Not covered, and I should have thought about it myself, as I have shorting plugs made up.
Quick test: there is still some buzz, but it seems somewhat muted to what I have been hearing, something like half? I'll do more later, as I don't think it is a good idea to power cycle the F5m, due to the thermistors doing some inrush duties.
That sums it up for me, and I feel my fine motor is very good.
Interesting to see your results on the small amount of AC on the offset. I guess I (we) should also try making that measurement with the RCA inputs shorted.
Amplifier setup and adjustment is always performed with inputs shorted (I'm guilty of not following my own advice sometimes). Sounds like you might have several factors contributing to the unwanted noise. Maybe not enough distance due to form factor, maybe layout and wire dress (real nasty big current spikes in those long green / blue secondary wires and rectifier), maybe some undesired ground current due to loop area, etc. Is the heatsink electrically connected (= zero ohms) to the chassis?
I started with everything at an ambient temp of 10C, remembered to set the logger within a few seconds of power up and did not see a repeat of this mornings weirdness. I have had my inputs shorted during all testing btw.
I wouldn't have considered half a mV problematic off the top of my head but will leave that to the more knowledgeable folks. Cheers
I wouldn't have considered half a mV problematic off the top of my head but will leave that to the more knowledgeable folks. Cheers
@william2001 the reading in question was a small amount of ac on the OPs output. When @Ben Mah stated that it looked a little high, I wondered if he had missed the decimal point but was standing at the bench so I spun the dial to mVac out of curiosity.
Remembering ZM's advice on the subject (but Remembering it as 50-60mv) I tweaked the dc offset beyond that necessary just because I could.
Remembering ZM's advice on the subject (but Remembering it as 50-60mv) I tweaked the dc offset beyond that necessary just because I could.
I have some curious info to report on the low-level noise (buzz) issue I’ve been experiencing. This morning, I’ve been able to determine that this noise is all but absent at the start-up of a dead-cold unit (off for several hours), becomes very faintly audible within 2 minutes, and builds slowly to a steady-state level by 20 min.
This discovery restores my sanity, as I was getting conflicting results and thought I was making mistakes. In post #1086 I reported a 90-95% reduction in the buzz when I removed the RCA inputs. In retrospect, this was because I checked this on a cold unit. In the middle of post #1096, I reported that disconnecting both RCA inputs resulted in no reduction of noise. This was because I restarted without the RCAs attached a few minutes after shutting down, and the unit was still warm. Lastly, in post #1101, I reported a buzz reduction in about 50% when using RCA shorting plugs, but this was done on a unit that had been off for only about 10 min (I was trying not to power cycle the F5m too quickly).
I haven’t followed this up with measurements of AC on the speaker terminals yet, perhaps I will soon, but when I reported differences in the ACV on the speaker terminals between a cold and warmed up unit (middle of post #1086), I misspoke, and should have described it as the difference between an not-fully warmed up unit and a fully warmed up one.
This discovery raises the question of the source of this noise. I now feel fairly confident that the noise source is in my F5m, as it is not dependent on what is on the RCA inputs: a source, nothing, or shorting plugs.
I don’t know the answer yet, but given the time-dependent behavior, I suspect it has something to do with the 2 thermistors on the rails of the PS, the ones that get hot in operation. I doubt they are the source of the noise itself, but they may be suppressing an existing noise when cold due to their modest resistance, and then unleashing it as they warm up and drop in resistance. The actual noise source is probably some detail in my particular build configuration (transformer location, wire dress, ground currents, etc.), as discussed in recent posts above.
This discovery restores my sanity, as I was getting conflicting results and thought I was making mistakes. In post #1086 I reported a 90-95% reduction in the buzz when I removed the RCA inputs. In retrospect, this was because I checked this on a cold unit. In the middle of post #1096, I reported that disconnecting both RCA inputs resulted in no reduction of noise. This was because I restarted without the RCAs attached a few minutes after shutting down, and the unit was still warm. Lastly, in post #1101, I reported a buzz reduction in about 50% when using RCA shorting plugs, but this was done on a unit that had been off for only about 10 min (I was trying not to power cycle the F5m too quickly).
I haven’t followed this up with measurements of AC on the speaker terminals yet, perhaps I will soon, but when I reported differences in the ACV on the speaker terminals between a cold and warmed up unit (middle of post #1086), I misspoke, and should have described it as the difference between an not-fully warmed up unit and a fully warmed up one.
This discovery raises the question of the source of this noise. I now feel fairly confident that the noise source is in my F5m, as it is not dependent on what is on the RCA inputs: a source, nothing, or shorting plugs.
I don’t know the answer yet, but given the time-dependent behavior, I suspect it has something to do with the 2 thermistors on the rails of the PS, the ones that get hot in operation. I doubt they are the source of the noise itself, but they may be suppressing an existing noise when cold due to their modest resistance, and then unleashing it as they warm up and drop in resistance. The actual noise source is probably some detail in my particular build configuration (transformer location, wire dress, ground currents, etc.), as discussed in recent posts above.
Yes, measures 0Ω. Multiple screws join them.
I think your other points are collectively spot on.
Would there be a benefit from shortening up the transformer secondary wires (I left them long with the idea I might need them if I migrate to a different chassis)?
Would there be a benefit from adding sub-µF caps to the rectifier to minimize noise, as illustrated here (from TI App Note AN1849), or does the design of the F5m PS obviate the need?
