That depends largely on the design of your amp. I find that I like the noise/ripple on the output of my POWER amp to be less than 1 mV and preferably below 500 uV. If you know the gain of your power amp, you can work your way back to how quiet the preamp should be. If there's 10 V ripple on the supply but none of it makes it to the amp output, would you care?
With a good B+ regulator, you can get way below 1 mV noise and ripple on the supply. I'm seeing 20 uV RMS of noise/ripple/EMI on the output of my 21st Century Maida Regulator. It doesn't get quieter than that...
~Tom
With a good B+ regulator, you can get way below 1 mV noise and ripple on the supply. I'm seeing 20 uV RMS of noise/ripple/EMI on the output of my 21st Century Maida Regulator. It doesn't get quieter than that...
~Tom
For conventional valve circuits significant amounts of power supply noise are coupled straight into the stage's output. Since you haven't said, let's guess half. So, 10mV p-p noise on an output signal of, say, 2V p-p is -46dB from full output. Not wonderful, fersure.
It would be instructive and a good test to actually measure the noise output.
All good fortune,
Chris
It would be instructive and a good test to actually measure the noise output.
All good fortune,
Chris
Noise at the preamp output is the main issue. That comes from PSU noise and preamp PSRR. You can't consider one in the absence of the other. Circuits vary quite a lot in their PSRR, and you can always improve a poor circuit by adding extra supply rail decoupling. Noise cancellation should always be regarded as a last resort, as it will often depend on valve characteristics, which can change with age.
Remember that measuring low levels of noise or hum can be difficult. You may just be measuring the ground noise in your measurement setup.
Remember that measuring low levels of noise or hum can be difficult. You may just be measuring the ground noise in your measurement setup.
Here is the noise.
Channel one is a direct out of the preamp to the direct in on the scope via a BNC adapter (i.e., 1:1).
Channel 2 is via a 10:1 probe of the power supply (+250 VDC) at the line amp PCB.
Picture 1 is the noise with the scope timebase set at high frequency; i.e., 20 nS per division.
Picture 2 is at 10 mS per division.
Obviously, there is a lot of noise and it appears to be coupled right into the output stage of the amp.
I see about 20 mV P-P noise from the power supply and about 15 mV of noise P-P on the output.
Here is the test setup. Input grounded, but it makes no difference to the noise. The inputs are inside the metal box on the upper left and switched in and out using relays.
Star ground is located on the metal divider between the power supply and line amp boards. The metal chassis is isolated from the analog ground via 4 diodes. The scope probe ground is attached to that star ground with the scope probe attached to the 250 VDC line where it enters the right PCB.
Note the only ground wire on the PCB is the one just above the 250 VDC input. The other signal grounds terminate at the end of the wires, but are not connected to the input or output jacks. The jack grounds are run directly to the star ground via separate wires.
The power supplies are International Power units. The left one is the HV and the right is 12 VDC. They are nice units, but obviously pretty crappy for audio work! I am thinking a choke on the output would be in order since all of the noise appears to be high frequency hash.
The reason for the original question was to determine what a good target for noise reduction should be for this project.
Probably too much information here, but I wanted to try to fend off as many questions as possible.
Channel one is a direct out of the preamp to the direct in on the scope via a BNC adapter (i.e., 1:1).
Channel 2 is via a 10:1 probe of the power supply (+250 VDC) at the line amp PCB.
Picture 1 is the noise with the scope timebase set at high frequency; i.e., 20 nS per division.
Picture 2 is at 10 mS per division.
Obviously, there is a lot of noise and it appears to be coupled right into the output stage of the amp.
I see about 20 mV P-P noise from the power supply and about 15 mV of noise P-P on the output.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Here is the test setup. Input grounded, but it makes no difference to the noise. The inputs are inside the metal box on the upper left and switched in and out using relays.
Star ground is located on the metal divider between the power supply and line amp boards. The metal chassis is isolated from the analog ground via 4 diodes. The scope probe ground is attached to that star ground with the scope probe attached to the 250 VDC line where it enters the right PCB.
Note the only ground wire on the PCB is the one just above the 250 VDC input. The other signal grounds terminate at the end of the wires, but are not connected to the input or output jacks. The jack grounds are run directly to the star ground via separate wires.
The power supplies are International Power units. The left one is the HV and the right is 12 VDC. They are nice units, but obviously pretty crappy for audio work!
I am thinking a choke on the output would be in order since all of the noise appears to be high frequency hash.The reason for the original question was to determine what a good target for noise reduction should be for this project.
Probably too much information here, but I wanted to try to fend off as many questions as possible.
An externally hosted image should be here but it was not working when we last tested it.
There is no simple blanket answer to this question. It depends on several factors. Is it push-pull, single ended series feed, or single ended parallel feed circuit? Are the tubes triodes or pentodes? What is the turns ratio of the output transformer? What is the physical layout of the amplifier/preamp? What is the sensitivity of your speakers? All these things will determine if any of the power supply ripple is becoming audible noise on the output.
Push-pull cancels the majority of power supply ripple. Upwards to a few volts of ripple is inaudible with most push-pull pentode output stages. Parafeed has better power supply ripple rejection than series feed due to the plate choke, or CCS used. Pentodes behave more like constant current sources so they are not affected by power supply ripple as much. Just look at the typical plate curves for pentodes and triodes. The pentode curves are almost horizontal. This means there is very little plate current change in relation to changing plate voltage. The triode on the other hand is quite different. Its plate curves are more vertical in nature. Therefore, a small change in plate voltage (ripple) has a more significant effect on plate current. Higher Mu triodes or more sensitive to power supply ripple than low Mu triodes. So, the traditional low Mu output triode is not as big an offender as some people might have you think. The turns ratio of the output transformer also impacts your final noise levels. Any noise on the output tubes gets stepped down by the turns ratio of the output transformer. Higher impedance transformers give better noise factors. I’ve had no problems with a 300B loaded with a 5K OPT and 200mV to 300mV power supply ripple. How you layout your amplifier is also very important. Some noise may be due to magnetic coupling and not from the power supply ripple. A good grounding design is also very important to a quite circuit. Lastly, how sensitive are your speakers? Are they 85dB cone and domes, or 105dB horns? The amplifier running 105dB horns has to be 100 times (20dB) quitter than the one running conventional 85dB speakers.
Push-pull cancels the majority of power supply ripple. Upwards to a few volts of ripple is inaudible with most push-pull pentode output stages. Parafeed has better power supply ripple rejection than series feed due to the plate choke, or CCS used. Pentodes behave more like constant current sources so they are not affected by power supply ripple as much. Just look at the typical plate curves for pentodes and triodes. The pentode curves are almost horizontal. This means there is very little plate current change in relation to changing plate voltage. The triode on the other hand is quite different. Its plate curves are more vertical in nature. Therefore, a small change in plate voltage (ripple) has a more significant effect on plate current. Higher Mu triodes or more sensitive to power supply ripple than low Mu triodes. So, the traditional low Mu output triode is not as big an offender as some people might have you think. The turns ratio of the output transformer also impacts your final noise levels. Any noise on the output tubes gets stepped down by the turns ratio of the output transformer. Higher impedance transformers give better noise factors. I’ve had no problems with a 300B loaded with a 5K OPT and 200mV to 300mV power supply ripple. How you layout your amplifier is also very important. Some noise may be due to magnetic coupling and not from the power supply ripple. A good grounding design is also very important to a quite circuit. Lastly, how sensitive are your speakers? Are they 85dB cone and domes, or 105dB horns? The amplifier running 105dB horns has to be 100 times (20dB) quitter than the one running conventional 85dB speakers.
Non-sinusoidal noise is normally a sign of circuit design issues, or a poorly designed power supply that has atypical periods and ringing. The first thing I would do is ditch that HV power supply and build my own.
First off, before you rip anything apart, clean up the grounding of your scope probe. The area of the loop formed by the probe, circuit under test, and the grounding clip needs to be as small as possible to minimize inductive coupling of "stuff" onto the ground of your scope probe. This means finding a ground near the B+ you're testing and wrapping the ground lead around the probe tip. You can also use one of the ground adaptors that come with the Tek probes. It's a little ground wire in an adaptor housing that slips onto the probe tip (after you remove the probe hat). It turns that long pigtail ground into maybe 15 mm of solid core wire. If you don't have that adaptor, remove the probe hat and wrap a couple of turns of solid wire around the ground sleeve on the probe tip. Bend it out and down so you can hit a nearby ground. Keep the ground lead short. You'll get much cleaner measurements that way.
Also, try limiting the measurement bandwidth (hit the 20 MHz button on your scope). In my neck of the woods, I get EMI from a local FM transmitter, so all my measurements have 20 mVpp of 91.7 MHz on them unless I hit the BWL.
If you're still seeing garbage on the supply, you'll need to figure out where it comes from and clean it up.
How much hum are you measuring on the amp output? What does the spectrum look like?
~Tom
Also, try limiting the measurement bandwidth (hit the 20 MHz button on your scope). In my neck of the woods, I get EMI from a local FM transmitter, so all my measurements have 20 mVpp of 91.7 MHz on them unless I hit the BWL.
If you're still seeing garbage on the supply, you'll need to figure out where it comes from and clean it up.
How much hum are you measuring on the amp output? What does the spectrum look like?
~Tom
Tom,
All I have are the measurements you saw. However, when connected to my power amp there is a lot of 120 Hz buzz or something. Sounds like a Fender amp with the input unshorted and the volume cranked up, but pretty loud.
I can't imagine a ground loop between them because the preamp connects via a short 3' lead to a active unbalanced to balanced converter box (using a OPA2134 & DRV134), which feeds my amp that has balanced line input (via Cinemag input transformers).
That unbalanced to balance box works perfectly with my PAS preamp.
I do not believe there is an internal ground loop in the amp, either. See the diagram below.
The DC 12 V heater is elevated by tying pin 9 of one input tube to +83 volts from the HV power supply.
Yes, I tried disconnecting the +83 volts and it did nothing. Actually, I pulled that input tube and no change in the buzz.
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.