I don't blow any more output transistors (I have blown enough :-( )
The driver stage has MJE15030/MJE15031 and output stage has NJW0281/NJW0302. Voltage across TP1 and TP2 is around 15 mV. Thank you for your help!
Now, when I power the amp and measure voltage across temporary 10 Ohm fuse replacements I see around 1V which gradually goes up to 1.6V and then down to 1.3V as VBE gets warmed up. Is this normal? The Build Guide says anything more than 0.5V at idle means I need to check my work....
I am using +/- 67V PSU and the longest I have powered the amp so far is a few minutes. I am afraid to keep it up longer so I don't blow any more output transistors (I have blown enough :-( )
The driver stage has MJE15030/MJE15031 and output stage has NJW0281/NJW0302. Voltage across TP1 and TP2 is around 15 mV. Thank you for your help!
Last edited: