Looking at the Speaker Amp Interaction .pdf again and noticed in Fig. 4 that the
difficult #55 cable loaded with the speaker or dummy load looks like it could be
approximated with a lumped tank circuit and Zobel to get the low 1.58 ohms at HF.
I wonder if a 1 - 1.6 ohm resistive load (not wirewound) might also cause the amp
to oscillate. Question is if it is the reactance or simply the low impedance value.
I've never SPICE simmed a Blameless, was never that interested, but I might take
a look at this with a simulated load.
difficult #55 cable loaded with the speaker or dummy load looks like it could be
approximated with a lumped tank circuit and Zobel to get the low 1.58 ohms at HF.
I wonder if a 1 - 1.6 ohm resistive load (not wirewound) might also cause the amp
to oscillate. Question is if it is the reactance or simply the low impedance value.
I've never SPICE simmed a Blameless, was never that interested, but I might take
a look at this with a simulated load.
Last edited:
Your question of whether it is just the low impedance load as opposed to the reactance is a good one. We need to bear in mind that as the load resistance goes down, the signal current flowing in the output stage goes up for a given signal amplitude. This means that if there is some feedback path via grounding or the power supply, the gain of that parasitic feedback path will go up as the load impedance goes down. Just speculation here, and I have not observed the phenomenon and have not in the recent past loaded an amplifier with 1 ohm (I regularly load them down to 2 ohms).
It might not be unreasonable to see the frequency response and transient response under small signal conditions with a 1-ohm load.
Cheers,
Bob
Cyril has completed sufficiently projects to be sure of his skills for building Self's amplifiers. I was in contact with him for the misfortune he met with them. The real reason remained unclear. I used a Self amp for 25 years. The boards were sold by Electronics World in 1994.
The input stage was exactly the same as on the schematics post #9790 :
https://www.diyaudio.com/forums/att...dells-power-amplifier-book-self_ch22_fig4-jpg
Note the absence of a resistor at the left of C1; cost of a connection to the input when the amp was on : a power transistor. I immediately soldered the missing resistor.
More important, there is no low pass filter. However, sensivity to parasitics never happened.
It is worth noting that this is a very old Blameless design, lacking many refinements to that design that came in subsequent years. Look at what transistors are used and that gives you an idea of how old the design is.
Cheers,
Bob
Adjusting the Klever Klipper from Bob's book
Bob, I wasn't sure whether to create a brand new discussion thread for your Klever Klipper, or to leave it in here with the other things in your book. If you wish, we can ask Moderators to split this off into a new thread, all its own.
I'm working on the user manual for Bob's Super Gain Clone (found in his chapter on Integrated Circuit Power Amplifiers) and, today, on the section about setting up and dialling in the correct potentiometer setting for the Klever Klipper.
I came across a bit of a surprise: If you set the Klever Klipper to smooth out and round off the amp waveform when it's just barely clipping, increasing the input signal some more (i.e. driving the amp harder into clipping) brings back the icky ugly P.U. unpleasant flatline clipped wave. I found that I had to set the input amplitude to 1.25X the input that juuuuust barely clips, and then adjust the Klever Klipper at that signal level, to get rid of all flatline non-smooth clipping at any and every input signal level.
Here's my data.
Figure 1 shows the amp with the KK threshold set waaaaay far away so that KK is doing absolutely nothing. An input signal level of 1.3V rms {don't imagine that my signal generator's rms amplitude readout is very precise!} gives a wee tad bit of clipping on the negative going peaks.
Figure 2 shows the amp with KK doing nothing, and input amplitude increased to 1.4V rms. Now there's clipping on both halves of the waveform.
Figure 3 shows what the Klever Klipper does, at 1.4V input level. I've dialled the KK potentiometer to just barely get rid of the flatline clipping and to make nice smooth rounded waveforms instead.
Figure 4 is where the fun starts. I left the KK pot at the same setting, but increased the input signal level from 1.4V to 1.5V. Alas, unwanted flatline clipping has returned on the negative peaks.
Figure 5 increases the input signal level to 1.6V, and dials the Klever Klipper to eliminate all flatline clipping at 1.6V. Peaks are smoothly rounded once again. I observed that with this pot setting, the KK successfully keeps the output smooth at all input signal levels, including 1.8V, 2.0V, 2.2V, and 1.3V.
I am struggling with this decision: Do I recommend that people should strive for smoothness at the onset of clipping (Figure 3), knowing that further increases in signal level will result in non-smoothness (Figure 5)?
Or do I recommend that people should (a) find the input signal amplitude that just barely clips; (b) set the input signal amplitude 1.25X higher; (c) dial the Klever Klipper to remove all non-smoothness at that level (Figure 5).
Math note: 1.6V (Figure 5) divided by 1.3V (Figure 1) equals 1.231.
Thanks for any critique or insights!
Mark
edit- in all photos, both channels are driven with the same 1 kHz input sinewave. Both channels are loaded with 8 ohm 150 watt resitors
_
Bob, I wasn't sure whether to create a brand new discussion thread for your Klever Klipper, or to leave it in here with the other things in your book. If you wish, we can ask Moderators to split this off into a new thread, all its own.
