Tried to find an answer in some of the books but got nowehere. So perhaps some of the guru's can (want to) help out here?
We have from Duncan's pages the wonderfull PSU designer software.
With the software it is easy to "go overboard" and try to desing a PSU with nanoVolts ripple but for a normal SE (EL34 / KT66 / 6L6) what is a figure to aim for and what is "good enough"? SHould the B+ have less than 1V ripple or is it more or is it less?
Thanks a lot, AM.
We have from Duncan's pages the wonderfull PSU designer software.
With the software it is easy to "go overboard" and try to desing a PSU with nanoVolts ripple but for a normal SE (EL34 / KT66 / 6L6) what is a figure to aim for and what is "good enough"? SHould the B+ have less than 1V ripple or is it more or is it less?
Thanks a lot, AM.
near field listening, background music while working on computer. speakers are approx 92db/1 meter/1 watt
Hello,
You want a book check out Morgan Jones’ book, Building Tube Amplifiers. Morgan goes into finite detail regarding ripple voltage. If you want to see ready to use technology read Tubelab.com’s power supply information for his Simple SE amplifier; buy his PCB. The part values are called out on the BOM. Use the recommended extra choke and motor start capacitor.
In my opinion if I can hear hum or buzz it is not High-Fi.
For small output watts I prefer regulated power supplies, check out TubeCad.com. John has some nice kits that make thing fast and easy.
BTW it is not just hum that is important for a SET power supply. By the time the power supply includes capacitors, resistors and chokes the power supply has impedance in the tens of ohms. The power supply impedance is in series with the load and consumes power that otherwise would go into the load, adds to the output impeadnce. Also this series impedance is not constant with frequency. Put it another way the power supply will color the sound.
DT
All just for fun!
You want a book check out Morgan Jones’ book, Building Tube Amplifiers. Morgan goes into finite detail regarding ripple voltage. If you want to see ready to use technology read Tubelab.com’s power supply information for his Simple SE amplifier; buy his PCB. The part values are called out on the BOM. Use the recommended extra choke and motor start capacitor.
In my opinion if I can hear hum or buzz it is not High-Fi.
For small output watts I prefer regulated power supplies, check out TubeCad.com. John has some nice kits that make thing fast and easy.
BTW it is not just hum that is important for a SET power supply. By the time the power supply includes capacitors, resistors and chokes the power supply has impedance in the tens of ohms. The power supply impedance is in series with the load and consumes power that otherwise would go into the load, adds to the output impeadnce. Also this series impedance is not constant with frequency. Put it another way the power supply will color the sound.
DT
All just for fun!
Well it also depends on how sensitive the amplifier is to power supply ripple. An amplifier with large amounts of feedback is much less sensitive to ripple than a simple no-feedback model.
Having said that, in my memory it seems I have seen 'normal' ripple of about 2%. So 350V would mean 7V of ripple.
So the answer to the original question is, it depends. I will stand by that.
Having said that, in my memory it seems I have seen 'normal' ripple of about 2%. So 350V would mean 7V of ripple.
So the answer to the original question is, it depends. I will stand by that.
Thanks for the thumbs up on the Jones' book, I'll look into that. I want to stay away from solid state parts, they seem to have a history of quickly becoming obsolete.
The plate in a pentode is not too sensitive to the voltage variation but the g2 is and indeed it is important to know the configuration (global feedback, no feedback or Schade feedback) and the way g2 is being fed.
edit: are you referring to Jones' book "Valve amplifiers" with the theory or the book "Building tube amplifiers" which deals with the physical construction and testing the amp? Thanks
AM
The plate in a pentode is not too sensitive to the voltage variation but the g2 is and indeed it is important to know the configuration (global feedback, no feedback or Schade feedback) and the way g2 is being fed.
edit: are you referring to Jones' book "Valve amplifiers" with the theory or the book "Building tube amplifiers" which deals with the physical construction and testing the amp? Thanks
AM
Last edited:
I can still get parts for my 20 year old Adcom 555 if I need them.
If you don't want to use solid state for the fun of it or some technical reason, fine, but because it might become obsolete? Even if the 317 regulator becomes obsolete there will be a replacement that will do as good or better job. Even MJ has no objection to sand in this application. 😉
Indeed !
Look at the power stage as three elements in serie:
- The (more or less noisy) PSU.
- The primary (Zp) of the output transformer
- The internal resistance (so called Rp) of the tube.
The PSU noise that appears at the OPT primary (and that will be transmitted to the LoudSPeaker) will divide itself by (Rp + Zp) / Zp. The larger this value, the lower the noise in the LSP.
