I tried my P3A with preamp section of my Yamaha AX592 integrated and ordinary PSU. The results were truly remarkable. Excellent sound. It turned out that my previous experience with ordinary PSU was influenced by the preamp that I used. No loss of detail and clarity with Yamaha's preamp section (my Yamaha is tweaked last year - NJM2068 opamps were replaced with NJM5532). Great soundstage and dynamics. Dynamite performance. This P3A circuit is the champ among voltage feedback simple circuits in AB class. So much fun with such simple circuit!
Member
Joined 2009
Paid Member
Member
Joined 2009
Paid Member
Sakis - do you have any theory why this is the case, is the 1503X higher distortion or too high capacitance etc ?
Seems like a good question, especially since I remember reading some different opinions about what the correct value might be, and whether it depends on the output transistor type.
It has probably been explained better but look at the current ratings of the MJE150xx series transistors. They are intended for driving with much heavier currents than seen in low voltage, single output pair amplifiers like P3a. None of the other drivers commonly used are rated like these monsters, and they will often be out of their "sweet spot" (most linear region) in CFP output stages, since operating conditions vary more widely than EF designs.
I've seen and used them in small PA amplifiers with 4 Pairs of output transistors and they really shine there IMHO, even if that's not considered Hi-Fi.
I think that Rod suggested MJE150xx drivers for reliability reasons. If one wants to squeeze max power from the circuit, using 42V rails, he can be sure that drivers won't overheat in any case because TO220 transistors have big tab, you don't need heatsinks for them. These driver are not suggested for "audiophile" qualities, but for high power.
there is some report of the bias in this thread.
DIY Audio Projects Forum • Elliot Sound Products - P3A 100W Amplifier
cheers
kp93300
Another comment of chlorophyll in that thread is revealing. The (dispre) preamp is just like a wire with gain - he's not sure whether he is disappointed or not. 
.
I'm not sure Ivan, about parts recommendations but Rod is very sensitive to builders using fakes and not only does he have to waste time on these problems (he sells thousands of PCBs and probably has to advise even more curious people but he gets a bad rap for sound with fakes which many customer countries are just flooded with - there may be no choice. Noobs have no clue - They see the type number looks OK and buy if it is very cheap. Often they have little spare money after buying PCBs so they are going to get ripped off more often than not.
Rod has changed part specs several times, not because of availability or new developments but the dominance of fakes in particular countries! Even BD139/40 has been faked and dunno why, with so many competitors like Fairchild, ST Micro etc. making different versions in China etc. Why would anyone seriously re-mark a pair of chips with about the same production cost as the ancient Philips types required, I'd think, only by DIYs or as a general purpose medium power transistor.
.I'm not sure Ivan, about parts recommendations but Rod is very sensitive to builders using fakes and not only does he have to waste time on these problems (he sells thousands of PCBs and probably has to advise even more curious people but he gets a bad rap for sound with fakes which many customer countries are just flooded with - there may be no choice. Noobs have no clue - They see the type number looks OK and buy if it is very cheap. Often they have little spare money after buying PCBs so they are going to get ripped off more often than not.
Rod has changed part specs several times, not because of availability or new developments but the dominance of fakes in particular countries! Even BD139/40 has been faked and dunno why, with so many competitors like Fairchild, ST Micro etc. making different versions in China etc. Why would anyone seriously re-mark a pair of chips with about the same production cost as the ancient Philips types required, I'd think, only by DIYs or as a general purpose medium power transistor.
I can confirm again the sound of 1503XX transistors , I will agree about the ratings and i will remind that Harman Kardon and many others like to run their drivers VAS and often LTP on the edge from aspects of ratings ( current /voltage and power ) since they think that this way play better and measure better . ( often used parallel drivers to manage the current and that never worked properly i think )
To do that in 1503X you need 300 v power supply ...I presume that this is the one reason but also has to do with the way the actual transistor is made . expirience from mix matching semis in the long journey of the P3A proved that similar semis from the Family of japan 2SA -2SC do not or suffer less from that issue .
Farther more my choice is to run my LTP out of specs and push it a bit harder together with some degeneration and seems that this part plays 2 clicks better .
Still didn't want to mess around with my VAS cause its bootstrapped.
Kind regards
sakis
To do that in 1503X you need 300 v power supply ...I presume that this is the one reason but also has to do with the way the actual transistor is made . expirience from mix matching semis in the long journey of the P3A proved that similar semis from the Family of japan 2SA -2SC do not or suffer less from that issue .
Farther more my choice is to run my LTP out of specs and push it a bit harder together with some degeneration and seems that this part plays 2 clicks better .
Still didn't want to mess around with my VAS cause its bootstrapped.
Kind regards
sakis
I only have a single set of Rod's boards assembled, so doing a-b type comparisons is difficult.That's my build
I settled on 250mA bias which gives a small amount of class A power for background listening sessions.
When I did try this with other circuits, it seemed to me that the ideal bias (and other things you can tweak), depended on the selection of output transistors. So for example, I built Rod's boards with MJL4302/MJL4281 (b/c I had lots of them on hand). I suspect those are overkill by a wide margin, and will have a different "best" bias than smaller devices, which I will probably use next time.
I think you said in the past, that you prefer using much faster VAS transistors than those specified by Rod. So, there are probably two variables here, one is the power rating, and the other one is speed/slew rate. I could easily build a second set of boards with KSA1381E/KSC3503E (which I have for another project). However, you are also using VERY large rail caps, and that is much more difficult to do in my case, or not with the supply voltage I would prefer.
By this you mean more current/hotter, or higher voltage?
I have grate respect for all these people that ""follow"" my thread and many things said here is a wonderful thing to learn for everybody .