I'm working on the user manual for Bob's Super Gain Clone (found in his chapter on Integrated Circuit Power Amplifiers) and, today, on the section about setting up and dialling in the correct potentiometer setting for the Klever Klipper.
I came across a bit of a surprise: If you set the Klever Klipper to smooth out and round off the amp waveform when it's just barely clipping, increasing the input signal some more (i.e. driving the amp harder into clipping) brings back the icky ugly P.U. unpleasant flatline clipped wave. I found that I had to set the input amplitude to 1.25X the input that juuuuust barely clips, and then adjust the Klever Klipper at that signal level, to get rid of all flatline non-smooth clipping at any and every input signal level.
Here's my data.
Figure 1 shows the amp with the KK threshold set waaaaay far away so that KK is doing absolutely nothing. An input signal level of 1.3V rms {don't imagine that my signal generator's rms amplitude readout is very precise!} gives a wee tad bit of clipping on the negative going peaks.
Figure 2 shows the amp with KK doing nothing, and input amplitude increased to 1.4V rms. Now there's clipping on both halves of the waveform.
Figure 3 shows what the Klever Klipper does, at 1.4V input level. I've dialled the KK potentiometer to just barely get rid of the flatline clipping and to make nice smooth rounded waveforms instead.
Figure 4 is where the fun starts. I left the KK pot at the same setting, but increased the input signal level from 1.4V to 1.5V. Alas, unwanted flatline clipping has returned on the negative peaks.
Figure 5 increases the input signal level to 1.6V, and dials the Klever Klipper to eliminate all flatline clipping at 1.6V. Peaks are smoothly rounded once again. I observed that with this pot setting, the KK successfully keeps the output smooth at all input signal levels, including 1.8V, 2.0V, 2.2V, and 1.3V.
I am struggling with this decision: Do I recommend that people should strive for smoothness at the onset of clipping (Figure 3), knowing that further increases in signal level will result in non-smoothness (Figure 5)?
Or do I recommend that people should (a) find the input signal amplitude that just barely clips; (b) set the input signal amplitude 1.25X higher; (c) dial the Klever Klipper to remove all non-smoothness at that level (Figure 5).
Math note: 1.6V (Figure 5) divided by 1.3V (Figure 1) equals 1.231.
Thanks for any critique or insights!
Mark
edit- in all photos, both channels are driven with the same 1 kHz input sinewave. Both channels are loaded with 8 ohm 150 watt resitors
_
Attachments
Last edited:
> An input signal level of 1.3V rms ...gives a wee tad bit of clipping...
> dials the Klever Klipper to eliminate all flatline clipping at 1.6V. ...keeps the output smooth at all input signal levels, including 1.8V, 2.0V, 2.2V, and 1.3V.
1.3V vs 1.6V is <2dB. If you are working within a few dB of clipping, you really need a bigger amp. I'd trim for 1.6dB. And I suspect for many users it would be best to design this in (make clip level a set fraction of supply, allowing for nominal diode drop).
> dials the Klever Klipper to eliminate all flatline clipping at 1.6V. ...keeps the output smooth at all input signal levels, including 1.8V, 2.0V, 2.2V, and 1.3V.
1.3V vs 1.6V is <2dB. If you are working within a few dB of clipping, you really need a bigger amp. I'd trim for 1.6dB. And I suspect for many users it would be best to design this in (make clip level a set fraction of supply, allowing for nominal diode drop).
This eye-burner is a hardcover. (It is louder in the hand than in this over-JPEGed image.)
A friend recently published a book. Book publishers have some very strange ideas about "art" and "marketing". After proposing 3 covers, 2 of which sucked, and liking the 3rd, they came back late in the game with a 4th really murky cover which does not "pop" on an Amazon page.
Attachments
As you and Bob probably know: "all" basic twinlead TP runs "about 100 Ohms" (85-120r). This is true back to the original telephone multi-pair paper-wrap cable. IIRC, TP of thin Kynar wirewrap gets to 70r, but PE CAT TP runs around 105r (by spec based on what a practical construction gives naturally). TV downlead has to be spaced w-i-d-e apart to approach 300 Ohms. Foil-film-foil sandwich may get to 50 Ohms.
75r across nominal 8r speaker takes 11% of the power of the speaker. A 100W amp doing unclipped speech/music will average <10W (as you say, tone may smoke it). Bet if you zone-out with loud rock you can warm those resistors. The neighbors may stop you before you get the 1,000 hour rated load/life.
I was not questioning Cyril's skills, but simply wanted to use the same design to duplicate
his results. I'd say that, unless the oscillation was due to parasitics, it should be possible
to duplicate his results in SPICE.
We use these guys, fast and inexpensive, with color hard covers and small runs. Ask for Jim Harris.
Bypass the publisher and sell direct via Amazon, ebay, or your web site. Shipping is cheap via USPS media mail.
G&H Soho
Last edited:
Resistors are one thing and 'Radio Frequency Characteristic Impedance' transmission lines are another.
