Transistor Substitutions

[wyn palmer]

Hello Again,

Continuing the question of FT, do you think it could hurt to replace Rotel RA-01 15Mhz output transistors (2SB817/2SD1047) with 30MHz toshiba or Fairchild 2SA1943/2SC5200 ?

It's highly unlikely that there will be problems simply because a transistor is replaced with another transistor with a 2x FT, all other aspects being equal, in an audio application.
FT simply means that the device maintains its beta/hfe/current gain as greater than unity up to
that "transition frequency"- at FT IC=IB.
In general a higher FT is a good thing and it's highly unlikely that a 2xFT will be problematic.
For a technical analysis see:
https://ocw.mit.edu/courses/electri...vices-spring-2007/lecture-notes/lecture37.pdf
Basically, FT is a function of the base width of the transistor.
Usually the current gain of the output devices is NOT a significant contributor to the gain, stability, or loop bandwidth of an amplifier. Amps with feedback usually behave like inductors at frequencies beyond where the open loop gain falls due to the increase in output impedance caused by the reduction in the amount of negative feedback, and this will not be affected by the increase in the FT of the output devices.
At frequencies above FT/beta (about FT/100) the output devices themselves start to look like they have inductive output impedances, but this rarely is problematic.

 

[dlucy]

Hello Again,

Continuing the question of FT, do you think it could hurt to replace Rotel RA-01 15Mhz output transistors (2SB817/2SD1047) with 30MHz toshiba or Fairchild 2SA1943/2SC5200 ?

From my limited experience, I'd say those are fine.

 

[chazix]

In general a higher FT is a good thing

I have no grounds to question the reasoning behind this statement, except this: it is contrary to my experience as a fix-it guy. That is, on more than one occasion, I have needed to replace output transistors in an actual, vintage, more-or-less consumer audio power amplifier, and found that the amp became unstable with my first choice of contemporary production transistors (which are almost exclusively higher in Ft by 2X or more than the originals). One fairly recent example was an Altec 9440, which originally used parts with, IIRC, about 4MHz Ft. It's hard to find any parts that slow these days, much less in TO-3 packages.

So, not to dispute the theory, but I can't help but wonder if there's a theory vs practice discrepancy here (where practice refers to repairs). Is there anybody out there who can relate an actual experience of making an audio power amp unstable by replacing its outputs with lower-Ft parts than the originals?

Thanks,

chazix

 

[wyn palmer]

I have no grounds to question the reasoning behind this statement, except this: it is contrary to my experience as a fix-it guy. That is, on more than one occasion, I have needed to replace output transistors in an actual, vintage, more-or-less consumer audio power amplifier, and found that the amp became unstable with my first choice of contemporary production transistors (which are almost exclusively higher in Ft by 2X or more than the originals). One fairly recent example was an Altec 9440, which originally used parts with, IIRC, about 4MHz Ft. It's hard to find any parts that slow these days, much less in TO-3 packages.

So, not to dispute the theory, but I can't help but wonder if there's a theory vs practice discrepancy here (where practice refers to repairs). Is there anybody out there who can relate an actual experience of making an audio power amp unstable by replacing its outputs with lower-Ft parts than the originals?

Thanks,

chazix

When you say unstable, what do you mean? Was it oscillating- if so what was the frequency? Was it at the closed loop -3dB bandwidth of the amp or elsewhere?
Stability theory is pretty straightforward for most older bipolar amps- there's usually a single cap that sets the dominant pole, and the basic design looks like this

CC is the dominant pole cap. D1,D2, Q11, Q12 comprise the unity gain (buffer) output stage.
The pole is set by the gm of the input stage and CC. The closed loop gain is set by the ratio of RF2 and RF1, feedback is taken from the output. Reducing the FT of the output devices can reduce the bandwidth of the gain stage (Q3) (something called pole splitting occurs in the gain stage and the extra cap load moves the higher frequency pole lower) and lower the phase margin of the amp- which could be problematic.
Increasing FT excessively could cause local instability in the output buffer due to interactions with various parasitics etc.
I confess to never have tried replacing output devices with lower FT types. I've replaced, in a prior life, devices with marginally higher FT devices and not experienced problems.

 

[chazix]

Was it oscillating- if so what was the frequency?

Yes, oscillating. I don't recall any frequency specifics. (Probably never measured, since one doesn't want to let amplifiers be oscillators for very long .) I use resistive loads for most testing.

