Bruno Putzeys said:
I think you'll find that beyond a certain threshold output stage distortion becomes the limiting factor.....
....the small power sections of a linear power amp. can be routinely persuaded to produce near-nil THD in the audio band with less than a handfull of components...
Well again it's not power amps I'm talking about, and the thd figures I'm aiming at are in the -140dB region. Because of that, the discussion we've been having is far from futile.
For power amps I certainly agree that the power stage is dominant once sufficient care has been given to the small parts.
But then, a power amp is a device with a known fixed gain, so you can throw high orders of feedback at the power stage. Because of that, I don't think of teasing -140dB THD out of a linear power amp as a serious challenge.
Unfortunately, one simply can't do that with a circuit intended as a "general purpose op amp" simply because you don't know what sort of feedback will be put around it. One is forced to live with a simple 1st order system. Especially if the output stage has local error correction, the input/gain stage that forms the outer (dominant) loop is going to dominate distortion.
Well we're having fun aren't we?
For power amps I certainly agree that the power stage is dominant once sufficient care has been given to the small parts.
But then, a power amp is a device with a known fixed gain, so you can throw high orders of feedback at the power stage. Because of that, I don't think of teasing -140dB THD out of a linear power amp as a serious challenge.
Unfortunately, one simply can't do that with a circuit intended as a "general purpose op amp" simply because you don't know what sort of feedback will be put around it. One is forced to live with a simple 1st order system. Especially if the output stage has local error correction, the input/gain stage that forms the outer (dominant) loop is going to dominate distortion.
Well we're having fun aren't we?
If it's op. amp. circuitry you're interested in...particularly the discrete variety, then life is vastly simpler....
A single pole roll-off to unity gain frequency can be arranged with output stage-inclusive minor-loop compensation, which should reliably give THD below 10ppm across the audio band...at least..
For the more adventerous, output stage-inclusive double-pole minor-loop compensation, which gives zero steady-state error for a ramp input, is recommended...Again, THD should reside well below noise across the audio band...
A single pole roll-off to unity gain frequency can be arranged with output stage-inclusive minor-loop compensation, which should reliably give THD below 10ppm across the audio band...at least..
For the more adventerous, output stage-inclusive double-pole minor-loop compensation, which gives zero steady-state error for a ramp input, is recommended...Again, THD should reside well below noise across the audio band...
"I don't think of teasing -140dB THD out of a linear power amp as a serious challenge."
You mean that a THD of 0.00001% over the complete audioband
is not a challenge ? You are some god to me...
Of course with wide bandwidth (-3db at ~1mhz) and stable + reliable...
For me, it's not a challenge, it's religion ! (Maybe i am some newbie,
but you are talking here about some numbers, phew!)
Mike
You mean that a THD of 0.00001% over the complete audioband
is not a challenge ? You are some god to me...
Of course with wide bandwidth (-3db at ~1mhz) and stable + reliable...
For me, it's not a challenge, it's religion ! (Maybe i am some newbie,
but you are talking here about some numbers, phew!)
Mike
There seems to be some consensus on this thread that achieving low distortion levels in the “small Power sections” is simple – would anyone be willing to share some circuit diagrams – with results?
I ask, as I’m currently developing an analogue input Class D amplifiers where the final THD, Noise & audio quality is now limited by the front-end opamps buffering the modulator section (5534 operating in unity gain).
FFT measurements show the same distortion and noise spurie on the input to modulator (from the Buffer), as on the output of the amplifier, so now its time for me to improve the input “buffers” performance.
Don’t be to quick to pan the 5534 in this application, they work well in the presents of “RF” without sounding too harsh – have low noise, unity gain stable (with Comp) & low THD, but admittedly are not THE best sounding.
This is the crux of my post; I need a Good (Great) sounding discrete buffer / opamp circuit that also has state-of-the-art measurements!
For a different project a couple of years ago, I tried the Super Breed opamps in comparison with the 5534 – yes there where small differences, little bit here, a little bit there etc… but when I designed a discrete OPAMP – the change was just ENORMOUS! Unfortunately at the time, the best I could achieve from the design was 0.00008% into a 100R load @ 1KHz 2Vrms.
Specifications in order of importance,
1. Sound quality!
2. Low RF sensitivity (both input and output)
3. Low Noise (not effected by RF)
4. Low THD (not effected by RF)
I’ve made a big point about RF sensitivity, as most “newer” opamps I’ve tried are very sensitive to RF, achieving good “Performance” in simple static THD+N tests, but completely fall to pieces when exposed to small RF levels – I hate to think what happens when you connect CD / DVD players to them (with there RF noisy outputs).
Also, what’s so wrong with a folded Cascade designs – they normally have balanced slew rates, low THD and sound very good to me – wide open sound stage, good bass smooth treble… is it just there relative complexity that’s the issue?
Cheers,
John
I ask, as I’m currently developing an analogue input Class D amplifiers where the final THD, Noise & audio quality is now limited by the front-end opamps buffering the modulator section (5534 operating in unity gain).
FFT measurements show the same distortion and noise spurie on the input to modulator (from the Buffer), as on the output of the amplifier, so now its time for me to improve the input “buffers” performance.
Don’t be to quick to pan the 5534 in this application, they work well in the presents of “RF” without sounding too harsh – have low noise, unity gain stable (with Comp) & low THD, but admittedly are not THE best sounding.
