Noise is definitely much much more than -100dB relative to maximum output Here's some test measurements of the 10B showing the noise floor:
Bryston 10B-SUB Active Stereo Crossover Network - HomeTheaterHifi.com
Bryston 10B-SUB Active Stereo Crossover Network - HomeTheaterHifi.com
Thanks a lot for finding that (and for your other posts). BTW do you mean "much less" than -100 dB? The graphs show noise at -130 dBV to -120 dBV.
I made some calculations below that show that what I'm observing is not in line with those measurements. Could you let me know if you can spot any errors?
OK, so, if the 10B's self-noise is at -130 dBV, and I had my 3B set to 1V input, then how does that relate to the fact that noise was definitely loud enough in the tweeter to be uncomfortable?
The 3B, set to 1V input sensitivity, has 29 dB gain, so the 10B's self-noise of -130 dBV after the 3B amplifies it should still be just -100 dBV.
My speakers are the PMC MB2 XBD. PMC rate them at 89 dB @ 1W @ 1m. The tweeter used is specified as the "Sonolex soft dome ferrofluid cooled 27mm tweeter", which can be found here:
TWEETERS
Though I am not sure which exact one it is. And maybe it's slightly different, as mine sits in a well, while the Sonolex are all flat up front.
The ones that fit the description most closely are:
H0831-06 27TFF
H0881-06 27TFFC
H1149-06 27TDC
H1189-06 27TDFC
They're all specified as 6 Ohm nominal impedance. (edit: previously I read the graph wrong and thought it showed 40 Ohm for some reason)
Sensitivity is specified as:
Characteristic sensitivity (2.83V, 1m): 90 dB SPL
(some are 91 dB SPL or 92 dB SPL)
2.83V is 9 dBV. This means the transducer has a gain of 83 dB.
The differences are mostly in power handling and HF extension, so not important here...
So based off those specs, what loudness should I be hearing out of the tweeter at 1m? At 2m?
The amplifier should be putting out -100 dBV into the tweeter. So then there's gain of 83 dB from the tweeter. That gives us -17 dB SPL at 1m. At 2m it should be -23 dB SPL. That's completely, absolutely inaudible, even in a perfect room, and I'm not even in a treated room, this is a residential area with currently a bunch of construction happening outside.
So whatever numbers there are in that article do not correspond to what I'm observing.
Do you agree with my calculations?
PS wow that color scheme, thank god for reader mode
I made some calculations below that show that what I'm observing is not in line with those measurements. Could you let me know if you can spot any errors?
OK, so, if the 10B's self-noise is at -130 dBV, and I had my 3B set to 1V input, then how does that relate to the fact that noise was definitely loud enough in the tweeter to be uncomfortable?
The 3B, set to 1V input sensitivity, has 29 dB gain, so the 10B's self-noise of -130 dBV after the 3B amplifies it should still be just -100 dBV.
My speakers are the PMC MB2 XBD. PMC rate them at 89 dB @ 1W @ 1m. The tweeter used is specified as the "Sonolex soft dome ferrofluid cooled 27mm tweeter", which can be found here:
TWEETERS
Though I am not sure which exact one it is. And maybe it's slightly different, as mine sits in a well, while the Sonolex are all flat up front.
The ones that fit the description most closely are:
H0831-06 27TFF
H0881-06 27TFFC
H1149-06 27TDC
H1189-06 27TDFC
They're all specified as 6 Ohm nominal impedance. (edit: previously I read the graph wrong and thought it showed 40 Ohm for some reason)
Sensitivity is specified as:
Characteristic sensitivity (2.83V, 1m): 90 dB SPL
(some are 91 dB SPL or 92 dB SPL)
2.83V is 9 dBV. This means the transducer has a gain of 83 dB.
The differences are mostly in power handling and HF extension, so not important here...
So based off those specs, what loudness should I be hearing out of the tweeter at 1m? At 2m?
The amplifier should be putting out -100 dBV into the tweeter. So then there's gain of 83 dB from the tweeter. That gives us -17 dB SPL at 1m. At 2m it should be -23 dB SPL. That's completely, absolutely inaudible, even in a perfect room, and I'm not even in a treated room, this is a residential area with currently a bunch of construction happening outside.
So whatever numbers there are in that article do not correspond to what I'm observing.
Do you agree with my calculations?
PS wow that color scheme, thank god for reader mode
BTW, even at -90 dBV noise, similar to what's specified in some places for the 10B, the tweeter should still be putting out only 7 dB SPL where I'm listening from. That's still completely inaudible. Robinson-Dadson curves specify the threshold of hearing at 1kHz-10kHz at roughly 5 dB SPL to 20 dB SPL (as you go up the frequency), and this isn't a treated room.
Anyways, really curious if you can see any issues in my calculations. Thanks.
Anyways, really curious if you can see any issues in my calculations. Thanks.
Last edited:
I've just woken up, so will think about your last posts later today. Two immediate thoughts to add some perspective though:
The threshold of hearing is taken to be 0 dBspl at around 1kHz, rising (falling?) to -6 dBspl or so at ~3kHz.
A high frequency driver in a passive design is typically attenuated by 3 - 10 dB, sometimes more, to match the sensitivity of the other drivers, reducing the noise from the amplifier at the driver by the same amount. In an active design there is no attenuation between the amplifier and the drivers. Nor is there any filtering, so unlike a passive design the noise signal presented to the driver is not band limited.
The threshold of hearing is taken to be 0 dBspl at around 1kHz, rising (falling?) to -6 dBspl or so at ~3kHz.
A high frequency driver in a passive design is typically attenuated by 3 - 10 dB, sometimes more, to match the sensitivity of the other drivers, reducing the noise from the amplifier at the driver by the same amount. In an active design there is no attenuation between the amplifier and the drivers. Nor is there any filtering, so unlike a passive design the noise signal presented to the driver is not band limited.
Last edited:
It doesn't really matter where /exactly/ the threshold of hearing is, this is a normal living room with stuff happening around, a PC (with a few silent fans but still), some loud hard drives, etc. This noise is obviously pretty loud but it shouldn't be audible here - unless I made a mistake in my calculations, which I don't think I have. So I don't think this crossover is at or below -90 dBV noise.
I don't think you calculations are wrong, but some of your assumptions might be. This is just "back of an envelope" musings:
The Seas Sonolex tweeters are rated as having a characteristic sensitivity of ~90dB. This is the output level adjusted for 1 watt of dissipation in the voice coil. and the nominal impedance is rated at 6 ohms.
None of this matters as what is important is the sensitivity against voltage input, not power. This can be taken from the response plot measured at a nominal 2.83 volts on the input terminals, which is 92 dBspl or so on SEAS data sheets I looked at.
Note that the sensitivity of the tweeter is measure with a high pass Butterworth filter @ 2500Hz 12 dB/octave so the real sensitivity of the tweeter in the 2-4kHz octave is likely to be 3-6 dB higher than specification. So lets say the sensitivity unfiltered is around 96 dBspl.
From there the relevant sensitivity is the output against a nominal 1 volt input which is -9.0dB below 2.83 volts can be calculated as 87 dBspl in the 2-4kHz band.
Assuming from Bryston’s specification the noise output from the 10B is -100dBv, the amplifier adds 29 dB of gain, the noise becomes -71 dBv fed to the tweeter terminals.
The acoustic output of the tweeter at this noise level becomes 87 - 71 = 16 dBspl at 1m or a little more* than 10 dBspl at 2m for one speaker (*-6 dB per distance doubling does not hold true in the near field).
At 2m this is 16 dBspl or more above the threshold of hearing in the 2-4 kHz octave (-6 dBspl), where human hearing is most efficient. Whether that is audible or not depends entirely on the environment and the listener, but under the right (is that optimal?) conditions it will be clearly audible.
If the crossover output noise level is -90 dBv, the noise level from the tweeter at 2m would be 26 dBspl above threshold in the 2 - 4 kHz octave, which would be enough to drive me out of a quiet room.
The Seas Sonolex tweeters are rated as having a characteristic sensitivity of ~90dB. This is the output level adjusted for 1 watt of dissipation in the voice coil. and the nominal impedance is rated at 6 ohms.
None of this matters as what is important is the sensitivity against voltage input, not power. This can be taken from the response plot measured at a nominal 2.83 volts on the input terminals, which is 92 dBspl or so on SEAS data sheets I looked at.
Note that the sensitivity of the tweeter is measure with a high pass Butterworth filter @ 2500Hz 12 dB/octave so the real sensitivity of the tweeter in the 2-4kHz octave is likely to be 3-6 dB higher than specification. So lets say the sensitivity unfiltered is around 96 dBspl.
