@Pingrs - JLH was not the only person to define a quasi output stage with base resistors and diode balanced. Mullard also published a circuit with the same concept (except for the capacitor across the diode which JLH added) in their second ed. transistor circuits book. Also in 1972.
Of course, JLH claims no originality for the diode/resistor combination. Did Baxandall have some input into Mullard's scheme?
Probably but that is what i wanted. As simpler as possible while not compromising too much. Again i want to thank you, John & others for that 🙂
Although i yet to build the Elektor amp but many who made that have nothing but only praise! No report of instability/ oscillations.
A complimentary version of this Elektor amp could be very interesting 😉
Regards
An old thread on quasi Elektor amp..
https://www.diyaudio.com/community/threads/amplifier-based-on-2n3055.10337/page-4
https://www.diyaudio.com/community/threads/amplifier-based-on-2n3055.10337/page-4
See "Symmetry in a Class B", P.J. Baxandall, Letters to editor, Wireless World, September 1969 (pdf available).
Btw i have couple of question, where should i put bias transistor for best thermal tracking? Is it necessary to use a capacitor across baxandall diode just like what JLH did? What's the approximate value? Or it is better to use a small ceramic(47-100p) between C & B of lower driver transistor T6 to stop possible oscillation?
T5 and T6 I used 243 and 253 which are fast enough to not have stability issues.
Capacitor Baxandall uses I have only had to use if the drivers are slow as well.
Like a quasi or even fully CFP that capacitor has been used to keep it from ringing
at high frequency. On the negative rail, typical behavior for many amps.
Mentioned earlier all the high frequency issues have been solved more effectively
with degen in the differential. R24 , R23 and using 243 / 253 higher Ft
likely with incredible slow drivers like 139/140 or countless vintage transistors
the cap might be necessary. but again even with simple amps, designers dont
use or not aware of how to do Degen in the differential to solve the problem.
Instead as usual just toss capacitors and base resistors everywhere to address
stability. When as mentioned instead of slewing or slowing things in usual riduculous
manners. Just use degen, and in this design very minimal degen 100 ohms made 22 Khz
squarewave absolutely ring free.
likewise Cdom C3 is low only 47p most amps would use up to 68 to 100p
Again the major clue how effective degen is . So effective the amp is stable and doesnt require
excessive compensation as usual. If I removed degen, I would have to increase C3 way up to 82p
to keep Tip 35 transistor amplifier from ringing. Or likely slew the on time of T6...hence the Baxandall
cap. All that slewing and pooing goes away with better design, and the majority of off
slewing which actually causes most the ringing. Is as mentioned solved with a quasi or CFP if the driver
transistor speed is fast as possible. likewise if available current is high. the drivers current is kept high hence
47 ohm r12 , r14 , r15 for around 8 to 9 ma. to were older amps with slow drivers and low current.
around 4 to 6mA it is a not huge mystery why things ring at high frequency
Capacitor Baxandall uses I have only had to use if the drivers are slow as well.
Like a quasi or even fully CFP that capacitor has been used to keep it from ringing
at high frequency. On the negative rail, typical behavior for many amps.
Mentioned earlier all the high frequency issues have been solved more effectively
with degen in the differential. R24 , R23 and using 243 / 253 higher Ft
likely with incredible slow drivers like 139/140 or countless vintage transistors
the cap might be necessary. but again even with simple amps, designers dont
use or not aware of how to do Degen in the differential to solve the problem.
Instead as usual just toss capacitors and base resistors everywhere to address
stability. When as mentioned instead of slewing or slowing things in usual riduculous
manners. Just use degen, and in this design very minimal degen 100 ohms made 22 Khz
squarewave absolutely ring free.
likewise Cdom C3 is low only 47p most amps would use up to 68 to 100p
Again the major clue how effective degen is . So effective the amp is stable and doesnt require
excessive compensation as usual. If I removed degen, I would have to increase C3 way up to 82p
to keep Tip 35 transistor amplifier from ringing. Or likely slew the on time of T6...hence the Baxandall
cap. All that slewing and pooing goes away with better design, and the majority of off
slewing which actually causes most the ringing. Is as mentioned solved with a quasi or CFP if the driver
transistor speed is fast as possible. likewise if available current is high. the drivers current is kept high hence
47 ohm r12 , r14 , r15 for around 8 to 9 ma. to were older amps with slow drivers and low current.
around 4 to 6mA it is a not huge mystery why things ring at high frequency
Last edited:
Mullard has the diode on different position.
