A question from a friend:
For repair of vintage amplifiers we need TO-3 replacement transistors.
Is there anyone who has experience with the influence on sound if one looks at the transistor data "Ft" in mHz ?
Without any oscillation and problems mentioned in another thread post 14# : transistor "Ft" what about it?
5 -6 amplifiers NAD3080, Luxman B12 & M12 and Sony ES has been repaired with new output transistors...all changed transistors with lower "Ft" specs than the original.
The amplifiers did all reach factory DC-specs , but the general opinion was that all the repaired amplifiers lost some quality in soundstage and perspective??
Several of the old Toshiba, NEC and Sanken TO-3 transistors were fast transistors with "Ft" data from 6 mHz up to and above 15 mHz.
We used MJ15003/4 2 mHz and MJ15022/23 approx. 3 mHz for repair.
The question is whether a low "Ft" transistor does have major impact on sound quality in a circuit designed for faster highspeed "Ft" transistors?
The available MJ21193G / MJ21194G from Onsemi does take quite high voltage and delivers sufficiant current, but they are not so fast as e.g. 2SD551 NPN 2SB681 PNP 150V / 12A / 15 mHz.
Any information out there?
__________________________________________________ ________
My friend: Google Sites: Sign-in
For repair of vintage amplifiers we need TO-3 replacement transistors.
Is there anyone who has experience with the influence on sound if one looks at the transistor data "Ft" in mHz ?
Without any oscillation and problems mentioned in another thread post 14# : transistor "Ft" what about it?
5 -6 amplifiers NAD3080, Luxman B12 & M12 and Sony ES has been repaired with new output transistors...all changed transistors with lower "Ft" specs than the original.
The amplifiers did all reach factory DC-specs , but the general opinion was that all the repaired amplifiers lost some quality in soundstage and perspective??
Several of the old Toshiba, NEC and Sanken TO-3 transistors were fast transistors with "Ft" data from 6 mHz up to and above 15 mHz.
We used MJ15003/4 2 mHz and MJ15022/23 approx. 3 mHz for repair.
The question is whether a low "Ft" transistor does have major impact on sound quality in a circuit designed for faster highspeed "Ft" transistors?
The available MJ21193G / MJ21194G from Onsemi does take quite high voltage and delivers sufficiant current, but they are not so fast as e.g. 2SD551 NPN 2SB681 PNP 150V / 12A / 15 mHz.
Any information out there?
__________________________________________________ ________
My friend: Google Sites: Sign-in
The amplifier can be designed around the output transistor's specifications. If you change it to a different type, those design decisions will no longer make sense, and you lose the intended specifications of the amplifier. This could be linearity, level of various distortion types, and so on.
This doesn't need to happen. If the amplifier is not designed directly around the output transistor used, you can put almost anything in there with no change in specifications.
However, it is not possible to repair an amplifier and hear any change in its "character", since the only way to do this would be to compare with another, working stock amplifier side by side.
This doesn't need to happen. If the amplifier is not designed directly around the output transistor used, you can put almost anything in there with no change in specifications.
However, it is not possible to repair an amplifier and hear any change in its "character", since the only way to do this would be to compare with another, working stock amplifier side by side.
Last edited:
I have a pair of Luxman B12s and on the schematic this is what it says about transistor replacement: "Transistors, ICs, and diodes may be replaced with any types having comparable ratings". Seems Luxman was fairly loose about replacements.
Craig
Craig
In general, substituting a faster OP transistor (within reasonable limits) doesn't pose too many problems: since it is mainly these transistors that govern the frequency behavior of the global loop, opting for faster types tends to improve the phase margin, which is good.
It can however lead to local instabilities, because the wiring or the driver are not compatible with such speeds, which is one of the reasons to remain reasonable.
If you use slower devices, you will degrade the phase margin, and this could lead to oscillations. That said, most amplifiers are prudently overcompensated, and they may not break into obvious instabilities, but they will work on the edge, and characteristics like slew-rate will probably be degraded too.
To summarize, if you cannot find an exact replacement, opt for the slightly faster near-equivalent.
I consider low Ft transistors to be 200 khz, like the original 2n3055. these were definitely audibly slow. 3 mhz is 150 x faster than the highest frequencies heard by undamaged teen ears. 3 mhz Ft should allow for a reasonable approximation of the proper sound.
Thank you for the inputs.
The problem is that specifications of the TO-3 transistors produced today does not equal the japanees types from the eighties.
...... the intention was to keep the vintage amplifiers as close to the original layout as possible, and use the TO-3's available.
