Hi Thimios,
Yes problem solved!
I installed the extra caps on the OPS board and increased the driver Re.
No more heating up or any sign of odd behaviour.
Now I´ll wait what will be the next spook version and build it.
This is the best thread for a long time in Diyaudio. You guys got me building even though I have many other diy amps which are basically all good enough.
Keep up the good work.
J
OK I was just taking some more measurements and I noticed that there are two 220R resistors added between Q6-Q7 and Q9-Q10. That may explain why the VAS seems to be running so hot and the output is so high. I guess no one is looking over the board pics I have been posting. I will try to add those and see how things look.
I don't really see why the H2/H3 balance matters when the distortion is so low that it is either A) buried within the noise of the system. B) of frequencies higher than we can hear or C) masked within the significantly higher distortion inherent to the rest of the system.
The ultra low distortion of these amplifiers is no joke and for it to become dominant you really need a signal source of the utmost quality using state of the art D/A converters. Nothing analogue comes close.
I do not see the distortion performance of the IPS of any of these amplifiers (providing they are correctly implemented) contributing to the overall sound of the end system. I would say it's more how the differences in architechture allow the amplifier to respond to the load they are connected to and to the systems they are implemented into (and by this I mean the hifi system as a whole and the amplifier system meaning case/psu/grounding/layout etc).
Edit - And yes Thimios your results do seem somewhat limited by the limitations of your sound card.
The ultra low distortion of these amplifiers is no joke and for it to become dominant you really need a signal source of the utmost quality using state of the art D/A converters. Nothing analogue comes close.
I do not see the distortion performance of the IPS of any of these amplifiers (providing they are correctly implemented) contributing to the overall sound of the end system. I would say it's more how the differences in architechture allow the amplifier to respond to the load they are connected to and to the systems they are implemented into (and by this I mean the hifi system as a whole and the amplifier system meaning case/psu/grounding/layout etc).
Edit - And yes Thimios your results do seem somewhat limited by the limitations of your sound card.
Correct on PD and ND output voltages. NFB to ground is -.450V, Bases of Q1,Q2, both .5V base to ground. R3 is 10R and has only a couple mV across it.
.489V across R8 and R9. Via is good beneath R10. Solid ground from star to R3. D1, D2.
If I adjust the pot to get 0 offset the NFB changes to 30mv and 1.02V on the NFB resistors.
Maybe I need to try and match the VAS transistors.
Terry, please check Q3 too, looks like collector current is higher than emitter current.
OS -> R114/115 + C110 -113 are the main driver/pre decoupling RC. Raising these
values to even 10R/68uf are easy and give 3-6db PSSR "out of the box".
R114, R115 (Going with 10R here ) 4R7 1WATT 10R CPF210R000BHE14 <-- LTspice best PSRR using 15R 47uF combo … have selected 10R / 68uf
C110, C112 (Does this value require adjustment cause of the 68uF bump up? ) .02uF .022uF MKP1837322161 <-- IS this bypass for the lytic below?
C111, C113 (Going with 68uF here ) 47uF 68uF 100ZLJ68MT78X20 <-- Raising these values to even 10R/68uf are easy and give 3-6db PSSR "out of the box".
C102, C105 22uF 22uF UPM2A220MPD-ND
Q101 ( Capmulti + More FT and greater HFE + point…might see full rails ie potential +/- NPN ) mje340 2SC5171
Q102 ( Capmulti + More FT and greater HFE + point…might see full rails ie potential +/- PNP ) mje350 2SA1930
C103, C106 (18mm 7.5 or 5.0 lead spacing as much UF as possible HFE X uF = Virtual C) 470uF 100YXG820MEFC18X40 Going with a 820uF Cap here.
C101, C104 ( Does this value need adjustment based on 820uF C103/106 ? ) .02u MKP1837322161 <-- IS .022UF OK ?
R109, R101 470R PRP-70813
R108, R102 22k PRP-50038
C114 .68uF B32671P4684K000
C108 1uF MKS2D041001K00JSSD
C107, C109 33p CD10ED330JO3F
Q103, Q104, Q105 qmje340 2SC4793
Q106 qmje350 2SA1837
R103 1.8k CMF601K8000BEBF
R104, R105 1.2k PRP-53808 1/4 watt PRP Non-Magnetic Resistors
R106 > 500R Trimmer 199R 3296W-1-501LF
R107 470R PRP-70813 1/4 watt PRP Non-Magnetic Resistors
R110 1k PRP-70167 1/4 watt PRP Non-Magnetic Resistors
R111, R112 22R PRP-50001 1/4 watt PRP Non-Magnetic Resistors
R113 68R CPF368R000GKB14 3 Watt 100PPM
R119, R116, R123, R127, R131, R135, R120, R124, R128, R132 4R7 MBA02040C4708FC100
R118, R117, R121, R122, R125, R126, R129, R130, R133, R134 R22 5 WATT MUNDORF-77129
R136 10R PRP-70058 PRP 1 Watt Resistor
R137 10R 3 WATT MP820-10.0-1% Caddock Thickfilm
C125, C127 470uF 100ZLH470MEFC16X31.5 Rubycon ZLH Series
C115, C117, C119, C121, C116, C118, C120, C122, C123, C122, C121, C124 22uF UPM2A220MPD-ND <-- According to Keantoken avoid film because of possible reasonance created by low ESR film cap and trace inductance.
