Mark - I got out my oscilloscope and the loudest, bassiest 45rpm 12" single from the disco era - Metropolis "I Love New York". I measured 2.4v peak at the output of my pre-amplifier (I can measure only with RIAA EQ). The pre-amplifier has a gain of 40dB at 1KHz. The cartridge has an output of 5mv (RMS) at 5cm/s and 1KHz.
The 2.4v peak corresponds to a velocity of 17cm/s and required headroom of 11dB (at 1KHz). This velocity is not directly comparable to the actual velocities on the link that you posted.
I also measured Dire Strait's "Money for Nothing" (from the 1985 "Brothers in Arms" LP). I saw peaks of 1.2v. This corresponds to a velocity of 8cm/s and required headroom of 5dB.
Ed
The 2.4v peak corresponds to a velocity of 17cm/s and required headroom of 11dB (at 1KHz). This velocity is not directly comparable to the actual velocities on the link that you posted.
I also measured Dire Strait's "Money for Nothing" (from the 1985 "Brothers in Arms" LP). I saw peaks of 1.2v. This corresponds to a velocity of 8cm/s and required headroom of 5dB.
That depends on the listening level and quietness of one's room. I listen at conversation-level loudness. I don't notice background noise on records.Mark Tillotson said:
Ed
"Loud" music often has the lowest peak-to-average ratios, partly because of the thermal limitations of cutting heads, and is instead just highly compressed. T. Holman did an excellent overview of peak levels on commercial records in the late 1970s, referencing Shure's studies. It's available as an Audio magazine pdf, and hopefully someone without their own razored out paper copy will post a link. Probably 1977.
All good fortune,
Chris
All good fortune,
Chris
I have Tomlinson Holman’s article exploring peak velocities and he did indeed find one record where the output ref 5cm/s was 70 cm/sec and many at 15-20 cm/sec. Most of the peaks are in the 2-4 kHz region where the preamp gain is maximum, so the opportunity for overload is indeed very real.
If you shoot for 26 dB O/load you should be ok. That will cover
dynamic range of 15dB
Hot carts giving >5 mV out 3dB
Hot recordings 6dB
Total 24 dB
John Atkinson at Stereophile has measured active-passive RIAA preamps with 9 dB overload margin. Way too little under any circumstances.
If you shoot for 26 dB O/load you should be ok. That will cover
dynamic range of 15dB
Hot carts giving >5 mV out 3dB
Hot recordings 6dB
Total 24 dB
John Atkinson at Stereophile has measured active-passive RIAA preamps with 9 dB overload margin. Way too little under any circumstances.
I only intend to emphasize that overload margin must vary with frequency, and that any topo other than long-loop RIAA EQ "op-amp" feedback to the input, or an input transconductance stage loaded by passive RIAA, will have less overload margin. (Also, more noise, all things being equal). Not popular, not fashionable, don't care.
Are these the only important criteria for phono equalizers? Of course not, but they're still mandatory. In the vacuum valve world, overload margins can be brute-forced away, but in the +/- 15VDC world we must play by the rules. You always do, but sometimes enthusiasts can get caught up in Internet schemes.
All good fortune,
Chris
Are these the only important criteria for phono equalizers? Of course not, but they're still mandatory. In the vacuum valve world, overload margins can be brute-forced away, but in the +/- 15VDC world we must play by the rules. You always do, but sometimes enthusiasts can get caught up in Internet schemes.
All good fortune,
Chris
My key design targets included
With the gain settings shown, using supply voltages of +/-17V as delivered from my hifisonix standard PSU (thanks Bonsai), the overload margin ref a nominal 5mV /1kHz input is 31dB. This has also been confirmed by measurement. RIAA component values for the time constants also achieve the <0.5% target.
The diagram below shows the resultant performance as simulated using Tina-TI (as opposed to LT-Spice). I've used Tina-TI here as it provides a signal/noise plot out-of-the-box. It also directly uses TI's spice model for the OPA1656.
As has been pointed out, the cartridge noise will exceed the phono stage noise. In real life listening, the phono is silent up until needle drop and subjectively is clean, precise and doesn't suffer from user fatigue.
- Low noise
- low distortion
- Inherent low RFI susceptibility
- High accuracy (<0.5%) for RIAA time constants using standard component values.
- >30dB overload margin
With the gain settings shown, using supply voltages of +/-17V as delivered from my hifisonix standard PSU (thanks Bonsai), the overload margin ref a nominal 5mV /1kHz input is 31dB. This has also been confirmed by measurement. RIAA component values for the time constants also achieve the <0.5% target.
The diagram below shows the resultant performance as simulated using Tina-TI (as opposed to LT-Spice). I've used Tina-TI here as it provides a signal/noise plot out-of-the-box. It also directly uses TI's spice model for the OPA1656.
As has been pointed out, the cartridge noise will exceed the phono stage noise. In real life listening, the phono is silent up until needle drop and subjectively is clean, precise and doesn't suffer from user fatigue.
Last edited:
geoffw1 - You got the gain structure right so that only the pre-amp's output can overload. The 17V rails may not lead to a headroom increase if the main amplifier overloads instead.
Ed
Ed
When I run the LT-Spice Sim with the gain structure given here, the initial gain stage is a low pass filter providing the 3180us pole, having max gain <50Hz and its overload margin is 30.5dB.
The following 318us and 75us pole / zero passive eq, together with the secondary low pass filter which compensated for the unity gain roll-off from the non-inverting first stage, provide the fully RIAA eq'd but roughly 14dB attenuated signal to the unity gain rumble filter. So it has around 44dB overload margin.
The final flat gain stage reinstates the 14dB loss from the 318us/75us passive eq so we're back to the 30dB margin.
