Working on a prototype 5V regulated supply for 300B filaments, using LM1086.
Schematic
PSUD output
Testing shows 5 mV RF hash on the output. This was traced down to rectifier switching / choke ringing, exactly like Morgan Jones' describes in Valve Amplifiers. This is a look at the transformer secondary:
I tried his Y-caps to ground method, to little avail. If anything, the frequency dropped, but the magnitude increased a little. Using schottky diodes with 0.01 uF bypass caps. Putting large (> 1uF) caps across the choke helps; I've considered playing around with much larger, possibly an RC network.
Any suggestions? I've read a number of other threads, but the consensus seemed to be throw snubbers in a whole lot of places. I've also read Jim Hagerman's article - I need to buy a few parts to do a little trial and error with his equations. I'm hoping someone has built a similar supply, and can ballpark me some values (this ain't the first DC supply for 300B's, though most seem to be cap input).
This RF business is a real ball buster.
Schematic
PSUD output
Testing shows 5 mV RF hash on the output. This was traced down to rectifier switching / choke ringing, exactly like Morgan Jones' describes in Valve Amplifiers. This is a look at the transformer secondary:
An externally hosted image should be here but it was not working when we last tested it.
I tried his Y-caps to ground method, to little avail. If anything, the frequency dropped, but the magnitude increased a little. Using schottky diodes with 0.01 uF bypass caps. Putting large (> 1uF) caps across the choke helps; I've considered playing around with much larger, possibly an RC network.
Any suggestions? I've read a number of other threads, but the consensus seemed to be throw snubbers in a whole lot of places. I've also read Jim Hagerman's article - I need to buy a few parts to do a little trial and error with his equations. I'm hoping someone has built a similar supply, and can ballpark me some values (this ain't the first DC supply for 300B's, though most seem to be cap input).
This RF business is a real ball buster.
Are the 0.01uF caps in parallel with the diodes and if so are they ceramic types?
Also how much load current were you drawing when you made this measurement?
How about a very small cap from the rectifiers to ground. (Not a large enough value to make a valid pi filter however.)
The ringing looks like it is relatively low frequency, not quite sure what you can do to damp it.. One other thought would be a resistor in parallel with the choke - say a few hundred ohms..
Also how much load current were you drawing when you made this measurement?
How about a very small cap from the rectifiers to ground. (Not a large enough value to make a valid pi filter however.)
The ringing looks like it is relatively low frequency, not quite sure what you can do to damp it.. One other thought would be a resistor in parallel with the choke - say a few hundred ohms..
The caps are in parallel with the diodes, very close proximity. They are polyester film.
The load current varied from 1.2A to 1.5A, depending on what mfr of 300B I elect to use. The waveform really didn't change at all between these current levels.
I'll try the damping resistor idea, also maybe an RC.
While it looks low frequency (whatever you classify as low frequency), I see a very clear 10 MHz sine wave on the output, and at the adjust pin of the regulator. Scope is only a 20MHz model.
The load current varied from 1.2A to 1.5A, depending on what mfr of 300B I elect to use. The waveform really didn't change at all between these current levels.
I'll try the damping resistor idea, also maybe an RC.
While it looks low frequency (whatever you classify as low frequency), I see a very clear 10 MHz sine wave on the output, and at the adjust pin of the regulator. Scope is only a 20MHz model.
Hi Zigzag,
Film caps don't make very good snubbers for rf ringing, ceramic types are really better for this spot, but are you certain that the regulator itself is not oscillating?
Some of these family of devices like some capacitance (>33uF) on the output particularly if they are driving an inductive load. I noted the presence of a resistor (R5) in series with the single C (C7) on the output of your regulator. This could be a recipe for RF oscillation with LT10XX series and LM317/350 series regulators. How far are the filter caps from the input of the regulator IC and what is the value and type of the cap right by the regulator input?
Kevin
Film caps don't make very good snubbers for rf ringing, ceramic types are really better for this spot, but are you certain that the regulator itself is not oscillating?
Some of these family of devices like some capacitance (>33uF) on the output particularly if they are driving an inductive load. I noted the presence of a resistor (R5) in series with the single C (C7) on the output of your regulator. This could be a recipe for RF oscillation with LT10XX series and LM317/350 series regulators. How far are the filter caps from the input of the regulator IC and what is the value and type of the cap right by the regulator input?
Kevin
Huh, always thought film caps were the way to go for snubbing. I've got some ceramics in the bin, maybe 0.1 uF. I'll give it a try.
Kind of followed the app notes from National on bypassing. Output is using a 1 ohm resistor (R5) in series with a 22 uF solid tantalum cap (C7). Out of curiosity, I shorted R5 to see if there was any difference; nothing.
Input cap C11 is a 22 uF tantalum. Bypass cap C9 is a 47 uF tantalum. BOM included for any further review. They are both mounted extremely close to the pins of the regulator. As the schematic indicates, I have three star grounding points for the three 'stages' of the power supply. The caps are grounded in this fashion.
I was assuming this ringing on the AC wave is working its way into the regulator, where it can't be attenuated, and passes through. You don't think this is the case?
Thanks for the help.
Kind of followed the app notes from National on bypassing. Output is using a 1 ohm resistor (R5) in series with a 22 uF solid tantalum cap (C7). Out of curiosity, I shorted R5 to see if there was any difference; nothing.
Input cap C11 is a 22 uF tantalum. Bypass cap C9 is a 47 uF tantalum. BOM included for any further review. They are both mounted extremely close to the pins of the regulator. As the schematic indicates, I have three star grounding points for the three 'stages' of the power supply. The caps are grounded in this fashion.
I was assuming this ringing on the AC wave is working its way into the regulator, where it can't be attenuated, and passes through. You don't think this is the case?
Thanks for the help.
Zigzagflux,
The ringing shown in the oscillogram is at about 10 KHz - nothing to do with rf. To work on that film caps would be quite OK, and closeness of components/short leads will have insignificant influence. The 10 MHz that you observed (but not shown in the oscillogram) is a different matter.
Exactly how is your audible output affected? It is quite unlikely that the 10 MHz will have any effect in this type of amplifier (I presume you saw it on the power supply and not the amp output). If it is the 10 KHz, it is most likely to interfere through induction/radiation, although the latter possibility is also remote. (You are probably not using the 300B as a low level pre-amp stage!) Although one may want to have a clean dc out of principle, I am a little puzzled as to what audible effect this could have, that is why I ask.
Without further information I would put wiring high on my investigation list - common wire resistance, that sort of thing. But as Kevinkr said, it can also quite likely be regulator chip oscillation. That will be easily curable, as you do not need regulation response at high frequency.
Regards
The ringing shown in the oscillogram is at about 10 KHz - nothing to do with rf. To work on that film caps would be quite OK, and closeness of components/short leads will have insignificant influence. The 10 MHz that you observed (but not shown in the oscillogram) is a different matter.
Exactly how is your audible output affected? It is quite unlikely that the 10 MHz will have any effect in this type of amplifier (I presume you saw it on the power supply and not the amp output). If it is the 10 KHz, it is most likely to interfere through induction/radiation, although the latter possibility is also remote. (You are probably not using the 300B as a low level pre-amp stage!) Although one may want to have a clean dc out of principle, I am a little puzzled as to what audible effect this could have, that is why I ask.
Without further information I would put wiring high on my investigation list - common wire resistance, that sort of thing. But as Kevinkr said, it can also quite likely be regulator chip oscillation. That will be easily curable, as you do not need regulation response at high frequency.
Regards
Good questions.
My concern with the ringing is primarily that it shouldn't be there; if it's on the secondary, it also finds its way into the primary, and therefore has the opportunity to migrate into other circuitry. This looks very similar to the effect of a light dimmer, which does produce an audible effect on my sound. It is not present on the DC output, but the 10Mhz is there, drives right through the LCL filtering.
There are lesser concerns, such as insulation stress due to the high transients. Morgan Jones delves into some of this in his book, along with modulation products.
As far as the 10MHz, I certainly don't claim it to be audible; however, I have had other amplifiers that were plagued with RF, and the whole soundstage was just depressed, dry, and cloudy. No, you couldn't hear it, but you could tell something just wasn't right. Sure, it could have been something else, but repairing the RF (different topology, thanks Kevin) drastically improved the sound, especially with fatigue.
I am working towards a Karna amp in about 3-4 years, and am in the first stages of establishing some of the building blocks (can I build a quiet filament supply, do I like the sound of xfmr coupling, etc). In agreeement with those such as K&K and Lynn Olson, it ain't easy to build a quiet DC supply, and I am learning that. They continue to use AC heating for a reason (except their active RAKK DAC stage), and since I know too little to debate their philosophies, I do follow their advice.
If I can't get a quiet supply (such that my 20Mhz scope can't see it both at the AC input and the DC output), I will default to AC heating, and deal with the hum issues.
