Well, no offence, but I think reaching at least 100-200 v/us should not be that difficult. I did that myself with simpler circuitry than this. Still very impressed with the distorsion figures though!
Actually , the very low distortion and 1-200V slew rates are very minor considerations in the audio band. Doug self's "blameless" only does 27-35v/uS according to his book.
I would rather have an unconditional stable , musical amp with .05% THD than a marginal stability amp with ultra - low THD / super high slew.
Absolutely not offended.
OS
That will be my GX , with full cascoded VAS just like the original. paired with the PB250 it should be a good clone. The apt used a triple ,, but with modern driver/output pairs 2SA1837/2SC4793 + MJL4302/4281's , the EF2 will be up to "snuff".
GX will be coming "soon" to a theater near you..
I am doing the "Blameless" now (AX1). The VAS was very easy , I kept it compact (below - 2) with it's 2- KSA3503 VAS CCS back - to - back. On the top end (+) the KSA1381 and it's EF driver (2N5401) are also , should I say .. "face to back". All 4 trannies will happily run in class A with a little 50 X 20 mm (2 x 3/4") heatsink which you could cut out of roof flashing. You don't need anything special since ALL my VAS's run 6-8mA per device. On my "supersym" , at 7mA X 5 = 35mA ... just slightly warm , regardless of output stage use.
I decided to use the ksc1845's on any differential without a cascode. At 120Vceo ,low COB (1.8pF) , high gain (Hfe100+) they are best. I will keep the 2N5551/2N5401's for CCS's and CM's.
Ahhh , I MUST DO A "CLASSIC" RIGHT. (Gets out the Doug Self "bible" to ponder a while
) OS
Attachments
For better clipping, put a protection diode between Q7's base and Q8's collector. This will help protect the VAS and decrease clipping recovery time.
The Blameless looks fine, and it seems you've carefully included an error made by doug himself. He said himself that R8 was a mistake, if only I could remember where. Anyways, notice that the 2N5551 has markedly worse Ft than the 2N5401 (PNP wins over NPN in this case). Increasing Vce decreases Cob. Including a resistor may or may not increase HF impedance in important places, but it will lower the RC corner frequency, where phase changes rapidly, possibly into the audio band (but probably in the ultrasonic). I try to steer clear of phase anomalies.
I have only simulated a few high-slewing amps that I was proud of and that were considerably stable. Are there any examples of high-slewing audio amps that I can look at?
- keantoken
The Blameless looks fine, and it seems you've carefully included an error made by doug himself. He said himself that R8 was a mistake, if only I could remember where. Anyways, notice that the 2N5551 has markedly worse Ft than the 2N5401 (PNP wins over NPN in this case). Increasing Vce decreases Cob. Including a resistor may or may not increase HF impedance in important places, but it will lower the RC corner frequency, where phase changes rapidly, possibly into the audio band (but probably in the ultrasonic). I try to steer clear of phase anomalies.
I have only simulated a few high-slewing amps that I was proud of and that were considerably stable. Are there any examples of high-slewing audio amps that I can look at?
- keantoken
Oh yeah, and
http://www.diyaudio.com/forums/everything-else/4182-wiki-direction-2.html#post2265161
Later when you have the time I think the WIKI would be the best place to put a Mongrel Guide.
- keantoken
http://www.diyaudio.com/forums/everything-else/4182-wiki-direction-2.html#post2265161
Later when you have the time I think the WIKI would be the best place to put a Mongrel Guide.
- keantoken
Changes in blameless have been assimilated (borg speak) (pix 1)
VoOdOo CcS ??????? GIMME...
It must have easily available , CHEAP parts, no more than 4 devices (semi's)
BTW , APT1 (the GX) is pix 2.
OS
Attachments
Let me finish them to at least the GX , they are hard (2 hours for a voltage board).
OS
Alright, I will post the CCS. It uses three BJT's, and optionally a schottkey (my personal improvement). No hard to find parts, just one extra transistor, and it's the best.
BTW, you put the diode in the wrong place I think, it's supposed to Q8's collector, not emitter. This is done often in opamps, for reference.
- keantoken
BTW, you put the diode in the wrong place I think, it's supposed to Q8's collector, not emitter. This is done often in opamps, for reference.
