Outstanding article!
One thing I would always like to see is a measurement of the frequency response as seen at the loudspeaker terminals with the given amplifier driving the cables.
The second thing I also allow for is that different amplifiers sound different because they are misbehaving differently.
Cheers,
Bob
I agree on this point. I'd really like to see if I can duplicate the failure with Self's amp and
then see if MIC (or anything else) helps the amp remain stable with the difficult load. I
wonder if the problem would show up in SPICE.
Sorry, I should have mentioned it was 100 Ohms. Part of the rationale was that most speakers (all I measured) were inductive above 20 KHz and a resistive load should keep amplifiers happy. And its something that seemed to consistently sound better. I still bridge 100 Ohm resistors across speakers. Even a 1W resistor will survive just fine for most people.
Jan has a link to a large .zip file of Cyril's work here:
Cyril Bateman's Capacitor Sound articles | Linear Audio NL
Direct link: Le Cloud d'Orange
Cyril Bateman's Capacitor Sound articles | Linear Audio NL
Direct link: Le Cloud d'Orange
And most twin/fig 8 cables have impedance of around 100R give or take.
I am running series 39R+36R 0.6W MF across each end of RG-59 75R coax antenna cable as termination resistances and I find this makes very pleasant improvement in clarity, detail and 'quietness'.
Surprisingly this 1.2W rating is sufficient for domestic usage with 100W rated amp playing music, pure tones might be a different matter though !.
Also surprisingly perhaps is subjective difference according to resistor type, ie I find MF clear and 'neutral; CF sounded strongly coloured.
Some may like the colouration of the CF, I didn't so they got the boot.
Dan.
Hi Jan,
They actually published a legit hardcover version of the second edition. That's why it costs so much.
Cheers,
Bob
This is a very good point, and it is pretty much the same concept of the Zobel network I suggest using at the end of the speaker cable, just that I put a capacitor on the order of 0.01 uF in series, which reduces the power dissipation in the resistor.
Cheers,
Bob
The thing that caught my eye in the schematic you showed was the lead capacitor in the feedback loop. Often used to improve phase margin, they can sometimes create more stability problems than they solve, especially if not used very sparingly. I never use feedback lead capacitors. However, I do use feedback networks with relatively low impedance so as to mitigate the effects of the pole formed by the feedback network impedance and the input capacitance of the input stage.
The use of a low-impedance feedback network also tends to reduce noise. The shunt resistor of the NFB network I use is usually 1k or 500 ohms. The price paid for this is the need to use feedback resistors with higher dissipation. For the series feedback resistor, I use two 1-watt metal film resistors in series. In some cases I'll use a pair of 2-watt resistors.
The center tap formed by the two feedback resistors also makes it possible to make a rough check for adequate phase margin and gain margin by enabling of the deliberate reduction of those margins by either shorting one of them (gain margin) or shunting the center tap with a small capacitance to ground to deliberately degrade phase margin for testing. This is described in Chapter 4 of the second edition where testing for the BC-1 amplifier is covered.
Cheers,
Bob
I kind of like a small lead cap to compensate for the pole formed by the input capacitance,
but wonder what might be the best way to "tune"/design the network without the need for an
expensive active probe. I don't think an exact solution is needed, just one that roughly
compensates to keep it stable.
I suppose one could split the shunt resistor into 10 or less ohms plus the rest of the resistance,
then feed a signal generator in with a known source impedance perhaps 10K and measure the
input impedance to determine Cin through measured frequency response.
Another issue - I'd expect the large shunt cap to really screw things up at HF but nobody seems
to talk about it. Just looked at some curves and while they usually self resonate below 1 MHz
the impedance remains low (less or much less than 10 ohms) below 10 MHz. The impedance is
mostly determined by the R in series and therefore it probably doesn't matter.
but wonder what might be the best way to "tune"/design the network without the need for an
expensive active probe. I don't think an exact solution is needed, just one that roughly
compensates to keep it stable.
I suppose one could split the shunt resistor into 10 or less ohms plus the rest of the resistance,
then feed a signal generator in with a known source impedance perhaps 10K and measure the
input impedance to determine Cin through measured frequency response.
Another issue - I'd expect the large shunt cap to really screw things up at HF but nobody seems
to talk about it. Just looked at some curves and while they usually self resonate below 1 MHz
the impedance remains low (less or much less than 10 ohms) below 10 MHz. The impedance is
mostly determined by the R in series and therefore it probably doesn't matter.
Last edited:
Hi Bob,
I think Jan and myself were wondering why the cover graphics on the hardcover was not the same as the one on the softcover?
BC-1 pcb uses the same film cap in parallel with the 10uF/50V (UES1H100MPM) at the input too.
Rick
I think Jan and myself were wondering why the cover graphics on the hardcover was not the same as the one on the softcover?
The pcb version of BC-1 uses a 1uF/100V stacked poly film cap (Kemet F612JT105J100R) in parallel with the large shunt 220uF/50V ecap (Nichicon UES1H221MHM). These are not shown in the BAF2019 slides nor the book, but are mentioned in the book. I assume they were left out to make it easier to read.
BC-1 pcb uses the same film cap in parallel with the 10uF/50V (UES1H100MPM) at the input too.
Rick
Last edited:
Small for that very reason is fine, but it usually needs to be very small, unlike what was in the schematic you posted. Think of the feedback network as a capacitance voltage divider in parallel with a resistance voltage divider of the same ratio, as is used in capacitance-compensated attenuators in test equipment. Suppose the stray capacitance at the IPS is 10 pf and the gain of the amplifier is 28. The appropriate compensating capacitor would then only be on the order of 0.4 pF. A different approach would be to go all-in with a capacitance-compensated attenuator for the feedback network. In this case, one might add 120 pF of shunt capacitance to ground and then all a series feedback capacitor on the order of 5 pf.
When a small-value feedback shunt resistor is used, like 500 ohms or 1k, it does necessitate a rather large shunt decoupling capacitor in a conventional arrangement, inevitably an electrolytic, inviting issues at high frequencies due to possible ESL and self resonances, plus electrolytic distortion. The first obvious step is to shunt the electrolytic with a 1 uF film capacitor of good quality. This will not do much for electrolytic distortion, however. The use of higher-voltage NP electrolytics designed for loudspeaker crossovers can provide a significant improvement, but take up more space. Ultimately, my preferred choice is to ditch the capacitor and use a well-designed DC servo with a quality integrating capacitor. Such a DC servo may end up taking no more space than a decent electrolytic shunt capacitor.
Cheers,
Bob
I suspect it is not practical or economical to put fancy and colorful graphics on a hard-cover book. Of course, sometimes hard-cover books come in paper sleeves where such graphics are entirely possible.
Yes, I left the film bypass capacitors off of the BAF slides for simplicity.
Cheers,
Bob
