Not causing a huge issue though it makes me wonder if I mucked something up. Would not expect to see noise peaking around -25dB. To be fair, it's hardly noticable since the freqs are so low. But since I'm mostly using this preamp on an audio interface, I'm constantly reminded of the issue since I'm looking at the mixer levels. But what you're saying about accepting the issue is a good thing to keep in mind before driving myself nuts!
If it is 1/f noise as I suspect the only thing you can do to deal with it is move the LF poles of both coupling caps up to around 6Hz, even at 10Hz I expect you will see a pronounced effect on LF reproduction accuracy. The other option is to try other tubes.
Unbypassed cathode resistors, a hallmark of my designs from 30+ years ago contribute significant excess noise as well as raising the cathode impedance and effective plate resistance contributing further to the noise problem, it also amplifies sensitivity to small fluctuations in filament voltage over time.
Unbypassed cathode resistors, a hallmark of my designs from 30+ years ago contribute significant excess noise as well as raising the cathode impedance and effective plate resistance contributing further to the noise problem, it also amplifies sensitivity to small fluctuations in filament voltage over time.
Looking at the plot in the opening post and assuming it is a power spectral density plot with a vertical scale in dB:
Instead of dropping, the noise PSD slightly increases with frequency from 1 Hz to just below 50 Hz
It drops at a remarkably high rate between 50 Hz and 100 Hz, much steeper than the approximately -9 dB/octave you would get from 1/f noise plus RIAA correction
The 120 Hz peak is a rather wide bump; apparently the frequency resolution is not very good
As a result, my first two comments may be nonsense
It would be nice if we could get a spectrum plot with better frequency resolution
Instead of dropping, the noise PSD slightly increases with frequency from 1 Hz to just below 50 Hz
It drops at a remarkably high rate between 50 Hz and 100 Hz, much steeper than the approximately -9 dB/octave you would get from 1/f noise plus RIAA correction
The 120 Hz peak is a rather wide bump; apparently the frequency resolution is not very good
As a result, my first two comments may be nonsense
It would be nice if we could get a spectrum plot with better frequency resolution
Last edited:
I am surprised that nobody asked for a schematic including supply from the beginning.
And nobody wanted to know - is it stereo and do you own an oscilloscope ? Now post 4
shows a stereo pcb and you should tell us the origin.
So my next question is : is the noise about the same on both channels ? This is easy to
check with a dual input scope.
If it is : the problem is supply related. Using a variac you can probably confirm that.
If the noise is not correlated between channels, it is a tube problem. You can also compare
the regulated supply output noise to that of the amp output by means of the scope.
Low frequency noise can be very different using different tubes. Try to check this as suggested
before. Sometimes vintage tubes need many hours of operation to balance out the initial noise.
I made this experience 30 years ago with phono input using a LOT of E88CC tubes and learned
that the most sought after types are not the best in terms of noise and it does not matter very
much to prefer "new" to used tubes.
I also do not trust the spectral plot - use a scope. It may be difficult to take a picture due to the
lf fluctuations, but today everybody owns a camera phone ?
Forgive me if I do not address everything that has been said before.
And nobody wanted to know - is it stereo and do you own an oscilloscope ? Now post 4
shows a stereo pcb and you should tell us the origin.
So my next question is : is the noise about the same on both channels ? This is easy to
check with a dual input scope.
If it is : the problem is supply related. Using a variac you can probably confirm that.
If the noise is not correlated between channels, it is a tube problem. You can also compare
the regulated supply output noise to that of the amp output by means of the scope.
Low frequency noise can be very different using different tubes. Try to check this as suggested
before. Sometimes vintage tubes need many hours of operation to balance out the initial noise.
I made this experience 30 years ago with phono input using a LOT of E88CC tubes and learned
that the most sought after types are not the best in terms of noise and it does not matter very
much to prefer "new" to used tubes.
I also do not trust the spectral plot - use a scope. It may be difficult to take a picture due to the
lf fluctuations, but today everybody owns a camera phone ?
Forgive me if I do not address everything that has been said before.
I built a 12AX7 based phono preamp with an RC 'passive' filtered PSU a couple years ago that had motorboating issues. I'd be playing a record and it would make these weird, intermittent, rolling rumbles from time to time. I thought the rumbles were from cargo planes flying over my place, until I realized the rumbles happened only when I was playing a record. Finally, with some help from good folks here at diyAudio, the problem was found to be that I had too short of a time constant in the decoupling to the input 12AX7. I think record warps were exciting 1Hz or 2Hz oscillations or something like that, and I was hearing (more like sensing) harmonics of those infrasonic frequencies being played by my poor woofers flopping around uncontrollably in their bass reflex cabs. The cure was to increase the value of the series R in the RC decoupling network between the first and second stages (increase the time constant). That fixed it, and the preamp is being enjoyed today.
Here's the schematic from Lenco Heaven. Is yours built like this one? If so, I think it's likely that there are infrasonic oscillations happening, causing you to hear harmonics of those in the extreme low audio range.
I would decouple the first stage plate supply feed by adding an RC network between the first and second stages -- especially if the 250V B+ isn't well regulated. Try 10k ohms and 100uF (for C12/C13), or 22k ohms and 47uF (for C12/C13).
Here's the schematic from Lenco Heaven. Is yours built like this one? If so, I think it's likely that there are infrasonic oscillations happening, causing you to hear harmonics of those in the extreme low audio range.
I would decouple the first stage plate supply feed by adding an RC network between the first and second stages -- especially if the 250V B+ isn't well regulated. Try 10k ohms and 100uF (for C12/C13), or 22k ohms and 47uF (for C12/C13).
Last edited:
I agree. I am using a DAW plugin to get that plot. I was using one before that couldn't show anything below 20Hz. I need to use an oscilloscope. I can share a sample of the noise for those who have better ways to plot. Unfortunately, at the moment, the only sample I have is with only one input connected to my DAW so it won't be a true representative stereo sample. I can re-sample with both channels later this weekend.
It is a stereo circuit based on the WAD Phono 3 circuit. PCB was designed by Craigtone audio. I do have a very old 1967 scope that my father dropped off at my place several months ago, but I have not confirmed that it is working properly and I need to order some probes for it. It worked last time he used it though, so fingers crossed! I'll try it out this weekend or early next week. However, I DO have a working variac that I can use per your suggestion. What should I do, bring it up gradually and observe impact on noise floor?
Yes, the noise is about the same on both channels and dances in lock-step. I've tried swapping out different tubes and it's about the same with some variation in the overall noise floor peaks.
No need for forgiveness, you've generously provided some great thoughts here + called for some clarity to my somewhat feeble beginning of a conversation. So thank you! I wish I had been a little more clear from the outset.
I'd feel a little sketchy about posting the schematic outright as it was provided upon purchase of a PCB. Though that Phono 2 schematic is very very close. Let me see if I can find a similar basis for comparison of the power supply.
I captured that plot in PreSonus Studio One and have observed left and right noise as correlated and moving together.
That's great. I'll have to track down that thread.
Yes, almost exactly like this. And heater supply is regulated DC. High voltage side is a bit less regulated. Transformer high AC voltage is rectified, filtered through a choke, and sent to the MOSFET. There are obviously some resistors and caps in between and in parallel, but that's the general idea. I need to study the high side of the power schematic a bit more to fully understand it and to provide a better frame of reference.
This seems like a good idea though I'm not entirely sure how to execute it. Are you saying to add an additional RC filter before the plate of the middle valve? Or are you saying to replace C12/C13? Where would the corresponding resistor go?
AS Audio raised a good point, the amplitude scaling seems to imply something other that 1/f noise and I would suggest looking for infrasonic oscillation since there is not any decoupling between stages in this design. (Note works fine with regulated supplies, otherwise not so much.)
At this point I don't know what is going on, so await further interesting developments. 😀
At this point I don't know what is going on, so await further interesting developments. 😀
Keep C12/C13.
Then I was thinking a new resistor would go here:
You don't want subsonic noise amplified by V2b to be fed to V2a inverted 180 degrees, since V2a is also an inverting amplifier, so the subsonic noise signals will now be in phase and thus amplified (positive feedback), and could oscillate, causing those intermittent low booming noises.
I see I explained that really poorly, so here's a better explanation from a Wikipedia article on the subject:
https://en.wikipedia.org/wiki/Motorboating_(electronics)#:~:text=As with all electronic oscillation,proper phase for positive feedback.
Don' t you need additional capacitor between V2 anode resistor and proposed 15K series resistor ?
As the designer of the final schematic and power supply, the HV side is a series pass regulator (IRF840). The HV supply is rock solid. He is having some serious issues with the heater supply. It is also a regulated (LT1085) DC supply. My original build had no noise or voltage issues so I have been trying to help troubleshoot but am a bit stumped since I no longer have a working unit on hand to compare. I did notice tube rolling affected things a good bit and recommended 12AX7LPS but would not recommend 5751 as the Rp is part of the RC RIAA EQ network and that will change the filter corners. I have tested the accuracy and it is within 0.2dB from 20-20kHz.
Also not sure what the builder's "reference" stage is but being a tube phono stage there is always an element of noise not found in solid state gear.
If anyone needs more specifics outside of sharing the full schematic I'd be glad to assist.
To put noise in perspective, when I built my unit and tested it, I was running a Tubelab SE into 104dB K-horns and all I ever got was tube rush. Granted that was with coveted painted tip Tele AX7's from old test equipment so probably some of the quietest AX7's that ever existed. 😎
Also not sure what the builder's "reference" stage is but being a tube phono stage there is always an element of noise not found in solid state gear.
If anyone needs more specifics outside of sharing the full schematic I'd be glad to assist.
To put noise in perspective, when I built my unit and tested it, I was running a Tubelab SE into 104dB K-horns and all I ever got was tube rush. Granted that was with coveted painted tip Tele AX7's from old test equipment so probably some of the quietest AX7's that ever existed. 😎
It's really off-topic as the thread starter's problem is not ordinary circuit noise, but anyway:
According to the data in Merlin Blencowe's ( @Merlinb ) AES article about triode noise, for which he measured the noise levels versus anode DC current of dozens of valves of various different types, ECC83 / 12AX7 is actually pretty good noise-wise.
I once calculated the ECC88 and ECC83 rows of this table based on Merlin Blencowe's data. The EF86 row is based on my own measurements of a very limited number of EF86's. The currents are the optimal anode DC currents and the number of nanovolts per root hertz is the white input noise voltage density that would result in the same integrated weighted noise as the noise of the actual valve.
According to the data in Merlin Blencowe's ( @Merlinb ) AES article about triode noise, for which he measured the noise levels versus anode DC current of dozens of valves of various different types, ECC83 / 12AX7 is actually pretty good noise-wise.
I once calculated the ECC88 and ECC83 rows of this table based on Merlin Blencowe's data. The EF86 row is based on my own measurements of a very limited number of EF86's. The currents are the optimal anode DC currents and the number of nanovolts per root hertz is the white input noise voltage density that would result in the same integrated weighted noise as the noise of the actual valve.