Ppl, don't you think that is a bit overkill for a headphone amp ?
The way you have cascoded the vas, you should also cascode the input diffamp, that heavily increases quality, giving a nearly constant Vds to the jfets. (exploding PSRR for lower freqs and increasing dynamics)
BTW, i do love that topology...
Mike [/B]
Yes, this topology is overkill, but I like it too. I will put up my version very soon, and that contains some solution which almost same. But the expected power is about 150W/8ohms, and it goes up to about 500W/2ohms
sajti
ppl said:
I somehow disagree, see my thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=74257&perpage=10&pagenumber=1
The same mechanism is valid for jfets. With a constant voltage cascode you increase the miller effect, with a floating one you "stop" it.
The xtra current introduced into the diffamp does not harm in any way, you have just more current through the ccs. Of course you need to be careful about introducing noise.
Mike
Thanks for the Link; however I was referring to the fact that the input stage is Cascode as you pointed out in your previous post. In this post you refer to where the Voltage reference is returned to ground or the Emitters. Moving along to this topic of where to return the reference to notice that while on the diff amp stage I use ground while on the second vas stage the V ref is returned to the Emiters.
Hi ppl, yes, we had some misunderstanding, i refered to the way the input diff amp is cascoded...
Mike
Mike
The previous post of mine re. Sym5.3.B is, as mentioned, used as a kind of reference to compare with this similar topology shown, (which leans heavily on MikeB’s Sym6 schematic) and carries four O/P transistors (only two fitted for initial test purposes). A dual slope SOAR protection scheme is included on the pcb (as fitted to my MiniLeach layout), but not shown on the schematic.
And regrettably, I have done no aural auditioning of either yet, purely bench testing so far.
Regards,
Brian
And regrettably, I have done no aural auditioning of either yet, purely bench testing so far.
Regards,
Brian
Hi!
that is me after vacation in Italy: 😎
today I bought MJE15030/31s and I am going to solder them today in the place of BD243/244! I believe that the oscillations will disapear!
that is me after vacation in Italy: 😎
today I bought MJE15030/31s and I am going to solder them today in the place of BD243/244! I believe that the oscillations will disapear!
And here are some more (boring) ‘scope shots. No cause for alarm there, I don’t think. Note the much more controlled damping to the ringing on the 8R//2u2 result in comparison to the Sym5.3.B. Due to the slower rise time?
How do these compare with other examples of this genre that are out there?-
Does anyone have an LTSpice file of the Symasym 5. My efforts seem to exhibit a ripple at the zero crossing on a sine wave, but this Sym7 is ok.
Regards,
Brian.
How do these compare with other examples of this genre that are out there?-
Does anyone have an LTSpice file of the Symasym 5. My efforts seem to exhibit a ripple at the zero crossing on a sine wave, but this Sym7 is ok.
Regards,
Brian.
MikeB said:
Mike you have again got me thinking about returning the reference for the cascodes on the diff amp to the Sources rather than Ground. This was done this way in the diff amp stage because the input voltage is in the mV range since this is a headphone amp with a rather High gain.
I got the idea for this type of Amp topology from the
opa627
Note that on this Op-Amp the reference is returned to the sources (As you sugest) as opposed to ground and since this amp is at present on a bread board i can change to this easy. I will do this and let you know how it works out. thanks for allowing me to revisit this idea, I thought about doing it at first but decided against it however am now re looking into this aspect.
Pingrs said:
Hi Brian, i have a few comments...
It looks like you doubled the vas current AND increased voltage supply, aren't the vas transistors getting too hot ? I think it will be necessary to select some nice to126 transistors here. (I prefer to keep to92 below ~150mw, <60°C)
I think the 3 transistor current mirror should do the job...
What is q11 ? Limiting device ? With the diodes in the inputdiffamp it shouldn't be necessary ?
You removed a lot of open loop gain, this might give reduced slew rate. (by opening up 2nd diffamp)
I believe, one of the main reasons why symasym sounds so good, is because of the giant gain in vas (~5000), keeping work of inputdiffamp at absolute minimum. (keeping it in linear operation)
The cascode in input diffamp should be referenced to the common emitters, not to ground. (An error from me in symasym6)
About the scope shots, what is actual slewrate ? I think it shouldn't get below 20v/us.
