Okay, about Q5 selection:
I've checked the following TO220 & TO126 transistors with a parts-express capacitance meter MS6013.
part # brand Cbc Cbe source
470 ceramic 450 pf Radio Shack grab bag 1970's
330 mica 320 tubeandmore antique radio supp
TIP41C Fa 270 735 newark TO220
MJE15028 ON 920 1880 newark TO220
MJE340 ON 720 970 newark TO126
TIP31c Fa 160 365 newark TO220
BD139 Fa 606 ^ newark TO126
KSC2690 Fa ^ ^ newark TO126
D44R2 GE 210 642 electronicsurplus gold pin TO220
2SC3423 y ^ ^ TO126 mcmelectronics datasheet toshiba, parts are not
up arrow means the reading kept rising ~60 pf per scan never stopped.
So perhaps the only reason my MJE15028 as driver in the AX6 sounds good on hi freq is that it is driven by a GE D44R2 VAS that doesn't care how much capacitance it drives. Both datasheet Ft are 30 mhz.
Happy shopping.
This exercise is interesting, but since the # of transistors is the same as AX6, I'll pass on this. Your thermal management is more careful than Mike Slavonic's (? apex) but I used 2 fans so I got away with it reliably.
I've checked the following TO220 & TO126 transistors with a parts-express capacitance meter MS6013.
part # brand Cbc Cbe source
470 ceramic 450 pf Radio Shack grab bag 1970's
330 mica 320 tubeandmore antique radio supp
TIP41C Fa 270 735 newark TO220
MJE15028 ON 920 1880 newark TO220
MJE340 ON 720 970 newark TO126
TIP31c Fa 160 365 newark TO220
BD139 Fa 606 ^ newark TO126
KSC2690 Fa ^ ^ newark TO126
D44R2 GE 210 642 electronicsurplus gold pin TO220
2SC3423 y ^ ^ TO126 mcmelectronics datasheet toshiba, parts are not
up arrow means the reading kept rising ~60 pf per scan never stopped.
So perhaps the only reason my MJE15028 as driver in the AX6 sounds good on hi freq is that it is driven by a GE D44R2 VAS that doesn't care how much capacitance it drives. Both datasheet Ft are 30 mhz.
Happy shopping.
This exercise is interesting, but since the # of transistors is the same as AX6, I'll pass on this. Your thermal management is more careful than Mike Slavonic's (? apex) but I used 2 fans so I got away with it reliably.
Last edited:
Frequency Compensation
The use of evergreen transistors is laudable since it makes the design more accessible for many observers.
I would not have chosen 2N3055H with that in mind, since apart from the low fT these are no longer in production. The TIP41 may be robust in the role of split phase driver but it too has a limited fT which puts a low ceiling on what you can get out of the circuit in gain bandwidth product.
Change in phase starts at a tenth of the -3dB frequency. You are having to use large values of capacitor for stability which load your split phase driver. I note a series resistor to moderate this.
You should not need such a compromise if you select more fT capable transistors as the value of the compensation capacitor for higher frequency will be lower and reduce the loading.
The use of evergreen transistors is laudable since it makes the design more accessible for many observers.
I would not have chosen 2N3055H with that in mind, since apart from the low fT these are no longer in production. The TIP41 may be robust in the role of split phase driver but it too has a limited fT which puts a low ceiling on what you can get out of the circuit in gain bandwidth product.
Change in phase starts at a tenth of the -3dB frequency. You are having to use large values of capacitor for stability which load your split phase driver. I note a series resistor to moderate this.
You should not need such a compromise if you select more fT capable transistors as the value of the compensation capacitor for higher frequency will be lower and reduce the loading.
There is something seriously wrong with your capacitances readings: they are much too high, and they should be reasonably stable.
Your meter probably forward-biases the junction it measures.
Most cheap meters apply a DC average voltage on their terminals, so the first thing to do is to reverse the polarity.
If it doesn't work, it means that the test voltage is pure AC or too high, in which case you can try to insert a 10nF or 100nF cap in series, in the hope that the junction will self-bias in the reverse mode.
Without any measurement, your MJE340, BD139 and KSC2690 should work well (but the compensation will differ from my build. When you have made your choice, I can try to help you with a sim, so that you land more or less in the ballpark; fine-tuning will be up to you)
Remember the purpose of this amplifier: it was an improvised build made ~20yrs ago, apparently solely to the test the viability of a topology.
I unearthed it, because I knew I had this kind of thing somewhere in a crate, and member Minek was interested in all-N topologies.
It had probably never been tested for frequency or step response: it was just stable enough (thanks in part to the high closed-loop gain) to carry out DC, thermal stability and linearity tests.
