The first circuit has two differential stages rather than one, and I call that a 4-stage amplifier. It appears to have the tone control wrapped around the power amplifier- a practice I have never understood. Designing power amplifiers with a fixed amount of NFB is hard enough, but if it varies with the tone control settings... any else got any thoughts on this? I know it was very common.
2 is not a Blameless because it has no input stage degeneration, no current-mirror, and only a simple VAS. (no emitter-follower or cascoding)
Do you have the full model numbers and dates for these circuits? That would be useful.
Last edited:
That sort of feedthrough is or was an issue in some ICs. I don't have it to hand, but if I recall correctly the mechanism was described in Erdi's 1979 paper "A 300V/us Monolithic Voltage Follower". Fascinating reading.
R23 was intended to aid fault-finding. Not my brightest idea.
I would have thought it must do.
I first got involved in serious power-amplifier design in 1975, and I don't recall current-mirrors being used at all then, outside of ICs.
This is a fascinating question; anyone got any more thoughts on it?
As usual you have to read the claims not look at the pictures. Widlar patented using a resistor between the devices to make microamp current sources. BTW the filing date was 1965. I can find no patent for the simplest basic mirror nor could I find the term "current mirror" in the patent records before 1968. George Wilson presented his improvement on the CM in 1968 showing the basic version as known but with no attribution. I have no access to R. D. Thornton's 1965 book on multi-transistor circuits which Wilson cites on an unrelated manner..
Last edited:
Sansui Z3900 and Sony -StrV5. (schema 1/2).
If my memory serves me right , the varying feedback on the Sansui was
rumble filter and loudness compensation associated .
OS
Bias feedthrough and bias instabilities are constant worries in the IC design world. At least in the group I worked in, we always had separate simulations for the bias networks which included stability simulations. Discrete design does have the distinct advantage that you can easily stab an o'scope probe on the circuit and see what's going on. In the IC world, that requires decapsulation and hours of focused ion beam machining to do.
If I was concerned about crosstalk between different circuit sections, I'd use a separate bandgap and current mirror. I've seen designs with a centralized bias and I've seen designs with many individual bias generators used across the chip. We were able to meet the specs (including die size) using either approach, so there doesn't seem to be a large drawback of one method or the other - unless your bias generator requires a pin to the outside world. Customers tend to frown at extraneous pins and external components.
Fair enough. I've seen the technique done on MOS mirrors to minimize noise. No gate current = no voltage drop across the gate resistor. Add a cap from the gate to the reference node and you have a quiet current mirror. Useful in the cases where a BJT can't be used.
Semiconductor processes offer poor absolute accuracy but stellar matching (provided both the circuit designer and layout designer pay attention). This makes it trivial to generate multiple identical currents from the same bias generator. It also means that if you can design the circuit to rely on the matching between components rather than the absolute value of components you can achieve amazing performance.
In the discrete world, absolute accuracy is stellar (0.1 % resistors for less than $1 anyone?) but the matching is poor, so the circuits are designed to rely on the absolute accuracy.
Another reason for current mirrors not being used so much in the discrete world could be cost. A simple mirror with emitter resistors would be four components. Add a beta-helper and the load for said helper and you're up to 6-7 components. That's a lot of components when a resistor in many cases will do the job.
Using resistors for biasing isn't an option in the IC world due to requirements of stellar PSRR. Also, the power supply voltage for a typical opamp varies by nearly a factor of 10, so setting the bias with a resistor off the power supply is a no-go.
My $0.02 worth.
Tom
Last edited:
More early diff amp power amp papers...
Here are two more early papers I found showing audio power amps with differential pair input stage:
http://updatemydynaco.com/HistoricDocuments/BrayDiffPair.pdf
http://updatemydynaco.com/HistoricDocuments/FreylingDiffPair.pdf
The Bray paper shows a date of August 1967, so most probably came after the Locanthi T-Amp paper, which was dated in January of 1967.
Unfortunately, the Freyling paper doesn't have a date...only my memory that I collected this paper in the late 60's or early 70's. It references a paper from 1962, so has to be later than that.
Here are two more early papers I found showing audio power amps with differential pair input stage:
http://updatemydynaco.com/HistoricDocuments/BrayDiffPair.pdf
http://updatemydynaco.com/HistoricDocuments/FreylingDiffPair.pdf
The Bray paper shows a date of August 1967, so most probably came after the Locanthi T-Amp paper, which was dated in January of 1967.