 

[wyn palmer]

Yes, oscillating. I don't recall any frequency specifics. (Probably never measured, since one doesn't want to let amplifiers be oscillators for very long .) I use resistive loads for most testing.

Was the oscillation in the audio band? If so was it at LF or HF?
If in the audio band then assuming the amp has a closed loop bandwidth not much higher than 20kHz- which was typical for older designs due to the lack of suitable devices- then the implication is that the phase margin fell not that the loop bandwidth increased which is hard to understand.
Do you know what the stable 3dB bandwidth of the amp was?
Was the output bias current correctly set?

 

[chazix]

Was the oscillation in the audio band?

Supersonic by a big margin, but that's as specific as I can be.

In an Altec 9440A manual, it says 100KHz for the high -3dB point. It specifies that at 1W, though, so I'm not sure how much the final output devices would be contributing.

I don't think it's possible to set output bias current for an amp that's oscillating.

Your mention of parasitics made me think of "base stoppers" - is it conceivable that faster output parts may have necessitated stoppers? (I'm pretty sure the Altec was not designed with stoppers.)

 

[wyn palmer]

Yes, oscillating. I don't recall any frequency specifics. (Probably never measured, since one doesn't want to let amplifiers be oscillators for very long .) I use resistive loads for most testing.

Another question- was the oscillation something that occurred at or near clipping, or was it there even without a signal source?

Basically, the people who designed elderly amplifiers often didn't really understand what they were doing and the stabilization was an ad hoc process- although often carefully crafted.
To set the bias current for a class A/B amp the current is ramped up from a low point- hopefully the output devices are near class B to start with.
If so, was the oscillation present with the output stage near cut off? Often the chances of local oscillation increase as the devices are biased on.
The "base stoppers" act as loss elements for the LC network created by the inductive output impedance of the transistors and any load capacitance. For a higher Ft device the inductance is proportionally lower, which moves the resonance up in frequency and tends to increase the chances of oscillation.

 

[sregor]

When you say unstable, what do you mean? Was it oscillating- if so what was the frequency? Was it at the closed loop -3dB bandwidth of the amp or elsewhere?
Stability theory is pretty straightforward for most older bipolar amps- there's usually a single cap that sets the dominant pole, and the basic design looks like this
View attachment 1330239
CC is the dominant pole cap. D1,D2, Q11, Q12 comprise the unity gain (buffer) output stage.
The pole is set by the gm of the input stage and CC. The closed loop gain is set by the ratio of RF2 and RF1, feedback is taken from the output. Reducing the FT of the output devices can reduce the bandwidth of the gain stage (Q3) (something called pole splitting occurs in the gain stage and the extra cap load moves the higher frequency pole lower) and lower the phase margin of the amp- which could be problematic.
Increasing FT excessively could cause local instability in the output buffer due to interactions with various parasitics etc.
I confess to never have tried replacing output devices with lower FT types. I've replaced, in a prior life, devices with marginally higher FT devices and not experienced problems.

My take (I have enough back ground to be dangerous but not enough to be authoritative) - designs with high speed output use a higher frequency (smaller CC) to increase feedback in the high frequency (like 20khZ) range. if you put too slow output (and driver) devices, there is HF stability problems. It may not oscillate, but generate parasitics when driven hard. Also, you don't show it but a lot of designs use lead compensation in form of a cap across the feedback resistor, which also is optimized for output stage. We're not talking 2 or 3 times Ft, but 10 or more.... IMO a good design should push the limit, but have a some room for variation.

 

[wyn palmer]

My take (I have enough back ground to be dangerous but not enough to be authoritative) - designs with high speed output use a higher frequency (smaller CC) to increase feedback in the high frequency (like 20khZ) range. if you put too slow output (and driver) devices, there is HF stability problems. It may not oscillate, but generate parasitics when driven hard. Also, you don't show it but a lot of designs use lead compensation in form of a cap across the feedback resistor, which also is optimized for output stage. We're not talking 2 or 3 times Ft, but 10 or more.... IMO a good design should push the limit, but have a some room for variation.

The purpose was not to show all possible variants- so I just included a simple dominant pole system- no lead/lag or multi pole designs were mentioned as that would just distract from the main point.
Lead compensation has the disadvantage of pushing the closed loop gain down at higher frequencies, so it can actually induce oscillation, so often a lead-lag network is used. I guess it might be possible for a lead feedback network to push the effective closed loop bandwidth up enough so that a stable amp would become an unstable one with a higher Ft output device, but it's not obvious to me why that would be the case.
Generally, whatever the compensation, is hard to see why having a lower phase shift element in the forward signal path will not improve things unless some form of local oscillation occurs.
The intelligent thing would be to actually measure the open loop gain/phase characteristic of the amp and figure out what is going on.
http://www.angelfire.com/electronic/funwithtubes/Amp-Compensation.html
shows how.

 

[chazix]

Another question- was the oscillation something that occurred at or near clipping, or was it there even without a signal source?

The oscillation occurred with no input signal.

Poring over the Altec service manual, I think the outputs are pretty close to cut off at idle. The adjustment procedure is based on a total idle consumption of 75mA per channel. Even if all of that 75mA were flowing through the 4.7 ohm resistor that's in parallel with the outputs' BE junctions, it wouldn't be enough to appreciably turn on the outputs. (It's certainly conceivable that the adjustment was so far off in my particular case that the outputs were turned on, but I think it is very unlikely.)

 

[wyn palmer]

The oscillation occurred with no input signal.

Poring over the Altec service manual, I think the outputs are pretty close to cut off at idle. The adjustment procedure is based on a total idle consumption of 75mA per channel. Even if all of that 75mA were flowing through the 4.7 ohm resistor that's in parallel with the outputs' BE junctions, it wouldn't be enough to appreciably turn on the outputs. (It's certainly conceivable that the adjustment was so far off in my particular case that the outputs were turned on, but I think it is very unlikely.)

Actually, Ft is a function of collector current up to a peak value- it's sort-of proportional to the square root of the current. Is it conceivable that there was actually too little collector current for the replacement device?

 

[depaj]

Well this is more answer than I was hoping for. I'm somewhat comforted in the idea of going with those outputs but most of all, you showed me once more that I've got some serious studying to do !

 

[avionic]

It's hard to find any parts that slow these days, much less in TO-3 packages.

No its not. MJ21193G & 94G --- 5 mhz

 

[chazix]

Actually, Ft is a function of collector current up to a peak value- it's sort-of proportional to the square root of the current. Is it conceivable that there was actually too little collector current for the replacement device?

'Spose so. If that was the problem, I was in a bind anyway, since the circuit appears to have been designed to run almost zero current in the outputs at idle.

No its not. MJ21193G & 94G --- 5 mhz

Thanks, avionic, I stand corrected.

I'm remembering more about the Altec job now. The original outputs, though I'm pretty sure they were not marked thus, are said by more than one source to have been 2N6259s. And some sources say that 2N6259s have an Ft as low as 0.2MHz. That might be a bit harder to undercut!

In fact, I think the whole reason that I ever thought that original outputs were 4MHz parts is that NTE currently offers something that they call 2N6259, and they specify an Ft of 4MHz for whatever it is.

It's probably about time I apologized for derailing the thread, but it sure beats working.

 

[chazix]

Dang - can't seem to stop myself.

I just realized one aspect that might have been important in my experience with replacing output transistors in the Altec - it is a quasi-complementary amp. So, the "bottom" transistor is in common-emitter configuration, where among things I think its Cob will act as a Miller capacitor. I should be, but am not, bright enough to know if that changes the analysis.

 

[wyn palmer]

Dang - can't seem to stop myself.

I just realized one aspect that might have been important in my experience with replacing output transistors in the Altec - it is a quasi-complementary amp. So, the "bottom" transistor is in common-emitter configuration, where among things I think its Cob will act as a Miller capacitor. I should be, but am not, bright enough to know if that changes the analysis.

Yes, it does somewhat, but I can't state exactly what the effect is without looking at it in some detail- however based on a cursory glance I think that the bottom output transistor is operating with a voltage gain of about 5 so the Miller cap is about 6x that of the true follower.

 

[Osibisa]

Is there a modern replacement for a 2SC644(T)?

 

[slateef]

What would be a good replacement for Sanyo D313E?

 

[rcs16]

Go to Mouser and use the parameter search to match specs.
For 2sc644,use ksc1845f (ECB pinout ), verify pin-outs against the original, I do not have a proper spec for that part.

For 2SD313
You have Vce max at 60Vmin, Ic max 3A min ,30W min, ft 8MHz min, TO-220 packages
I think your closest match is 2N6487G