A 5mV nominal input would provide approximately 350mV output from all the above.
In an overall amplifier system, the phono stage will typically be followed by a signal selector (e.g Phono / CD / Tuner / Streamer), attenuator, line level pre-amplifier combination. If we ran the output from all this into our power amp without applying a decent measure of attenuation, we would seriously overload it, blow our speakers and our eardrums.
The following 318us and 75us pole / zero passive eq, together with the secondary low pass filter which compensated for the unity gain roll-off from the non-inverting first stage, provide the fully RIAA eq'd but roughly 14dB attenuated signal to the unity gain rumble filter. So it has around 44dB overload margin.
The final flat gain stage reinstates the 14dB loss from the 318us/75us passive eq so we're back to the 30dB margin.
A 5mV nominal input would provide approximately 350mV output from all the above.
In an overall amplifier system, the phono stage will typically be followed by a signal selector (e.g Phono / CD / Tuner / Streamer), attenuator, line level pre-amplifier combination. If we ran the output from all this into our power amp without applying a decent measure of attenuation, we would seriously overload it, blow our speakers and our eardrums.
That is somewhat low. I use 0.5v. The line-level 0dB reference is 0.775v.geoffw1 said:
A line-level pre-amp usually has stages preceding the volume control. That is where the overload will occur.geoffw1 said:
Ed
Using 0.5v output would reduce the overload margin here by about 4dB. Line level 0.775v would reduce overload margin by 7dB. The prime limitation with any op-amp based device (and valve based ones too) is the supply voltage together with just how close can the device remain linear as the peaks of the signal approach the rails.
As you say, where a line-level pre-amp has gain stages preceding the volume control, this is where the overload will occur first. Taking a look at the overall pre-amp system I have in use at home, I currently have a Putseys balanced volume control design with a MAS6116 digital volume controller integrated into it in place of the 10Kohm potentiometer. The PSU for this is the same one as for the phono amp. This has a classic unity gain balanced input stage feeding the Baxandall active volume control. By my reconning, the overload margin will still be at the level presented to it from the phono stage. At usual listening levels (when the wife is around), the volume controller is set for around 40dB of attenuation. In this scenario the limiting point for overload will still be the phono output stage.
As an aside, this makes the 2v output levels that the likes of CD and Media players provide seem silly!
Geoff
Last edited:
It's not quite apples to apples, because digital sources specify output as their peak output, rather than the 0VU of analog sources, which are expected to accommodate (larger) peaks. Yours would be spec'd at about 10V by digital standards.
All good fortune,
Chris
All good fortune,
Chris
Thanks for that, Chris. I always thought the output levels quoted out of digital stuff was out of kilter with analogue.
Isn't clamping just overload?
I think the confusion arises because overload can occur at multiple points - before the EQ, in-between a multi-stage EQ, and after EQ. IMO, a good design overloads only at its output for any input 20Hz-20KHz. The input may have to tolerate 1 volt briefly.
Ed
I think the confusion arises because overload can occur at multiple points - before the EQ, in-between a multi-stage EQ, and after EQ. IMO, a good design overloads only at its output for any input 20Hz-20KHz. The input may have to tolerate 1 volt briefly.
Ed
For an excellent, comprehensive overview I can really recommend Andrew's (Bonsai's) page:
which includes everything except MarcelvdG's recent news-to-everybody-else that a DC servo's pole effects all the time constants, but a capacitor in series to ground leg of feedback resistor does not! Been trying, but I'm not smart enough to understand why - interesting though.
All good fortune,
Chris
which includes everything except MarcelvdG's recent news-to-everybody-else that a DC servo's pole effects all the time constants, but a capacitor in series to ground leg of feedback resistor does not! Been trying, but I'm not smart enough to understand why - interesting though.
All good fortune,
Chris
Last edited:
What is the purpose of the 49860 opamp below the OPA1656 in the 2nd schematic?
Also, how does one calculate various adjustments to the filters in order to make the different turnover and roll off characteristics for non standard eq curves? I’ve had some luck with trial and error, but would like to understand better how to calculate values for the turnover zero and the second pole for the roll off when they are in the same passive network after the first opamp. All the info I can find calculates the entire eq network including the first 50hz pole.
Also, how does one calculate various adjustments to the filters in order to make the different turnover and roll off characteristics for non standard eq curves? I’ve had some luck with trial and error, but would like to understand better how to calculate values for the turnover zero and the second pole for the roll off when they are in the same passive network after the first opamp. All the info I can find calculates the entire eq network including the first 50hz pole.
The one labelled "20Hz high pass filter" in the OP? That's a rumble filter to knock down arm resonances and warped-record infrasound which in bad cases can push woofers to the end-stops...
I think anchorman is referring to the electronic cooling circuit.
anchorman - You are probably not going to find a calculator for the RIAA EQ method in geoffw1's pre-amp. It does straight integration above 50Hz followed by a second network to introduce the 500Hz zero and 2122Hz pole. The component values can be easily calculated.
Ed
anchorman - You are probably not going to find a calculator for the RIAA EQ method in geoffw1's pre-amp. It does straight integration above 50Hz followed by a second network to introduce the 500Hz zero and 2122Hz pole. The component values can be easily calculated.
Ed
Last edited:
Something has to limit scratch transients, before the ideal power amplifier rips the tweeter leads off. I argue that doing this before bass boost leads to a quicker non-linearity and less "smearing"
Ed got this exactly right. The 49860 is performing "electronic cooling". The way I have it shown which uses the virtual earth of the electronic cooling op-amp as the ground for R2 doesn't work too well in practice.
Like Ed says, apart from the secondary low pass filter to compensate for the non-inverting op-amp's 0db roll-off, the frequency calculations are straightforward filter calculations.