Unfortunately, the Freyling paper doesn't have a date...only my memory that I collected this paper in the late 60's or early 70's. It references a paper from 1962, so has to be later than that.
Doesn't sound like a very good idea but in practice, to my surprise, works very good. Subjectively, I almost like it better than when tone control circuit is built as a separate circuit in front of the power amp section. Additional fet input stage in classic Sansui AU317 power amp section serves as some kind of impedance matching device for the transistor based tone control circuit, but some listeners who have both AU317 and AU217 actually prefer AU217 which has tone control wrapped around power amp.
I also think they were tailoring the amplifier to "play well" with the (included) loudspeakers.
Many of these classics were part of a full "package" (source/amp/speakers).
Doug was right , compensating the 4 stage was difficult enough ...
I would not of wanted another dynamic to fuss over , like a tone control.
But today we have more advanced sources , so the amp's compensation should be
static.
OS
Last edited:
One of the early amps that had a differential input pair was the Crown DC-300, released in 1967. Appendix D of the Service manual goes into the theory of operation - D-Series | Crown Audio - Professional Power Amplifiers
Many of the legacy (late '60's through to early '80's) hi-fi products employed high gain power amplifiers that could take 100-150mV source signals like tuners, tapes and RIAA EQ preamplifiers directly in. There was no pre-amplifier stage as we know it today and this was primarily driven by cost concerns. A small signal transistor then in volume cost 100x what it costs in 2016. A BC847C/BC857C in high volumes is about 0.1c today - i.e. you get 10 of them for a penny. Of course, if you go onto one of the online catalog houses, you are still paying an arm and a leg.
So, putting tone controls around a power amp stage was driven by as much economic concerns as by design choices. Trust me - I've worked with these people and I've lost deals on .01c per part (high volume of course).
Of course, we are now seeing source signals at 1-2V levels driven by digital media and the traditional 15 dB gain line stage is no longer required - we simply need source selection with volue control and perhaps some buffering to drive today's power amplifiers.
Let me add, doing things the old way was crap. As SY has remarked elsewhere, we are living in the golden age of hi-fi now, except for class D of course
So, lets all enjoy it!
So, putting tone controls around a power amp stage was driven by as much economic concerns as by design choices. Trust me - I've worked with these people and I've lost deals on .01c per part (high volume of course).
Of course, we are now seeing source signals at 1-2V levels driven by digital media and the traditional 15 dB gain line stage is no longer required - we simply need source selection with volue control and perhaps some buffering to drive today's power amplifiers.
Let me add, doing things the old way was crap. As SY has remarked elsewhere, we are living in the golden age of hi-fi now, except for class D of course
So, lets all enjoy it!
Last edited:
Thanks for two very interesting app notes.
The function of R5 in the first one is mysterious; some sort of built-in subsonic filter?
The second is remarkable because it has a folded cascode structure- perhaps the first ever in a power amp?
I would be interested to hear of examples.
Where were the volume and balance controls?
Where was the input select switching?
Last edited:
Good point, Scott.
It ought to be possible to work out when the standard current mirror was invented. I've got some old books and I'll root through them when I get a moment.
Googling 'current mirror inventor' and 'current mirror invention' has so far got me nowhere.
The 1967 ( I think ) Phillips Technical Library book "Precision Electronics" is full of very complex input stages, mostly all tube, using mirrors and other things like current sources, etc... anything in that book (probably has citations in it) predates the publication by quite a few years... thought I had a scan of it, but it may have been lost in a disc crash last year (this fool did not back up properly).
Of course the LTP is represented there as well.
I actually have one.
was at one time (may still be) in the original box!Grey plastic base, two dual triodes, iirc.
_-_-
I've got nothing solid except vague recollections and "feeling", but I suspect that the "folded cascode" was used fairly regularly in VHF and up applications? It may be that its application to audio freq amps, such as this one, may in fact be unique. Hope I am not barking up the wrong tree?
_-_-
http://updatemydynaco.com/HistoricDocuments/BrayDiffPair.pdf
R5 pre-biases C1 and C2. Probably to reduce turn-on transients.
Tom
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.