I don't have intention to keep secrets about the P3A still there is a couple of things about LTP and use of specific transistors together with degeneration that i use and id like to confirm better before posting my results
The transistors i use in the LTP are lower voltage (and far lower noise ) than supposed to be ( which is against safety ) but working flawless for 10 hours or more per day for more than 3 years now
The newer board that features also similar transistors and a couple of more tweaks hasn't put to extensive test yet .
I may confirm partially the bias versus choice of semis but i cannot share my results or double cross with any of you since i use original 1302-3281 which are unobtainable ,still i expect old style slow outputs to behave better with plenty of bias while newer faster types the opposite . Yet again this will go under the line together with the rest of the circuit, pcb and implementation .
Kind regards
Sakis
I don't have intention to keep secrets about the P3A still there is a couple of things about LTP and use of specific transistors together with degeneration that i use and id like to confirm better before posting my results
The transistors i use in the LTP are lower voltage (and far lower noise ) than supposed to be ( which is against safety ) but working flawless for 10 hours or more per day for more than 3 years now
The newer board that features also similar transistors and a couple of more tweaks hasn't put to extensive test yet .
I may confirm partially the bias versus choice of semis but i cannot share my results or double cross with any of you since i use original 1302-3281 which are unobtainable ,still i expect old style slow outputs to behave better with plenty of bias while newer faster types the opposite . Yet again this will go under the line together with the rest of the circuit, pcb and implementation .
Kind regards
Sakis
@Sakis: I was not referring to your latest version, but some of your old posts were suggesting much faster VAS transistors (which I have not tried yet). The original boards do not have plated holes, and my working set has had as much rework as it can stand, so that will be for the next build...
Running the front end closer to the limit is new to me, I would have thought that 25~50% of rated power, and 75% of rated voltage would be conventional wisdom here.
The output transistors I used I knew to be excessive when I assembled the board. They are new-ish and I had them already.
NJW1302/3281 seem closer to a typical P3A power range, and they cost around $1 each (not much worry about fake parts there!), so that is about what I was aiming for next time.
Running the front end closer to the limit is new to me, I would have thought that 25~50% of rated power, and 75% of rated voltage would be conventional wisdom here.
The output transistors I used I knew to be excessive when I assembled the board. They are new-ish and I had them already.
NJW1302/3281 seem closer to a typical P3A power range, and they cost around $1 each (not much worry about fake parts there!), so that is about what I was aiming for next time.
Last edited:
Increasing BIas and comparison to DOZ.
I have a distinct prference for higer bias. Music is more fluid and the treble shows a very obvious improvement . in fact treble quality has bettered the class A DOZ with TIP 3055 .
Mid is very close to the DOZ --mod by Shaan and dc coupled version .
Bass is better and more controlled with P3a.
Must thank Sakis for the inspiration to build this.
Can you share where is the point in the circuit you take the speaker ground /
return ?
kp93300
I play around with bias increasing from about 45ma to 75ma.
I have a distinct prference for higer bias. Music is more fluid and the treble shows a very obvious improvement . in fact treble quality has bettered the class A DOZ with TIP 3055 .
Mid is very close to the DOZ --mod by Shaan and dc coupled version .
Bass is better and more controlled with P3a.
Must thank Sakis for the inspiration to build this.
Can you share where is the point in the circuit you take the speaker ground /
return ?
kp93300
@Sakis: I was not referring to your latest version, but some of your old posts were suggesting much faster VAS transistors (which I have not tried yet). The original boards do not have plated holes, and my working set has had as much rework as it can stand, so that will be for the next build...
Running the front end closer to the limit is new to me, I would have thought that 25~50% of rated power, and 75% of rated voltage would be conventional wisdom here.
The output transistors I used I knew to be excessive when I assembled the board. They are new-ish and I had them already.
NJW1302/3281 seem closer to a typical P3A power range, and they cost around $1 each (not much worry about fake parts there!), so that is about what I was aiming for next time.
Been trying many .. and each and every one of them will sound slightly different issue that will depend on the miller cap ...Lately i have settled with BD829-830 combo and 82pf compensation silver mica 5%
Running the front end closer to the limit is new to me, I would have thought that 25~50% of rated power, and 75% of rated voltage would be conventional wisdom here.
Sorry i dont get the point here .Using lower voltage transistor and lower noise is a risk that pays ...Bringing transistors closer to their limits will add a marginal improvement if one is willing to take the risk ( always as said extremely carefully matched )
Kind regards
Sakis
My latest stereo pcb features a real star ground from both topology and length of ground trace for both ch ... its a tribute to symmetry !!! so CT is the center of star ground and speaker ground starts from there .
still i have zobel located in the binding post with a separate return to the star ground , additionally a clean ground for the small current things .
On the other hand in earlier versions of pcb before version 2525.1.25
i have been operating ground otherwise without problems so keeping pcb rules for that also will add something to the overall design but how much this is i have never calculated Kind regards
Sakis
I tried it with the 10k stereo log pot on the pic and Marantz CD57MK2. Don't know who is producer of the pot but it looks like exact replica of Alps RK16.
I usually don't like power amps with passive preamps (pot at the input) but P3A works fine in that configuration. The sound is detailed and open, not dull at all. But I still prefer active preamps. With passive preamps there is a certain lack of ultimate dynamics, the sound has less grip and authority, everything is a bit softened and bland. Passive preamps are for people that prefer gentler presentation of music but I like max drama and emotion, even if the price is somewhat more aggressive sound overall. If one doesn't have active preamp it is possible to use pot for some time if that's the only solution. Very few, if any, amps will give their best with passive preamps. I would always go for active and use small gain of 6-10dBs in the preamp.
I think that P3A will work excellent directly driven from PC sound card output, but using software controls is not practical for some people. The feel is much better when using the real knob than when sliding faders on the monitor.
