the sim is just a sketch of the idea, the diode is ccs biased by the bootstrap RC so it becomes a shunt Vref - which many here worship
up the R4,C1 time constant and replace R3 with a crd if you feel the need
my actual implementation used a pre and post RC filtered, ccs biased precision bandgap ref in place of the sim's diode
while I don't claim golden ears some headfier's who swear that op amp types are clearly audible to them couldn't tell when the floating Vbias source was switched on or off - changing the bias from a few mA to ~ 220mA with Senn HD-600 or AKG K701 headphone loads and SACD source
likewise RMAA distortion numbers with a esi Juli@ card don't change with bias or 32 Ohm to 300 Ohm load and are less than ~50% greater than the loopback #s - except for channel crosstalk which is dominated by the common ground impedance of the TRS jack-plug contact resistance
an earlier multiloop prototype's distortion was unmeasurable below the ~ -160 dB test setup resolution with an indirect IMD test looking for 1KHz IMD with 8V pk 19KHz+20KHz 1:1 driving 300 Ohms with LM6171 output op amp (after I replaced the 2x series 50V NP0 filter caps with polystyrene - the NP0 gave ~ -110 dB 1KHz distortion )
up the R4,C1 time constant and replace R3 with a crd if you feel the need
my actual implementation used a pre and post RC filtered, ccs biased precision bandgap ref in place of the sim's diode
while I don't claim golden ears some headfier's who swear that op amp types are clearly audible to them couldn't tell when the floating Vbias source was switched on or off - changing the bias from a few mA to ~ 220mA with Senn HD-600 or AKG K701 headphone loads and SACD source
likewise RMAA distortion numbers with a esi Juli@ card don't change with bias or 32 Ohm to 300 Ohm load and are less than ~50% greater than the loopback #s - except for channel crosstalk which is dominated by the common ground impedance of the TRS jack-plug contact resistance
an earlier multiloop prototype's distortion was unmeasurable below the ~ -160 dB test setup resolution with an indirect IMD test looking for 1KHz IMD with 8V pk 19KHz+20KHz 1:1 driving 300 Ohms with LM6171 output op amp (after I replaced the 2x series 50V NP0 filter caps with polystyrene - the NP0 gave ~ -110 dB 1KHz distortion )
It isn't really a static distortion reduction contest. That has already been done on other threads. It is really making the 'open loop distortion' effectively lower, especially removing higher order components. These components appear to 'map' into the listening process, even if the negative feedback reduces them to insignificant levels. (Now this is going to get me into real trouble)

certainly my specific multiloop implementation was focused on maximizing loop gain over the audio band
the push-pull op amp class A bias scheme is seperable from that approach and an arbitrarily good floating Vref to offset the op amp outputs is just a matter of implementation
the push-pull op amp class A bias scheme is seperable from that approach and an arbitrarily good floating Vref to offset the op amp outputs is just a matter of implementation
This is how I bias opamp in class A, works like a charm in my cd player as I/V converter and LP filter with various opamp, you just need to know optimal class A bias for opamp in question😉

There is AD711 in DC servo of SCP Vendetta, is not it? I have also realized that one needs a JFET opamp for DC servo of MC pre, rather than BJT opamp.
PMA, I have a DC gain of 10,000 from input to output. I certainly need something to control the dc offset.
Certainly, I mentioned current noise problem of BJT input opamp used as DC servo. I need a dc servo too, I have quite same DC gain.
Scott,
Seems to me, what we need in audio is an high-performance opamp with an extra pin that let you set the output stage bias. Does that exist?
Jan Didden
Seems to me, what we need in audio is an high-performance opamp with an extra pin that let you set the output stage bias. Does that exist?
Jan Didden
You can use well AD844 with output from compensation pin (5). AD744 can be used this way, too. Just get rid of opamp low bias output stage. Or insert unity gain buffer biased at some 20mA, as I have been doing for years, it helps.
hermanv said:Just thinking out loud, forgive me if this is stupid.
Can you sum two Op-Amps, whose outputs are biased so that the N stage is conducting in one and the P stage in the other? Seems you could use a moderate value summing resistor pair (few hundred Ohms?) because the way op-amps work only a small offset voltage will bias one half pretty much off. Since the two outputs will be nearly identical there will be only a small drop on the summing resistors. Take the feedback for both Op-Amps from the summing node this should return the output Z to near zero (watch out for compliance issues).
Parts count will be up, but at least no matching would be needed and any output stage crossover distortion should approach zero.
I don’t quite get the DSL reference. DSL is either multi-phase and multi amplitude or discrete frequency bins. In either case the signal is decoded into a digital bit stream. Crossover distortion should be decent but doesn’t need to be excellent. What am I missing?
I believe (Rupert) Neve did this in some products. Years ago I was
looking at a 9098 channel strip and from memory there were a
bunch of dual opas that appeared to be working together as such.
As PMA suggested a better approach is just bung on the high
current discrete OP stage.
