There are few areas that need attention with CFB amplifiers. Keeping them stable is not easy... I suppose some experience with RF designs could help. Apart from outstanding sound in I/V application, they are quite educational in a way that everything has to be almost perfect to get them perform adequately.... and that is a challenge. Latter, VFB amplifier can be dropped in if that is preferable; they will also benefit from optimised “surrounding environment”... but, once you try CFB you will not go back to VFB... and very son you will realise that various OP amplifier will not be considered any more as devices that will colour the sound, but as a silicon gain stages…
My suggestion is to start with LME(can’t remember the number…) CFB Op, then move to 844 and finish with 811. Each step forward will require additional effort in circuit design optimisation to keep them stable. If you can get 811 to run stable -> you’ll know you’ve done (a fair few) things right.
Boky
My suggestion is to start with LME(can’t remember the number…) CFB Op, then move to 844 and finish with 811. Each step forward will require additional effort in circuit design optimisation to keep them stable. If you can get 811 to run stable -> you’ll know you’ve done (a fair few) things right.
Boky
Thanks for the info.
On a side note I guess some have taken offense with my comments on the first post that says: "Folks have resorted to high distortion high bjt count open loop transimpedance I/V that really never caught on."
Now About the best you can do with these circuits is -80db distortion from what I can tell the first BJT is -90db , maybe you can cancel some even harmonics subsequently. But I run a tube headamp after the DAC (which is about -70 db at my listening levels), Just seems that starting at -80db isn't ideal for my setup.
I guess in the past consumers didn't "like" the sound of CFB opamps in thier DAC's, but for feeding probably the most linear singleended feedback free single active device amplifier possible they may prove to be a good match.
Everyone has a different situation, I try to keep that in mind.
Also it looks like I have a line on a JT no3 PCB
On a side note I guess some have taken offense with my comments on the first post that says: "Folks have resorted to high distortion high bjt count open loop transimpedance I/V that really never caught on."
Now About the best you can do with these circuits is -80db distortion from what I can tell the first BJT is -90db , maybe you can cancel some even harmonics subsequently. But I run a tube headamp after the DAC (which is about -70 db at my listening levels), Just seems that starting at -80db isn't ideal for my setup.
I guess in the past consumers didn't "like" the sound of CFB opamps in thier DAC's, but for feeding probably the most linear singleended feedback free single active device amplifier possible they may prove to be a good match.
Everyone has a different situation, I try to keep that in mind.
Also it looks like I have a line on a JT no3 PCB
I think that theory actually supports your ears in this case. Two independent DAC outputs, after conversion to voltage, will produce a net 3dB improvement in SNR whether configured differentially (serially) or in parallel. In the differential config., the signal will be doubled (+6dB), but the noise will also increase by 3dB for a net change of +3dB. In the parallel config. the signal remains the same (0dB), but the noise is reduced by -3dB giving a net SNR change of +3dB.
Okay, so then, how might theory support the difference you hear? I'd argue that it has to do with the way harmonic distortion is handled in each configuration. Differentially, distortion will be correlated with itself, much the way the desired signal voltage having a 100% positive correlation with itself produces a doubling in signal amplitude. With harmonic distortion, however, the even order products will have a negative correlation and tend to cancel each other. A fact which helped popularize push-pull and symmetrical amplifier topologies. The odd order components wil have some degree of positive correlation and will tend to reinforce each other.
In the parallel DAC configuration, although distortion is correlated, the even order harmonics are not cancelled but neither are the odd order harmonics reinforced. Remember how the signal voltage itself does not increase in this configuration? So, I'll speculate
that you are experiencing an affinity for even order distortion products over odd order.Certainly, however, there are many other system variables which could have accounted for your listening preference, depending on whether those variables were kept constant in your comparative evaluations.
Last edited:
The even order products should be out of phase because the DACs are out of phase in the differential configuration. This is akin to what occurs with push-pull and complementary symmetry amplifier topologies.
Well, the ones that make those harmonics... Fhe fundamental signal. I don't see why the harmonics won't be reversed too, if the base signal is reversed. Especially why some (odd order) of the harmonics would be in phase and the other (even order) would be out of phase.
They will be all in reverse phase with respect with each other, because they follow the fundamental signal.
Remember, this is not statistic noise!
They will be all in reverse phase with respect with each other, because they follow the fundamental signal.
Remember, this is not statistic noise!
Last edited:
I think that theory actually supports your ears in this case. Two independent DAC outputs, after conversion to voltage, will produce a net 3dB improvement in SNR whether configured differentially (serially) or in parallel. In the differential config., the signal will be doubled (+6dB), but the noise will also increase by 3dB for a net change of +3dB. In the parallel config. the signal remains the same (0dB), but the noise is reduced by -3dB giving a net SNR change of +3dB.
Okay, so then, how might theory support the difference you hear? I'd argue that it has to do with the way harmonic distortion is handled in each configuration. Differentially, distortion will be correlated with itself, much the way the desired signal voltage having a 100% positive correlation with itself produces a doubling in signal amplitude. With harmonic distortion, however, the even order products will have a negative correlation and tend to cancel each other. A fact which helped popularize push-pull and symmetrical amplifier topologies. The odd order components wil have some degree of positive correlation and will tend to reinforce each other.
In the parallel DAC configuration, although distortion is correlated, the even order harmonics are not cancelled but neither are the odd order harmonics reinforced. Remember how the signal voltage itself does not increase in this configuration? So, I'll speculate
Certainly, however, there are many other system variables which could have accounted for your listening preference, depending on whether those variables were kept constant in your comparative evaluations.
I don't agree with this, the last BB design with these DAC's, the PCM1702 reference board has 4 PCM DAC's in parallel per channel, almost all high end designs parallel pcm1704's, the minimal increase in linearity is certainly not audible (we aren't talking about tda1543s here.)
It is obvious that the increased current output helps the output stage performance. The BB engineers knew that the low current combined with low output impedance current source was a bottleneck only helped by using more chips.
Even TI app notes can be misleading...
He writes that:
Inverting the input signal will NOT change the spectrum of output distortion, like that paper wants to imply. A square wave will remain a square when inverted, it will not change into a triangle.
How about using the correct formula instead:
He writes that:
Really? That means: Vout- = -k1(Vin) + k2(Vin)^2 - k3(Vin)^3 + . . . Obvious wrong!
Inverting the input signal will NOT change the spectrum of output distortion, like that paper wants to imply. A square wave will remain a square when inverted, it will not change into a triangle.
How about using the correct formula instead:
Last edited:
They are talking after the DAC (k1,k2) are the harmonics of the opamp not the DAC. An example would be a push pull stage could be used would would cancel 2H of the amplifiying device not the intrinsic 2H of each DAC chip ( agreeing with you). And I agree that differential summing is only going to give CMMR of the original DAC chip output, it helps if your powersupplies aren't perferct or your opamps have significant 2H.
Its not as big of an impact in improved performance as paralleling the DAC's, because as a current source these DAC's are hard to design around.
Its not as big of an impact in improved performance as paralleling the DAC's, because as a current source these DAC's are hard to design around.
yes, the doc refers to difefrential OP amplifier....two identical transistors (ideal situation!) that carry same quiescent currents, will completely cancel second harmonic distortion…. (generated for example by emitter-base diode)
Boky
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.