It would be interesting to see the noise spectrum of each channel. If it is identical on both, reinstalling the origional OP-amps should be the next test.
If there are huge differences, this should point to some defective part. I had noisy small signal FET's, which made strange sounds and produced anoying hiss.
It makes fault finding with a scope pretty easy, if you have a working and a distorted channel.
Even at that time, decades ago, these FET's were hard to source.
If there are huge differences, this should point to some defective part. I had noisy small signal FET's, which made strange sounds and produced anoying hiss.
It makes fault finding with a scope pretty easy, if you have a working and a distorted channel.
Even at that time, decades ago, these FET's were hard to source.
Exactly. J-Fets tend to go bad one at a time unless something bad happened in the circuit. Therefore I would expect one channel and not both if that were the case.
Since both are an issue, and it is broadband noise from the sounds of it ... check ground integrity, and also the supplies should be clean.
Since both are an issue, and it is broadband noise from the sounds of it ... check ground integrity, and also the supplies should be clean.
I assumed a voltage gain of 50, so that's 34 dB.
Domestic background sounds are not necessarily wideband, so they do not necessarily mask white noise well.
152 W per channel into 8 ohm and 1.5 V sensitivity according to https://www.hifiengine.com/manual_library/marantz/300dc.shtml
That's a gain of only 23.25, or 27.328 dB.
Your amplifier gain is about 7 dB less than I assumed and the loudspeaker sensitivity about 7 dB higher, so you get about the noise level I calculated in the free field at 1 metre distance from both speakers. With a QUAD 405 or QUAD 405-2 as amplifier, you would get about 7 dB more. What you get in a living room depends on how reverberant it is. It's all irrelevant anyway, because you are not Pat.
You could very well be right. I find it a bit peculiar that both channels are equally affected, though.
I don't know about that. Sounds that I don't even notice frequently drive my partner up the wall.
@PatAllen The noise levels we are discussing are around 100 nV/√Hz if your preamplifier should be in perfect working order and much more than that if @anatech is correct and you have a malfunctioning preamplifier. In either case, it should be high enough to measure with the FFT function of a digital oscilloscope when you use a 1:1 passive probe and set the scope to its most sensitive setting.
Could you try to do so and also measure with the exact same settings and shorted 1:1 probe, to check the scope noise floor? The result should proof without doubt whether anatech is correct and on top of that, the shape of the spectrum you measure may give some clues as to where the problem is.
I've never used the FFT function of a digital scope myself, I only have an analogue scope at home and at work we use spectrum analysers, but I guess you have to capture 200 ms or so of data and then look through the menu structure to find where they hid the FFT.
Addition to post #62: 'proof' should have been 'prove' and the preamplifier input should be shorted.
It's just a rough indication of the noise level I'm after, not a signal-to-noise ratio.
Hi Marcel,
I'll have to measure the gain of these amps some day. They may well be approx. +34 dB, which puts them on the higher gain end of the spectrum. I test amplifiers day in and day out, the results I posted are averaged, so the indicated noise level should be higher. I am normally more concerned with distortion and types of distortion. More importantly (and my point) is I own one of these preamps and used it daily. The speakers are about 1.5 metres from my ears. The overall gain structure of this setup (including speakers) is higher than average and noise level lower since my lab is below ground in a very quiet semi-rural area. At 65, my hearing is surprisingly good, including higher frequencies.
I have some expensive DSOs, and the FFT functions are disappointing. I use spectrum analyzers and the RTX 6001. For quicky measurements I use ARTA (paid) and also M.I. To use an inexpensive DSO with a 1:1 probe you've got to be very careful you don't damage the input. Just like using a spec-an. Pat's 54622D has better input protection and voltage range. I do find myself doubtful these scopes have the dynamic range needed.
Marantz products of this era and earlier put out approx. 25% more power then specified into the specified load. That was a Marantz internal requirement, I was trained at Marantz.
Yes, I am wondering about the issue. A noisy channel is not unheard of, but both is unusual. He has my curiosity raised. I would LOVE to see that on my bench, too bad he isn't closer.
lol!
Yep! I hear all kinds of things my partner doesn't, or some other folks. I can sometimes hear telephones ringing in recordings way in the background. I had a nixie clock in my living room that drove me nuts, I had to move it out. Some reviewers and other "golden eared" experts never once complained or heard it. Same for horizontal oscillators on CRT TVs, very annoying.
Let's see where this goes ...
I'll have to measure the gain of these amps some day. They may well be approx. +34 dB, which puts them on the higher gain end of the spectrum. I test amplifiers day in and day out, the results I posted are averaged, so the indicated noise level should be higher. I am normally more concerned with distortion and types of distortion. More importantly (and my point) is I own one of these preamps and used it daily. The speakers are about 1.5 metres from my ears. The overall gain structure of this setup (including speakers) is higher than average and noise level lower since my lab is below ground in a very quiet semi-rural area. At 65, my hearing is surprisingly good, including higher frequencies.