What are THD values at 1khz/20khz, high order harmonics ? I typically simulate with 20khz,12v peak into 4ohm.
If you continue your good work, you will have a symasym with PSRR >90db over the whole audioband.
Padamiecki, welcome back !
Mike
pinkmouse said:
I got oscilations, and the diagnose was because of use slow BD243/244 instead of MJE 15030/31.
Padamiecki, that would have been too easy... 🙁
To be sure if it is caused by the output devices, you should install emitter resistors in the diffamp, like 47ohms or even 100, this lowers the openloop gain.
Do you have a scope ? It would be helpful to know magnitude and frequency of the oscillatons. A normal oscillation should be 4-7mhz, any frequency higher than that would be local oscillation in the outputstage.
Also check all resistor and caps, maybe you got some cold solder or wrong value somewhere. Especially check feedback network.
Mike
To be sure if it is caused by the output devices, you should install emitter resistors in the diffamp, like 47ohms or even 100, this lowers the openloop gain.
Do you have a scope ? It would be helpful to know magnitude and frequency of the oscillatons. A normal oscillation should be 4-7mhz, any frequency higher than that would be local oscillation in the outputstage.
Also check all resistor and caps, maybe you got some cold solder or wrong value somewhere. Especially check feedback network.
Mike
Mike,
Many thanks for taking the time to offer your expert comments on my attempt.
“It looks like you doubled the vas current AND increased voltage supply, aren't the vas transistors getting too hot ?”
Well, I immediately dragged it out and connected to a sig. gen. to clipping on no load. No sign of anything getting warm apart from the drivers, of course. I then ran at 15Vrms into 8R (my heatsink limit). Again, fine.
I do remember being aware of the effect those resistors would have on the Vas transistors, and my albeit very limited spice knowledge seemed to suggest it was ok.
Easy to lower them, though, no?
“What is q11 ? Limiting device ? With the diodes in the input diffamp it shouldn't be necessary ?”
This is a protection feature as described by Leach. I’ve also seen it applied to a JLH mosfet amp from the early eighties (at least, I think it was for protection). In my ignorance, I simply added it along with the SOAR protection. Relevant Leach section attached, to save you looking.
”You removed a lot of open loop gain, this might give reduced slew rate. (by opening up 2nd diffamp)”
Did I? My knowledge of the impact that opening up the 2nd diff. to provide the above protection is sadly lacking, i.e. forgive me, I don’t understand the mechanism, but the slew rate certainly went down compared to the Sym5.3.B. I knew I was right to make a compare reference!
”I believe, one of the main reasons why symasym sounds so good, is because of the giant gain in vas (~5000), keeping work of input diffamp at absolute minimum. (keeping it in linear operation).”
Honestly, I’ll do whatever is needed to keep it that way!
”The cascode in input diffamp should be referenced to the common emitters, not to ground. (An error from me in symasym6).”
I obviously wasn’t aware that this was an error. I assumed rather that this was an improvement on the zener reference used as in, for example, the Holton amp, (see attachment) which also references ground, insofar as the boostrap gave a kind of “constant current” feature to this resistor string? thus avoiding zener noise. Is your recommendation a result of your “riddle” post some time ago, because I was under the impression that the massive improvement in distortion was merely the result of spice anomalies. Guess I got that wrong.
”About the scope shots, what is actual slewrate ? I think it shouldn't get below 20v/us.”
It certainly looks lower than that, but, to be honest, the intention of that ‘scope view was to show the clean slope, with no h.f. nasties or kinks that may not be visible at a lesser magnification. I’ll leave the figure out in future,’cos I don’t believe it
I’ve read, though, that the slew rate can’t be gleaned from a benign power level such as this; it should be running close to clipping into a load to see the true figure. Is that true? Anyway, the slew rate tends to vary depending on the square wave generator I use. The three, one of which has 50R or 600R o/p impedance; all give different results. The 100kHz square wave, though, looks quite respectable to me. Better than other amps I’ve built, including the Self and the Leach.
”What are THD values at 1khz/20khz, high order harmonics ? I typically simulate with 20khz,12v peak into 4ohm.”
