Hi,
I'd like to know the important parameters to choose the active device of a Vbe multiplier, aka bias spreader. Most important thing seems to be ease of mounting it in very close thearmal contact to the output devices of a power amp. So we nowadays choose SOT32/TO126 devices, as they have become rather cheap. Formerly TO92 transistors, plugged into a 5 mm hole by using thermal glue, were more common.
Next would be the collector current rating, which has to exceed the current through the Vas or predriver stage.
Transition frequency, Vce and Ptot don't seem to matter at all, as the spreader uses to be in parallel with a large capacitor and bias voltage commonly is some volts only (and collector current in the range of some few mA's).
But what about hfe? Can I, e.g., use those ancient video transistors with their rather modest hfe, such as BF457, that I have laying around in large quantities?
Best regards!
I'd like to know the important parameters to choose the active device of a Vbe multiplier, aka bias spreader. Most important thing seems to be ease of mounting it in very close thearmal contact to the output devices of a power amp. So we nowadays choose SOT32/TO126 devices, as they have become rather cheap. Formerly TO92 transistors, plugged into a 5 mm hole by using thermal glue, were more common.
Next would be the collector current rating, which has to exceed the current through the Vas or predriver stage.
Transition frequency, Vce and Ptot don't seem to matter at all, as the spreader uses to be in parallel with a large capacitor and bias voltage commonly is some volts only (and collector current in the range of some few mA's).
But what about hfe? Can I, e.g., use those ancient video transistors with their rather modest hfe, such as BF457, that I have laying around in large quantities?
Best regards!
I don't think that is the way the Multiplier passes current.
The main current through the Vbe multiplier transistor is near constant DC.
The low AC impedances in parallel with the transistor suck most of the AC current rather than passing all the AC through the transistor.
If you have a 10mA VAS and allow for upto ±10mApk as the AC signal, then you would probably find that the maximum current peak through the Vbe multiplier transistor never exceeds 10mApk (110% of the 9mA transistor quiescent current).
If the Multiplier voltage is 3.6Vdc to drive a triple EF output stage, then the highest transient dissipation is likely to be around 36mW. Well within the capability of a 225mW sot23.
That sot23 can be glued to the collector lead for quickest response time and highest response temperature to Tj of that output device.
Is there an LTspice expert out there willing to simulate the worst case dissipation in the Vbe multiplier transistor?
Last edited:
The main point of a bais spreader is to create a voltage difference between the positive and negative drivers to create a bit of class a operation. The second main function is to keep this class a operation constant as the output stage changes temperature. The transistor(s) sensing the temperature must change their current flow the same amount proportionally as the output devices and drivers to achieve this.
The main function of the Vbe Multiplier is as a Shunt Voltage Regulator. It tries to keep a constant DC voltage across the following stages two inputs. Not necessarily maintaining ClassA. The next stage could be switching off on alternate waveform halves.
The secondary purpose is to add a temperature compensation to the Vbias so that output bias current does not change significantly as temperatures around the output stage change.
Last edited:
Okay. Feel free to nit pick about some ridiculous miniscule point then if it makes you feel better.
Thanks. To my understanding a Vbe multiplier equals an adjustable Zener diode, and shares it's low dynamic impedance.
And now, what about hfe? Is it really necessary e.g. to use a 2SC3503C (hfe=40, ~40 €ct), or a MJE340 (hfe=30, ~50 €ct), or a BD139 (hfe=25, ~50 €ct), or can I take my BF457 (hfe>25) that don't cost me anything? Oh, I see, hfe of these don't differ all too much, so there's no need to worry about?
Best regards!
And now, what about hfe? Is it really necessary e.g. to use a 2SC3503C (hfe=40, ~40 €ct), or a MJE340 (hfe=30, ~50 €ct), or a BD139 (hfe=25, ~50 €ct), or can I take my BF457 (hfe>25) that don't cost me anything? Oh, I see, hfe of these don't differ all too much, so there's no need to worry about?
Best regards!