That's why pentodes having high Rp are "less sensitive" to PSU noise than triodes.
But, applying some amount of feed back results in a reduction of the effective Rp. This is routinely done to improve the damping factor.
So, the first question should be "wich damping factor will make my LSP happy".
The second one should be "wich noise level may I accept" expressed as a ratio with "normal listening level".
(Not less than 50 dB, that represents a 1/100000 power ratio -about 1/300 voltage ratio- , or also 10µW for 1 W "normal listening level" !)
In a pratical 8 ohm LSP, this gives roughly 2.8 V for 1 W and less than (2.8 / 300) around 9 millivolts of residual noise (not easy to trace even with a good o'scope !).
The answers will give the necessary internal output impedance (the reflected effective Rp) to reach, and will allow to compute the allowable PSU residual noise from the above formula.
Yves.
I know of a 10K US transceiver that 3 years after its introduction had unobtainable parts: the RF output finals were no longer available and there was no easy substitution without having to do extensive work. Basically if you blew the output power transistors then the rig was good for the rubbish tip as far as transmitting was concerned. Other issues are often with LCD control screens that more often than not are unobtainable after a couple of years.
For the Simple SE amp from tubelab.com one cannot find in our country the CCS mosfet and it is very hard to obtain it even worldwide. Some of the places won't ship outside their country, others charge ridiculous amounts when shipping and others just send fakes and not even the originals.
AM
not quite true AM - the 10M45 has been done as a group buy to NZ and I'd be interested if anyone was doing another one!
Having said that, there is plenty of scope to use VERY common sand devices to build shunt regulators and the like to tame ripple. MJE340/350 series BJTs come to mind...
Having said that, there is plenty of scope to use VERY common sand devices to build shunt regulators and the like to tame ripple. MJE340/350 series BJTs come to mind...
Thanks but the maths will likely be a bit above my capabilities. However using a Schade feedback means that the Rp gets a lot less than the published figures resulting that perhaps the (Rp+Zp)/Zp may become something like 2 or 3.
From the 50db or 1/300 I would deduce that when using 350V on the plate of a 6L6GC we may not have anything above 1V ripple. The Schade feedback (10% recommended by Schade?) means that it would drop to more than 50dB but from the Morgan Jones book I gather we should be aiming more towards a lot higher figure like 90 dB.....
Looks like I may have to install some self referencing series regulator that will smooth the PSU. I saw something with two mosfets on a French speaking website.... USe google to translate but even the picture speaks for itself: Rgulateur Alimentation Haute Tension
"The emperor's new amplifier" also has some relative regulators. This allows for sagging mains and still regulates. The Emperor's New Amplifier
I do not like silicon! Yikes!
AM
I used a 'Maida' style reg in my latest amp. MJ mentions it in his book. LM317 regulator and a MOSFET pass device. The devices were available locally. Having said that, however, my previous valve amps just used CLC or CRC filters with very little hum.
I managed to get some of the current regs for the Simple SE from one of the US suppliers, but yes, the shipping is expensive. Last time I got stuff from the US I had 20 of them shipped for future experiments/projects. I am using a few of them in my current amp with CCS diff amps in the driver. If you are building from scratch, some of the CCS schematics mentioned in Morgan Jones can be built using easy to obtain parts.
I managed to get some of the current regs for the Simple SE from one of the US suppliers, but yes, the shipping is expensive. Last time I got stuff from the US I had 20 of them shipped for future experiments/projects. I am using a few of them in my current amp with CCS diff amps in the driver. If you are building from scratch, some of the CCS schematics mentioned in Morgan Jones can be built using easy to obtain parts.
You got it !
This is an highly respectable goal 😉
I suggested 50 dB below a "normal listening level" of about 1W, but if you refer to a "max power lower" of, says 10 W, this number becomes 60 dB (add 10 dB each time you multiply the power by 10).
90 dB below 10W represent 10 nanowats !
I'm the author of this paper, ask me for more if needed . . . even if my english is not really better than Google translator 🙄
If so, use more iron for better smoothing 🙂I do not like silicon! Yikes!
"Shade like" feedback (also know as "plate to plate") has the specificity of lowering Rp and thus making the power stage more sensitive to supply noise.
Global feed back is better on this aspect because it returns the noise back in amplifier in such a way that it tend to cancel itself.
In other words, you reinject some "out of phase" noise component.
There are other ways like "noise compensation" wich deliberatly inject some noise. Look at this:
http://www.dissident-audio.com/SE_6V6/Da33-4.gif
Shade like feedback is applied by R5 + R12 to the plate of the lower triode of the SRPP (where the impedance is high enough for the feedback voltage not being derived to ground).
