The commonly available 3-6 page datasheets for BJT power transistors usually don't go into or characterise behaviours like quasi-saturation, other than to display the collector current versus collector-emitter voltage curves without comment. Since it's important to whether power semis are suitable or otherwise, how should we or where can we derive this info?
Most if not all epitaxial devices using several epi layers (base as well as collector) will have some degree of quasi-saturation. Without data from the datasheets, it will have to be measured. A clue from some datasheets can be extracted from differences between 20V and 5V characteristics which ON semi published on their MJL3281A's for example, but 5V and above is often the voltage at which Q-S is not affecting gain very much.
Another clue might be extracted from the saturation curves. Look at the base current needed for a given Ic as a function of Vce and note how the current has to increase as Vce falls, starting at a point which is not obviously in saturation (eg 5V). THe difficulty there is that at higher voltages we don't know if the charactersitics will flatten to a more or less constant line (Early effect) or are still in the Q-S region.
I would have to say measure it on a sample of devices you might want to use.
Another clue might be extracted from the saturation curves. Look at the base current needed for a given Ic as a function of Vce and note how the current has to increase as Vce falls, starting at a point which is not obviously in saturation (eg 5V). THe difficulty there is that at higher voltages we don't know if the charactersitics will flatten to a more or less constant line (Early effect) or are still in the Q-S region.
I would have to say measure it on a sample of devices you might want to use.
Thanks for the clues. I can see this is likely to be heavy going and a little expensive for DIYs. It suggests test gear capable of at least 30V/5A and heavier duty hardware than might suit a beginner. Running experiments without suitable quality gear can also be a waste of time and money though. As a lab. tech. in both chemical and electrical R&D testing, I had to find out the hard way and several times over. Nonetheless, nothing ventured-nothing gained.
Last edited:
If DIY-ers have an oscilloscope a simple curve tracer can be created using it as the display. All I use is a simple oscillator (not particularly low distortion, typically like the one now hosted on sound-au's projects using a filament lamp stabiliser) feeding into a (robust) 50W (A.C. coupled) amplifier which then feeds a transformer 2x115V to 2x 25V either step down (for lower voltage/higher current) or step up (for high voltage low current) measurements, bridge rectified and feeding the transistor under test. Scope needs two channels which can be set to X-Y mode so that Y plots the current (AKA voltage across a 1 ohm resistor in the emitter) and X the Vce. As you turn the volume control up the sweep extends from zero to whatever. Not suitable for testing Vceo (although it can, if the transistors can stand being tested this way - many datasheets now suggest using a switched inductor). Separate isolated and also variable PSU used to set base current using a suitable series resistor. Have used this to measure quasi-saturation and Early voltages (but you have to make sure the transistors don't heat up and need good heatsinking. A DSO would capture single pulses, which would be ideal, and I'm guessing some already have DSO's. Some are now as cheap as my ageing CRT scope.
There is plenty of info on curve tracer on the web some very simple to one duplicating a Tektronix's. Long video but this gives a run down of some options and he does have the real thing.
Small signal though and if you want more detail on what the last method is "measuring" you need to watch the last 10min or so of one he calls a big crazy transistor one. These are more about matching transistors.
He also mentions the catch with buying some curve tracers built or in kit form when used on power transistors.
There are several simple designs out there and they often mention the problem with hooking them up to a DSO. That might be partly down to too much bandwidth and nvg conversion to digital. There have been some high sampling bit depth ones around with a limited bandwidth, No point if they are noisy but they have been costly. PC sound gear may have sufficient bandwidth for this sort of application. The real things don't use a high bandwidth for the usual measurements.
He also mentions the catch with buying some curve tracers built or in kit form when used on power transistors.
There are several simple designs out there and they often mention the problem with hooking them up to a DSO. That might be partly down to too much bandwidth and nvg conversion to digital. There have been some high sampling bit depth ones around with a limited bandwidth, No point if they are noisy but they have been costly. PC sound gear may have sufficient bandwidth for this sort of application. The real things don't use a high bandwidth for the usual measurements.
This thread is getting polluted (again). Perhaps discussions on measurements of transistors needs to be in a different one.
But for those of you interested in the method I use, read what I wrote and see if you can work out what I did. There are indeed plenty of examples "out there" of how to hook the actual device to the measurement system.
But for those of you interested in the method I use, read what I wrote and see if you can work out what I did. There are indeed plenty of examples "out there" of how to hook the actual device to the measurement system.
@lisoformio
I've started a curve tracer thread to give more info about my setup.
I've started a curve tracer thread to give more info about my setup.
So I took an old JLH build and set it up with 2SC5200's in the circuit I posted (#9070 back in 454).
First results (with some minor changes to the components) - bandwidth <6Hz-300kHz.
Stable into inductive and resistive loads.
example square wave at 50kHz into resistive load attached.
Really seems that Bode and Tian can show stability. But if anyone does try this and finds some problems I'd be interested to hear.
Changes: base resistors 47 ohms (was 22)
PNP input shunt: 220pF, 47 ohm resistor omitted as showed little effect in simulation.
Output capacitor only 6.8mF as was used when first built.
Measured without the output inductor. May need this for capacitive loads.
First results (with some minor changes to the components) - bandwidth <6Hz-300kHz.
Stable into inductive and resistive loads.
example square wave at 50kHz into resistive load attached.
Really seems that Bode and Tian can show stability. But if anyone does try this and finds some problems I'd be interested to hear.
Changes: base resistors 47 ohms (was 22)
PNP input shunt: 220pF, 47 ohm resistor omitted as showed little effect in simulation.
Output capacitor only 6.8mF as was used when first built.
Measured without the output inductor. May need this for capacitive loads.
Attachments
...and some distortion figures:
1kHz 1W 2nd .029% 3rd .004%
8W 2nd .08% 3rd .02%
20kHz 1W 2nd .025% 3rd .006%
8W 2nd .075% 3rd .06%
Note that with the base resistors the output is limited and the circuit needs a 32V supply to achieve 10W output (higher dissipation!)
The PSU I used for testing had a 30V limit, and at 9V rms the distortion increased - evidence of clipping. Load was 7.5 ohms (which also needs a higher current).
Not sure why 1kHz is slightly worse than 20kHz but I haven't calibrated my LDO for ages. But as expected I think for a JLH.
If you want lower distortion it will need some additional modifications.
1kHz 1W 2nd .029% 3rd .004%
8W 2nd .08% 3rd .02%
20kHz 1W 2nd .025% 3rd .006%
8W 2nd .075% 3rd .06%
Note that with the base resistors the output is limited and the circuit needs a 32V supply to achieve 10W output (higher dissipation!)
The PSU I used for testing had a 30V limit, and at 9V rms the distortion increased - evidence of clipping. Load was 7.5 ohms (which also needs a higher current).
Not sure why 1kHz is slightly worse than 20kHz but I haven't calibrated my LDO for ages. But as expected I think for a JLH.
If you want lower distortion it will need some additional modifications.
Last edited:
veysel - hope your kit works. In my testing the PSU I had was limited at 30V and 2A, so any short term current over 2A caused a limit (which may have also been a problem for measuring at full power (10W) - I need to look into that.)
It does concern me that the JLH PSU is only a capacitor multiplier. If that is the PSU circuit you are using it may not provide as good a regulation as expected. JLH specified 5mF in his original circuit and ran it from a 30V transformer. The ripple voltage on a 2.4A load is about 5V, and if your amplifier is to run down to 20Hz or lower then it needs a PSU which can provide double the standing current for at least the peak of the low frequencies. (The PSU current is not constant as some may think as the upper transistor current varies with the signal and therefore demands a varying current from the PSU with a signal).
I would have recommended JLH to have used a more standard PSU. With just 2 more transistors he could have got better stabilisation and lower ripple!
Also, I very much doubt that your kit will be using the frequency compensation I have suggested. Would you have the circuit for your kit?
It does concern me that the JLH PSU is only a capacitor multiplier. If that is the PSU circuit you are using it may not provide as good a regulation as expected. JLH specified 5mF in his original circuit and ran it from a 30V transformer. The ripple voltage on a 2.4A load is about 5V, and if your amplifier is to run down to 20Hz or lower then it needs a PSU which can provide double the standing current for at least the peak of the low frequencies. (The PSU current is not constant as some may think as the upper transistor current varies with the signal and therefore demands a varying current from the PSU with a signal).