For a snubber, you'll want to look into the "Quasimodo" stuff (warning - rabbit hole). For the transformer itself, physical distance is certainly the most effective "fix". If that isn't possible, you're down to a GOSS band / belly band, steel cover, etc. Those problematic secondary wires... If someone threw this amp in front of me I might try to rotate that PS board 90 deg. So the secondary and DC power wires aren't in parallel anymore. DC power wires could just exit stage left there, out and around a bit. With a gun to my head about not cutting secondary wires, I would just pull those back away to the other side of the transformer, cable tied / secured and tucked out of harms way kinda back towards the corner, I guess.
Snubber - M. Johnson's .pdf attached here has some valuable information. I think guys are using 10 ohms for Rs for the Antek 4218. Assuming the standard Cx=10nF and Cs=150nF. With a center tapped secondary arrangement you would of course need two of these networks, as shown in the pdf.
Following up on William's Quasimodo suggestion, there's also the "Quasimodo results" thread (https://www.diyaudio.com/community/threads/quasimodo-results-only.313202/) where members post their measured snubber values.
This is very informative. As the power supply thermistors between the capacitors warm up, their resistance drops from 10 Ohm to less than 1 Ohm. This resistance is part of the CRC filter of the power supply. As the resistance drops, the power supply AC ripple increases., and amplifier output noise increases.
Another factor for CRC power supply AC ripple is power supply current. As current decreases, AC ripple decreases. You can experiment by reducing the bias current, re-zeroing the offset, and listen/measure the amplifier output noise.
One thing to note is that the typical First Watt amplifier power supply has a CRC filter of 2x15mF - 0.12 Ohm - 2x15mF. The F5M CRC filter is 4x4.7mF - approx 0.5 Ohm - 2x4.7mF. Simulating the power supply on PSUD2 and comparing the two assuming each channel draws 1A, the FW supply had 92mV of ripple per rail and the F5M supply had 134mV of ripple per rail.
The much higher power supply ripple may be the reason why you hear some noise from you amplifier output. If you have some extra electrolytic capacitors, you can experiment by adding them to the power supply. For instance adding another 4.7mF capacitor per rail at the power supply output reduces the ripple to 92mV, according to PSUD2.
Thanks william and Dennis. I had looked through MJ's Quasimodo post some time ago, out of curiosity. I note that Rs is part of the CRC snubber between the transformer secondaries and the rectifier. On the F5m PS, there is nothing at all there, so presumably the suggestion is to add one.
I also note that the thermistor in the F5m PS is after the rectifier. Yet, if I am correct in my preliminary interpretation of the effect of the latter thermistor suppressing the noise while cold (certainly an if here, because I am beyond my depth here), I wonder what detriment (other than a small drop in rail voltage) comes from replacing it with a standard resistor of suitable ohm and power rating. I'm sure Nelson had very specific reasons for designing the F5m PS as he did, with the network of caps and the thermistor, so substituting it with a plain resistor is most likely a poor idea. (As I'm typing, Ben Mah posted an interesting note related to this, focused on capacitance rather than resistance; I'll answer separately).
I also realize that what I am saying above may be a little like a band aid, and you are probably suggesting the additional CRC (and the Quasimodo to calculate it) because the real problem is lies with the transformer/wiring in my example, and it would be better to take care of it there.
I also note that the thermistor in the F5m PS is after the rectifier. Yet, if I am correct in my preliminary interpretation of the effect of the latter thermistor suppressing the noise while cold (certainly an if here, because I am beyond my depth here), I wonder what detriment (other than a small drop in rail voltage) comes from replacing it with a standard resistor of suitable ohm and power rating. I'm sure Nelson had very specific reasons for designing the F5m PS as he did, with the network of caps and the thermistor, so substituting it with a plain resistor is most likely a poor idea. (As I'm typing, Ben Mah posted an interesting note related to this, focused on capacitance rather than resistance; I'll answer separately).
I also realize that what I am saying above may be a little like a band aid, and you are probably suggesting the additional CRC (and the Quasimodo to calculate it) because the real problem is lies with the transformer/wiring in my example, and it would be better to take care of it there.
@Ben Mah -- thanks for this info, and the two good suggestions. Particularly for taking the time to run the parameters for the FW and F5m PS through a simulator to get the ripple values, and the suggestion on how to reduce them.
I do have a selection of suitably sized caps with low ESR for PS duty on hand, and will probably try adding them. It will probably be after I finish my taxes, because I've been having fun with the F5m rather than doing what I should.
I do have a selection of suitably sized caps with low ESR for PS duty on hand, and will probably try adding them. It will probably be after I finish my taxes, because I've been having fun with the F5m rather than doing what I should.
With monoblocks the power supply supplies only one half of the current compared to the stereo amplifier, so the power supply ripple would be substantially lower, and the amplifier should be much quieter.
Another factor is that with lower current, the thermistor temperature would be lower and its resistance would be higher. That would reduce the ripple even more. So the ripple would probably be much less than half of the stereo setup.
Another factor is that with lower current, the thermistor temperature would be lower and its resistance would be higher. That would reduce the ripple even more. So the ripple would probably be much less than half of the stereo setup.
Last edited:
One thing to note about the F5m kit PS boards is that there are two different types of thermistors recommended in the CRC section. One is a 10Ω cold start device for ~1.0A bias current in each channel, and the other is a 4.7Ω cold start device for higher bias currents of around 1.4A per channel. A third type is used on the primary side of the transformer as a soft start to reduce initial inrush of current. The remaining thermistor location is used to connect audio GND and chassis GND. This last one is a 10Ω cold start device, same as used in the CRC section.
It will be helpful to be mindful of the different types and their effects on system performance.
It will be helpful to be mindful of the different types and their effects on system performance.
Check that decimal place if you're thinking of the Epcos B57236S0479M000