What do you mean by “simulate”?
Using THD equipment, I measure the residual of 1kHz at –89dB, and 10kHz at-81dB at no load. My oscillator isn’t up to much at 20kHz. I’ll post those shots when I can find them. I’ve taken some X-Y readings, too. To me, the phase response implied by them looks good.
As for higher order harmonics, this forum has made me aware of their crucial importance, but I haven’t got the equipment to measure them.
“If you continue your good work, you will have a symasym with PSRR >90db over the whole audioband.”
Thank you, with your invaluable input, I hope so.
Regards,
Brian.
Many thanks for taking the time to offer your expert comments on my attempt.
“It looks like you doubled the vas current AND increased voltage supply, aren't the vas transistors getting too hot ?”
Well, I immediately dragged it out and connected to a sig. gen. to clipping on no load. No sign of anything getting warm apart from the drivers, of course. I then ran at 15Vrms into 8R (my heatsink limit). Again, fine.
I do remember being aware of the effect those resistors would have on the Vas transistors, and my albeit very limited spice knowledge seemed to suggest it was ok.
Easy to lower them, though, no?
“What is q11 ? Limiting device ? With the diodes in the input diffamp it shouldn't be necessary ?”
This is a protection feature as described by Leach. I’ve also seen it applied to a JLH mosfet amp from the early eighties (at least, I think it was for protection). In my ignorance, I simply added it along with the SOAR protection. Relevant Leach section attached, to save you looking.
”You removed a lot of open loop gain, this might give reduced slew rate. (by opening up 2nd diffamp)”
Did I? My knowledge of the impact that opening up the 2nd diff. to provide the above protection is sadly lacking, i.e. forgive me, I don’t understand the mechanism, but the slew rate certainly went down compared to the Sym5.3.B. I knew I was right to make a compare reference!
”I believe, one of the main reasons why symasym sounds so good, is because of the giant gain in vas (~5000), keeping work of input diffamp at absolute minimum. (keeping it in linear operation).”
Honestly, I’ll do whatever is needed to keep it that way!
”The cascode in input diffamp should be referenced to the common emitters, not to ground. (An error from me in symasym6).”
I obviously wasn’t aware that this was an error. I assumed rather that this was an improvement on the zener reference used as in, for example, the Holton amp, (see attachment) which also references ground, insofar as the boostrap gave a kind of “constant current” feature to this resistor string? thus avoiding zener noise. Is your recommendation a result of your “riddle” post some time ago, because I was under the impression that the massive improvement in distortion was merely the result of spice anomalies. Guess I got that wrong.
”About the scope shots, what is actual slewrate ? I think it shouldn't get below 20v/us.”
It certainly looks lower than that, but, to be honest, the intention of that ‘scope view was to show the clean slope, with no h.f. nasties or kinks that may not be visible at a lesser magnification. I’ll leave the figure out in future,’cos I don’t believe it
I’ve read, though, that the slew rate can’t be gleaned from a benign power level such as this; it should be running close to clipping into a load to see the true figure. Is that true? Anyway, the slew rate tends to vary depending on the square wave generator I use. The three, one of which has 50R or 600R o/p impedance; all give different results. The 100kHz square wave, though, looks quite respectable to me. Better than other amps I’ve built, including the Self and the Leach.
”What are THD values at 1khz/20khz, high order harmonics ? I typically simulate with 20khz,12v peak into 4ohm.”
What do you mean by “simulate”?
Using THD equipment, I measure the residual of 1kHz at –89dB, and 10kHz at-81dB at no load. My oscillator isn’t up to much at 20kHz. I’ll post those shots when I can find them. I’ve taken some X-Y readings, too. To me, the phase response implied by them looks good.
As for higher order harmonics, this forum has made me aware of their crucial importance, but I haven’t got the equipment to measure them.
“If you continue your good work, you will have a symasym with PSRR >90db over the whole audioband.”
Thank you, with your invaluable input, I hope so.
Regards,
Brian.
Hi Brian,
About Slewrate, you can see slewratelimit at low powerlevels, where the slew starts to form a straight line, it has reached the slewratelimit. Of course it's easier to see at higher output levels.