This is the crux of my post; I need a Good (Great) sounding discrete buffer / opamp circuit that also has state-of-the-art measurements!
For a different project a couple of years ago, I tried the Super Breed opamps in comparison with the 5534 – yes there where small differences, little bit here, a little bit there etc… but when I designed a discrete OPAMP – the change was just ENORMOUS! Unfortunately at the time, the best I could achieve from the design was 0.00008% into a 100R load @ 1KHz 2Vrms.
Specifications in order of importance,
1. Sound quality!
2. Low RF sensitivity (both input and output)
3. Low Noise (not effected by RF)
4. Low THD (not effected by RF)
I’ve made a big point about RF sensitivity, as most “newer” opamps I’ve tried are very sensitive to RF, achieving good “Performance” in simple static THD+N tests, but completely fall to pieces when exposed to small RF levels – I hate to think what happens when you connect CD / DVD players to them (with there RF noisy outputs).
Also, what’s so wrong with a folded Cascade designs – they normally have balanced slew rates, low THD and sound very good to me – wide open sound stage, good bass smooth treble… is it just there relative complexity that’s the issue?
Cheers,
John
Bruno Putzeys said:
I am not certain i understand....'minor-loop' and 'outer loop' may be the source of confusion here.....they do not mean the same thing.....
Nevertheless, if you mean output stage distortion is still the limiting factor, even with the output stage enclosed within the compensation loop, then i agree...
lynx soundcards can do this directly, add some cleverness from jerald graeme's "optimizing op amp performance" book and with a dedicated low noise +30 dB amp in front of the lynx (and a lot of averageing to get below the noise floor) i can see to -160 dB - as evidence that i have really measured distortion, i found that 2 series 200V np0 caps gave ~ -110 dB distortion with 8Vrms while relpacing them with polystyrene dropped the harmonic to the measurement noise floor (caps in Av +4 sallen-key 30KHz low pass, 1 Kohm Rs in a compound/buffered op amp filter circuit) - the fundmental is also visible at the correct level calculated from loop gain
Just out of curiosity, how much resistance do you need to have in order to minimize noise and swamp the nonlinearities of the transistors in a current mirror?
For example, using the double current mirror suggested by someone here, if I put resistance at the emitters of each transistor, and add some resistance between the output collectors of the mirror and the collectors of the differential, what values should I use for the 6 resistors involved in order to get optimal performance? Voltages below +/- 120Vdc are not an issue. Suggestions?
For example, using the double current mirror suggested by someone here, if I put resistance at the emitters of each transistor, and add some resistance between the output collectors of the mirror and the collectors of the differential, what values should I use for the 6 resistors involved in order to get optimal performance? Voltages below +/- 120Vdc are not an issue. Suggestions?
thanh said:In Orcad, a symmetry differential amp with current mirror has 6uV 2nd ,8uV 3rd at 21 V 1khz output .
I could not building it. So where can I buy it ?
Hi thanh !
Try cfp'd inputstage with cascoded vas, this can give ~4uv 2nd,2uv 3rd !
And still below 20uv for 20khz...
With tripledarlington output using mje15030/15031+mjl1302a/mjl3281
I think you can get better result with even faster output-bjts.
But i believe you don't need these low thd values, i prefer bandwidth
and reliability. This somehow excludes using high OL-gain monsters
with that low THD.
My amp i am working on at the moment has 100times bigger thd, but
sounds absolutely sweet and undistorted... (And all biasing is steady)
You can get even lower values, i've built one. It was asymetrical,
using jfet+cfp+currentmirror. OL-gain was > 1:1.000.000. It did
not oscillate, but would need shielding as it revealed any electrical
disturbance somewhere in the house. At these ratios, receiving a
single uV from the air can give a well audible peak at the output.
I had some fun with a lightning still far away !
And it didn't sound better than my actual amp.
Don't forget: If you are simulating, take the time to slightly modify
some Vbe's of transistors to simulate thermal drift.
Forgot: You can buy these ! Get some pioneer/yamaha/sony chipamp,
they tend to have that low thd-numbers... (good for marketing)
Mike
Now this is a case where gain is fixed, so there's no reason to get hung up on a "single op amp" type circuit.JohnW said:
Nothing should stop you from making a higher order system with n op amps as integrators (the last doubling as output driver)? If high order loops sound good on power amps, they should be fine for buffers as well.
Upupa Epops said:As I heard, John have realy good measuring instruments
Hi Pavel,
Good to hear from you again….
True – but shame they cannot measure sound quality!
I’m having a SMALL personal crisis, I just directly compared a NEW DAC design which is the indirect result of say 10 years of work, 127dB dynamic range, THD 0.000028% FS etc – and compared it to a CD player I designed back in 1995 (had the unit back to repair for a friend) – the difference is enormous - I mean HUGE – the 1995 design is way better!
So PMD100 and SAA7350 on steroids (external I-V and VRef circuits) beats DAC with 32Bit internal data path – 4th order noise shaper with Quad array 128 level PWM output, 512 tap length digital filter, sub 3pS clock jitter
Oh well, new DAC design seems destined to be used as the “Analogue Output” section on my new test set-up.
Also, “Friends” not getting his CD player back until I source “that hard to find component” – what are friends for
Circuits anyone?
John
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.