So looks like drop from 30,3db to 27,6 if i'm correct ,thats about 3db ,as author mentioned .What means Va in your diagram ?
If you look at the .asc file you will see that V(a) is the output before the inductor.
I just increasing the size of the output inductor untill I got a 3db drop with a 50uH
inductor normaly they are about 1uH and the output would be the same as V(a).
(input filter response)
I just increasing the size of the output inductor untill I got a 3db drop with a 50uH
inductor normaly they are about 1uH and the output would be the same as V(a).
(input filter response)
Attachments
Last edited:
I use different software for sim , it don't open .asc files .So -0,77db according to last picture is input filter response at the end .
Staring to recover - phew. Allow me to catch up. Mesanwhile, FYI, here's the simplified schematic of their previous design, which has some notable differences.
1. The 33 pf // 47K connects to the amplifier output, rather than just the cap into the driver offset / bias.
2. It doesnt have that second feedback from output into the upper driver transistor (C19//C20 to R41-R42)
Curious if these design choices - while maybe allowing the amplifier to better, more reliably, drive some transformer coupled constant voltage audio distribution system - are what's affecting its sound, to my ears. (like someone threw a towel over the speaker)
1. The 33 pf // 47K connects to the amplifier output, rather than just the cap into the driver offset / bias.
2. It doesnt have that second feedback from output into the upper driver transistor (C19//C20 to R41-R42)
Curious if these design choices - while maybe allowing the amplifier to better, more reliably, drive some transformer coupled constant voltage audio distribution system - are what's affecting its sound, to my ears. (like someone threw a towel over the speaker)
C19//C20 provide "bootstrap" drive to R41, making it behave as if it's a constant current source: note that C19//C20 are fairly large and will tend to have a fixed DC voltage across its terminals (roughly half the 60V supply voltage, about 30V) even though the output may be a large AC signal; similarly, the voltage at the lower terminal of R41 will be be around three diode drops above the output (~ 2.1V) and also tend to track the output signal. Thus, the voltage across R41 will be ~28VDC and the current through will be ~28V/2.4K, or about 12mA. Current source drive helps ensure the Q2-Q8 drivers have adequate drive, especially as output swing approaches the positive supply and R41 current would otherwise drop to near 0. It would seem that current in R34 design above might suffer this issue.
Another big difference is the lack of feedback resistor from output back to the Q1 emitter. I think gain from opamp to output will resemble an integrator and drop at 20dB/decade. I don't yet understand how open-loop gain is tailored/compensated to achieve stability. BTW, what type of opamp is Z3?
With R35 a bit smaller than R39, the bias spreader Q4 would have less than two base-emitter drops, not enough to bias the Q3-Q9 output pair into class A/B conduction. Maybe this design has some crossover distortion? Of course, the large negative feedback will help mitigate, but will be less effective at higher frequencies. Is it possible that distortion at higher frequencies could make the above design inordinately bright re the newer amp?
Very interesting design differences.
Clearly taken with a PC sound-card, which are near brickwall.
The jaggy is because it was taken with random hiss. Hiss is proper for large systems with multi-path: room acoustics. Hiss is not normally used for amplifiers because we know (usually) than an amplifier can only have simple errors. However room-acoustics software has become cheap/free. (I remember when it was a year's income.) Oddly, amplifier-specific sine-sweep software has also become free (RightMark RMAA is popular) but I guess it was not handy?
The jaggy is a distraction. I'd want to see the "loopback" to know how much bass and treble loss is in the sound card and how much more is due to amplifier. However I assume his "another SS amp" tells us what we need to know.
Cool, and that´s my point.
Something with such a strong lowpass filter may very well start dropping above 10kHz, and in that case, the dropping curve can very well be caused my the Soundcard (measuring instrument) itself and not be present on the amp per se.
To say in a blunter way: I am not certain the Amplifier has a frequency response problem, certainly not enough to turn it from silk purse to sow´s ear as described.
I felt ambitious and took the amplifier apart today so as to access the 6th card at the end of the chassis. Through hole components are so nice - I can actually just clip on probe leads, then start a sweep with both hands free! I changed to an actual frequency sweep, to avoid further confusion.
First, across C31;
So, this FR is whatever's upstream - and that's the part of the schema I left out, their balanced to unbalanced converter;
One would think it's simply C31's LP effect, but not so fast. Putting the probe on the J2, or Z1 pin 7;
You've...got to be kidding me. It's a simple circuit! Are the input beads, C2 having something to do with it? Let's try across R8;
C'mon. I'll never need balanced input, so why not just get rid of the whole thing? Make the input look like this;
To wit, after doing that - pulling the chip out of the (lucky me) socket, jumping a wire from the center post of J2 to pin 3 the whole amp now measures;
Well, that's something I can actually do across the six channels without pulling the boards / 5 devices off the heat sink for each. I'll do a second channel and have a listen soon. Here's what's involved;
Q: Why on earth would that simple, common, instrumentation op-amp design for a balanced to SE do that to the FR? When half of it works "fine"? I'm lost.
This isnt the end...C10 and C31 are accessible - with a little soldering iron skill. I think I could eliminate C6 via jump, cause there's always C7 right on down the line - and I'll never connect this amp to anything that does not already have a DC blocking cap on its outputs 🙂
First, across C31;
So, this FR is whatever's upstream - and that's the part of the schema I left out, their balanced to unbalanced converter;
One would think it's simply C31's LP effect, but not so fast. Putting the probe on the J2, or Z1 pin 7;
You've...got to be kidding me. It's a simple circuit! Are the input beads, C2 having something to do with it? Let's try across R8;
C'mon. I'll never need balanced input, so why not just get rid of the whole thing? Make the input look like this;
To wit, after doing that - pulling the chip out of the (lucky me) socket, jumping a wire from the center post of J2 to pin 3 the whole amp now measures;
Well, that's something I can actually do across the six channels without pulling the boards / 5 devices off the heat sink for each. I'll do a second channel and have a listen soon. Here's what's involved;
Q: Why on earth would that simple, common, instrumentation op-amp design for a balanced to SE do that to the FR? When half of it works "fine"? I'm lost.
