The automotive biz has been the mainstay, although JBL Pro has been profitable.
It was joked about Gina Harman, Sid's daughter from a first marriage, that when she left heading up the Consumer Group to direct a non-profit foundation in NYC, that she was a good fit, having had lots of experience not making a profit.
The brands of course were used wherever it made sense in the Automotive Group, although they had nothing to do with the original businesses. In some ways Corporate would have considered shutting some of the originals down, but they needed to have brands to help sell car systems.
When Multimedia (under various names) started to make little powered speaker systems for Compaq and Dell, both the JBL and harman/kardon brands were deployed. It was remarked that regardless of whether the division made money (and it did, a little, for a while) the brand recognition was a considerable benefit --- many computer users had never heard of harman/kardon, although their parents and grandparents had.
Scott interesting circuit in post 135, something different to the usual seen to death circuits seen here. I did some simulating with it although I admit 15 min is not enough but Im a bit busy with other things. Maybe its the jfet models Im using or some other mistake but I dont get even near the distortion figures posted here by other members. At 20 Khz the only way I can get to close -100db is by increasing the vas current substancially to around double the figure. What does one call that cascode used in the vas, first time Ive seen such a setup with BJTs. Ive been using composite cascodes for the last 5 years with mosfets for vas when seeking very low THD and it seems to me your cascode does the same function. My sim though showed that the circuit actually performs better without it at 20Khz ?
Ill have a second go at it tonight if I can spare some time.
Ill have a second go at it tonight if I can spare some time.
These are for low input impedance resistive feedback applications, if you scroll down the data sheet the high input impedance application requires a lot more stuff.
Dick, this discussion of non-switching class B might be of interest.
http://www.diyaudio.com/forums/solid-state/122007-margan-non-switching-class-b-2.html#post1956690
http://www.diyaudio.com/forums/solid-state/122007-margan-non-switching-class-b-2.html#post1956690
I see, open loop, about 136ppm, 2nd -81dBr, 3rd -80dBr, 1kHz, 14.5V out p-p, 200uV p-p in, with a perfect buffer, zero source impedance, R29 = 100 ohms. To make things settle more quickly I reduced C13 and C14 to 10nF from the suggested 100nF. Sims ran for 52ms with 1us steps.
Open loop d.c. gain shows as 123dB, -3dBr at 50Hz; phase is well-behaved at the canonical 90 degree shift to about 135 degrees at 22MHz, with C4 = 30pF as shown. The gain at that frequency is 9.2dB.
At 20kHz, 4mV p-p in, 14.5V p-p out, about 150ppm, again 2nd and 3rd dominating. Of course closed loop the higher loop gain for closed loop gains of a few will drastically reduce distortion, and at 1kHz especially.
Scott has of course warned us about the output buffer's contributions, so these results are rather idealized, but nonetheless interesting. And as usual perfect matching of the input devices is assumed.
Depends on what newspaper you read
Scott refers to the approach as base current recapture, which is a nice neutral description that elucidates the functionality.
My money is on E. E. Aldridge, who showed it around 1962 with bipolars, in amplifiers for telcom etc.
Hawksford published unaware of prior art and the configuration has most frequently been associated with him.
By the way, the closely-related Boxall circuit from 1957, and which became known much later as the Baxandall-Swallow superpair, Walt Jung just noticed has propagated under a misspelling of the second author! It should be known as the Baxandall-Shallow superpair!
further results for the post 135 schematic
Again with perfect buffer, 20kHz 7V rms out, closed loop gain of 4 (1k Rf, 333 ohm divider), harmonics to 5th: 3.6ppm, slight edge to 2nd harmonic. This is when driven from a zero impedance source.
Driven from 1kohm, the distortion at the input is 9.5ppm, at the output 13ppm. Undoubtedly, bootstrapped cascoding of the input devices would reduce this, while cutting into the common-mode range.
With the ideal output buffer, there is a latchup problem on turnon, but this is probably due to the unlimited output swing of the buffer. A real buffer run from the same rails (in my case +/- 15V) I doubt would have the issue.
Again with perfect buffer, 20kHz 7V rms out, closed loop gain of 4 (1k Rf, 333 ohm divider), harmonics to 5th: 3.6ppm, slight edge to 2nd harmonic. This is when driven from a zero impedance source.
Driven from 1kohm, the distortion at the input is 9.5ppm, at the output 13ppm. Undoubtedly, bootstrapped cascoding of the input devices would reduce this, while cutting into the common-mode range.
With the ideal output buffer, there is a latchup problem on turnon, but this is probably due to the unlimited output swing of the buffer. A real buffer run from the same rails (in my case +/- 15V) I doubt would have the issue.
That is very helpful. Especially re. thermals.
I'm having my friend Demian Martin out here do accurate backup base line measurements for me before any possible changes/tweeking. Do you have any other data from SIM? Thx. Dick
Last edited:
PS: just noticed (my printout had cut it off) that SW's rails are +/-30 V, not 15. So my results should be re-run for these values.
I did verify the absence of latchup with a real output buffer, and the enhancements with cascoding of the input devices. More at 11.
Actually upon reflection, cascoding of the input devices is advised with 30V rails, as the 862 is rated for a maximum of 20V. A process 51 part sorted for ~50V breakdown would work fairly well, although not something one could count on for production. The NXP version of the J111 is rated for 40V.
Another alternative would be one of the Supertex deplection mode DMOS parts, the smallest of which is the DN1509, which has a specified 90V breakdown.
Depends on what newspaper you read
Scott refers to the approach as base current recapture, which is a nice neutral description that elucidates the functionality.
My money is on E. E. Aldridge, who showed it around 1962 with bipolars, in amplifiers for telcom etc.
Hawksford published unaware of prior art and the configuration has most frequently been associated with him.
By the way, the closely-related Boxall circuit from 1957, and which became known much later as the Baxandall-Swallow superpair, Walt Jung just noticed has propagated under a misspelling of the second author! It should be known as the Baxandall-Shallow superpair!
I should have looked at that angle at the cascode. Composite cascode is very different.
I didnt use a perfect buffer so can understand the difference experienced in performance. Still boggled though why at 20Khz I got better figures without that cascode. Round 2 simulating this circuit needed.
Long before Hawksford published his paper the japanese companies were producing amps using the so called hawksford cascode, this in late 70s. Tanberg was also using it in their amps around 1982.
Is it possible there's a fortuitous cancellation of distortions between the non-cascode stage and the buffer?
Another inventor in and around the various cascodes, whom Walt J. uncovered and with whom he's been communicating, is a modest gentleman named Csanky. We shall hear more about him
Scott is doing at least a couple of things in this circuit that are important for distortion reduction. One of them is the translinear drive to the gm transistors, the other the base current recapture.
Where one ought to see the cascoding improvement in the gm stage is in the appreciable reduction in output capacitance, which in turn reduces its propensity for variation with voltage and hence distortion generation. The buffer following can spoil this in a hurry. I spliced in the diamond buffer I showed a while back in here to get additional results, and it does limit the performance a bit even though its input devices are bootstrapped.
Another of the problems with no cascode in the output gm stage is severe swings of dissipation for a non-cascode part as well as significant Early effects on both beta and Vbe. The thermal stuff is tough to model so tends to become the elephant in the living room for evaluating sweet Spice lies.
The input stage also sees thermal swings with signal, but device matching takes almost all of this out, especially with high loop gain. If, in addition, both the input devices and their current sources are cascoded, this all gets even better. As well the input capacitance drops more and the distortion at the input for a more-than-zero Z source gets very small indeed.
Some sim results using an ideal buffer, for cascoded BF862-J111 pairs at the input: 1k source R, 20kHz to 5th harmonic, 20V p-p out (feedback 333 ohm divider R, feedback R 1k): output distortion 242ppb 20V p-p, 24.7ppb at the input. Again the internal gain-setting R is 100 ohms and the compensation C is still 30pF. Pretty impressive for the component count!
Well, I might as well interject something more about HK, just for 'completeness'.
My experience with HK goes back to 35 years ago, when Sidney Harman was not directly with the company, and HK Intl consisted of HK,JBL, and Ortofon, primarily. In 1977, I was hired by HK Intl to be a consultant for both HK, and JBL. I spent about 1 1/4 years there.