- keantoken
I was just thinking , ANOTHER reason I am driven to do this .. besides irking Carlos 
...is ALL the PCB's I have wasted just building an amp .. getting the itch to try a new one .. DIY is addictive .
It just seems to be a waste to go through so many boards. The "frugalamp 1" went to a neighbor , "frugalamp 3 " went to "Danielwritesback" (forum member).
I have only kept 2 DX style amps (35-0-35) and my large "supersym" (77-0-77).
KEAN, have you tried the "super pair" (it is below).
Superpair is now officially the "FX1.1" (next board after the AX)
OS
...is ALL the PCB's I have wasted just building an amp .. getting the itch to try a new one .. DIY is addictive .It just seems to be a waste to go through so many boards. The "frugalamp 1" went to a neighbor , "frugalamp 3 " went to "Danielwritesback" (forum member).
I have only kept 2 DX style amps (35-0-35) and my large "supersym" (77-0-77).
KEAN, have you tried the "super pair" (it is below).
Superpair is now officially the "FX1.1" (next board after the AX)
OS
Attachments
Attachments
The baxandall super pair? The super secret voodoo CCS is based around it...
I like the idea of the Baxandall pair as a VAS, but look at C6. The purpose of the Baxandall pair is to decrease effective Cob, yet you've just added a whopping 68p. I may argue that it does help cure problems by canceling nonlinear Cob effects, but how much Cob does the KSA1381 have? Is it enough to make a difference? Second, the MPSA42 is a very slow transistor with whopping Cob (IIRC), and Cob is at it's largest with Vcb=~0mV. Even a faster transistor may have significant Cob in this position. I also don't know if saturation could be an issue, though it could help to put a diode in series with Q8's emitter, raising Vcb(Q7), decreasing Cob (negligibly?), and steering clear of saturation.
Now all that aside, if you can get it stable, I think it still may be better than the Darlington VAS, but how much so, and do the pitfalls outweigh the benefits?
The best situation I can see in using the Baxandall VAS is where C6 and C5 are kept well away from the VAS output, quarantining the output switching high order distortions, and another, albeit adventurous, stability compensation scheme is used. Contact me if you're adventurous.
In the situation just described, I think the Baxandall CCS is a very good choice.
One last thing to consider. For the Baxandall CCS R16 is necessary. This causes Vcb(q3) to vary according to VAS current, and this invokes Early affect, decreasing the gain of the CM+LTP stage. It also makes the Cob of Q3 and Q1 less negligible because Vcb is no longer constant.
You may be able to get around this, if you can find a clever way to get Q7 to sink all the current Q8 takes so that current through R16 is constant.
Next, power supply ripple through R17 will inject straight into the VAS, decreasing PSRR. This may not be a problem but if it is I would bootstrap R17 to positive rail, filtering out hum.
I have had some really good ideas though, that I'm testing in the sim as I write.
EDIT: didn't see your latest post, yes, that is what I mean. Now that I look, I don't know if a 1N4148 will be able to take that voltage drop, and a larger diode might have worrying parasitic capacitance... Hmm...
- keantoken
I like the idea of the Baxandall pair as a VAS, but look at C6. The purpose of the Baxandall pair is to decrease effective Cob, yet you've just added a whopping 68p. I may argue that it does help cure problems by canceling nonlinear Cob effects, but how much Cob does the KSA1381 have? Is it enough to make a difference? Second, the MPSA42 is a very slow transistor with whopping Cob (IIRC), and Cob is at it's largest with Vcb=~0mV. Even a faster transistor may have significant Cob in this position. I also don't know if saturation could be an issue, though it could help to put a diode in series with Q8's emitter, raising Vcb(Q7), decreasing Cob (negligibly?), and steering clear of saturation.
Now all that aside, if you can get it stable, I think it still may be better than the Darlington VAS, but how much so, and do the pitfalls outweigh the benefits?
The best situation I can see in using the Baxandall VAS is where C6 and C5 are kept well away from the VAS output, quarantining the output switching high order distortions, and another, albeit adventurous, stability compensation scheme is used. Contact me if you're adventurous.
In the situation just described, I think the Baxandall CCS is a very good choice.
One last thing to consider. For the Baxandall CCS R16 is necessary. This causes Vcb(q3) to vary according to VAS current, and this invokes Early affect, decreasing the gain of the CM+LTP stage. It also makes the Cob of Q3 and Q1 less negligible because Vcb is no longer constant.
