Yup - that is the Baxandall circuit. And the Monte Carlo plots show the insensitivity of the design to practical modern component tolerances. Saved me from uploading the spice model. The schematic as shown has a gain of 40dB, which is rather high. At 5mV at 1kHz that is 0.5V output. Add 20dB boost at 20Hz, and a 20dB overload margin and you quickly run out of headroom.
The only difference in the original is a three position switch to give 40dB (100x), 30dB (31.6x) and 20dB (10x) while still presenting a 6k8 load.
It is a difficult circuit to beat - punishingly accurate RIAA, switchable gain, and all with standard component values.
The only thing I'd change is the 10uF capacitor in the lower arm. Given the tolerance of electrolytics this is not a good position to provide a matched low frequency knee between channels. I prefer (shock horror) use a non-polar in series with the input. That prevents any input offset voltage from the opamp from putting a current into the cartridge.
And for the record, I warn anyone in the future that criticizes me for drifting off-topic as this thread of mine has gone.
Nitpicked and analized to death as is common on here, turning a relatively simple question into a massive web of confusion.
I showed you almost your circuit in #47 with simulations in that thread at #242 and #245.
That should answer your questions, right ?
Hans
It all depends on the individual extent of the conversation, which of course cannot be determined beforehand.
Some topics tend to sway from the original post, it's a normal human thing to expect.
But this thread has gotten utterly rediculous in its wandering.
And I appreciate that, thanks!
The 'simple' circuits like the one I originally posted get beatened down and laughed at all the time.
Yet, in fact, they've satisfied millions for decades being used in designs by all manufacturers.
It's just that much of the obsessed-with-perfection crowd insists on their preferences/designs to the point of sounding like annoying elites.
Mind you, there's nothing wrong with 'improving' something to gain benefits, but after a while it becomes annoying.
I doubt that a deviation of a fraction of a percent from ideal would be heard or noticed by human ears.
This site attracts people interested in high-end audio because designing and building one's own equipment makes sense only at the high-end.
Some improvements are just engineering feats, having long surpassed the point of being inaudible.
Ed
Bear in mind that potential channel differences are doubled, and so are more easily audible.
This is even more important than the absolute accuracy of either channel.
That's been clear to me, and I'm sure others as well.This site attracts people interested in high-end audio because designing and building one's own equipment makes sense only at the high-end.
Some improvements are just engineering feats, having long surpassed the point of being inaudible.
Ed
But I'm of the ones who finds the inaudible 'nitpicking' of things only some sort of holy grail attempts beyond being reasonable in the real world of human hearing.
The thankfully small community of the 'obsessive nerds' as opposed to the whole human society gives me hope.
Nonsense......
First of all, jfet´s for riaa is absolutely fine.
Second..... B&O did NOT use jfet´s in their riaa´s previous to year 2000.
From the 1960´ies throughout the 1990´ies, they did with the same circuitry of 2 transistor pr. channel.
Used to be BC149 or similar, but that seemed good enough for all their radio´s/amplifiers for 3 decades.
You should stop making statemens about things, to which you have no knowledge
Last edited:
The Quad 33 used a 2 transistor cartridge stage. Baxandall, who consulted for Quad, and was instrumental in design of the 303 and ESL57, wanted to include a unity gain stage and used that to drive the the stage feedback from a low impedance. But Quad said no and kept with a two transistor stage. By plugging a circuit card in on its four edges you get two MM sensitivities, a ceramic cartridge, and a blank position that enabled a user to configure their own EQ. Often set up for CD or something similar.
It is worth noting that the transistor count in the 33 was a mere 6 transistors per channel. That took care of MM or ceramic input, tape in and out with selected gains, a Baxandall tone control, and a switchable HF filter with slope control. Comprehensive mono/stereo switching, tone and filter control cancel, and four inputs.
It was a landmark product, and still works exceeding well even today, many many decades on.
It is worth noting that the transistor count in the 33 was a mere 6 transistors per channel. That took care of MM or ceramic input, tape in and out with selected gains, a Baxandall tone control, and a switchable HF filter with slope control. Comprehensive mono/stereo switching, tone and filter control cancel, and four inputs.
It was a landmark product, and still works exceeding well even today, many many decades on.
Here is one from Beomaster 3500. All digital functions, but still the "OLD" 2 transistor riaa pr. channel.
This was in production from 1990 to 1993. Exactly the same circuit, you could buy and mount as an accessory in the Beogram 1000 from the 1960´ies.
Last edited:
So I asked Chat GPT to create a RIAA spice model using transistors and 18v power rail.
This is what it spat out. This was my first attempt. Im sure a few more minutes refining the prompts will give me better modells.
Here is my prompt. I have not tried to load the model in LT spice just yet.