The book is dated august 1969, one month before the Baxandall WW article.
see
https://frank.pocnet.net/other/sos/Mullard_TransistorRadioAudioCircuits_1969.pdf
I think D1 is the most important one, the one that reduces the asymmetry between the upper and the lower part. The other two diodes with their resistors turn the output stages into a kind of current mirrors.
Some points from the previous posts -
BD139/BD140 are not slow, BD139 typ. 250MHz, BD140 typ. 75MHz.
You may be thinking of the similar sounding BD237/BD238 which are epi base and slow. I've never used these for that reason, but have used many BD139/BD140's with no problems. Including measuring ST, Motorola and Philips devices at 1MHz to check the ft. There may be some fake devices out there which are slow, of course.
In the circuit White Dragon has posted (#70) R7 and R17 provide some high frequency oscillation stopping capability. Simulations suggest that 47 ohms in this design (along with the input degen as WD mentioned) are adequate. I see no need for a capacitor across the Baxandall diode. The 1N4002 device type is probably a reasonable frequency match to an epibase transistor like TIP35 (or the recent 2N3055 for that matter). Though I might see if simulations show anything up - but that rather depends on how accurate the SPICE models of the devices would be.
Regarding the circuit in post 71, that is not the circuit I was referring to. I suspect the diodes in series with the local 100 ohm resistors will effectively disconnect the 100 ohm resistor when the relevant BDY20 is turning off, leaving only a 1k resistor to shunt any base current. BDY20's (and 2N3055's for that matter (back then, when they were 1MHz ft)) needed all the reverse base current they can get to keep them from "lingering" on at high frequencies. (And that meant 20kHz back then!) I was specifically referring to the second edition of the book where a 50W amplifier has exactly Baxandall's configuration as has been discussed.
Frequently reported ringing in the PNP/NPN stage depends on adequate supporession. I have not seen any sign in simulation with the circuit WD posted. I have built several circuits which have needed sprog stopping capacitors, though, but in every case I can think of, that also showed up in sim.
Though if anything does show up in sims I'll mention it. As I posted, it seems to be well behaved.
Edit - at least with TIP35/2N3055 epi transistors. There may be a case to check using higher frequency devices like 2SC5200.
For thermal tracking I have always put the bias stabiliser on the main heatsink holding the output transistors. There are two options here. In a fully complementary amplifier I would put the drivers on the same heatsink as well. I see no reason why that should not apply here either. The second option I have sometimes used (traditionally with TO-39's) is to calculate the worst case power dissipation in the drivers and use a heatsink which keeps the drivers getting no hotter than the output transistors. In theory that should mean that the output heatsink temperature/bias current is overcompensated, but that does not usually happen with a single bias transistor since the exact number of junctions it represents is often not quite enough due to the bias stabiliser Vbe being typically greater than the driver+output transistor voltage drops, meaning that the multiplier ratio is slightly below what would really be required.
BD139/BD140 are not slow, BD139 typ. 250MHz, BD140 typ. 75MHz.
You may be thinking of the similar sounding BD237/BD238 which are epi base and slow. I've never used these for that reason, but have used many BD139/BD140's with no problems. Including measuring ST, Motorola and Philips devices at 1MHz to check the ft. There may be some fake devices out there which are slow, of course.
In the circuit White Dragon has posted (#70) R7 and R17 provide some high frequency oscillation stopping capability. Simulations suggest that 47 ohms in this design (along with the input degen as WD mentioned) are adequate. I see no need for a capacitor across the Baxandall diode. The 1N4002 device type is probably a reasonable frequency match to an epibase transistor like TIP35 (or the recent 2N3055 for that matter). Though I might see if simulations show anything up - but that rather depends on how accurate the SPICE models of the devices would be.