Calculation from a friend in Uk showed that the 21193G (Ft=4mHz) has a Hfe max. DC current gain of 75, while min. 25 DC current gain at Ic= 8 A and Vce = 5Vdc ...which gives approx 50khz roll-off...
The 2SB681 (Ft=13mHz) has a Hfe max. DC current of 140 and min. 20 DC current gain at Ic= 5A and Vce=5Vdc
...which according to the above calculation should give 13mHz/140= by approx 93kHz the current gain should start to roll-off.
We took the chance tonight and replaced the old 2SD/2SB with the Onsemi MJ TO-3's , and there are no sign of oscillation on the scope so far.........listen test tomorrow between the two 3080's.....and welcome has a point : we have a second 3080 available with the original opt's working.....
It looks like Craig was right...there are some margin with replacement transistors after all........
Gudmund
_________________________________________________________________________
Last edited:
I do agree with you on the point that the highest frequencies can be detected by young undamaged ears. I will assume that my upper hearing threshold is located around 10kHz.
The point is whether the "low" Ft transistors will effect the original factory made sound quality of the amplifier?
I can repeat as follows:
As suggested it has been tried to test with the RC dummy-load at 40Hz and there were no sign of instability ....after first test the amplifier was connected to AR3a with a 40Hz signal.
Above the resonance frequency AR3a's impedance curve gets close to 2 Ohm...... By 2/3 of power the amplifier got very hot but kept the 40Hz frequence signal steady for 15 minutes without any audible change from the output terminals .
So far the amplifier seems to be stabil.
It is possible that distortion is higher by the "low" Ft transistors? ....we have two identical amplifiers...except for the opt's.....we could not detect any difference in sound between them.
Conclusion must be: If the circuit has been designed with a margin ,it is possible to switch to the "low ft" transistors without affecting the soundstage dramatically.
Gudmund
__________________________________________________________________________________________________________
Link to NAD3080: https://sites.google.com/site/httpstubeamp/home/nad-3080
Last edited:
Most standard circuit topologies (ie, EF2, EF3, etc, with a Lin-ish front end) are perfectly stable with outputs as 'slow' as 2 MHz. Above that, fT isn't nearly as important to sound quality as how well the beta holds up at high current. The MJ2119x, being ring emitter transistors, actually do quite well in that regard. There are moden ones that do "better", but not back in the day of the D551 there weren't. If there are any changes in the distortion 'character' it will be in the third decimal place - assuming that the bias has been optimized for the replacement outputs. Optimum AB bias may be different, but probably not by more than a factor of two.
Thanks. The last part follows the thoughts we had about the good standard from japanese transistors back in the eighties.
About the MJ21193G/MJ21194G from Onsemi being RET transistor types I am not quite sure whether this is right?
If I remember correct Fujitsu made the first old RET transistors around 1978/1979:
http://www.classiccmp.org/rtellason/transdata/2sa1075.pdf
Some of the typical specs from RET transistors are very high Ft , which for 2SA1075 is 60 mHz , high collector to base voltage and extended frequence linearity.....just like LAPT transistors from Sanken: 2SA1216/2SC2922 180V 17A 200W hfe: 30min 40MHz MT200
As I understand the description of the MJ21193G/MJ21194G the design is a perforated emitter which should perform as a "rugged" high voltage version of the older MJ15022/23 with slightly higher Ft and good linearity. However, the mentioned transistors can't reach specifications like RET transistors.
I might be wrong?
The text below taken from the MJ21193 (PNP) and MJ21194 (NPN) datasheet:
Publication Order Number:
MJ21193 / D
MJ21193? -? PNP
MJ21194? -? NPN
Silicon Power Transistors
The MJ21193 (PNP) and MJ21194 (NPN) utilize Perforated Emitter
technology and are Specifically designed for high power audio output,
disk head positioners and linear applications.
Gudmund
_________________________________________________________________________
Fujitsu RET transistor specs: https://sites.google.com/site/httpstubeamp/ ...........and search for Luxman L11 at Picasaweb sites.
Last edited:
people should bear in mind that ft is the frequency at which hfe falls to 1
for an EF stage it is the frequency when the driver is pushing as much current into the load through the output Q BE diode junction as the output Q is adding with its collector current
the pole, rolloff, effective max usable frequency is much lower
for an EF stage it is the frequency when the driver is pushing as much current into the load through the output Q BE diode junction as the output Q is adding with its collector current
the pole, rolloff, effective max usable frequency is much lower
If you operate your audio amplifier at 1 MHz, you deserve to blow your driver transistors. If you put out full power at 50 KHz you deserve to blow your tweeters. Hfe falling by half at say 50 KHz just isn't a real problem. A much more real problem is Hfe falling to 15 at 20A, while playing bass. That's how most driver transistors get taken out.
