Well, I'll disagree a little. Here in North America, the best era extended from the 1950's until later 1980's. 1980 saw some of the best (on average) equipment and people were excited about music. That did die out and certainly in the 2000's there wasn't much excitement for music. You are correct, these days people hardly know what music is, and I doubt the "feel it" like we did as kids.
Music may be making a comeback as the "kids" discover how good old equipment was and how junky new popular stuff is.
Music may be making a comeback as the "kids" discover how good old equipment was and how junky new popular stuff is.
That s a really nice design, they enclosed the output capacitor in the feedback loop and there s a triple for the OS, i ve seen
integrated amplifiers from the era that were much less well thought, and the rest is no slouch, guess that it sounded very well.
Last edited:
Hi Anatech, by triple i m talking of the OS, in the positive branch that s two serialised compound while in the negative branch that s a compound with the driver and power transistor connected as dual EF, beta wise that s three transistors in serial for each branch of the OS, generaly this type of design used only a driver for each 2N3055, so that s quite remarkable given that it s not a very high power system, one of the rare case where there was a triple arrangement at the time was the Quad 405 IIRC.
I think they were forced to do that given the low beta of the 2N3055, especially at higher currents and maybe frequencies. The voltage amp stage doesn't have a lot of drive or total transconductance. Given it's a singleton and not a diff pair, it is behind another 8 ball.
Like I said, not the worse. That doesn't make it very good compared to some other designs of the same period. I think the swing might be 15 ~ 18 volts peak with a 40 VDC supply. So it can be asked to supply some current into 4R loads, although the average European use wouldn't be high power.
Anyway, any design using the 2N3055 is a very cost conscious type system. Quality was not the design objective, clearly. Even the bias current control is "not great" ... okay, terrible. Cheap. While I'm sure they work and make okay sound, it does not compare to decent amplifiers of the time. The failure rate would probably be high in North America. Once the bias pot gets dirty - kaboom.
Like I said, not the worse. That doesn't make it very good compared to some other designs of the same period. I think the swing might be 15 ~ 18 volts peak with a 40 VDC supply. So it can be asked to supply some current into 4R loads, although the average European use wouldn't be high power.
Anyway, any design using the 2N3055 is a very cost conscious type system. Quality was not the design objective, clearly. Even the bias current control is "not great" ... okay, terrible. Cheap. While I'm sure they work and make okay sound, it does not compare to decent amplifiers of the time. The failure rate would probably be high in North America. Once the bias pot gets dirty - kaboom.
Agree that the bias circuitry is quite random and that s quite unhappy because they could had made something really great with
just a few more cheap components.
The only big mistake other than the erroneous biasing is the VAS huge degeneration that increase distorsion by almost an order a magnitude, if they shunted the degeneration resistance with a 100uF cap then the amp would had no more than 0.01% distorsion at 1KHz for 15V peak output even in 4R, i made a sim to check the numbers, really very few was needed to yield something
exceptional price/perf wise for the time.
That being said i ll add that a differential at the input wouldnt reduce distorsion, that s a common urban legend as
actually a singleton provide as good linearity, it s just that it s no more relevant once one want a DC output coupling.
just a few more cheap components.
The only big mistake other than the erroneous biasing is the VAS huge degeneration that increase distorsion by almost an order a magnitude, if they shunted the degeneration resistance with a 100uF cap then the amp would had no more than 0.01% distorsion at 1KHz for 15V peak output even in 4R, i made a sim to check the numbers, really very few was needed to yield something
exceptional price/perf wise for the time.
That being said i ll add that a differential at the input wouldnt reduce distorsion, that s a common urban legend as
actually a singleton provide as good linearity, it s just that it s no more relevant once one want a DC output coupling.
Last edited:
For a 40 volt single power supply, two pair of TIP41/2 in parallel will probably outperform a pair of 2N3055’s.
Unless they were the TI versions. Anybody else’s, they’d be ok.
IME, a singleton input stage, properly done, beats a badly mismatched diff pair. I’ve seen some bad ones, with 2:1 difference in current. Having it reasonably balanced at quiescent but going out of whack when the supply drops under load often happens too. When that happens you’re better off with a singleton.
Unless they were the TI versions. Anybody else’s, they’d be ok.
IME, a singleton input stage, properly done, beats a badly mismatched diff pair. I’ve seen some bad ones, with 2:1 difference in current. Having it reasonably balanced at quiescent but going out of whack when the supply drops under load often happens too. When that happens you’re better off with a singleton.
Agreed on the balanced diff pair, I do run into unbalanced pairs, but badly matched ones are found on amps that had gone DC most times. Guess why?
Absolutely, a pair of TO-220 outputs would perform better, but can be blown. How many amplifiers have we seen using TO-220 outputs in this power class? Tons!
Hi wahab,
It isn't the singleton on it's own that performs badly, but it doesn't usually come with good design in the rest of the circuit (witness early NAD, and it is direct coupled). You can have higher transconductance with a diff pair circuit, and I rarely see high transconductance in single type circuits, in fact they are normally used for "low feedback" designs. English designers have clung to singleton inputs much longer than anyone else.
As for a sim of the circuit, doubtful in real life. Depends on the linearity of the models used, and also introducing the wide transistor variations you encounter in real life. The 2N3055 isn't a poster child for flat gain curves and rises and drops a lot over the current range. If the sim doesn't take at least that into account, the results will be way off. Model carbon composition resistors while you are at it. Absolutely, degenerating the AC gain in the VAS drops transconductance.
A few extra cheap parts ... these designs typically avoid using parts as much as possible. That's why this looks like it does and uses the parts it does. It was very clearly designed for cost. So there are worse, yes. But this isn't really very good either. I guess you could say, Performance is very good for the money. Wasn't that the tag for NAD products?
Absolutely, a pair of TO-220 outputs would perform better, but can be blown. How many amplifiers have we seen using TO-220 outputs in this power class? Tons!
Hi wahab,
It isn't the singleton on it's own that performs badly, but it doesn't usually come with good design in the rest of the circuit (witness early NAD, and it is direct coupled). You can have higher transconductance with a diff pair circuit, and I rarely see high transconductance in single type circuits, in fact they are normally used for "low feedback" designs. English designers have clung to singleton inputs much longer than anyone else.
As for a sim of the circuit, doubtful in real life. Depends on the linearity of the models used, and also introducing the wide transistor variations you encounter in real life. The 2N3055 isn't a poster child for flat gain curves and rises and drops a lot over the current range. If the sim doesn't take at least that into account, the results will be way off. Model carbon composition resistors while you are at it. Absolutely, degenerating the AC gain in the VAS drops transconductance.
A few extra cheap parts ... these designs typically avoid using parts as much as possible. That's why this looks like it does and uses the parts it does. It was very clearly designed for cost. So there are worse, yes. But this isn't really very good either. I guess you could say, Performance is very good for the money. Wasn't that the tag for NAD products?
True. Many small Japanese amps died on low impedance loads for sure. But then we'd go to a TO-218 part instead of two pair of TO-220.
Car amps used a 7 ampere TO-220, then later we got 10 ampere TO-220 transistors. Car amplifiers often ran +/-20 ~ 27 VDC rails. 4 ohm loads were the standard and heat sinks ran hot. Those seemed to survive okay. Everything had at least over current protection, SOA protection was generally reserved for the bigger and better home power amps.
Car amps used a 7 ampere TO-220, then later we got 10 ampere TO-220 transistors. Car amplifiers often ran +/-20 ~ 27 VDC rails. 4 ohm loads were the standard and heat sinks ran hot. Those seemed to survive okay. Everything had at least over current protection, SOA protection was generally reserved for the bigger and better home power amps.
Hi Anatech, the model i have for the 3055 exhibit a gain of 41 at 3.6A, that s in line in respect of the datasheets,
sims can be accurate providing one use proved models of ubiquitous transistors that are much used in industrial
applications.
When it comes to topologies it s quite surprising how well a singleton based amplifier can behave, with of course the usual shortcomings that are the lower PSRR and capacitive coupling but this can be remedied, anyway when it comes to linearity
an input differential will be no better unless it use a current mirror, indeed if there s not such a circuit it s useless
to chase this kind of design because at low level linearity without CM is not that good due to second order effects.
As for transistors i m vaccinated against TO220 when it comes to OSs unless we re talking something like 5-10W RMS at the
very most, those casings just do not cut the mustard and better to look at some TOP3 even if they are apparently overkill
for say 15-20W, one never know what kind of load they may encounter in their destiny.
sims can be accurate providing one use proved models of ubiquitous transistors that are much used in industrial
applications.
When it comes to topologies it s quite surprising how well a singleton based amplifier can behave, with of course the usual shortcomings that are the lower PSRR and capacitive coupling but this can be remedied, anyway when it comes to linearity
an input differential will be no better unless it use a current mirror, indeed if there s not such a circuit it s useless
to chase this kind of design because at low level linearity without CM is not that good due to second order effects.
As for transistors i m vaccinated against TO220 when it comes to OSs unless we re talking something like 5-10W RMS at the
very most, those casings just do not cut the mustard and better to look at some TOP3 even if they are apparently overkill
for say 15-20W, one never know what kind of load they may encounter in their destiny.
Hi wahab,
Have you modeled the extreme beta vs Ic curve?
