That is very good news. I just finished a phono pre design with a bootstrapped cascode front end with 862s. The noise levels are low enough that despite MM cartridges being the intended sources, with 20dB more gain it is quite acceptable for MC. Probably I will do a dedicated step up stage with some paralleled 862s, but more for specsmanship than much of an audible benefit.
I could use some bipolars too for MC.
I could use some bipolars too for MC.
Someone here actually found the reference design schematic (AM front end for car radios). Thanks to the fact that the emergency broadcast system will be AM probably forever there will be an ample supply.
Don't count on long term supply . . .
I figure eventually it will get designed out no matter how cheap it is. Though you could probably still buy more than you would ever need for the price of a nice dinner for two.
You may want to have a quick glance at Horowitz and Hill 3rd edition Table 8.2 (page 516) in which the BF862 noise measurements finish in third place, behind two current-production JFETs which do not appear in the Electronic Design table or figures.
Those parts also have 750pF or more input capacitance, the gm/C product is all that matters. You can parallel FET's ad nauseum 0.3 nV phono stages have been designed with BF862's for "pennies".
Let's look more closely at Scott's 'challenge'. 0.3nV/rt Hz for pennies is quite difficult to do well.
First let us look at what we have to work with: 0.3nV/rt Hz is equivalent to 5 ohms at room temperature. That is all we can have from both the jfet(s) and any biasing/feedback resistor we use. We could go full Idss and open loop, so let's look at the number of 862 jfets that we must parallel to get an honest 0.3nV/rt Hz: 0.8nV/rt Hz is about 40 ohms, so we would have to parallel at least 8 devices, even without a bias resistor. Now the Idss range of the 862 is 10ma-25 ma. How much current do we need for the input stage. The answer is: 80-250ma. What a range! Or, how many 831 devices do we have to buy to find parts of 15ma or less? 10-100. How the pennies fly! Now let's do something else, how about 10 paralleled 831 devices with a 1 ohm resistor for biasing or 10 ohms for each device? Maybe we can get away with it, but will it be LINEAR enough? What about the next stage? AND we might be forced to add negative feedback, but we have to drive a 1 ohm load. So X100 means the following stages have to drive 100 ohms, almost a complete headphone amplifier. Very few IC's will do it without an added fast buffer. Pennies? No way. '-)
First let us look at what we have to work with: 0.3nV/rt Hz is equivalent to 5 ohms at room temperature. That is all we can have from both the jfet(s) and any biasing/feedback resistor we use. We could go full Idss and open loop, so let's look at the number of 862 jfets that we must parallel to get an honest 0.3nV/rt Hz: 0.8nV/rt Hz is about 40 ohms, so we would have to parallel at least 8 devices, even without a bias resistor. Now the Idss range of the 862 is 10ma-25 ma. How much current do we need for the input stage. The answer is: 80-250ma. What a range! Or, how many 831 devices do we have to buy to find parts of 15ma or less? 10-100. How the pennies fly! Now let's do something else, how about 10 paralleled 831 devices with a 1 ohm resistor for biasing or 10 ohms for each device? Maybe we can get away with it, but will it be LINEAR enough? What about the next stage? AND we might be forced to add negative feedback, but we have to drive a 1 ohm load. So X100 means the following stages have to drive 100 ohms, almost a complete headphone amplifier. Very few IC's will do it without an added fast buffer. Pennies? No way. '-)
Bcarso, 0.3nV/rt Hz is a realistic low limit given that you might want other qualities like direct coupling and reasonably low input capacitance. I knew 'how' to achieve ultra low input noise with a jfet input (or bipolar for that matter) by 1968, but it took me years to develop PRACTICAL circuits, especially open loop ones, that made the grade. It is the circuit that gives the least distortion, and is direct coupled throughout that is the best. IC's are second best in this effort.
Syn08 did some very good work on practical phono amps. That should be acknowledged in my view. A few things I would have done differently, but that's just a designers perogative.
My investigations lead me to belive that with an opamp (5534) you can get to 78 dB ref 5 mV at 1 kHz, while with a JFET design similar to The HPS 5.1, something in the order of 88 to 90 dB at 1 kHz is possible.