T
jcx said:like this?
I am surprised to hear form Scott that some have poor open loop performance - maybe just the low quiescent current versions?
in any event the easiest way to apply them in a multiloop amp is with plenty of local feedback - since most are "faster" than some common audio transistors I would count the cfa output stage local feedback as nearly equivalent to the local feedback degeneration of an emitter follower
I guess I wasn't clear the current designs use local feedback either as a Sziklai pair or CFA output stage. It sort depends on what you consider open loop when you have a multiloop amp.
The actual performance to kill the spec is more like -75dBc but the BW creeps toward 30MHz
I was thinking in terms of the depth of modulation and delta-Vbe of the output devices. It turns out it's not the voltage (Vbe modulation) but the current (Vbe/Cje) injected into the loop that causes the most problems.
The first gen used a VAS added to a single stage CFA. To make it all complementarily symetric it used CMCF. Haven't had any luck with EC in any of these projects (unless one considers the multiloop amp EC), though some of our guys have some patents on input and output EC on some impressive A/D drivers.
janneman said:Scott,
Seems to me, what we need in audio is an high-performance opamp with an extra pin that let you set the output stage bias. Does that exist?
Jan Didden
I think PMA got it. Just one thing back in 1980 when Walt was here we got away with a lot of fun stuff. You could even sneak RIAA pre-amp articles past the publicity department and we made some quirky parts with hooks for all kinds of things. That's then and this is now.
Scott, both LT and NS have applications articles with phono preamps. What is the problem? Politics? Even if phono stages are specialized and minimum potential, profit-wise, they are interesting and challenging examples of circuit design. Personally, I find it just as challenging to make a quaility phono preamp, as I first did 35 years ago, when first I designed the Levinson JC-2 preamp. At the time, I didn't have a main frame computer, just my HP-35, so I didn't check the RIAA as carefully as I should have and was a little off. Since I just copied Dick Burwin's EQ, that he did for Mark in the LMP-2 preamp previously, the final circuit was a little off, to my consternation, when this was pointed out by Bascom King, years later.
In any case, I find both the LT and the NS phono circuits to be well done, and if I had to do it over again, I might just use one of their designs.
Of course, it is amusing to watch engineers become concerned with things like power supply noise and provide methods to buffer circuits from standard power supplies, something that I was forced to do in 1973, as well.
In any case, I find both the LT and the NS phono circuits to be well done, and if I had to do it over again, I might just use one of their designs.
Of course, it is amusing to watch engineers become concerned with things like power supply noise and provide methods to buffer circuits from standard power supplies, something that I was forced to do in 1973, as well.
Scott:
With the advent of complementary CMOS digital devices, it became possible to integrate many analog functions into single chip solutions. Fewer and fewer external parts. Less and less market for "Jelly Bean" Op-Amps. I recently built (can't really call it design) a single IC AM/FM radio, no Op-Amps
Doesn't the industry need to specialize to survive? Us audio nuts will pay for exotic parts, yes low volumes, but still beats hand sorting and discrete device characterization matching.
If you don't make them, people like John will happily work around you. I'm guessing that now you sell him next to nothing.
With the advent of complementary CMOS digital devices, it became possible to integrate many analog functions into single chip solutions. Fewer and fewer external parts. Less and less market for "Jelly Bean" Op-Amps. I recently built (can't really call it design) a single IC AM/FM radio, no Op-Amps
Doesn't the industry need to specialize to survive? Us audio nuts will pay for exotic parts, yes low volumes, but still beats hand sorting and discrete device characterization matching.
If you don't make them, people like John will happily work around you. I'm guessing that now you sell him next to nothing.
Just to make shure we speak same language
http://www.national.com/nationaledge/jul07/national_analogedge_jul07.pdf
In case you confirm you mean the same circuits, I will tell you.
http://www.national.com/nationaledge/jul07/national_analogedge_jul07.pdf
In case you confirm you mean the same circuits, I will tell you.
The problem is noise contribution of U3 - LME49710 used at position of DC servo. NS probably do not produce low noise JFET opamp. Current noise of LME49710 adds several nV/sqrtHz, due to high values of R15, R16. Even though they are shunted by C13, C14, the noise contribution is not small.
hermanv said:
If you don't make them, people like John will happily work around you. I'm guessing that now you sell him next to nothing.
Unfortunately, from where Scott sits, Johns are literally invisible. To figure out what means today a sucessfull IC busines, here's an example from my day job. Do you have any idea on how many ST19 chips does ST manufacture and deliver per year? Answer: 5,000,000,000 yes, that is corect: five billions.
When you need every bit of capacity to manufacture at this output levels, do you really think they care about what you me John or Dalai Lama would like to have on our breadboards?
Edit: at a second thought, I think Dalai Lama may though get some attention.
- Status
- Not open for further replies.
- Home
- Amplifiers
- Solid State
- John Curl's Blowtorch preamplifier