Hi guys,
After a while I've been dabbling around here again and have been picking up certain things again 🙂
Something of interest I heard about how it generally is the 1st Watt power level that's pretty important, as that's the power level what you'll be listening to on most occasions (apart from your happy hours where the sky (and neighbours) are the limit).
So I've been running some plots and came up with this hi-res graph from LT spice which you can see at the bottom.
No. It is not an opamp. Hmm.. what kind of amp could do this. Note, it's a simulation and realworld prototyping comes with its own limits like layouting etc 🙂
Anyways, you can add your own 1K/1W graphs here and perhaps with a little description about the amp, so we could look at and compare harmonic profiles. If you have actual listening experience, you might want to add that too - maybe we can correlate certain experience to types of harmonic content and at what dB level things seem to become moot.
The 1KHz / 1W spectrum might be an interesting benchmark to get some data to compare against the "sound" of the schematic / amp.
After a while I've been dabbling around here again and have been picking up certain things again 🙂
Something of interest I heard about how it generally is the 1st Watt power level that's pretty important, as that's the power level what you'll be listening to on most occasions (apart from your happy hours where the sky (and neighbours) are the limit).
So I've been running some plots and came up with this hi-res graph from LT spice which you can see at the bottom.
No. It is not an opamp. Hmm.. what kind of amp could do this. Note, it's a simulation and realworld prototyping comes with its own limits like layouting etc 🙂
Anyways, you can add your own 1K/1W graphs here and perhaps with a little description about the amp, so we could look at and compare harmonic profiles. If you have actual listening experience, you might want to add that too - maybe we can correlate certain experience to types of harmonic content and at what dB level things seem to become moot.
The 1KHz / 1W spectrum might be an interesting benchmark to get some data to compare against the "sound" of the schematic / amp.
Attachments
That is nice !
I can't seem to get my noise floor quite as low as yours. 😱
And , I rarely test at 1 W !
But , here is the new CFA with the crazy Baxandall/Hawksford VAS.
(below)
Both my present CFA's seem to be preferred ,sound wise.
But some prefer the VFA's ??
The "lines" have blurred ... all my amps have similar harmonic
structures and are sub-ppm at 1w .
PS - on one VFA ... simulated at 5-8ppm , a member built it and
actually tested it - he read 3-5ppm ... 20k!
OS
I can't seem to get my noise floor quite as low as yours. 😱
And , I rarely test at 1 W !
But , here is the new CFA with the crazy Baxandall/Hawksford VAS.
(below)
Both my present CFA's seem to be preferred ,sound wise.
But some prefer the VFA's ??
The "lines" have blurred ... all my amps have similar harmonic
structures and are sub-ppm at 1w .
PS - on one VFA ... simulated at 5-8ppm , a member built it and
actually tested it - he read 3-5ppm ... 20k!
OS
Attachments
Last edited:
Which are the start conditions for simulating?
1. Which load at output?
2. Whit or whitout input capacitor?
3. Which type of FFT analysis (visual comparisation).
....?
1. Which load at output?
2. Whit or whitout input capacitor?
3. Which type of FFT analysis (visual comparisation).
....?
That's nice! You note the downward slope in your graph? Unhook the DCservo output and it becomes flat and you'll see the harmonics better 🙂
I can't say I have listened to my schematic yet but I am working towards building it (designing supply unit and case).
At every power level, even and odds have an even distribution, this may sound a little harsher if one could actually hear the harmonics contribution. I think at 80db, harmonic content may still make an audible difference.
Last edited:
1) Resistive 8Ohm or 4 OHm, as long as you attune the input signal to produce 1W of power in your chosen load. It's 4V peak-peak into 8Ohm.
2) Signal generator connected to the amp input, no capacitor.
3) LTspice standard settings, no windowing function.
To obtain an accurate graph, make sure your timestep setting is 8192 steps per cycle. Simulate 32 cycles. This yields 256K data points, which is the datapoints default for the FFT.
Here are the settings to run the required transient analysis:
PS: startup cycles before recording simulation.
PC: count of measurement cycles
FFT: The number of points per cycle.
Code:
.param Freq 1KHz
.param FFT 8192
.param PS 2
.param PC 32
.param TStart (PS)/Freq
.param TEnd (PS+PC)/Freq
.param TStep 1/(FFT*Freq)
.tran {TStep} {TEnd} {TStart} {TStep}PC: count of measurement cycles
FFT: The number of points per cycle.
Thank you , now I can see the 5th.
Better than expected ... but being a symmetrical I knew the even order
would be fully cancelled (no 4th at all). 🙂
edit - below 2 , NOW I can see that 4th (and the EF3 X-over distortion) ... more "juice".
OS
Attachments
Last edited:
Looks good, the x-over sits at 150dB+, the 1st even harmonic @ ~110dB if I read the graph correctly (the one with the full-power signal) 🙂
You'd say that at full power, both evens and odds are distributed about equally.
You'd say that at full power, both evens and odds are distributed about equally.
Edit: To put in some listening experience, I can definitely correlate cross-over distortion to listening fatigue. There's also an immediate recognition: 'S' and 'T' letters are much sharper and more pronounced in voice. Basically any pronounciation that generates noise (air through lips, tongue etc rather than voice muscles) is sharper and much more pronounced. This is repeatably testable. Once you know what to listen for you'll know it instantly.
(p.s. couldn't edit the previous post any more)
(p.s. couldn't edit the previous post any more)
Last edited:
I have this 2 watt into 4 ohm load 1Khz FFT data on my computer.
FFT: Prototype 2 watts into 4 ohm load at 1Khz. and 200 watts into 4 ohms at 20Khz.
Symmetrical bipolar input with cascode + current mirror. Hawksford VAS. TripleEF output with 8 bipolar pairs.
FFT: Ayre MX-R 2 watts into 4 ohm load at 1Khz. Stereophile data.
Symmetrical JFET input with JFET self-bias cascode. No Global feedback.
FFT: Prototype 2 watts into 4 ohm load at 1Khz. and 200 watts into 4 ohms at 20Khz.
Symmetrical bipolar input with cascode + current mirror. Hawksford VAS. TripleEF output with 8 bipolar pairs.
FFT: Ayre MX-R 2 watts into 4 ohm load at 1Khz. Stereophile data.
Symmetrical JFET input with JFET self-bias cascode. No Global feedback.
Attachments
Thank you. Much nicer FFT than with the parameters I had been using.
Attachments
Impressive noise baseline you have there! I wonder what attributes to that, given that the plots so far all seem to have a baseline surrounding -180dB. I'm guessing your OPS at that level stays in class A fully; and from the few plots I've now seen, I'm beginning to think BJTs do better than MOSFETs with respect to harmonic distribution and amplitude. So here's hoping to see a few more but with MOSFET OPS' (the one I posted uses MOSFETs, though still class A at the 1st watt).
The Ayre looks as if it will color the sound noticably. Anyone (listening)experience with that amp?
"spooky" w/ EC class A OPS
Using a simple "moskido"class A with the common
"spooky amp" (new leach symmetrical IPS).
Parts per billion.
Before the expressions - I just assumed there were no harmonics. 😱
Class A OPS uses "dirty" ol' Vertical MOSFETS with simple EC at the driver stage.
OS
Using a simple "moskido"class A with the common
"spooky amp" (new leach symmetrical IPS).
Parts per billion.
Before the expressions - I just assumed there were no harmonics. 😱
Class A OPS uses "dirty" ol' Vertical MOSFETS with simple EC at the driver stage.
OS
Attachments
- Status
- Not open for further replies.