Regarding the circuit in post 71, that is not the circuit I was referring to. I suspect the diodes in series with the local 100 ohm resistors will effectively disconnect the 100 ohm resistor when the relevant BDY20 is turning off, leaving only a 1k resistor to shunt any base current. BDY20's (and 2N3055's for that matter (back then, when they were 1MHz ft)) needed all the reverse base current they can get to keep them from "lingering" on at high frequencies. (And that meant 20kHz back then!) I was specifically referring to the second edition of the book where a 50W amplifier has exactly Baxandall's configuration as has been discussed.
Frequently reported ringing in the PNP/NPN stage depends on adequate supporession. I have not seen any sign in simulation with the circuit WD posted. I have built several circuits which have needed sprog stopping capacitors, though, but in every case I can think of, that also showed up in sim.
Though if anything does show up in sims I'll mention it. As I posted, it seems to be well behaved.
Edit - at least with TIP35/2N3055 epi transistors. There may be a case to check using higher frequency devices like 2SC5200.
For thermal tracking I have always put the bias stabiliser on the main heatsink holding the output transistors. There are two options here. In a fully complementary amplifier I would put the drivers on the same heatsink as well. I see no reason why that should not apply here either. The second option I have sometimes used (traditionally with TO-39's) is to calculate the worst case power dissipation in the drivers and use a heatsink which keeps the drivers getting no hotter than the output transistors. In theory that should mean that the output heatsink temperature/bias current is overcompensated, but that does not usually happen with a single bias transistor since the exact number of junctions it represents is often not quite enough due to the bias stabiliser Vbe being typically greater than the driver+output transistor voltage drops, meaning that the multiplier ratio is slightly below what would really be required.
Last edited:
Probably i was thinking too much because ltp degeneration & base stopper resistors are there to tame possible oscillations. If still oscillation occurs with transistor like C5200 we have many ways to deal with but because i don't own an oscilloscope i should stop overthinking. For me all depends on trial & error. As driver transistors i have BD139/40 & C4793/A1837 pairs which i believe adequate for this amplifier.
Choice of output transistor - for 50W I would use 2N3055 metal can rather than any of the plastic wannabe's. The TIP3055 (not TIP3055T) is rated at 90W and had (once) a second breakdown break point at 30V/3A and is supplied in case TO-247 rather than TO-219 I think these days; the MJE3055T is rated at 75W with a break point at around 35V and shipped in TO-220. The ST datasheet no longer gives SOA data for the TIP3055. On Semi's datasheet for the 2N3055 shows a break point at 40V/2.875A and an improved characteristic from earlier data sheets with 60V at about 0.8A compared with only 200mA on the old 1970's version. Possibly because it is now manufactured on a similar process to MJ15003? I'm wondering if it should be called 2N3055G3 for a third generation of manufacturing process!
I would suggest the TIP3055 or MJE3055T suitable for up to 30W or parallel for 50W.
For the drivers I would recommend BD139-16 BD140-16 if you can select the option for a higher minimum gain. C4793/A1837 are good too. THough your devices must be old stock as I think they are obsolete now. MJE243/253 are similar to BD139-140 but with higher voltage and current ratings, though minimum gain of 40 as is the standard (un-grouped) BD139-140. (measured at slightly different currents, 200mA rather than 150mA but choosing the higher current in the pre-driver stage means this difference is not particularly significant).
I would suggest the TIP3055 or MJE3055T suitable for up to 30W or parallel for 50W.
For the drivers I would recommend BD139-16 BD140-16 if you can select the option for a higher minimum gain. C4793/A1837 are good too. THough your devices must be old stock as I think they are obsolete now. MJE243/253 are similar to BD139-140 but with higher voltage and current ratings, though minimum gain of 40 as is the standard (un-grouped) BD139-140. (measured at slightly different currents, 200mA rather than 150mA but choosing the higher current in the pre-driver stage means this difference is not particularly significant).