I would have recommended JLH to have used a more standard PSU. With just 2 more transistors he could have got better stabilisation and lower ripple!
Also, I very much doubt that your kit will be using the frequency compensation I have suggested. Would you have the circuit for your kit?
Hi.Sorry i do not have a circuit. I hope it is not an advertising.It is here
https://tr.aliexpress.com/item/1005...st_main.5.5e083d12AWcUeR&gatewayAdapt=glo2tur
I want to run it with my EI transformer which have 16v output voltage and about 12A current.The kit has ttc5200 and i will use my used 2sc5200 and sell new TTC5200 ones( i think 2sc5200 may be better because of larger chip)
I will build it for my 4 ohm fullrange speakers.
https://tr.aliexpress.com/item/1005...st_main.5.5e083d12AWcUeR&gatewayAdapt=glo2tur
I want to run it with my EI transformer which have 16v output voltage and about 12A current.The kit has ttc5200 and i will use my used 2sc5200 and sell new TTC5200 ones( i think 2sc5200 may be better because of larger chip)
I will build it for my 4 ohm fullrange speakers.
Hi John, very interested in your latest posts with the slightly modified circuit you showed back in post #9070 - can I ask if you used a pcb for this latest design or did you breadboard it up on matrix board or similar? If a pcb, is it available or are gerbers around for it, thanks.. Also your comments on the PSU are food for thought. Would a way overspec simple diode bridge, CRC PSU with heaps of capacitance work well do you think, over a cap multiplier or regulated supply?
@Gary - no, I did not use a PCB. I actually used plain SRBP drilled as though it were a PCB, but there were some other mods I did not show. For example instead of a variable resistor in the driver stage I used a fixed 100 ohm to the +rail and binary weighted resistors of 470, 220, 100, 47 and 22 ohms with shunts to short out some until I achieved the quiescent current. Now it seems to work it would be useful to construct a PCB.
A CRC filter would work I think. I simulated JLH's PSU and two 10mF with 1 ohm which gave similar results but obviously the 1 ohm is an additional voltage burden. Ripple voltage (if I recall correctly- would have to check notes) was about 10mV when supplying 2.4A.
With a CCS transistor (PNP) instead of a potentiometer (JLH seems to like building things with tweaks to set currents etc) much of the ripple is reduced, which only needs a fourth transistor to shunt the current. I fed this from a Zener diode at the operating voltage (note -only in simulation!) with a 220 ohm base resistor. Zener shunted by a 47uF capacitor (and a diode connected in reverse on the base of the NPN shunt to prevent damage if the rail voltage collapses quickly). Ripple reduced to about 1mV.
What is a subtle effect in the JLH is that between signal and no signal, if the transistors used in the output do not have absolutely linear gain then the D.C. current will change with signal - because the average current in the upper transistor (and the lower for that matter) is different when the current sweeps from about zero to about twice Iq. A capacitor multiplier won't keep the voltage constant so between a large voltage signal passage and quiescent silence the voltage may change. It may not be audible but it might create some low frequency content which might cause some effects. I'd recommend a true regulator rather than a CM.
A CRC filter would work I think. I simulated JLH's PSU and two 10mF with 1 ohm which gave similar results but obviously the 1 ohm is an additional voltage burden. Ripple voltage (if I recall correctly- would have to check notes) was about 10mV when supplying 2.4A.
With a CCS transistor (PNP) instead of a potentiometer (JLH seems to like building things with tweaks to set currents etc) much of the ripple is reduced, which only needs a fourth transistor to shunt the current. I fed this from a Zener diode at the operating voltage (note -only in simulation!) with a 220 ohm base resistor. Zener shunted by a 47uF capacitor (and a diode connected in reverse on the base of the NPN shunt to prevent damage if the rail voltage collapses quickly). Ripple reduced to about 1mV.
What is a subtle effect in the JLH is that between signal and no signal, if the transistors used in the output do not have absolutely linear gain then the D.C. current will change with signal - because the average current in the upper transistor (and the lower for that matter) is different when the current sweeps from about zero to about twice Iq. A capacitor multiplier won't keep the voltage constant so between a large voltage signal passage and quiescent silence the voltage may change. It may not be audible but it might create some low frequency content which might cause some effects. I'd recommend a true regulator rather than a CM.