Aka holes in the emitter to raise the ratio between circumference/area of the emitter. (dates back as far as the '60s)
A lot of things were invented back in the 60's that weren't practical till much later. The computer on board the missions to the moon wouldn't be able to load this web page.
...as I understood the explanation.
The transistor Ft is the frequency at which the common emitter current gain (hfe) has fallen to 1, normally the gain rolls off at 6dB/octave... the roll-off starts at a frequency of Ft/hfe.
..the phase margin or free space-area is not as wide as it looks like in the first place, and the three poles of the transistors might come too close so the phase margin becomes inadequate....and we run into oscillation.
Gudmund
In a Miller compensated amplifier, the non-dominant poles do not move 1 for 1 with the output transistor fT. And it's dominated more by the driver transistors' characteristics. Put in a 3 MHz driver transistor and you're in trouble. Good thing 100 MHz 20 watt TO-220's are a dime a dozen. The go obsolete and get replaced more often than cell phones these days, but there is always SOMETHING comparable available.
I will add just a test done some years ago that i actually haven't verified with instruments to find out why ....
Most P3A i make is made with 1302-3281 this happens because i have them available and original ...
Though i have made one with no precision resistors with no matched transistors and for outputs i used Tip 2955-and 3055 ( in a bit lower voltage ) it sounded horrible next to the 1302-3281 1% resistors and all transistors matched
Yet again in a board with all the above goodies i installed 1943-5200 ( all tests at the exact same bias ) and the sound was worst than 1302-3281
I have to measure why one day to see what the spectrum will say about that ....
Most P3A i make is made with 1302-3281 this happens because i have them available and original ...
Though i have made one with no precision resistors with no matched transistors and for outputs i used Tip 2955-and 3055 ( in a bit lower voltage ) it sounded horrible next to the 1302-3281 1% resistors and all transistors matched
Yet again in a board with all the above goodies i installed 1943-5200 ( all tests at the exact same bias ) and the sound was worst than 1302-3281
I have to measure why one day to see what the spectrum will say about that ....
Does anybody still make TO-3 (TO-204-2) transistors with Ft's on the order of 5 to 20 Mhz these days? Like the ones the Japanese use to make that would be suitable for high power amplifiers around 200 watts per channel+ into 8 ohms both channels driven?
An example of a high Ft transistor could be the Sanken 2SA1116/2SC2607's (at 20Mhz) that Sansui had custom made for them into the NMA/NMC series output transistors.
When someone replaces output transistors in an amplifier that had high slew rate, low rise time figures with a MJ series (i.e. MJ15024, MJ21194, MJ21196 with its PNP complement) do you subsequently measure the amplifiers rise time and slew rate? I will bet most people do not and still say it meets factory specs when they return it to the customer.
An example of a high Ft transistor could be the Sanken 2SA1116/2SC2607's (at 20Mhz) that Sansui had custom made for them into the NMA/NMC series output transistors.
When someone replaces output transistors in an amplifier that had high slew rate, low rise time figures with a MJ series (i.e. MJ15024, MJ21194, MJ21196 with its PNP complement) do you subsequently measure the amplifiers rise time and slew rate? I will bet most people do not and still say it meets factory specs when they return it to the customer.
Don´t overthink it.
I bet a good design is somewhat flexible on (good) parts used.
Doubt people can detect rise time and slew rate differences by ear anyway.
I bet a good design is somewhat flexible on (good) parts used.
Doubt people can detect rise time and slew rate differences by ear anyway.
The slew rate is not directly affected by the output transistors. It is usually set by the speed of the VAS stage, which is intentionally slowed down for stability. When the outputs affect speed is when you have to recompensate the amp to make it stable. With a good design you don’t have to.
The issue of cross conduction does come up. It is more of an issue with “slower” outputs, but due to their ruggedness they can usually take some. Maybe not a sustained full power 300kHz square wave, but that’s not real world usage for an audio amp. That will take out the zobel and your tweeters long before it fries a 21193/4 pair anyway.
The issue of cross conduction does come up. It is more of an issue with “slower” outputs, but due to their ruggedness they can usually take some. Maybe not a sustained full power 300kHz square wave, but that’s not real world usage for an audio amp. That will take out the zobel and your tweeters long before it fries a 21193/4 pair anyway.