They apparently were looking for 'new blood' and had a succession of 'gurus' in which they hoped to help to successfully produce new product ideas.
In previous years (decades) HK had made some noteworthy audio equipment, such as the Citation A preamp (germanium), and huge, high bandwidth, tube power amplifiers.
In 1977, they were making the Citation solid state series of audio products, including quasi-complementary output stages, and discrete preamp designs that showed a lot of potential. Somewhere in their history they became enthralled with 'wide bandwidth' extremely wide bandwidth, in fact, due to some listening tests that they made in the past. (quite a departure from today)
The place was run by middle managers, who seemed to be immune to layoffs, or any specific task. The engineers, techs and production workers were the 'expendable' ones, apparently, and while they were competent engineers, etc, they were NOT dedicated to audio quality in any significant way, and ultimately they moved on, mostly to aerospace.
In any case, they were always looking for a competitive edge, in this case, it might be my contributions, if any, to give them an edge on the competition. They were completely opposite of objective testing, at the time, including even personal persuasion as to what was best to build, instead of what sounded best, measured best, or had the most hope to make into a successful product. In a way, it was 'subjective testing' run wild, without any real conclusions as to what was the best design approach.
The company, the president of HK told me over dinner, one evening, at a restaurant that I could have never afforded, was losing $100,000/mo and with little hope of change, so eating out, etc at the best places seemed a good idea at the time. So much for company spirit. (more later)
My experience with HK goes back to 35 years ago, when Sidney Harman was not directly with the company, and HK Intl consisted of HK,JBL, and Ortofon, primarily. In 1977, I was hired by HK Intl to be a consultant for both HK, and JBL. I spent about 1 1/4 years there.
They apparently were looking for 'new blood' and had a succession of 'gurus' in which they hoped to help to successfully produce new product ideas.
In previous years (decades) HK had made some noteworthy audio equipment, such as the Citation A preamp (germanium), and huge, high bandwidth, tube power amplifiers.
In 1977, they were making the Citation solid state series of audio products, including quasi-complementary output stages, and discrete preamp designs that showed a lot of potential. Somewhere in their history they became enthralled with 'wide bandwidth' extremely wide bandwidth, in fact, due to some listening tests that they made in the past. (quite a departure from today)
The place was run by middle managers, who seemed to be immune to layoffs, or any specific task. The engineers, techs and production workers were the 'expendable' ones, apparently, and while they were competent engineers, etc, they were NOT dedicated to audio quality in any significant way, and ultimately they moved on, mostly to aerospace.
In any case, they were always looking for a competitive edge, in this case, it might be my contributions, if any, to give them an edge on the competition. They were completely opposite of objective testing, at the time, including even personal persuasion as to what was best to build, instead of what sounded best, measured best, or had the most hope to make into a successful product. In a way, it was 'subjective testing' run wild, without any real conclusions as to what was the best design approach.
The company, the president of HK told me over dinner, one evening, at a restaurant that I could have never afforded, was losing $100,000/mo and with little hope of change, so eating out, etc at the best places seemed a good idea at the time. So much for company spirit. (more later)
Last edited:
The 30V rails are a holdover from some power amp experiments, 18 max or maybe in some cases 24 would make more sense for line level stuff. The 4401/03 have large capacitances there should be better choices in some places (just have them lying around in case I want to actually build one). In 2009 I posted a crazy ultra symmetrical circuit that had Aol and offset trims (Aol trims to "infinite" and Vos to 0) for which I bought bags of SK170/74 and 4401/4403.
Can we then use /-15 and +/-24 volts as rails for simulation and circuit performance reporting then? If someone has a desire to see low voltage performance perhaps +/-8V would be reasonable?
This is not to limit any rail voltage usage or application but to make compairsons easy between designs for conversation sake. These two nicely cover integrated op amps and the pro audio modules so they are typical usage.
Dave
This is not to limit any rail voltage usage or application but to make compairsons easy between designs for conversation sake. These two nicely cover integrated op amps and the pro audio modules so they are typical usage.
Dave
Last edited: