I am repairing a Class A amplifier with a blown channel. I found out that one of the dual transistors used as a differential pair got shorted. I decided to replace the input stage with KSC1845/KSA992 transistors. I understand that matching the input stage transistors is an absolute must. However, I am not sure what the case is when there is when there is both PNP and NPN side.
Is it necessary that all 4 transistors be matched with one another?
What would happen if the two sides were matched separately?
I attached the circuit, and this is the complementary differential pair in question:
Is it necessary that all 4 transistors be matched with one another?
What would happen if the two sides were matched separately?
I attached the circuit, and this is the complementary differential pair in question:
The two NPNs and the two PNPs need to be vbe and hFE matched for differential balance. Normally (like in a Leach amp) the NPN would need to be hFE matched to PNP as well to minimize input bias current. This circuit also uses FET source followers and intentionally runs the bases off of zero volts, which make the latter unnecessary.
Thank you for the replies! 🙂
Based on what Ed pointed out, does anyone know if there is a chance to use KSC1845/KSA992 instead of the original twin transistors if the discrete transistors are very closely matched? I now understand that could be a problem since the emitters are tied together... Alternatively, do modern equivalents exist for uPA74/uPA75 ?
Based on what Ed pointed out, does anyone know if there is a chance to use KSC1845/KSA992 instead of the original twin transistors if the discrete transistors are very closely matched? I now understand that could be a problem since the emitters are tied together... Alternatively, do modern equivalents exist for uPA74/uPA75 ?
Last edited:
I agree that it's desirable to have matched transistors, but I'm going to point out some contrarian considerations:
Note that the design incorporates a DC servo (IC 1) that should drive the amplifier output to near 0V.
The input FETs are source followers, so they will tend to absorb any net base current mismatch from NPN and PNP (Q2, Q3) pairs. More significant is offset voltage mismatch in the Q2, Q3 pairs, as it adds to offset error in FET pair. The FET pair has 30mV max offset. It's common wisdom that low offset error is desirable to minimize distortion in the input differential stage.
Given the DC servo, I suggest looking at servo output at IC1 pin 7. The closer it is to 0V, the better the inherent matching of the input stages. You can measure offset in the bipolar and FET pairs for insight to where any offset errors reside. Maybe you'll be content with your replacement parts.
Good luck!
Note that the design incorporates a DC servo (IC 1) that should drive the amplifier output to near 0V.
The input FETs are source followers, so they will tend to absorb any net base current mismatch from NPN and PNP (Q2, Q3) pairs. More significant is offset voltage mismatch in the Q2, Q3 pairs, as it adds to offset error in FET pair. The FET pair has 30mV max offset. It's common wisdom that low offset error is desirable to minimize distortion in the input differential stage.
Given the DC servo, I suggest looking at servo output at IC1 pin 7. The closer it is to 0V, the better the inherent matching of the input stages. You can measure offset in the bipolar and FET pairs for insight to where any offset errors reside. Maybe you'll be content with your replacement parts.
Good luck!
Thanks BSST!! I've learned so much in this thread already! Given the DC servo, does that mean that using discrete matched transistors (in my case KSC1845/KSA992), could work well despite not having the emitter degeneration resistors?
I have an idea that might work. I'd like to hear some thoughts on this.
I found DMMT5401/DMMT5551 SMD devices that might be good replacements for the original uPA74/uPA75. The transistors in these little devices are matched within 2%... which is pretty good? My idea is to design an adapter PCB to retrofit them. Would this be better than using hand-matched KSC1845/KSA992 pairs?
That's a good point, Ed.
I have an idea that might work. I'd like to hear some thoughts on this.
I found DMMT5401/DMMT5551 SMD devices that might be good replacements for the original uPA74/uPA75. The transistors in these little devices are matched within 2%... which is pretty good? My idea is to design an adapter PCB to retrofit them. Would this be better than using hand-matched KSC1845/KSA992 pairs?
Hi Ed,
I don't understand what you're saying. Would you elaborate?
I made the point that a servo can't correct offset error imbalance and attendant distortion. It can only force output error to 0V.
Thanks.
I don't understand what you're saying. Would you elaborate?
I made the point that a servo can't correct offset error imbalance and attendant distortion. It can only force output error to 0V.
Thanks.
Hi BSST - The problem goes beyond distortion. Any difference in the offset voltage of one differential pair versus the other will lead to incorrect quiescent current in the VAS. For this reason, emitter degeneration is usually applied to both differential pairs in designs that have two pairs.
Ed
Ed
I do not believe the original n/p pairs were required to be matched by the manufacturer or they would have said so. Look at the offset spec for the NEC parts, it's probably a few mV max so just use your DMM in diode test mode and select two that match within a few mV and your fine. Buying the ksa992 and ksc1845 on tape, I find that they Vbe match very well. You should be fine. Since the circuit has a DC servo it’s not necessary to thermal couple them.
The NPN pair does not need to be matched against the PNP pair. The two NPNs need to be matched against each other. The two PNPs need to be matched against each other. The matching includes a thermal connection since these transistors will heat up.
Ed
Ed
If the diff pair are in balance then they will heat up equally. This is not a DC amp so these thermal effects are not important.
Maybe. Any temperature difference leads to an offset voltage at 2mV/C. The transistors are dissipating 120mW each and the thermal resistance is ~200C/W.
The need for tight matching is unique to the fully complementary topology. I would agree with you if this were a single differential pair.
Ed
The need for tight matching is unique to the fully complementary topology. I would agree with you if this were a single differential pair.
Ed
I used many KSC/KSA, and i have good experience with matching. The hfe is usually between 400-500, so it is not big issue to find npn/pnp pairs.
As an example, the old Hafler designs DH-200,220,500 are all discrete complementary IPS designs. They did not specify device matching back then when device variation was not as good as it is today. Dc gain was unity because they used a ecap in the feedback. DH-200 did not have a DC offset trim either.
@rsavas I looked at the design briefly and it seems like the Hafler DH-200 has emitter resistors on each transistor in the complementary pair, whereas, the circuit I posted does not. It has the emitters tied together. Do you think that could present an obstacle in using discrete transistors in this design?