For 50watts i'm thinking about one pair of TIP35C because i've many of those. I do not have higher hfe grade BD139/40-16 so i may use c4793/a1837 instead.
A 3055 is a 3055, from any given manufacturer. You get slightly different guaranteed s/b breakpoints because the die temperatures run higher in the smaller packages. You can’t get heat out of a TO-220, and TO-3 is still better than TO-247.
The 3055 always was run on the same process as the 15003. Technically, the other way around. There have been improvements over the decades and I’m sure most of them have been passed down. Not everybody’s TIP35 are made by the same process. Some are epi-base like 3055’s, some are a cheap/slow version of epitaxial planar. Not in the same league as Toshiba or Sanken (not even close), but it does help with the beta hold up at 15 amps. ST’s used to be something else entirely (they called it “base island”, which is sort of a ring emitter) but that got dropped and it moved to EP. IIRC, they had a PCN on it, and it was probably 20 years ago now.
You have to be careful with TIP35’s. People seem to get the idea that they have a lot more SOA than 3055’s. They don’t. The 100 volt rating doesn’t help anything, other than not having to select 3055’s for >60V operation. If you raise voltage, you still need to parallel them so the extra current capability may be wasted. THE advantage of TIP35’s is the fact that the gain holds up much much better at currents above 2 amps. Above 10 the 3055 is useless. Low voltage, low ohms, is where TIP35’s have significant advantage.
The 3055 always was run on the same process as the 15003. Technically, the other way around. There have been improvements over the decades and I’m sure most of them have been passed down. Not everybody’s TIP35 are made by the same process. Some are epi-base like 3055’s, some are a cheap/slow version of epitaxial planar. Not in the same league as Toshiba or Sanken (not even close), but it does help with the beta hold up at 15 amps. ST’s used to be something else entirely (they called it “base island”, which is sort of a ring emitter) but that got dropped and it moved to EP. IIRC, they had a PCN on it, and it was probably 20 years ago now.
You have to be careful with TIP35’s. People seem to get the idea that they have a lot more SOA than 3055’s. They don’t. The 100 volt rating doesn’t help anything, other than not having to select 3055’s for >60V operation. If you raise voltage, you still need to parallel them so the extra current capability may be wasted. THE advantage of TIP35’s is the fact that the gain holds up much much better at currents above 2 amps. Above 10 the 3055 is useless. Low voltage, low ohms, is where TIP35’s have significant advantage.
I don't know much but Rod elliot strongly recommends TIP35/36C as a low budget option for his P3A amplifier (+/-35vDC max). On the other hand i have seen TIP3055 used in many 40-50w low fi amplifier (single pair).
The TIP35 is fine at +/-35V. So are 3055’s, usually, unless impedance is low enough where they run out of gain. Either is iffy at 40, and Just Plain A Bad Idea at 50.
This is what Rod said about TIP35/36C___
"For a budget system, you can use TIP35C (NPN) and TIP36C (PNP) output transistors. If you can get the 'full-pack' TO-247 case versions they can be mounted under the board in the same way as the MJL devices. These are limited to 125W dissipation (25°C case temperature), but despite that apparent limitation they can still drive a 4 ohm load from ±35V supplies. In theory the peak dissipation may be exceeded, but these are extremely rugged transistors and handle abuse with ease. Don't push your luck though - the maximum unloaded supply voltage is ±35V!"
But here my plan is not to use more than +/-30vDC so i guess one pair of TIP35C will suffice. But i understand your point, derating is always a good idea.
Regards
"For a budget system, you can use TIP35C (NPN) and TIP36C (PNP) output transistors. If you can get the 'full-pack' TO-247 case versions they can be mounted under the board in the same way as the MJL devices. These are limited to 125W dissipation (25°C case temperature), but despite that apparent limitation they can still drive a 4 ohm load from ±35V supplies. In theory the peak dissipation may be exceeded, but these are extremely rugged transistors and handle abuse with ease. Don't push your luck though - the maximum unloaded supply voltage is ±35V!"
But here my plan is not to use more than +/-30vDC so i guess one pair of TIP35C will suffice. But i understand your point, derating is always a good idea.
Regards