This isnt the end...C10 and C31 are accessible - with a little soldering iron skill. I think I could eliminate C6 via jump, cause there's always C7 right on down the line - and I'll never connect this amp to anything that does not already have a DC blocking cap on its outputs 🙂
Are you sure the rolloff isn't from your test setup? Was the measurement done with a white noise input or with a sine sweep?
Tom
Tom
Juan, Tom: again, he claims to have run the same setup "using another SS amp which goes right out to 20k flat".
Yes, a steep digital cut-off "can" slope an octave down from the corner, except sound-card filter designers have been doing this for years. Even in 1997 I measured flat to 15kHz, +0.4/-1dB to 19kHz. On a non-audiophile card. (Sound Blaster 16 ISA) JJ's plot is closer to 3dB at 19kHz. On THIS amp, apparently not on "another".
The input L-C filter is interesting. But does not seem to explain it all.
Yes, a steep digital cut-off "can" slope an octave down from the corner, except sound-card filter designers have been doing this for years. Even in 1997 I measured flat to 15kHz, +0.4/-1dB to 19kHz. On a non-audiophile card. (Sound Blaster 16 ISA) JJ's plot is closer to 3dB at 19kHz. On THIS amp, apparently not on "another".
The input L-C filter is interesting. But does not seem to explain it all.
@jjasniew , what did you do with the negative side of the input? Short it to ground, leave it open or drive it in antiphase?
If you left it open, the input filter will cause some treble loss; the 330 pF will couple some of the signal from the positive to the negative input and the circuit then takes the difference.
If you left it open, the input filter will cause some treble loss; the 330 pF will couple some of the signal from the positive to the negative input and the circuit then takes the difference.
Interesting and thanks for all the detailed testing.
Like I obsessively say, guessing is fine (we do it all the time), simulation is a great tool, but actually testing/measuring rules.
IF anything, because it includes things which "are not in the schematic", such as grounding, layout, wiring, stray capacitance, etc.
Going deeper into the hole, can´t you lift C6´s right leg (as seen on schematic) and inject signal straight there?
So as to REALLY bypass the whole balanced input?
And resweep the power amp only?
Extra test: inject Music there ... is amp still "lifeless"?
Like I obsessively say, guessing is fine (we do it all the time), simulation is a great tool, but actually testing/measuring rules.
IF anything, because it includes things which "are not in the schematic", such as grounding, layout, wiring, stray capacitance, etc.
Going deeper into the hole, can´t you lift C6´s right leg (as seen on schematic) and inject signal straight there?
So as to REALLY bypass the whole balanced input?
And resweep the power amp only?
Extra test: inject Music there ... is amp still "lifeless"?
Even after pulling the opamp you're still down 1.5-2 dB at 20 kHz. I would have expected more bandwidth from the amp.
I would also have expected more bandwidth from the balanced input (which is worth retaining). It's a 14 MHz (typ.) opamp (10 MHz w/c).
If you decide to make this modification permanent, you might want to solder a wire onto the bottom of the board instead of relying on wires stuck into IC sockets.
Tom
I would also have expected more bandwidth from the balanced input (which is worth retaining). It's a 14 MHz (typ.) opamp (10 MHz w/c).
If you decide to make this modification permanent, you might want to solder a wire onto the bottom of the board instead of relying on wires stuck into IC sockets.
Tom
To be specific: leaving the negative input open theoretically causes a drop to -2.2734 dB at 20 kHz with respect to the low frequency transfer.
Your measurement with bypassed differential-to-single-ended converter shows that the rest drops to about -1.2 dB at 20 kHz, so when you add that to the -2.2734 dB, you are quite close to -3.5 dB.
If my hypothesis is correct, the solution is simply to short the negative input to ground, preferably at the source.
Last edited:
First check the frequency response of your measurement device in loop configuration. soundcard output to soundcard input
Here's the loopback FR of the USB interface I'm using;
With distortion;
I hope this shows that it's not the measurement system. I'm curious how the classic differential input op-amp circuit can do what it does, even when fed single ended? Remember, all 6 amplifiers do this. My solution is to remove the entire circuit from the signal path, which is easily done and the measured effect of doing so is shown in my post above.
My future post will hopefully relay "how it sounds" w/o this collection of op-amps in the signal chain. It would be interesting for curiosity's sake to know why they do what they do, but afaic, they're out. Particularly on finding their behavior makes no sense to me. One op-amp output (buffering my input signal connection) is fine; copy of my input signal FR wise. Output of the summing amplifier (with zero signal going into the other input) has this HF rolloff effect. Ciao, op-amps - dont need you and your interference.
With distortion;
I hope this shows that it's not the measurement system. I'm curious how the classic differential input op-amp circuit can do what it does, even when fed single ended? Remember, all 6 amplifiers do this. My solution is to remove the entire circuit from the signal path, which is easily done and the measured effect of doing so is shown in my post above.
My future post will hopefully relay "how it sounds" w/o this collection of op-amps in the signal chain. It would be interesting for curiosity's sake to know why they do what they do, but afaic, they're out. Particularly on finding their behavior makes no sense to me. One op-amp output (buffering my input signal connection) is fine; copy of my input signal FR wise. Output of the summing amplifier (with zero signal going into the other input) has this HF rolloff effect. Ciao, op-amps - dont need you and your interference.