Its cathode receive some residual 100 Hz ripple picked in the PSU trough R4 and C6.
R4 must be adjusted for the best rejection.
However, as for Global Feed Back, this may introduce some IMD products btwn the signal and the 100 Hz.
Nobody's perfect 😎
Yves.
Iron has been bought and it is too expensive to get more. The chassis has been made and no more room for more iron on top and chassis is internally 48 mm high so not much room there.
Question: if I have 1V ripple @ 350V and then use your regulator: how much ripple rejection can be expected?
I noticed that in the "Emperor's New Amplifier" the author wrote about noticing a slight hum so wanting more ripple rejection. I suppose that a very small (low current) many Henries choke can be used which can be very cheap. I do have a number of IRF820's here so can experiment with it....
AM
If I remember correctly: I've seen a schematic where there wasa second triode across the output(plate) of of the driver to the power tube and it was being fed from (if I remember correctly) a variable voltage divider from the 6.3V heater.
Another system I've seen was feeding the ripple from B+ into a tap on the grid resistor to the power tube.
In my un-educated opinion I think using the B+ ripple is the way to go since not necessarily the heater ripple will identical in shape. Plus using the B+ will help reducing crosstalk that is introduced by the other power amplifier onto the B+.
AM
Another system I've seen was feeding the ripple from B+ into a tap on the grid resistor to the power tube.
In my un-educated opinion I think using the B+ ripple is the way to go since not necessarily the heater ripple will identical in shape. Plus using the B+ will help reducing crosstalk that is introduced by the other power amplifier onto the B+.
AM
In a recently built 300B amplifier, I have some 5 volts ripple at the recto output (a simple bridge followed by a simple 300µF cap) and less than 100mV after the regulator.
http://www.dissident-audio.com/SE_300B/KISS/2.gif
http://www.dissident-audio.com/SE_300B/KISS/1.gif
The Rp of a 300B is roughly 500 ohms.
I can hear some hiss with the "head in the loudspeaker" but no hum, at least with 92dB speakers.
Keep in mind that, as any regulator, it must have some voltage to drop, here, even if the IRF820 itself can reach very low value of serie resistance, you must drive its gate some 4 volts above its source (output) voltage.
Too low voltage drop and you loose regulation, too high and you must enlarge the heat dissipator 🙄
Power supply noise IS NOT the unique source of hum 😉
Yves.
Just looking back at your requirements, I have built about three or four valve amps that have been used in very similar circumstances. All have used CLC or CRC power supply filtering successfully with hum that could only barely be heard with ear right inside the speaker cone. With the size of capacitors now days, the simple option might just be to add another RC section. Morgan Jones book 'Valve Amplifiers' has a really good chapter on power supply design and effective use of filters. For a couple of my amps I just blatantly copied the power supply of Tubelab's Simple SE (I had already purchased the board and built the amp, hence I knew it worked).
As mentioned by Yves, remember too that there are a bunch of other noise problems in the amp to do with layout. You may have a great power supply, but poor grounding will ruin the result, as will bad signal and heater wiring. This is a great rundown on grounding Star Grounding Also, I have had ground loop and noise issues trying to hook up amps to computers. As much as I like Apple, the worst was an iMac.
Cheers,
Chris
As mentioned by Yves, remember too that there are a bunch of other noise problems in the amp to do with layout. You may have a great power supply, but poor grounding will ruin the result, as will bad signal and heater wiring. This is a great rundown on grounding Star Grounding Also, I have had ground loop and noise issues trying to hook up amps to computers. As much as I like Apple, the worst was an iMac.
Cheers,
Chris
Hello,
The Morgan Jones book I was referring to is the 3rd edition of “Valve Amplifiers” of the pair of books this is the theory book. Speaking of g1 and g2 you can go from full pentode to full triode or in between with UL. You can use g2 as an off specification sheet inverting input to inject a % of raw B+ for noise cancelation.
Get out the kitchen breadboard and prototype using tubes that you do mind destroying. Put this week’s best result into the chassis. There is no final project.
Using the breadboard approach I found that if the power supply transformers were near signal parts of the amplifier hum would creep into the output. My thought is to put the power supply iron, rectification, smoothing capacitors in a separate chassis and regulator (if there is one), or final power supply capacitors in the amplifier chassis.
DT
All just for fun!
The Morgan Jones book I was referring to is the 3rd edition of “Valve Amplifiers” of the pair of books this is the theory book. Speaking of g1 and g2 you can go from full pentode to full triode or in between with UL. You can use g2 as an off specification sheet inverting input to inject a % of raw B+ for noise cancelation.