You may be able to get around this, if you can find a clever way to get Q7 to sink all the current Q8 takes so that current through R16 is constant.
Next, power supply ripple through R17 will inject straight into the VAS, decreasing PSRR. This may not be a problem but if it is I would bootstrap R17 to positive rail, filtering out hum.
I have had some really good ideas though, that I'm testing in the sim as I write.
EDIT: didn't see your latest post, yes, that is what I mean. Now that I look, I don't know if a 1N4148 will be able to take that voltage drop, and a larger diode might have worrying parasitic capacitance... Hmm...
- keantoken
Last edited:
You lost me , Kean. I assume you are speaking of the FX VAS (R16,17 etc. ???) where is the voodoo CCS ??, you can email it to me.
On the real amp (the coda with the super pair), they set the dominant pole between the base of the first superpair tranny to the base of the non-inverting input transistor (56pF). I used a mpsa42 because I have 150v between rails , I was scared of the 2SA733's Vceo.
Schema is below , amp really works .. I have the real one in the house.
You could send me the "voodoo" CCS as a .ASC file !!
OS
Edit .. Kean ,you are also right about isolating the input stage from the "dirty switching" of the OPS. By where they place the miller and the fact they take NFB from the drivers ..NOT the OPS , this seems to be the koda designers goal.
I assume you are speaking of the FX VAS (R16,17 etc. ???) where is the voodoo CCS ??, you can email it to me.On the real amp (the coda with the super pair), they set the dominant pole between the base of the first superpair tranny to the base of the non-inverting input transistor (56pF). I used a mpsa42 because I have 150v between rails , I was scared of the 2SA733's Vceo.
Schema is below , amp really works .. I have the real one in the house.
You could send me the "voodoo" CCS as a .ASC file !!
OS
Edit .. Kean ,you are also right about isolating the input stage from the "dirty switching" of the OPS. By where they place the miller and the fact they take NFB from the drivers ..NOT the OPS , this seems to be the koda designers goal.
Attachments
Last edited:
Also, It is fine to use A733 there or anything better than MPSA42, even the 2N5771 (I would recommend the 2N5089). This transistor will die of overcurrent in the event of a failure far before it succumbs to overVce.
Here is the modfied Baxandall VAS. I was a bit optimistic in my last reply, there are certain tradeoffs between it and the Darlington VAS. I don't know if the Baxandall VAS can perform better than the Darlington in an LTP+CM configuration. The merit of it is that Vout is not reflected on Vin through Early affect, and so there is greater isolation of the OPS switching transients.
- keantoken
Here is the modfied Baxandall VAS. I was a bit optimistic in my last reply, there are certain tradeoffs between it and the Darlington VAS. I don't know if the Baxandall VAS can perform better than the Darlington in an LTP+CM configuration. The merit of it is that Vout is not reflected on Vin through Early affect, and so there is greater isolation of the OPS switching transients.
- keantoken
Attachments
This schematic is in short a common emitter followed by a
common base...Some sort of cascode....
Thanks for the PDF.
I never used a Baxandall pair in a practical amplifier,
since i had results good enough with traditionnal
cascodes , but i ll dig the idea in my simulator...
The only case i saw by there is Apex s schematic,
and according to his sayings , it work well , although
i wouldn t myself venture in such builds without
heavy investigations...
Great you posted that, I had a idea for a new type vas 2 3 years back which involves folded cascode but as I was very busy building 2 engines at the time I totally forgot about it and never even simmed it to have a look if its any linear. Compact folded cascode has just reminded me of the idea
Time to play a little with ltspice again Os I simmed the sym amp, but the one using Jfets, by changing to 2sk170 jfets and making 3 resistor changes the result is a amp that measures just as good as the BJT version with the advantage of higher speed. With Jfets dont use degeneration, their gm is very low, drop the degeneration resistors to 10 ohm, raise the ltp current to 2ma at least through the jfets and use 2sk170 fet, and youll get much better performance from it. THD drops from your 0.3 level to 0.004 which brings it in line with the Bjt model. If you use better jfets the performance is even better, you can also ease on the compensation as I found it to be a bit overcompensated which would improve things further.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.