The actual gain of the device isn't as important as the actual gain and VBE change with temperature. Ostripper put a lot of time into engineering the Honeybadger, including the VBE multiplier tempco. A $0.50 transistor is pretty cheap. Why reinvent the wheel to try to save a few cents while risking thermal runaway problems?
hFE matters. The idea is sensing a voltage-divider to make a voltage multiplier. If hFE is low, base current loads the resistor divider, spoiling the intended factor. You trim this out, then hFE changes with temp, and it is off-factor again.
I would lean to low current (100mA), insignificant Vceo rating (5V would be ample but you don't see <12V), and, in a basic scratch design, high hFE.
Figure the Hie of the transistor and make the divider resistors much smaller.
However the semi-predictable change of hFE with temperature *can* be used by a very-clever designer to add a 2nd-order compensation for a mis-match between bias and output devices. So if you are using someone else's work, use their parts.
I would lean to low current (100mA), insignificant Vceo rating (5V would be ample but you don't see <12V), and, in a basic scratch design, high hFE.
Figure the Hie of the transistor and make the divider resistors much smaller.
However the semi-predictable change of hFE with temperature *can* be used by a very-clever designer to add a 2nd-order compensation for a mis-match between bias and output devices. So if you are using someone else's work, use their parts.
If you look at many designs on this forum, and in books, they use the BD139/140 in a to-126 package, because it is cheap and easy to use. the demands on this part are not very critical from what I can tell.
using a sm sot-23 is a pita really, so I suggest to use the to-126 until they become obsolete which I think is a long way off. a sot-223 is more equiv to a to-126 as far as thermal performance goes. if you are stuck using smt for this function a small double sided pcb with a lot of thermal vias and copper pours would do the trick.
As with any design, try to minimize the amount of different parts, unless economics /profits say otherwise.
jw if have not figured out, at is a stickler for details = a good EE
using a sm sot-23 is a pita really, so I suggest to use the to-126 until they become obsolete which I think is a long way off. a sot-223 is more equiv to a to-126 as far as thermal performance goes. if you are stuck using smt for this function a small double sided pcb with a lot of thermal vias and copper pours would do the trick.
As with any design, try to minimize the amount of different parts, unless economics /profits say otherwise.
jw if have not figured out, at is a stickler for details = a good EE
any amp design that you have captured in ltspice, will instantly tell you device static Pd, just run a .op simulation. the vas current is constant, is it not? and the Vdrop across c-e is constant too.
the VAS current is not constant.
The source or sink for the VAS tries to be constant.
But the drivers source or sink current to or from the VAS line.
The AC voltage across the Vbe multiplier should be near zero, so should have little effect.
But I can't get my head around using LTspice.
Identifying the worst case power dissipation is probably not easy and may take looking at a large variety of quite different operating conditions and abuses to find that worst case. It is likely to be a transient effect and the instantaneous peak dissipation may be what becomes one of the worst cases.
If one were to look at DC conditions, then Pq is very easy and it's done in one's head rather than using a simulation of a complex model to predict an answer.
The source or sink for the VAS tries to be constant.
But the drivers source or sink current to or from the VAS line.
The AC voltage across the Vbe multiplier should be near zero, so should have little effect.
But I can't get my head around using LTspice.
Identifying the worst case power dissipation is probably not easy and may take looking at a large variety of quite different operating conditions and abuses to find that worst case. It is likely to be a transient effect and the instantaneous peak dissipation may be what becomes one of the worst cases.
If one were to look at DC conditions, then Pq is very easy and it's done in one's head rather than using a simulation of a complex model to predict an answer.
Last edited:
I got better at ltspice after going through the tutorials in Bob Cordell's book.
To me, the book makes this circuit design more understandable.
I also have learnt using user's astx and dado submitted designs and simulations of their designs.
I think you know mooly has a thread on ltspice too.
So if you say the VAS current is not constant, then you can put voltage and current probes across the device to show Pd.
There is a special transistor from SANKEN therefore - go to post #41 under
https://www.diyaudio.com/community/threads/optimizing-the-vbe-multiplier.216385/page-3