Discussing the merits of the components chosen makes therefore little sense: it is painfully obvious that they are completely inadequate, but the principles of the thermal compensation should readily apply to better devices, because they will also comply with the first principles of semiconductor theory.
This build has a merit though: however weak it may be, it shows that the principle works even for a very low grade amplifier, built with low-grade components.
It is very much in the early seventies style, and many members are fond of such retro tech.
That is true, if you look around the city streets you will see cars old and new that enthusiasts, young and old, have customised to make them feel a demonstrable sense of achievement among their peers.
Also able to be built with whatever is on hand. Doesn’t require that one spend money on MJL4281’s, C3503’s, or any high voltage PNP’s. Would work equally well with a couple TIP3055’s salvaged from a 40 year old car stereo amplifier or a couple of “Je ne sais quoi” TO-3’s that read as NPN Si but you have no other information.
I have some Toshiba 2SD665 O suffix transistors and heat sinks already drilled to take these and a suitable toroidal transformer - leftovers from a lot of bits and pieces I was given by a friend whom I had helped out with some repairs before he moved his family overseas. My friend was the proprietor of an amplifier brand that was later absorbed by a larger overseas concern. This would be 30 or more years ago now.
What if you delete the bootstrap at Q2 collector, connect said collector directly to the pos. rail; and instead bootstrap directly at Q2 base, and "harder" i.e. with higher current? This would throw away some positive swing headroom, but may be "faster" in the end?
Wow! 15MHz Ft
Compared to the 3055H, they are in completely different league.
Use a matching VAS/splitter, and you should have an almost modern amplifier
Here is the idea tested in sim, but the TIP41 stubbornly resists all tentatives....
Attachments
That's a little harsh, imagine the Q3 fearing for its life ... what I meant was a "classic" bootstrap with a split resistor (R3 -> R3a+R3b), like f.e. 2k2+2k2 and the C1.
So the TIP41 is the reason for the slow pos. slewrate? What if (in sim) you replace it with the BD139 model? Should be faster. Or is the slow-ness elsewhere?
So the TIP41 is the reason for the slow pos. slewrate? What if (in sim) you replace it with the BD139 model? Should be faster. Or is the slow-ness elsewhere?
I still stand by my past remark - derate the amp slightly, to say a 50-watter/4ohm; drop the rails a bit, use faster BD139-16's in splitter/driver positions; and use TIP3055/TIP35 as outputs. That would make it a old-boy with a facelift.
Now let's discuss the bootstrap section. The R3 goes directly to Q2 collector as it should. Keep this connection. But what happens if you insert a diode between this R3-Q2c junction and the pos. rail - and drop the value of R3 to ... say 220R (should be somewhat similar to what you had in the first place - only "backwards"!!) Now this should both "lift" the Q2c and also tickle its base drive harder than before (which in theory at least should be "faster, better, more").
Now let's discuss the bootstrap section. The R3 goes directly to Q2 collector as it should. Keep this connection. But what happens if you insert a diode between this R3-Q2c junction and the pos. rail - and drop the value of R3 to ... say 220R (should be somewhat similar to what you had in the first place - only "backwards"!!) Now this should both "lift" the Q2c and also tickle its base drive harder than before (which in theory at least should be "faster, better, more").
No, I personally won't do anything, at least on the physical prototype: it has done its job, and I published it only to be helpful to Minek and other similarly-minded people, and as a source for inspiration for the general community: there is a life outside complementary designs, even if you don't adopt transformer based topologies or beasts like circlotrons.
If Indianajo spells out his specs and his preferences about OP devices, I'll try to come up with a virtual design (that will need to be refined in the real world), but that will be it (unless I stumble on something really interesting).
It is clear that with better devices, even your vintage 2SD665, it will be possible to achieve vastly better dynamic performance.
I am less optimistic regarding the linearity: it is probably possible to go below 0.1% (the MOS example showed a possibility), but not much lower: with such a simplistic topology, options are limited once you have explored the usual alleys.
I think such a project (if it turns into one, which is not presently the case) will appeal to a number of members, because amplifiers showing a "strong character" combined with simplicity generate a lot of interest.
This thread is open to all contributors, like Anti who has provided valuable leads
Last edited:
The thread would be more open if you would post an .asc for people additional to Anti to experiment with.
For instance the closed loop gain of the circuit is (470R/10R) +1. They may want to improve the THD a bit by reducing that ratio.
While I could draw up a simulation I don't want to hijack your thread.
For instance the closed loop gain of the circuit is (470R/10R) +1. They may want to improve the THD a bit by reducing that ratio.
While I could draw up a simulation I don't want to hijack your thread.