Can you create a spice model for a moving magnet RIAA compliant preamp using npn transistors and 18 volts dc power supply
=================
* RIAA Preamp Spice Model * Transistors .model NPN NPN(IS=1E-14 BF=100) * Power Supply Vcc 10 0 DC 18 * Input Capacitor Cin 1 0 10uF .model Cideal CAP IS=1E-12 * Input Stage Q1 1 2 0 NPN R1 2 3 100k C1 3 0 100pF Re1 3 0 1k * RIAA Network C2 3 4 330pF R2 4 5 47k C3 5 6 100nF R3 6 0 47k C4 6 7 47nF R4 7 8 1k C5 8 0 470pF * Output Stage Q2 8 9 0 NPN Rout 9 0 1k * Capacitive Load Cout 9 0 100uF Rload 9 0 47k * Transient Analysis .tran 0.1ms 100ms .end
====================
In this Spice model, the input stage consists of transistor Q1, biased by resistors R1 and capacitor C1. The RIAA network is implemented using capacitors C2, C3, C4, C5, and resistors R2, R3, and R4. The output stage is implemented using transistor Q2, with a load resistor (Rload) and an output coupling capacitor (Cout).
To simulate the preamp, you can use any Spice simulator such as LTspice, Proteus, or PSpice. Connect the input signal source to node 1 and observe the output at node 9. Adjust the component values as needed to meet the RIAA equalization requirements.
Please note that this is a basic example, and the model may require further refinement and tweaking based on your specific requirements and the characteristics of the transistors you are using. Additionally, other considerations such as grounding, shielding, and noise suppression may need to be addressed in a real-world implementation.
(Ok you have to tell the ai to generate for ltspice version 17) It does that also.
This is what it spat out. This was my first attempt. Im sure a few more minutes refining the prompts will give me better modells.
Here is my prompt. I have not tried to load the model in LT spice just yet.
Can you create a spice model for a moving magnet RIAA compliant preamp using npn transistors and 18 volts dc power supply
=================
* RIAA Preamp Spice Model * Transistors .model NPN NPN(IS=1E-14 BF=100) * Power Supply Vcc 10 0 DC 18 * Input Capacitor Cin 1 0 10uF .model Cideal CAP IS=1E-12 * Input Stage Q1 1 2 0 NPN R1 2 3 100k C1 3 0 100pF Re1 3 0 1k * RIAA Network C2 3 4 330pF R2 4 5 47k C3 5 6 100nF R3 6 0 47k C4 6 7 47nF R4 7 8 1k C5 8 0 470pF * Output Stage Q2 8 9 0 NPN Rout 9 0 1k * Capacitive Load Cout 9 0 100uF Rload 9 0 47k * Transient Analysis .tran 0.1ms 100ms .end
====================
In this Spice model, the input stage consists of transistor Q1, biased by resistors R1 and capacitor C1. The RIAA network is implemented using capacitors C2, C3, C4, C5, and resistors R2, R3, and R4. The output stage is implemented using transistor Q2, with a load resistor (Rload) and an output coupling capacitor (Cout).
To simulate the preamp, you can use any Spice simulator such as LTspice, Proteus, or PSpice. Connect the input signal source to node 1 and observe the output at node 9. Adjust the component values as needed to meet the RIAA equalization requirements.
Please note that this is a basic example, and the model may require further refinement and tweaking based on your specific requirements and the characteristics of the transistors you are using. Additionally, other considerations such as grounding, shielding, and noise suppression may need to be addressed in a real-world implementation.
(Ok you have to tell the ai to generate for ltspice version 17) It does that also.
Last edited:
Peter Baxandall's RIAA circuit may be accurate, but he made life easy for himself by separating the two higher time constants. That avoids the main problem where both are combined into one network, when the overall impedance has to be evaluated and component values determined accordingly. Many have copied that idea since. Those who keep to a single network can either use Lipshitz's equations or some other approach to obtain the accurate values. One point is that with only two transistors, the impedances and gains of the devices have a significant impact and have to be included in the evaluation.
Even with three, a final HF roll-off may be needed because of stability problems when large capacitor values are used in the feedback loop, or even if stability is achieved because the usual non-inverting circuit converges to unity gain at high frequencies rather than continuing to fall off.
JLH tried an inverting design but overlooked a rather serious noise problem. Published noise data for a BC109 suggests that with a 47k input impedance the current should be 10uA. I've not looked into whether a better inverting RIAA amplifier could be achieved, as excellent results can be obtained with three transistors or op amps in the non-inverting mode, with whatever network components are required.
Even with three, a final HF roll-off may be needed because of stability problems when large capacitor values are used in the feedback loop, or even if stability is achieved because the usual non-inverting circuit converges to unity gain at high frequencies rather than continuing to fall off.
JLH tried an inverting design but overlooked a rather serious noise problem. Published noise data for a BC109 suggests that with a 47k input impedance the current should be 10uA. I've not looked into whether a better inverting RIAA amplifier could be achieved, as excellent results can be obtained with three transistors or op amps in the non-inverting mode, with whatever network components are required.