I find sim models do not throw variability into the mix, not the ones the average person uses. So they end up with all transistors behaving the exact same if they are the same number. Then variations in characteristics vs current are not modeled well, including re and capacitance. So while they can give you a very rough dynamic picture of a circuit, real results often differ quite a bit.
Singletons have enough problems that "can be overcome", why not sidestep the lot? CCS tail currents are an easy fix, cheap too. Cascoding the drains / collectors is also easy and then you start doing tricks to enhance things. I don't know, a diff pair is superior in many ways and if you want to linearize it more, it responds amazingly well. That's why op amps normally use diff pairs.
I'll agree TO-220 case styles are best for low power. However I have seen enough very reliable TO-220 designs of this power class through history, and we have much better / tougher devices today. Sure, TO-3P (TO-218 is a type) are much better due to thermal resistance, but the TO-220 can do many jobs very reliably. I am not advocating we use them in applications that push their limits, but they aren't that bad. One issue I have seen is that over tightening the mounting screw may well crack the die. This is well known, and it comes down to the assembler being ignorant. That is the reason the plastic TO-3 packages these days are full mold, no longer is there a metal tab. motorola has an application note on mounting power devices. It's a good read and should be required for anyone who works with semiconductors.
As for protection against overload, easily done. You should do that for any output type anyway. Maybe designers don't like to think, but it is easily done. Want "bulletproof"? Use TO-204 devices, the larger packages tend to get discontinued so stay with something that will be around long time.
Have you modeled the extreme beta vs Ic curve?
I find sim models do not throw variability into the mix, not the ones the average person uses. So they end up with all transistors behaving the exact same if they are the same number. Then variations in characteristics vs current are not modeled well, including re and capacitance. So while they can give you a very rough dynamic picture of a circuit, real results often differ quite a bit.
Singletons have enough problems that "can be overcome", why not sidestep the lot? CCS tail currents are an easy fix, cheap too. Cascoding the drains / collectors is also easy and then you start doing tricks to enhance things. I don't know, a diff pair is superior in many ways and if you want to linearize it more, it responds amazingly well. That's why op amps normally use diff pairs.
I'll agree TO-220 case styles are best for low power. However I have seen enough very reliable TO-220 designs of this power class through history, and we have much better / tougher devices today. Sure, TO-3P (TO-218 is a type) are much better due to thermal resistance, but the TO-220 can do many jobs very reliably. I am not advocating we use them in applications that push their limits, but they aren't that bad. One issue I have seen is that over tightening the mounting screw may well crack the die. This is well known, and it comes down to the assembler being ignorant. That is the reason the plastic TO-3 packages these days are full mold, no longer is there a metal tab. motorola has an application note on mounting power devices. It's a good read and should be required for anyone who works with semiconductors.
As for protection against overload, easily done. You should do that for any output type anyway. Maybe designers don't like to think, but it is easily done. Want "bulletproof"? Use TO-204 devices, the larger packages tend to get discontinued so stay with something that will be around long time.
The TO-218’s got expensive. TIP41/2 were cheap. And more importantly, full power handling to 40 volts VCE - something the TO-218’s can’t claim. One pair doesnt cut it average power wise but two will. I guess that extra 4 emitter resistors, mounting screws, and mica washers were going to break the bank, eh?
Back in those days I guess they just wouldn’t use parallel pairs unless absolutely necessary. But it helps with that thermal resistance, and the problem of beta droop - something the 2N3055 suffered from in spades.
Back in those days I guess they just wouldn’t use parallel pairs unless absolutely necessary. But it helps with that thermal resistance, and the problem of beta droop - something the 2N3055 suffered from in spades.
Hi,
Yes, and I agree 1000% with you. If you used a non-TI TIP series, you would be fine in most cases. I can't remember, but there were some nice 2N numbers and MJE of course. I think the Japanese parts went up to 7 amperes maximum. 2SA771 and the compliment comes to mind. 2SD613/2SB633 was 6 amperes.
Beta droop with early power transistors was a real issue. 2N3055 normally saw power supply duty without a great deal of collector current change, so they weren't improved and stayed on as their terrible selves for a long time. Only when the foundries discontinued that process did they improve. There was only one reason a 2N3055 ever saw the inside of an audio component. They were the least expensive power device you could buy (for good reason). This explains the designs using them.
Yes, and I agree 1000% with you. If you used a non-TI TIP series, you would be fine in most cases. I can't remember, but there were some nice 2N numbers and MJE of course. I think the Japanese parts went up to 7 amperes maximum. 2SA771 and the compliment comes to mind. 2SD613/2SB633 was 6 amperes.
Beta droop with early power transistors was a real issue. 2N3055 normally saw power supply duty without a great deal of collector current change, so they weren't improved and stayed on as their terrible selves for a long time. Only when the foundries discontinued that process did they improve. There was only one reason a 2N3055 ever saw the inside of an audio component. They were the least expensive power device you could buy (for good reason). This explains the designs using them.
Hi Anatech, the gain collapse gradually from about 110/120mA, 93/840mA, 75/1.5A, 41/3.6A, and so on to 31/5.2A,
the dive at high current is well respected, max current should be 4A, and the same with all 2N3055/MJ2955 in disguise such as the TIP33/34 and proelectron equivalent BD249/250, as well as derived darlingtons such as MJ2501/3001 and their european siblings BDX66/67 and BDV64/65 FI, unless of course that pre drivers (and drivers for darlingtons) are used.
In the early 80s one could use cheap darlingtons like the 100V BDW93C/94C to drive such lazy transistors as the 2N3055/MJ2955
or MJ15003/15004 up to +-45V voltages, that was a convenient PA solution that i used circa 1987/88 with my musical band,
heck, we even bridged as stage monitor for our singer a Citation 12 from Harman Kardon that had 2N3055s, it survived
the heavy duty courageously for years !
the dive at high current is well respected, max current should be 4A, and the same with all 2N3055/MJ2955 in disguise such as the TIP33/34 and proelectron equivalent BD249/250, as well as derived darlingtons such as MJ2501/3001 and their european siblings BDX66/67 and BDV64/65 FI, unless of course that pre drivers (and drivers for darlingtons) are used.
In the early 80s one could use cheap darlingtons like the 100V BDW93C/94C to drive such lazy transistors as the 2N3055/MJ2955
or MJ15003/15004 up to +-45V voltages, that was a convenient PA solution that i used circa 1987/88 with my musical band,
heck, we even bridged as stage monitor for our singer a Citation 12 from Harman Kardon that had 2N3055s, it survived
the heavy duty courageously for years !
In 1966 USA RCA 2n3055 were the only silicon device available for 2 amps up, 35 v up. No electronic supply available to me (sterling, electrotex, allied radio) carried BD anything. Newark would not sell to individuals without a tax number. Digikey was a surplus house with a mail order flyer. Previous TO3 transistors popular in organ amps etc were PNP, used at 30 v or less, 1/2 amp Ic or less. See Wurlitzer 4500 organ of 1966. Hammond was using 7189 tubes in their 60 watt organ amps through 1972. There still were a lot of AD161 and Delco flying saucer germanium transistors sold in 1966. Germanium were reliability disasters. My employer in 1975 hated germanium, preferred vacuum tubes for Vickers solenoid drive. I remember allied radio catalog price of 2n3055 was over $10 or 6 hours pay. TO5 transistors like 40409/10 were over $1 or 40 minutes pay, which is why 6 transistor designs like Leak 30, armstong 621 and dynakit ST120 were sales leaders. No source of leak & armstrong in the USA, at least I never saw them in Houston. I bought a used vacuum tube ST70 in 1970; could not afford ST120 until blown up ones showed up in giveaway newspapers about 1985. Really good McIntosh amps I drooled at in 1972 cost more than a years college tuition. I coveted SWTC tigersaurus about 1975, was saved from that disaster by necessity to buy radiator, starter, tires for my car.
Last edited:
Indeed and always liked the triple for Quasi but was little hard to tame for stability.
Without the differential input, could be much easier to make a stable amplifier.
Feedback loop after output cap should improve bass , be able to use shown smaller cap.
I could see potential to get to much lower distortion with fine tune.
And just keep it single supply for fun
Just make that upper bank a EF3. Stability problem gone. That stupid CFP of a CFP is just a bad idea, even though it saves about a volt and a half of signal swing. The QC triple bottom half is stable if you follow a few simple precautions, which are easier to do on a single supply amp than they are on dual rails. Proper biasing and usual loop gain improvements in the front end always apply.
This is just a quick shout-out to thank all those participating in this thread and to say how much I have appreciated the sheer amount of expertise and been-there-done-that knowledge that has been shared. The types of applications, considerations and effects being discussed are not those that have applied to my work in the past and I have very much enjoyed learning from the posts here - posts which, for the most part, are well grounded in experience and reality. In Metrology I have always enjoyed working my way through the ways that seemingly insignificant properties or effects can greatly change the outcome of a measurement and to see the exact same principals have the same magnitude of effects on what, for me, are completely new (in both type and magnitude) applications is very educational and enjoyable. So, Thanks all!
Hal
Hal
I preferred the Motorola version of 2n3055 along with their MJ2955 used in amps with 70V rails. 3 pairs of them would do 200W rms into 4 ohms all day with the right heat sink, but there was a little sag built into the design on purpose along with big CRC filtering and .33R emitter resistors. I cut the top off of a Motorola 3055 to look at the dye. It was pretty big by today's standards.