So, I want to at least try to get this supply (input and output) quiet. A good learning experience, even if it fails. I'll poke around with more solid grounding, and maybe even put in a DC source as a preregulator; maybe a car battery. That would certainly prove the issue is within the regulator.
My concern with the ringing is primarily that it shouldn't be there; if it's on the secondary, it also finds its way into the primary, and therefore has the opportunity to migrate into other circuitry. This looks very similar to the effect of a light dimmer, which does produce an audible effect on my sound. It is not present on the DC output, but the 10Mhz is there, drives right through the LCL filtering.
There are lesser concerns, such as insulation stress due to the high transients. Morgan Jones delves into some of this in his book, along with modulation products.
As far as the 10MHz, I certainly don't claim it to be audible; however, I have had other amplifiers that were plagued with RF, and the whole soundstage was just depressed, dry, and cloudy. No, you couldn't hear it, but you could tell something just wasn't right. Sure, it could have been something else, but repairing the RF (different topology, thanks Kevin) drastically improved the sound, especially with fatigue.
I am working towards a Karna amp in about 3-4 years, and am in the first stages of establishing some of the building blocks (can I build a quiet filament supply, do I like the sound of xfmr coupling, etc). In agreeement with those such as K&K and Lynn Olson, it ain't easy to build a quiet DC supply, and I am learning that. They continue to use AC heating for a reason (except their active RAKK DAC stage), and since I know too little to debate their philosophies, I do follow their advice.
If I can't get a quiet supply (such that my 20Mhz scope can't see it both at the AC input and the DC output), I will default to AC heating, and deal with the hum issues.
So, I want to at least try to get this supply (input and output) quiet. A good learning experience, even if it fails. I'll poke around with more solid grounding, and maybe even put in a DC source as a preregulator; maybe a car battery. That would certainly prove the issue is within the regulator.
zigzagflux said:
Looks to me like the bypass caps you are using are resonating with the leakage inductance of your transformer. C10 is the worst culprit. C5 & C6 are not helping either.
Go back and re-read Jim Hagerman's articles on snubbers.
When you are reading, please realize that there are two different definitions of "snubbers" that get thrown around on these boards. While they are both series RC networks, the ones that Hagerman talks about are associated with the transformer and its leakage inductance. There are other people that use a "snubber" in parallel with an electrolytic cap as a "high-frequency bypass".
In your case, you want to focus on Hagerman's definition. Skip the "snubbers" that bypass the electrolytics.
Try adding resistors in series with C6 and C5, and making them bigger. Maybe 0.1uF and 47 ohms or so.
The film caps by themselves don't dissipate the energy during reverse recovery; they are returning it to the circuit. So you either need resistance, or a really bad cap. That's why ceramics are sometimes used here - they are more lossy.
I agree that C10 might hurt, not help.
What is Q1?
Why the resistor ini series with C7?
Pete
The film caps by themselves don't dissipate the energy during reverse recovery; they are returning it to the circuit. So you either need resistance, or a really bad cap. That's why ceramics are sometimes used here - they are more lossy.
I agree that C10 might hurt, not help.
What is Q1?
Why the resistor ini series with C7?
Pete
GOT IT !!!
Whooohooooo !!!!
All very good suggestions, everyone. First, I will answer the latest questions, since you took the time to help. Then, I'll explain my solution.
C10 was added per someone else's suggestion at a different forum. In both this case, and the C5-C6 case, I was always a little confused as to why resistance was never added in series. Your suggestions confirm why, as well as Jim Hagerman's article. Removed C10, with no change. It will be left out. Fewer parts are always better.
Q1 is the LM1086, linear regulator.
The resistor in series with C7 was used to prevent the parasitic inductance of the regulator from resonating with the capacitance of C7. Being that tantalums have fairly good HF performance, a small, 1-5 ohm resistor is recommended by National and Morgan Jones. Shorting it out made no difference, anyway.
So, the solution was........(drum roll, please)...........an RC snubber. What a surprise, huh?
Tried numerous cap sizes and R sizes. In the end, a 0.47uF in series with 100 ohm worked best. Completely removed the ringing. What was interesting is that it worked best in two locations, identically. Right across the choke terminals was one method. The other method was per Kevin, from output of rectifiers to circuit common. In this position, though, cap alone did not work; it still wanted the series R of 100 ohm.
To be honest, that solution doesn't surprise me, and I kind of expected that to be the easier solution. Fixing the megahertz, tho, I expected to be very difficult. Once again, wrong was I (I'm getting used to being wrong).