From there the relevant sensitivity is the output against a nominal 1 volt input which is -9.0dB below 2.83 volts can be calculated as 87 dBspl in the 2-4kHz band.
Assuming from Bryston’s specification the noise output from the 10B is -100dBv, the amplifier adds 29 dB of gain, the noise becomes -71 dBv fed to the tweeter terminals.
The acoustic output of the tweeter at this noise level becomes 87 - 71 = 16 dBspl at 1m or a little more* than 10 dBspl at 2m for one speaker (*-6 dB per distance doubling does not hold true in the near field).
At 2m this is 16 dBspl or more above the threshold of hearing in the 2-4 kHz octave (-6 dBspl), where human hearing is most efficient. Whether that is audible or not depends entirely on the environment and the listener, but under the right (is that optimal?) conditions it will be clearly audible.
If the crossover output noise level is -90 dBv, the noise level from the tweeter at 2m would be 26 dBspl above threshold in the 2 - 4 kHz octave, which would be enough to drive me out of a quiet room.
Last edited:
The second speaker will add 3 dBspl (a doubling of power). The exception is if the two speakers are connected to a mono source, then their outputs will be correlated (not random) with respect to each other; then the addition at the listing point would be 6 dB.
I think that you are overlooking the fact that when you attenuate the output of the crossover, not only are you attenuating noise, you are attenuating distortion as well. I.e. you will be driving the crossover circuitry into a more linear region and THD will be lower as a result. this premise does not hold true for all audio equipment, but it does hold true for competently designed equipment, like that from companies such as Bryston.
If you have a look at the output TN+D graph for the crossover output amps below, you want your signal to be operating where the curve is at a minimum at nominal listening levels.
By far the biggest improvement you can achieve for your crossover's performance is to pad the output ahead of the amplifier to optimise its crossover's TN+D. This is true irrespective of how good the circuitry is inside the crossover. As you have it, your system is operating in the left hand third of the graph.
I think that you are overlooking the fact that when you attenuate the output of the crossover, not only are you attenuating noise, you are attenuating distortion as well. I.e. you will be driving the crossover circuitry into a more linear region and THD will be lower as a result. this premise does not hold true for all audio equipment, but it does hold true for competently designed equipment, like that from companies such as Bryston.
If you have a look at the output TN+D graph for the crossover output amps below, you want your signal to be operating where the curve is at a minimum at nominal listening levels.
By far the biggest improvement you can achieve for your crossover's performance is to pad the output ahead of the amplifier to optimise its crossover's TN+D. This is true irrespective of how good the circuitry is inside the crossover. As you have it, your system is operating in the left hand third of the graph.
Last edited:
You're right about the +3dB.
Sorry, I'm confused, how does attenuating the output of the crossover do anything to distortion that happens inside the crossover, before the crossover's output?
Cool graph - where's it from? I wasn't able to locate it.
Sorry, I'm confused, how does attenuating the output of the crossover do anything to distortion that happens inside the crossover, before the crossover's output?
Cool graph - where's it from? I wasn't able to locate it.
The graph is for the output of a TI OPA2604 which is the op amp used in the 10B. The lowering of THD+N as signal level rises is a common feature all practically all signal and power amplifiers.
Without attenuation on the output of your crossover the signal level in the filter circuits is well below optimum for THD+N, in the 0.05% region. By attenuating the output of the crossover you are able to move the signal levels up where THD+N is ~20dB lower relative to signal level in the .005% region.
When you attenuate the output to match the power amplifier input reuirment, the attenuation drops the signal level back to what's needed for the amplifier whilst maintaining the ~20dB improvement in THD+N.
Or were asking why THD+N lowers as an active stage or power amplifier nears its full unclipped output?
Without attenuation on the output of your crossover the signal level in the filter circuits is well below optimum for THD+N, in the 0.05% region. By attenuating the output of the crossover you are able to move the signal levels up where THD+N is ~20dB lower relative to signal level in the .005% region.
When you attenuate the output to match the power amplifier input reuirment, the attenuation drops the signal level back to what's needed for the amplifier whilst maintaining the ~20dB improvement in THD+N.
Or were asking why THD+N lowers as an active stage or power amplifier nears its full unclipped output?
THD+N is a relative fixed amount of low level garbage. Split supply circuits always have a degree of low level crossover distortion (crossover as in the change over from the positive side output device to the negative side output device).
In high quality audio circuits feedback takes care of signal error aka distortion as signal levels rise, leaving the crossover distortion and noise behind. As the output approaches the voltage rails, nonlinearity in the circuits will rise, which is the cause of the kick up in THD+N just before clipping; actually it is all THD at this point as noise stays at the same low level irrespective of signal.
Good quality power amplifiers have a similar THD+N versus level. This is the Bryston 4B:
In high quality audio circuits feedback takes care of signal error aka distortion as signal levels rise, leaving the crossover distortion and noise behind. As the output approaches the voltage rails, nonlinearity in the circuits will rise, which is the cause of the kick up in THD+N just before clipping; actually it is all THD at this point as noise stays at the same low level irrespective of signal.
Good quality power amplifiers have a similar THD+N versus level. This is the Bryston 4B:
Last edited:
Thanks PRR. I wondered as much, but thought there may be some residual crossover or non-linear distortion as well.
Found a discussion here:
Understanding total harmonic distortion and noise curves - Precision Hub - Archives - TI E2E support forums
Found a discussion here:
Understanding total harmonic distortion and noise curves - Precision Hub - Archives - TI E2E support forums
Last edited:
Oh right, yeah I heard of that once. Thanks for explaining it in easy terms.
Regarding the OP 2064 A. You said it's probably the source of hiss. However, here's the thing. It exists in my circuit, but I think it's not being used. My 10B has an LFR input, but I haven't connected anything to it. If you look at the schematic, the OP 2064 A is only used on the BBSB board. It's hard to read the schematic because of the configuration switching, but it looks to me like this is only ever used to sum the left and right channel for mono bass. However, my crossover already has only got one input, because it's in Mono 3-Way mode. It has a line input and an LFR input. Maybe in the PMC the BBSB board sums the line input and LFR input and puts it through to the LF filters?
Regarding the OP 2064 A. You said it's probably the source of hiss. However, here's the thing. It exists in my circuit, but I think it's not being used. My 10B has an LFR input, but I haven't connected anything to it. If you look at the schematic, the OP 2064 A is only used on the BBSB board. It's hard to read the schematic because of the configuration switching, but it looks to me like this is only ever used to sum the left and right channel for mono bass. However, my crossover already has only got one input, because it's in Mono 3-Way mode. It has a line input and an LFR input. Maybe in the PMC the BBSB board sums the line input and LFR input and puts it through to the LF filters?
I didn't say the OPA1602 was the source of the hiss. I was using the OPA1602 as an example of how the level an amp operates at affects the THD+N in the output.
Electrical noise is generated by electrons buzzing around in resistance. There is a theoretical minimum amount of noise in a given circuit with is a function of the source resistance. Good op amps in a properly designed circuit get within a fraction of a dB of the theoretical minimum noise.
In a real circuit the noise performance is limited by the source resistance plugged into the input of the circuit. If that value is known then the design engineer can design around the value and get a circuit to perform very close to theoretical minimum.
Try this: listen to how noisy a phono input is with a turntable connected, then unplug the turntable and see how the phono input is a LOT noisier. That's because a cartridge has a source resistance of around 500Ω; but when you unplug it, the preamp sees the 47,000Ω input load resistance as the source resistance, so now there is a lot more noise! Likewise if you short circuit the phono inputs with shorting plugs there will be less noise than with a cartridge connected.
Electrical noise is generated by electrons buzzing around in resistance. There is a theoretical minimum amount of noise in a given circuit with is a function of the source resistance. Good op amps in a properly designed circuit get within a fraction of a dB of the theoretical minimum noise.
In a real circuit the noise performance is limited by the source resistance plugged into the input of the circuit. If that value is known then the design engineer can design around the value and get a circuit to perform very close to theoretical minimum.
Try this: listen to how noisy a phono input is with a turntable connected, then unplug the turntable and see how the phono input is a LOT noisier. That's because a cartridge has a source resistance of around 500Ω; but when you unplug it, the preamp sees the 47,000Ω input load resistance as the source resistance, so now there is a lot more noise! Likewise if you short circuit the phono inputs with shorting plugs there will be less noise than with a cartridge connected.
The 10B is not a phono input with so much gain as to make it obvious. In fact the crossover is probably unity gain, which means that the noise floor will be likely be the quiescent noise of the gain amp fliers at unity gain. You might hear a dB or two difference when you short the inputs.
Is your unit balanced in balanced out, or unbalanced in and out?
Is your unit balanced in balanced out, or unbalanced in and out?