Get out the kitchen breadboard and prototype using tubes that you do mind destroying. Put this week’s best result into the chassis. There is no final project.
Using the breadboard approach I found that if the power supply transformers were near signal parts of the amplifier hum would creep into the output. My thought is to put the power supply iron, rectification, smoothing capacitors in a separate chassis and regulator (if there is one), or final power supply capacitors in the amplifier chassis.
DT
All just for fun!
When I simulate a simple CLC filter on PSUD II for my upcoming amp, I have 450V, 10u, 8H/30R, 100u only and the calculated ripple is already less than 0.14V
I am not sure it is worth much more effort to improve this. I will see how close this is when up and running, but I can cut it in half simply by doubling C2.
Is more complex ripple reduction going to make a difference? Especially since I am new at this and likely will have other sources of noise simply due to my construction/design inexperience.
I am not sure it is worth much more effort to improve this. I will see how close this is when up and running, but I can cut it in half simply by doubling C2.
Is more complex ripple reduction going to make a difference? Especially since I am new at this and likely will have other sources of noise simply due to my construction/design inexperience.
How about connecting a signal generator set on 120 Hz (or 100 Hz) to the speaker through a 1000 ohm resistor (to reduce the load on the generator). Then measure the voltage at the speaker terminals when the generator is set to produce an acceptably low hum level in a quiet room.
When I do this with my 92db JBL 4312s when the room is quiet, I can hear 10mV of 120 Hz hum in a single speaker in the room. For near-field listening I can hear about 2mV when I'm 3 feet away from the speaker.
Based on this I would say my requirement for studio monitoring amps driving box speakers is on the order of 1mV at the speaker for near field and 5mV for mid-field or room monitoring.
Hum threshold seems more to be related to the quiet level of your listening room vs. the music level you'll be playing, so it's less meaningful to me to relate hum power to the amp's output power.
The real problem is that it depends on the speaker, the room , and the listener, making it hard to come up with a universal standard. My standard is if I can ever hear the hum in normal use, it's too much.
There is also the question of inaudible hum becoming audible through intermodulation with the music. This could be measured with a spectrum analyzer but that tells you nothing about audibility in itself. Presumably one would not hear the hum per se but it would cause a decrease in transparency and clarity.
I like the idea of the regulator or floating ripple reducer using SS. The ripple reducer does have a settling time, much like a damped RC but less droop relative to the ripple reduction.
Beware of ripple cancelling schemes that only suppress the ripple at idle. These may induce copius intermodulation products into the output.
I think the output ripple voltage can be estimated by multiplying the B+ ripple by Zp/(Rp+Zp) and then dividing by the turns ratio of the OPT.
With Schade feedback, Rp is the Schade-effective Rp so yes, local feedback increases ripple sensitivity by lowering Rp, but most output stages with decent damping factors are already between 0.7 and 1 ripple divider ratio, so it's not going to increase output ripple by more than a db or so vs. a 500 ohm triode. A pentode amp with global feedback could do better of course...
When I do this with my 92db JBL 4312s when the room is quiet, I can hear 10mV of 120 Hz hum in a single speaker in the room. For near-field listening I can hear about 2mV when I'm 3 feet away from the speaker.
Based on this I would say my requirement for studio monitoring amps driving box speakers is on the order of 1mV at the speaker for near field and 5mV for mid-field or room monitoring.
Hum threshold seems more to be related to the quiet level of your listening room vs. the music level you'll be playing, so it's less meaningful to me to relate hum power to the amp's output power.
The real problem is that it depends on the speaker, the room , and the listener, making it hard to come up with a universal standard. My standard is if I can ever hear the hum in normal use, it's too much.
There is also the question of inaudible hum becoming audible through intermodulation with the music. This could be measured with a spectrum analyzer but that tells you nothing about audibility in itself. Presumably one would not hear the hum per se but it would cause a decrease in transparency and clarity.
I like the idea of the regulator or floating ripple reducer using SS. The ripple reducer does have a settling time, much like a damped RC but less droop relative to the ripple reduction.
Beware of ripple cancelling schemes that only suppress the ripple at idle. These may induce copius intermodulation products into the output.
I think the output ripple voltage can be estimated by multiplying the B+ ripple by Zp/(Rp+Zp) and then dividing by the turns ratio of the OPT.
With Schade feedback, Rp is the Schade-effective Rp so yes, local feedback increases ripple sensitivity by lowering Rp, but most output stages with decent damping factors are already between 0.7 and 1 ripple divider ratio, so it's not going to increase output ripple by more than a db or so vs. a 500 ohm triode. A pentode amp with global feedback could do better of course...
Last edited: