Most of the "Circlophonists" seem to have been struck by the remarkable silence of this amplifier.
I didn't pay a lot of attention to this aspect, since the design goals of the topology had nothing to do with the noise floor, but this is recurrent observation from those who have built it.
Myself, I do not have high efficiency speakers, and I found the silence normal anyway.
Now that I have been intrigued by these reports, I have decided to investigate the subject more deeply.
I have measured (rather crudely for the time being) the output noise, and I find 50µVrms wide band. That is -112dB with respect to 20V. Not bad, but not extraordinary either.
I have also tried to connect an earphone directly to the output, and strangely, I found it dead silent. Very unusual (my input transistors are banal BC556).
I am beginning to wonder if the servoing of the standing current of all active elements could have the effect of removing part of the noise, by eliminating second order effects like modulation.
What are your opinions?
I didn't pay a lot of attention to this aspect, since the design goals of the topology had nothing to do with the noise floor, but this is recurrent observation from those who have built it.
Myself, I do not have high efficiency speakers, and I found the silence normal anyway.
Now that I have been intrigued by these reports, I have decided to investigate the subject more deeply.
I have measured (rather crudely for the time being) the output noise, and I find 50µVrms wide band. That is -112dB with respect to 20V. Not bad, but not extraordinary either.
I have also tried to connect an earphone directly to the output, and strangely, I found it dead silent. Very unusual (my input transistors are banal BC556).
I am beginning to wonder if the servoing of the standing current of all active elements could have the effect of removing part of the noise, by eliminating second order effects like modulation.
What are your opinions?
I'm not an electronics engineer, have not reiterated blameless looking standard designs, Circlophone is my first experience with some success in building a high performance discrete parts audio amplifier on copperless phenolic perfboard, and as far as I know, high current global feedback plus multi feedback should be doing as described. The performance is not unexpected but it is definitely entertaining and beneficial.
Elvee's Circlophone is capable of faithfully replaying a 64th rest at full halt despite also running at full blast into an 8 ohm sealed box speaker. It seems to punch the bottom out of the noise floor. I am grateful to Elvee for providing education about transistors and I am grateful to Elvee for providing a design that is both doable and richly performing. This time is far too premature for me to write a fully detailed review as I have not yet confirmed that my sample represents Elvee's design as accurately as necessary for that activity. As for potentially addictive dynamics amplifier category, the list is short, includes Circlophone, and you should have one so that you never get bored.
References:
Here is the schematic that was used for my sample: The inexpensive Circlophone© ♫♪.
Here is the main thread for Circlophone, with much information and many options. http://www.diyaudio.com/forums/solid-state/189599-my-little-cheap-circlophone.html#post2582392
Elvee's Circlophone is capable of faithfully replaying a 64th rest at full halt despite also running at full blast into an 8 ohm sealed box speaker. It seems to punch the bottom out of the noise floor. I am grateful to Elvee for providing education about transistors and I am grateful to Elvee for providing a design that is both doable and richly performing. This time is far too premature for me to write a fully detailed review as I have not yet confirmed that my sample represents Elvee's design as accurately as necessary for that activity. As for potentially addictive dynamics amplifier category, the list is short, includes Circlophone, and you should have one so that you never get bored.
References:
Here is the schematic that was used for my sample: The inexpensive Circlophone© ♫♪.
Here is the main thread for Circlophone, with much information and many options. http://www.diyaudio.com/forums/solid-state/189599-my-little-cheap-circlophone.html#post2582392
So, I'd like to know if you listened to some music through that earphone/headphone?
I'm not sure if this circuit Headphone Adaptor for Power Amplifiers is fully compatible, but if it is, then this would probably lead to some extra entertainment.
In a related question, I'd like to know how low the operating voltage needs to be prior to using 1w BCP53 as drivers, what's the related gain setting for that (with computer/ithing source), and what would be some suggested output devices for such a small amp?
Thanks!
P.S.
Topic is: If we're going to listen very carefully, might as well use headphones for that so you get a much smaller amount of variables. Then all you have to do is use whatever headphones you think have a flat response, and although that is up to the individual, at least speaker and room variances got subtracted. In a related issue, a jfet at the amp input seems like it could reduce some source variety, since not every source drives ideally. However, this proposal is from purely selfish motives, since a tiny Circlophone could be so very useful.
Elvee - If the effect of noise is to cause the output to wiggle a bit resulting in current changes in R8 and R24/R11, then it is possible that the servo detects these disturbances and cancels them out. I think as long as the noise level is below the schottky's Vf the servo will react to cancel them. What do you think?
Yeah, precisely what I thought.
In a conventional output stage, the OP transistors are biased with a fixed voltage, and have a negligible emitter degeneration resistor, which makes them powerful noise current sources.
Seen from the ouput, the two are in parallel, but this change very little from the noise perspective. A mere 3dB.
The amplifier has to rely solely on its global NFB to reduce the noise by a factor equal to the loop gain.
I have made a quick test, to mimic the conditions of operation of an output transistor: C1 presents a negligible input impedance between B and E, just like a good bias network would.
The collector current is adjustable with I1.
The noise generated across the 56 ohm wirewound resistor is amplified 1000 times.
The jig was first tested without a collector connection, to make sure the noise floor was OK: it was, under 1mV rms.
Then, various audio power transistors were tested (MJ15015, MJ15024, BDX95, 2N3055, 2N3772, 2SD428, ..). As expected, the noise is heavily dependent on Ic. The tests were made at 75mA, a normal Iq for class AB.
On average, the measured noise voltage was ~13mV rms. Some were slightly below, some were way higher, but 13mV/75mA is a reasonable average figure.
That is 13µV across the 56 ohm, or 0.23µA.
Not very much, but drivers could also contribute significantly.
In the Circlophone, this type of noise is also dealt with by the servo loop, in addition to the regular NFB.
This might partly explain the relative quietness of the Circlophone.
Attachments
I am fairly certain that output noise in your Circlophone is entirely input stage dominated at AF, like in most any amplifier with plenty of open-loop gain to spare. With moderate gain (27 dB), a reasonably low-impedance feedback network, the usual moderate input stage currents (like 1..3 mA x2) and a low-impedance source, I don't see why it shouldn't achieve perfectly decent noise performance, even with relatively modest BC556Bs. (See e.g. Douglas Self's discussion of input stage noise. He's not using dedicated low-noise types either, and typically gets -124 dBu (0.5 µVrms) input-referred in a 22k b/w.)
No rocket science involved at all.
I wish I could say the same about the output stage - I still can't wrap my head around that. If there's any "academic" explanation floating around, I'd appreciate a link.
If there is no audible noise on earphones, noise performance has to be quite decent at the very least. Even ~20 µVrms of noise across the audio b/w (~0.9 µVrms input-referred with the given gain) tends to be inaudible on any but the most discerning IEMs with sensitivities of 125+ dB / 1 Vrms. Noisy speaker amps may have 300 or even 500 µV (usually at gains of ~45 dB, which makes keeping noise down a lot harder).
Have you considered building a little 20 kHz lowpass for noise measurements (2nd order or so)?
No rocket science involved at all.
I wish I could say the same about the output stage - I still can't wrap my head around that. If there's any "academic" explanation floating around, I'd appreciate a link.
If there is no audible noise on earphones, noise performance has to be quite decent at the very least. Even ~20 µVrms of noise across the audio b/w (~0.9 µVrms input-referred with the given gain) tends to be inaudible on any but the most discerning IEMs with sensitivities of 125+ dB / 1 Vrms. Noisy speaker amps may have 300 or even 500 µV (usually at gains of ~45 dB, which makes keeping noise down a lot harder).
Have you considered building a little 20 kHz lowpass for noise measurements (2nd order or so)?
Last edited:
Yes, I agree and that's why I didn't particularly care about that aspect: I expected it to be "normal".
But a recurrent observation from most of the builders was that this amplifier was exceptionally quiet.
I am not in a good position to judge that by myself, as I live in relatively noisy environment: I can never enjoy a dead silence.
That's why I resorted to measuring instruments and earphones.
The measured level looks normal, but it is not ponderated, and extends to ~100KHz, and there might be something there: the secondary loop happens to have a high gain in the audio range.
Output stages using same sex transistors are pretty well covered, Tiefbass has collected a number of links about that, and I have tried to explain the auto-bias in the thread.
That's something I am considering, as my current noise-measuring capabilities are rather limited.
A different practical usage. . .
So, you're discussing a Circlophone set as conservative as possible, but what if it was set differently?
I have:
Matched all of the BC556B, because there was supposedly some benefit and it was very easy to do.
A 22k feedback resistor because I wanted high gain (470R with 22k divider)--I guess 20k would have been fine and not too many tracks come up short.
Matched 2SB649 drivers, because they are available (unlike the Philips drivers of the schematic).
Turned down the computer source to pass RMAA quality control without blaring/distorting itself. This source output strength is also similar to a portable player, and some of those have useful fidelity when not run all the way to max.
After setting Circlophone to higher gain/lesser feedback, here is what changed:
1). The previously dynamic amplifier now more dynamic (totally expected).
2). Somewhat more useful power than expected (slight disproportion, yet favorable).
I believe that these are ordinary consequences to increasing gain.
Result:
Very loud playback, enjoyable in the absence of blare (source not straining), for example:
On a 78 RPM live recording of Ain't Misbehavin' which was noticeably at a pub, I could hear conversations located at different tables even though it was actually a compressed MP3 dub of the 78 RPM record. Wow. Yes, that was very loud replay, and it was up that loud because it is nice since the computer isn't straining to push a high gain Circlophone. In this very real example, the noise reduction benefit did not require a moderate gain setting. It was high gain and the noise reduction feature did not change.
So, you're discussing a Circlophone set as conservative as possible, but what if it was set differently?
I have:
Matched all of the BC556B, because there was supposedly some benefit and it was very easy to do.
A 22k feedback resistor because I wanted high gain (470R with 22k divider)--I guess 20k would have been fine and not too many tracks come up short.
Matched 2SB649 drivers, because they are available (unlike the Philips drivers of the schematic).
Turned down the computer source to pass RMAA quality control without blaring/distorting itself. This source output strength is also similar to a portable player, and some of those have useful fidelity when not run all the way to max.
After setting Circlophone to higher gain/lesser feedback, here is what changed:
1). The previously dynamic amplifier now more dynamic (totally expected).
2). Somewhat more useful power than expected (slight disproportion, yet favorable).
I believe that these are ordinary consequences to increasing gain.
Result:
Very loud playback, enjoyable in the absence of blare (source not straining), for example:
On a 78 RPM live recording of Ain't Misbehavin' which was noticeably at a pub, I could hear conversations located at different tables even though it was actually a compressed MP3 dub of the 78 RPM record. Wow. Yes, that was very loud replay, and it was up that loud because it is nice since the computer isn't straining to push a high gain Circlophone. In this very real example, the noise reduction benefit did not require a moderate gain setting. It was high gain and the noise reduction feature did not change.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.