Slewratelimit is not unimportant, otherwise you get TIM when trebles are played at higher levels.
About the heat up of Q5/6/..., this already happens when amp is idle, no need to torture the amp for this check. They shouldn't be that hot that they become uncomfortable to touch.
The limiting device (q11), skip it ! It does not work like that in this topology, it would only protect the positive clipping. Revert the 2nd stage back to diffamp. The 2 diodes are protection enough. These diodes reduce the maxswing at 1st diffamp output to +/- 0.3v (from +/- 1v) Still enough to saturate 2nd stage.
Even without the diodes q11 would never activate. (max voltage from r6 is 2v, minus 0.6v gives 1.4v through the 69ohms, divided over the 22ohms you get vbe of 0.32v to q11, not enough to open it)
The casode in input diffamp is result from this riddle thread, it's not spice anomaly. With wrong cascoding you can explode the miller effect.
Ideally, the input devices have constant Vce and current, then they cease to distort. With the bootstrapped cascode Vce gets nearly constant, the current was already nearly constant. (because of high gain in 2nd stage)
The 3bjt current mirror like i used in the riddle thing should be more than adequate.
Mesuring the harmonics is important tool, the THD value only is useless. You can have an amp with -60db 2nd harmonic sounding great, but having a 7th harmonic at -80db is a desaster...
For measurings at 1khz, a decent soundcard works very good.
Last but not least, listen to the amps... 😉 Measurings can only proove proper operation.
Mike
About Slewrate, you can see slewratelimit at low powerlevels, where the slew starts to form a straight line, it has reached the slewratelimit. Of course it's easier to see at higher output levels.
Slewratelimit is not unimportant, otherwise you get TIM when trebles are played at higher levels.
About the heat up of Q5/6/..., this already happens when amp is idle, no need to torture the amp for this check. They shouldn't be that hot that they become uncomfortable to touch.
The limiting device (q11), skip it ! It does not work like that in this topology, it would only protect the positive clipping. Revert the 2nd stage back to diffamp. The 2 diodes are protection enough. These diodes reduce the maxswing at 1st diffamp output to +/- 0.3v (from +/- 1v) Still enough to saturate 2nd stage.
Even without the diodes q11 would never activate. (max voltage from r6 is 2v, minus 0.6v gives 1.4v through the 69ohms, divided over the 22ohms you get vbe of 0.32v to q11, not enough to open it)
The casode in input diffamp is result from this riddle thread, it's not spice anomaly. With wrong cascoding you can explode the miller effect.
Ideally, the input devices have constant Vce and current, then they cease to distort. With the bootstrapped cascode Vce gets nearly constant, the current was already nearly constant. (because of high gain in 2nd stage)
The 3bjt current mirror like i used in the riddle thing should be more than adequate.
Mesuring the harmonics is important tool, the THD value only is useless. You can have an amp with -60db 2nd harmonic sounding great, but having a 7th harmonic at -80db is a desaster...
For measurings at 1khz, a decent soundcard works very good.
Last but not least, listen to the amps... 😉 Measurings can only proove proper operation.
Mike
my Symasym works!!!
Hi!
The reason of oscillations of my Symasym was mismathing R29 and R30 on the board. This caused huge gnfb and oscillations.
Now my amp works fine with FJR4315/4215 as output devices and with 165mA idle current.
C14 = 3,3 pF (unfortunately ceramic), C3,4 = 220 pF.
Rest of the transistors are BC550C, BC546B, 2N5400/5550, MJE15030/31.
Offset is equal in both channels -3.3mV
Cheers to Michael
and
PS. Probably I will never make new amp after more than 3 beers!
Hi!
The reason of oscillations of my Symasym was mismathing R29 and R30 on the board. This caused huge gnfb and oscillations.
Now my amp works fine with FJR4315/4215 as output devices and with 165mA idle current.
C14 = 3,3 pF (unfortunately ceramic), C3,4 = 220 pF.
Rest of the transistors are BC550C, BC546B, 2N5400/5550, MJE15030/31.
Offset is equal in both channels -3.3mV
Cheers to Michael
and
PS. Probably I will never make new amp after more than 3 beers!