I was riffling through 200+ Watt transistors last night looking for two pairs for OldFashMax. The MJW21194 I have are all defective, Pw is 250 mw not a misprint. So the MJL21195 are orphans. The MJ21194 & MJ21195 are TO3 and I have a heat sink with a big flat bar, but it won't fit in any chassis I have that I can ventilate. The MJ15015 were destined for clone ST120's for Hammond H182 organs and I'm tired of that idea. MJ15015 could go in archosaurus , but they have 2 amp dcsoa @ 70 v. (this is +-35) I'd like 3. Also no ventilated case for TO3 heatsink. Drilling 80 mm holes for fan/exit in metal ??? I do not have any center tap transformer under +-45 @ 1.5 A except +-17@100 ma for Archosaurus. Delete center tap connection, double the filter cap voltage, would the Archisaurus centerline stay centered? No DC servo that I see. So archosaurus is theoretical/desirable at under 50 wires, but power supply/case not in stock. No asc file necessary for point to point builds of <50 wires, IMHO mjona.
OlfashMax is going to be MJW1302/NJW3281 outputs since I have 6 pairs of those. Olfashmax bare board is being drilled tonight for ~70 wires. triple 1n4148 series will be 2.1 v zener. I do have BD139/140, MJE340/350 MJE15028/29 and KSA1220/KSC2690 for drivers or VAS on either project. Olfashmax board is 4"x4" and heat sink is 1.5"x3.75" per channel with fan, two will fit in bud 19"x5.5x3.25" chassis from the failed LM1875 project. Olfashmax transformer will be 44 v 7 A no CT if it works when it gets here, 36 v 2 A no CT if 44v tfmr fails current leak test. Want about 70 W peaks out of either with 1.6 v input, average output 1/4 W. Low crest factor classical music mostly, a little rock. 8 ohm SP2-XT load 5.5 ohm minimum @ 2000 hz. Thanks for simming, I don't have the op system for spice software.
BTW will be off internet 10/12-10/22.
Last edited:
I think if the .asc file was to be posted it would shift the focus away from people making mistakes since from post 1 up to post 17 the basic structure of the circuit remained the same. There was a modification to this at post 30 and a further one at post 32. That leaves the early .raw files in need of an update.
Last edited:
I threw a quick sim in online falstad sim and it confirmed my "theory" outlined in p.#34. (the diode thing etc). One thing that should be obvious is that the present ratio of R1:R6 isn't optimal, R1 should be somewhere around 2/3 of the R6 (needs evaluation ???). Like 6.8R for R1 and 10R at R6 ... or similar.
My ..err.. "masterpiece" used a opamp to drive the NPN "quad-totem"; and the bootstrap section per my recommendation and it seemed to cope well with 2 Ohm load at +/-15V volts supply, and cca 16mA standing current (with the marvelous falstad BJTs that never fail). Delivered +14.1x/-14.4x Vp before flat topping.
So the assumtion seems valid and awaits furher polishing from dedicated developers.
My ..err.. "masterpiece" used a opamp to drive the NPN "quad-totem"; and the bootstrap section per my recommendation and it seemed to cope well with 2 Ohm load at +/-15V volts supply, and cca 16mA standing current (with the marvelous falstad BJTs that never fail). Delivered +14.1x/-14.4x Vp before flat topping.
So the assumtion seems valid and awaits furher polishing from dedicated developers.
I shall post a single-supply version of the circuit (which is not going to be difficult, since it is natively so), and I'll try to come up with something that suits you (but it will be virtual, and other members may have other ideas)
The mods are relatively minor, and their main purpose is to allow a good time resolution (to see the edges in detail), without having to wait for ever for the average conditions to settle (there are certainly smarter ways to do it, but reducing large capacitors value is quick and easy).
Anyway, here is the "switching" asc file, also including a (dangerous!) speed-up trick.
The modified version suggested by Anti in post #32 is not included, because although it seems to work well, it does so by forcing the circuit to operate in class A, which is not desirable for a lean, class AB amp.
Using a BD139 in the initial circuit does markedly improve the slew performances, but not as much as with pure class A
That was one of my first ideas, because R1 is in series with the dynamic emitter resistance of Q2, and to restore symmetry, it should be made lower (and one of the sims I posted earlier uses a smaller resistor) , but the tests I made do not seem very conclusive. The issue remains open.
Using an opamp in this circuit is definitely off-limits for me, but feel free to post any variation on the theme, even if it looks somewhat extravagant: some people (including myself) could be interested
I have the same message for Mjona: don't fear to hijack the thread in this way.
Some people do it in a more egoistic and less constructive manner, and since this thread is not an actual project (yet), just a basic framework, any positive contribution is welcome.
Here is the asc and the BJT list of models I use (found on a wiki). You have to remove the bogus asc extension needed to bypass the forum limitation on txt size
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.