I have some expensive DSOs, and the FFT functions are disappointing. I use spectrum analyzers and the RTX 6001. For quicky measurements I use ARTA (paid) and also M.I. To use an inexpensive DSO with a 1:1 probe you've got to be very careful you don't damage the input. Just like using a spec-an. Pat's 54622D has better input protection and voltage range. I do find myself doubtful these scopes have the dynamic range needed.
Marantz products of this era and earlier put out approx. 25% more power then specified into the specified load. That was a Marantz internal requirement, I was trained at Marantz.
Yes, I am wondering about the issue. A noisy channel is not unheard of, but both is unusual. He has my curiosity raised. I would LOVE to see that on my bench, too bad he isn't closer.
lol!
Yep! I hear all kinds of things my partner doesn't, or some other folks. I can sometimes hear telephones ringing in recordings way in the background. I had a nixie clock in my living room that drove me nuts, I had to move it out. Some reviewers and other "golden eared" experts never once complained or heard it. Same for horizontal oscillators on CRT TVs, very annoying.
Let's see where this goes ...
I used to hear the 15625 Hz of our PAL CRT television during the first few minutes after switch-on until I was in my mid-30's. I haven't heard it after that, and we kept using the same television until about a decade later.
Have you ever heard the FM stereo pilot residue? I haven't, but a former colleague of mine could hear it when he was young.
Anyway, regarding noise floor measurements, we know from your measurements that a healthy unit has an output noise of about -136 dBV RMS in 192000/65536 Hz, or 92.6 nV/√Hz.
Ideally, the noise of the scope should be less than that; theoretically you can always measure the noise of the scope separately and subtract it in power from the total, but that soon gets inaccurate when the scope noise dominates. 92.6 nV/√Hz is 414 uV RMS in 20 MHz. The width of the trace you see on a scope (when really used as an oscilloscope) is roughly six times the RMS value (three sigma below to three sigma above average), so 2.48 mV, almost half a division at 5 mV/div.
All the oscilloscopes I have seen have much less than half a division of noise at 5 mV/div, 20 MHz bandwidth limit, so I think it should work. Assumptions: enough noise to properly dither the ADC, reasonably flat noise spectrum instead of a huge noise density increase in the audio band.
Have you ever heard the FM stereo pilot residue? I haven't, but a former colleague of mine could hear it when he was young.
Anyway, regarding noise floor measurements, we know from your measurements that a healthy unit has an output noise of about -136 dBV RMS in 192000/65536 Hz, or 92.6 nV/√Hz.
Ideally, the noise of the scope should be less than that; theoretically you can always measure the noise of the scope separately and subtract it in power from the total, but that soon gets inaccurate when the scope noise dominates. 92.6 nV/√Hz is 414 uV RMS in 20 MHz. The width of the trace you see on a scope (when really used as an oscilloscope) is roughly six times the RMS value (three sigma below to three sigma above average), so 2.48 mV, almost half a division at 5 mV/div.
All the oscilloscopes I have seen have much less than half a division of noise at 5 mV/div, 20 MHz bandwidth limit, so I think it should work. Assumptions: enough noise to properly dither the ADC, reasonably flat noise spectrum instead of a huge noise density increase in the audio band.
Hi Marcel,
My measurements are averaged. That kills random noise pretty effectively, so the actual noise is higher. I do this to see distortion and other things hiding below the noise floor.
Some day I'll have to put the Beta 30 on the bench and measure it without averaging. The main fact is that I don't hear noise or hiss with mine.
Our TVs ran at 15,750 Hz. Yes, I could always hear it until we didn't own a CRT TV anymore. I doubt I could hear it today.
My measurements are averaged. That kills random noise pretty effectively, so the actual noise is higher. I do this to see distortion and other things hiding below the noise floor.
Some day I'll have to put the Beta 30 on the bench and measure it without averaging. The main fact is that I don't hear noise or hiss with mine.
Our TVs ran at 15,750 Hz. Yes, I could always hear it until we didn't own a CRT TV anymore. I doubt I could hear it today.
Averaging just reduces the fluctuations you see in the noise floor, assuming you mean a normal power average and not a coherent average. That's why the measured noise floor doesn't keep dropping with increasing number of averages, but just starts looking more like a thin line.
In any case, 100 nV/√Hz is perfectly plausible considering the gain and impedance level of the preamplifier. It's the thermal noise of about 20 kohm amplified five times.
In any case, 100 nV/√Hz is perfectly plausible considering the gain and impedance level of the preamplifier. It's the thermal noise of about 20 kohm amplified five times.
To be honest, I haven't paid that much attention to how it averages. My goal was to see and deal with distortion mechanisms. I"ll pay attention next time I run tests. I typically start the test and ignore it until it's done. I paid attention years ago, long forgotten now.
gawd i lost the track on this thread.
i did put back the original opamp and the noise is identical if not worse. OPA2134 and LM4562 give same results while ne5532 seems to make some random pop corn noise (as described in mike self book). So ill stick with the Burson for now.
i have rew in one of my laptop so i might give it a try. my scope is not really able to measure anything usefull, so i am using one of my lab amplifier that is dead silent, the nikko preamp connected to it with inputs shorted and a set of headphones to compare opamp noise residue. i can also hear a slight background humm no matter what. as soon as the preamp is turned off, thus activating the mute circuit, the noise goes away totaly.