When the needle hits the groove, I doubt the noise of either approach is an issue since its swamped by the record surface noise.
The best performance for the $ is still the 5534 in my book. This is a sub $1-00 part and you can build a complete 78 dB S/N phono stage with it and get >35 dB overload performsnce across the audio band. Beyond this, it rapidly starts to get expensive.
My investigations lead me to belive that with an opamp (5534) you can get to 78 dB ref 5 mV at 1 kHz, while with a JFET design similar to The HPS 5.1, something in the order of 88 to 90 dB at 1 kHz is possible.
When the needle hits the groove, I doubt the noise of either approach is an issue since its swamped by the record surface noise.
The best performance for the $ is still the 5534 in my book. This is a sub $1-00 part and you can build a complete 78 dB S/N phono stage with it and get >35 dB overload performsnce across the audio band. Beyond this, it rapidly starts to get expensive.
Last edited:
... which is exactly what Samuel Groner did in his Vol 3 article to get 0.35nV/RtHz.
This article presentsthe design of a low noise laboratory preamplifier, which is particularly aimed at noise
measurement tasks. Its design features a single-ended JFET input stage, an active drain load for good
powersupply rejection and a novel input AC coupling network which attains, together with aDC servo, a
second-order response. The achieved performance includes an equivalent input voltage noise density of
390 pV/rtHz (1 kHz) and 500 pV/rtHz (10 Hz), a frequency response of ±0.1 dB (10 Hz–100 kHz), a power
supply rejection of at least 95 dB (10 Hz–100 kHz) and a quiescent current of just 17 mA. Detailed theo-
retical background on low noise design along with concise construction and application notes are in-
cluded.
John if you don't have this article I can send you a PDF.
Jan
Last edited:
Bonsai, why would DIYers care whether the ICs in their self built phono stages cost $1 or $30? Compared to the other expenses in home building a finished piece (including chassis, power supply, PCB fab) it's a very small uptick. And when you consider cost per hour of listening enjoyment, before discarding this gear and building its replacement, the extra few dozen dollars of initial outlay are insignificant.
Mark, that seems a logical fallacy: why should a diyers, regardless the other expenses, pay any more for any specific part than absolutely what is needed to achieve their objective? I certainly wouldn't spend a couple hundred extra on a TV/appliance to get exactly the same result.
And, most diyers here are assemblers, not designers, so there's something valuable knowing a competent 5534 design isn't leaving much on the table. Not that I've spent a moment designing a phono preamp. It certainly does have decent current noise performance and can be run from higher rails than most other opamps.
And, most diyers here are assemblers, not designers, so there's something valuable knowing a competent 5534 design isn't leaving much on the table. Not that I've spent a moment designing a phono preamp. It certainly does have decent current noise performance and can be run from higher rails than most other opamps.
My objection to the author is that (in his Linear Audio article) he suggested that around 300pV/sq rt Hz was some intrinsic limit (not some "practical" limit). I believe that he even invoked heating of the JFETs, which is pretty silly. This provoked my one and only LTE to Linear Audio, which was particularly motivated by the tutorial flavor of the article. The guy who got traction was Vogel, whom the article author seemed wowed by (and in reaction, engaged in significant degree inflation). Vogel was far less charitable than I. He included low frequency noise and pages of equations, which I enjoyed even if he declared triumphantly that I was wrong as well! In fact I sent him a thank-you note. However, I was aiming for a pedagogical corrective, not a comprehensive and unassailable treatise. I failed, despite providing some notes to give the author a little wiggle room.
Provided we are dealing with amplifying devices with sufficient gain-bandwidth product, reducing input capacitance within the audio range and fairly far above, almost arbitrarily, is feasible, provided we don't eschew feedback. Series feedback can nearly eliminate Cgs, and bootstrapping the drain(s) to the source(s) can nearly eliminate Cgd. The advantages include input distortion reduction for moderate source impedances. Direct coupling is a bit tougher for high closed-loop gain, but paralleling devices doesn't make things that much harder.
- Status
- Not open for further replies.