Two words: Airport Express
The noise was being picked up by the flying leads used in the test setup. Keep the leads short and tight, noise goes way down. Turn off the Airport Express, silence. In actual use, I will be installing with shielded twisted pair, so I don't see this as being a big problem when actually installed in the amp chassis.
So, all is clean, and I'm ready to build three more of these for the other filaments. Not that this is a stellar design or anything, but the schematic and BOM is in this thread for anyone's use. It's been troubleshot enough that I think it can be classified as a fairly low noise design.
Thanks, everyone, for your input.
Whooohooooo !!!!
All very good suggestions, everyone. First, I will answer the latest questions, since you took the time to help. Then, I'll explain my solution.
C10 was added per someone else's suggestion at a different forum. In both this case, and the C5-C6 case, I was always a little confused as to why resistance was never added in series. Your suggestions confirm why, as well as Jim Hagerman's article. Removed C10, with no change. It will be left out. Fewer parts are always better.
Q1 is the LM1086, linear regulator.
The resistor in series with C7 was used to prevent the parasitic inductance of the regulator from resonating with the capacitance of C7. Being that tantalums have fairly good HF performance, a small, 1-5 ohm resistor is recommended by National and Morgan Jones. Shorting it out made no difference, anyway.
So, the solution was........(drum roll, please)...........an RC snubber. What a surprise, huh?
Tried numerous cap sizes and R sizes. In the end, a 0.47uF in series with 100 ohm worked best. Completely removed the ringing. What was interesting is that it worked best in two locations, identically. Right across the choke terminals was one method. The other method was per Kevin, from output of rectifiers to circuit common. In this position, though, cap alone did not work; it still wanted the series R of 100 ohm.
To be honest, that solution doesn't surprise me, and I kind of expected that to be the easier solution. Fixing the megahertz, tho, I expected to be very difficult. Once again, wrong was I (I'm getting used to being wrong).
Two words: Airport Express
The noise was being picked up by the flying leads used in the test setup. Keep the leads short and tight, noise goes way down. Turn off the Airport Express, silence. In actual use, I will be installing with shielded twisted pair, so I don't see this as being a big problem when actually installed in the amp chassis.
So, all is clean, and I'm ready to build three more of these for the other filaments. Not that this is a stellar design or anything, but the schematic and BOM is in this thread for anyone's use. It's been troubleshot enough that I think it can be classified as a fairly low noise design.
Thanks, everyone, for your input.
zigzagflux said:
If you do some listening tests, you will find that any Wi-Fi network will degrade the sound of your audio system. So if you really care about sound quality, I would recommend replacing your wireless Ethernet with a wired Ethernet setup. I just had my house wired up by an installation company for around $300.
Plus Wi-Fi operates at the exact same frequency as a microwave oven. Turning your house into a low-level microwave oven that is on 24/7 doesn't seem like the best idea in the world to me from a health standpoint.
zigzagflux said:
That sounds (as Charles Hansen suggested) like transformer leakage inductance. In one position (from rectifier output to ground) your snubber stopped the ringing, and in the other (across the choke) it prevented the kick that would have started the ringing. I often think of this sort of problem as being like a bell; you can stop it ringing by damping the bell and by not striking it...
I would agree, it is the transformer that determines the ring characteristic. I have implemented the same solution in two different DC supplies I have (one in a linestage, and this one for the power amp).
The linestage transformer is sized differently than the power amp transformer - other than that, the front end of each DC supply is almost identical. Yet, they required vastly different values of RC. Linestage required 4uF, 10 ohm, and the power amp required 0.47uF, 100 ohm. In the end, the results of each were the same, no more ringing. And essentially independent of load, as long as critical inductance is met, I suspect.
Biggest difference was the power transformers.
The linestage transformer is sized differently than the power amp transformer - other than that, the front end of each DC supply is almost identical. Yet, they required vastly different values of RC. Linestage required 4uF, 10 ohm, and the power amp required 0.47uF, 100 ohm. In the end, the results of each were the same, no more ringing. And essentially independent of load, as long as critical inductance is met, I suspect.
Biggest difference was the power transformers.
zigzagflux said:
It would be interesting to know the winding inductances, leakage inductance, parasitic C, series R, and some things about the core (magnetic) behavior of those transformers. This isn't something I have paid enough attention to over the years, but this doesn't seem too surprising. Knowing these things you might be able to simulate some of this behavior in spice..
Good that you were able to get to the bottom of all of these issues, it is always a good feeling knowing that your design is technically sound, and any compromises made are deliberate and not forced on you due to a lack of understanding..
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.