The only thing so far that really made a big difference is by changing the values of the feedback network 330k/56k by a magnitude of 10. so 33k and 5.6k.
hiss has now became acceptable to my ears, maybe has reduced 50% but sure enough the tone control is now not working right, well it does but with 75% less effect, probalby +- 3db. for now i am satisfied. thanks for all the inputs.
i did put back the original opamp and the noise is identical if not worse. OPA2134 and LM4562 give same results while ne5532 seems to make some random pop corn noise (as described in mike self book). So ill stick with the Burson for now.
i have rew in one of my laptop so i might give it a try. my scope is not really able to measure anything usefull, so i am using one of my lab amplifier that is dead silent, the nikko preamp connected to it with inputs shorted and a set of headphones to compare opamp noise residue. i can also hear a slight background humm no matter what. as soon as the preamp is turned off, thus activating the mute circuit, the noise goes away totaly.
The only thing so far that really made a big difference is by changing the values of the feedback network 330k/56k by a magnitude of 10. so 33k and 5.6k.
hiss has now became acceptable to my ears, maybe has reduced 50% but sure enough the tone control is now not working right, well it does but with 75% less effect, probalby +- 3db. for now i am satisfied. thanks for all the inputs.
what would be a decent piece of equipment to acquire so i can make scientific noise measurements ? something that doesnt require a mortgage please.. 🙂
Usually the most accurate way to measure noise is to compare it to a known noise source; the exact gain and filter shapes of the measuring equipment then tend to drop out of the equation. For audio, you can often use a resistor with an accurately known value and temperature that has no DC flowing through it as the known noise source.
Using this method, you can do a lot with a TL071, a few resistors and something that can measure RMS values over a given bandwidth. I either use a home-made A-weighting filter and an old Radiometer BKF6d for that, or a field memory recorder and the GoldWave digital audio editing program
Using this method, you can do a lot with a TL071, a few resistors and something that can measure RMS values over a given bandwidth. I either use a home-made A-weighting filter and an old Radiometer BKF6d for that, or a field memory recorder and the GoldWave digital audio editing program
You'd be further ahead with something better than a TL071, improved op amps are not what I would call expensive.
To get the known current, you need good test gear and controlled temperature. You and I can do this easily Marcel, others who aren't familiar with how various factors can affect circuits are up against a lot of variables.
REW with a decent sound card should work. You need an accurately known oscillator amplitude to calibrate your setup. You can make ratiometric measurements without this.
My bench is unfortunately more than a mortgage, and that is what is needed to make very accurate measurements. This is why we have to buy expensive equipment, it isn't a fashion name tag. So with a great deal of knowledge and experience you can achieve this more inexpensively than others, but there still is a substantial cost involved. So loosen your specs, accept however much uncertainty you will have and simply keep it in mind.
To get the known current, you need good test gear and controlled temperature. You and I can do this easily Marcel, others who aren't familiar with how various factors can affect circuits are up against a lot of variables.
REW with a decent sound card should work. You need an accurately known oscillator amplitude to calibrate your setup. You can make ratiometric measurements without this.
My bench is unfortunately more than a mortgage, and that is what is needed to make very accurate measurements. This is why we have to buy expensive equipment, it isn't a fashion name tag. So with a great deal of knowledge and experience you can achieve this more inexpensively than others, but there still is a substantial cost involved. So loosen your specs, accept however much uncertainty you will have and simply keep it in mind.
@anatech What known current?
@PatAllen For clarity, the noise measurement I proposed was only meant to check if the noise is about normal or way off, and in the latter case, if it is mostly white or mostly flicker noise. High accuracy isn't needed for that. I think your present digital scopes could do the trick, although I haven't managed to convince anatech of that.
@PatAllen For clarity, the noise measurement I proposed was only meant to check if the noise is about normal or way off, and in the latter case, if it is mostly white or mostly flicker noise. High accuracy isn't needed for that. I think your present digital scopes could do the trick, although I haven't managed to convince anatech of that.
Last edited:
You would generally run a known, stable current through your reference resistance.
Nope. Cheaper DSOs are noisy. I wouldn't trust my current Keysight DSO for this compared to an analogue oscilloscope. Maybe I should, maybe not.
Nope. Cheaper DSOs are noisy. I wouldn't trust my current Keysight DSO for this compared to an analogue oscilloscope. Maybe I should, maybe not.
When you want to use an ordinary resistor as a known noise source, you can best run no current through it. Its noise is then simply thermal noise that depends only on resistance and temperature and not on peculiarities of the resistor.
True, but then it is a very low noise voltage.
I don't know, doing it this way is more complicated for beginners. Not a first time project in other words. So many sources of uncertainty with the following amplification that hasn't been characterized.
I don't know, doing it this way is more complicated for beginners. Not a first time project in other words. So many sources of uncertainty with the following amplification that hasn't been characterized.
I'm curious if you've tried these gadgets that adapt modern plastic packages to TO3 mountings?
I only learned of this a week ago.
https://hoppesbrain.com/product/to3-to-to3p-adapters/