C126, C129, C128 .1uF MKP1837410161F
L101 1u5
D101, D102 1N914 1N914 Fairchild Semi
D103, D106, D105, D104 MUR460 MUR460-E3/54 Vishay
Q108, Q110, Q112, Q114, Q116, Q118 qnjw0302g 2SA1186
Q107, Q109, Q111, Q113, Q115, Q117 qnjw0281g 2SC2837
values to even 10R/68uf are easy and give 3-6db PSSR "out of the box".
R114, R115 (Going with 10R here ) 4R7 1WATT 10R CPF210R000BHE14 <-- LTspice best PSRR using 15R 47uF combo … have selected 10R / 68uf
C110, C112 (Does this value require adjustment cause of the 68uF bump up? ) .02uF .022uF MKP1837322161 <-- IS this bypass for the lytic below?
C111, C113 (Going with 68uF here ) 47uF 68uF 100ZLJ68MT78X20 <-- Raising these values to even 10R/68uf are easy and give 3-6db PSSR "out of the box".
C102, C105 22uF 22uF UPM2A220MPD-ND
Q101 ( Capmulti + More FT and greater HFE + point…might see full rails ie potential +/- NPN ) mje340 2SC5171
Q102 ( Capmulti + More FT and greater HFE + point…might see full rails ie potential +/- PNP ) mje350 2SA1930
C103, C106 (18mm 7.5 or 5.0 lead spacing as much UF as possible HFE X uF = Virtual C) 470uF 100YXG820MEFC18X40 Going with a 820uF Cap here.
C101, C104 ( Does this value need adjustment based on 820uF C103/106 ? ) .02u MKP1837322161 <-- IS .022UF OK ?
R109, R101 470R PRP-70813
R108, R102 22k PRP-50038
C114 .68uF B32671P4684K000
C108 1uF MKS2D041001K00JSSD
C107, C109 33p CD10ED330JO3F
Q103, Q104, Q105 qmje340 2SC4793
Q106 qmje350 2SA1837
R103 1.8k CMF601K8000BEBF
R104, R105 1.2k PRP-53808 1/4 watt PRP Non-Magnetic Resistors
R106 > 500R Trimmer 199R 3296W-1-501LF
R107 470R PRP-70813 1/4 watt PRP Non-Magnetic Resistors
R110 1k PRP-70167 1/4 watt PRP Non-Magnetic Resistors
R111, R112 22R PRP-50001 1/4 watt PRP Non-Magnetic Resistors
R113 68R CPF368R000GKB14 3 Watt 100PPM
R119, R116, R123, R127, R131, R135, R120, R124, R128, R132 4R7 MBA02040C4708FC100
R118, R117, R121, R122, R125, R126, R129, R130, R133, R134 R22 5 WATT MUNDORF-77129
R136 10R PRP-70058 PRP 1 Watt Resistor
R137 10R 3 WATT MP820-10.0-1% Caddock Thickfilm
C125, C127 470uF 100ZLH470MEFC16X31.5 Rubycon ZLH Series
C115, C117, C119, C121, C116, C118, C120, C122, C123, C122, C121, C124 22uF UPM2A220MPD-ND <-- According to Keantoken avoid film because of possible reasonance created by low ESR film cap and trace inductance.
C126, C129, C128 .1uF MKP1837410161F
L101 1u5
D101, D102 1N914 1N914 Fairchild Semi
D103, D106, D105, D104 MUR460 MUR460-E3/54 Vishay
Q108, Q110, Q112, Q114, Q116, Q118 qnjw0302g 2SA1186
Q107, Q109, Q111, Q113, Q115, Q117 qnjw0281g 2SC2837
Hi BV,,
I checked R2, it is 10k. I did notice that during start up it goes from zero drop and gradually ramps up to about .489mv after about ten seconds. I don't know how to measure for beta. I did match hFE on Q1/Q2.
Yeah, I just replaced Q3 and Q4 with matched units thinking it might help. No change. I think the small variances in amps is due to resistor tolerances. I didn't measure the current. I measured voltage and then used ohms law based on the resistor value. It was meant more to be a visual aid.
Blessings, Terry
Member
Joined 2009
Paid Member
Yes indeed - the simulated distortion is indeed so tiny that it's not going to tell us what the as-built distortion is (pcb layout, real world components), or how the amplifier will interact with it's power supply, or with it's speaker loads.
But it's fun to see such super-low distortion be possible, in principle, with this topology and good engineering.
Well the low frequency distortion isn't too hard to arrive at once decent design practises have been followed.
The difficult part is getting what 'should' theoretically be possible at the high frequencies. This is mainly due to induction distortion in my opinion. Self mentions that probably all amplifiers suffer from this to at least some extent and this gets worse with rising frequency. Not only this but it gets worse the higher the power of the amplifier.
As you mentioned though you do need an amplifier of a certain performance level for this to show up at all and then to know that it's that distortion mechanism that's causing it in the first place.
Originally I was getting distortion numbers for 8 and 4 ohm loads that you would expect for an EF amplifier, at least what Self describes as an EFs typical performance. This was a decent amount worse than OS predicted, but measured results tend to differ from the perfect environment of the simulator. I wasn't too put out at the time, I just figured this was normal, but I did want it to be better. So I reread Self, figured out what it was most likely to be and had a play around with the wires. I managed to push the induction distortion down a decent amount, now we're talking 0.005% down to 0.001% @ 10kHz, it was already low to begin with, but this could have easily been masked in an amplifier with lower overall performance.
It's so easy to do too, you just set up the spectrum analyser and start moving wires around and see what happens, you just need the hardware with the capability to be able to measure it. I mean we can spend all the time optimising designs in CAD work, both in terms of PCB layout and component values, all to have some of the hard work reduced to naught by a piece of wire >.<
It makes me wonder when some people are comparing the sound of various amplifiers if their results are being dominated by wire instead of circuits as some misplaced connections can be really harmful to an amplifiers performance.
I suppose it's nice that these amplifiers give a decent amount of class A performance, especially the 4/5 pair versions, when optimally biased. If you're more concerned with the first watt or two, you're going to get simulator like performance for those first couple of watts at least.
This brings up another question though. Should, when comparing these amps subjectively, the listening be done such that the amplifiers are not operating in class A, or are? Having them in class A would level the playing field, but it's possible that some of the designs sound better than others when operating in class AB. OS mentioned that the inherent speed and wide bandwidth of the current feedback amp was better at compensating for nasties created by under biasing the output stage. It stands to reason perhaps that this ability might give better sonics when the amp leaves class A?
I guess you'd need to do listening sessions at low and high volume play back.
The difficult part is getting what 'should' theoretically be possible at the high frequencies. This is mainly due to induction distortion in my opinion. Self mentions that probably all amplifiers suffer from this to at least some extent and this gets worse with rising frequency. Not only this but it gets worse the higher the power of the amplifier.
As you mentioned though you do need an amplifier of a certain performance level for this to show up at all and then to know that it's that distortion mechanism that's causing it in the first place.
Originally I was getting distortion numbers for 8 and 4 ohm loads that you would expect for an EF amplifier, at least what Self describes as an EFs typical performance. This was a decent amount worse than OS predicted, but measured results tend to differ from the perfect environment of the simulator. I wasn't too put out at the time, I just figured this was normal, but I did want it to be better. So I reread Self, figured out what it was most likely to be and had a play around with the wires. I managed to push the induction distortion down a decent amount, now we're talking 0.005% down to 0.001% @ 10kHz, it was already low to begin with, but this could have easily been masked in an amplifier with lower overall performance.
It's so easy to do too, you just set up the spectrum analyser and start moving wires around and see what happens, you just need the hardware with the capability to be able to measure it. I mean we can spend all the time optimising designs in CAD work, both in terms of PCB layout and component values, all to have some of the hard work reduced to naught by a piece of wire >.<
It makes me wonder when some people are comparing the sound of various amplifiers if their results are being dominated by wire instead of circuits as some misplaced connections can be really harmful to an amplifiers performance.
I suppose it's nice that these amplifiers give a decent amount of class A performance, especially the 4/5 pair versions, when optimally biased. If you're more concerned with the first watt or two, you're going to get simulator like performance for those first couple of watts at least.
This brings up another question though. Should, when comparing these amps subjectively, the listening be done such that the amplifiers are not operating in class A, or are? Having them in class A would level the playing field, but it's possible that some of the designs sound better than others when operating in class AB. OS mentioned that the inherent speed and wide bandwidth of the current feedback amp was better at compensating for nasties created by under biasing the output stage. It stands to reason perhaps that this ability might give better sonics when the amp leaves class A?
I guess you'd need to do listening sessions at low and high volume play back.
"Unplugged" for 1-2 months..
I'm leaving , hope nobody gets shocked when I'm gone.
I do see these IPS's all converging into PPM, the only differences
being slew and overload behavior.
Hopefully , without the taxman stealin' all my dough ... I can
build again in the serenity of the mountains.
OS
I'm leaving , hope nobody gets shocked when I'm gone.
I do see these IPS's all converging into PPM, the only differences
being slew and overload behavior.
Hopefully , without the taxman stealin' all my dough ... I can
build again in the serenity of the mountains.
OS