* * * * * * *
A transmission line does NOT dissipated power! It just acts as an infinity long cable. The only loses are due to imperfections in a real world system. Also as Cyril Bateman points out that rated impedance only above 1 Megahertz.
Last edited:
Have a look at the Klever Klipper schematic, either in Bob's book or in post #9345 of this thread. It does set the clip level to a fraction of supply.
As the book says,
Diodes D1 and D2 provide the soft-clip function in combination with R1. The soft-clip threshold voltages are created at op amps U1A and U1B. These voltages track the short-term average power supply rail voltage. Adaptive soft clipping occurs just shy of output amplifier hard clipping at any given rail voltage condition. As a result, little or no dynamic headroom is sacrificed.
This implements a trimmer-adjustable clipping threshold; the trimmer modifies the "fraction"
- Clip the input signal when it exceeds [(Rail Voltage) * fraction]
- Clip the input signal when it exceeds [(Rail Voltage - OPSdrop Voltage) * (1 / AmplifierGain)]
- Clip the input signal when it exceeds [(Rail Voltage * fraction) - (OPSdrop Voltage * fraction)]
And, fortunately, the second term is small compared to the first term. The LM3886 datasheet provides many different datapoints from which you can back-calculate an estimate of OPSdrop Voltage. For example, max undistorted output into 8R load with 28V supplies is 38W on the datasheet. Thus Vout = 17.44V rms = 24.66V peak and OPSdrop Voltage = 3.34V. The first term is (28 * fraction) and the second term is (3.3 * fraction). The error term is small but perhaps not negligibly small.
One way to go, of course, is to use a combination of measurements and datasheet calculations, to find the worst (largest) possible value of OPSdrop Voltage, and use that as a conservative over-estimate. Maybe the KleverKlipper will activate a smidgen too soon for folks with 6 or 8 ohm speakers, but it certainly won't active too late and thus allow the 3886 chip to enter clipping. And there are probably a few other approaches one could take, which end with the same result: the cat has been skinned.
Hi Mark,
These are nice, insightful measurements. Although I can't explain the behavior when the signal is set to be just into clipping where harsh clipping occurs at a higher level, it could have something to do with power supply sag interactions. I completely agree with your suggestion of setting it up so that non-rounded clipping does not occur at any signal level.
Cheers,
Bob
Send Power into an infinitely long cable, it never comes back. No, of course the cable does not heat (except by parasitic loss, which can be significant). Power just goes-away. (Related to the question of where you get an infinite cable.)
Long telephone lines do have characteristic impedance. Unlike the world above 1MHz, it isn't real constant. For PE insulation it gets near infinite at sub-audio. By 200Hz it may be a couple kOhms. It may not approach 110r by the top of even the hi-fi band. (The very long wide-space pairs on the first cross-continent line came to 800-500r over much of the top of the speech band, hence "600 Ohms".)
But I thought the topic here was very ultra-sonic effects in very wide-band "audio" amps driving cable with a nearly non-terminating box across the room. (Dynamic loudspeaker inductance will tend to rise well above 100 Ohms by 1MHz.) Adding a simple 100r resistor at the end will limit the rise and approximately terminate the line. It also steals like 8/100th the amplifier power as heat. We could avoid that with a 0.01uFd series cap, letting things go loose below 160kHz. Various tunings are possible; I'd look at 0.05uFd to intersect the loudspeaker line slightly outside the audio band.
Last edited:
I don't want to derail this thread so,
For commercial product yes it is wise to limit the resistor dissipation with series cap, in practice I find directly connected 1.2W of MF resistors do not get even warm to the touch with 100W amp and various music genres and not sparing the volume control.
Today I fitted 26R shunts (2 x 13R 0.6W MF) across speakers wired with Cat5 wired as four parallel pairs (each pair rated 100R) to my housemates system, tomorrow I will fit at amp end also.
This is a reasonable quality 8yo Samsung AV amp with SMPS and Class D outputs and the result is immediate improvement in overall clarity, detail, depth and musicality and is now infinitely more listenable on music and TV.
This is similar/same as improvements to my Class AB analogue amp with 75R coax speaker cable and 75R terminating resistors each end.
Can anybody here do measurements including RF impedance on a say 3m length of Cat 5 please ?.
Dan.
For commercial product yes it is wise to limit the resistor dissipation with series cap, in practice I find directly connected 1.2W of MF resistors do not get even warm to the touch with 100W amp and various music genres and not sparing the volume control.
Today I fitted 26R shunts (2 x 13R 0.6W MF) across speakers wired with Cat5 wired as four parallel pairs (each pair rated 100R) to my housemates system, tomorrow I will fit at amp end also.
This is a reasonable quality 8yo Samsung AV amp with SMPS and Class D outputs and the result is immediate improvement in overall clarity, detail, depth and musicality and is now infinitely more listenable on music and TV.
This is similar/same as improvements to my Class AB analogue amp with 75R coax speaker cable and 75R terminating resistors each end.
Can anybody here do measurements including RF impedance on a say 3m length of Cat 5 please ?.
Dan.
Last edited: