Lazy!!!! You know how heavy is that stupid thing?!!!
Maybe it's easier to roll the scope out. Still very lazy to do that as it's going to be replaced soon.
Effect of speaker cable on THD
- Thread starter Alan0354
- Start date
Maybe it's easier to roll the scope out. Still very lazy to do that as it's going to be replaced soon.
I am hot on the trod with my amp, I am hoping to replacing the Nakamichi soon. I ordered another 200' of the 20 gauge speaker wires to make more cables. I can't even do much of the THD testing as I already put the two 10 X 20 gauge red/black cables into the system. It's some work to fish them in through the hole, run behind the big home entertainment furniture. I am not going to take them out anymore. When I build the new cables, it would be interesting to see how the Nakamichi and the Acurus fare with the high capacitance cable. they both don't have output inductor, so there's nothing protecting the amp from the capacitance.
I don't even want to try the cable on my original prototype amp as I don't think it like the high capacitance either. I had to work on the compensation on the new amp to stabilize it for high capacitance.
I don't even want to try the cable on my original prototype amp as I don't think it like the high capacitance either. I had to work on the compensation on the new amp to stabilize it for high capacitance.
I am just thinking out loud here.
My 10ft 12 gauge monster gave 0.14% THD at 20KHz as the thread starter here. Then I did parallel 3 section of the 3ft 12 gauge Monster cable and pretty much reduced the THD by half. I can't find the specific post on the number. But if we follow this trend, If I parallel 4, I should get about 0.14%/4= 0.035% for 10ft. So it should be comparable to my 10 pairs of 20 gauge red/black wires twisted together.
I measured the capacitance of that cable, it is about 240pF. So 4 pairs in parallel will give me about 1000pF. The 10 pairs of re/black 20 gauge measured 2200pF. This means if I use 4 pairs of 12 gauge monster, I should get the same THD but HALF the capacitance!!!!
I am thinking about ordering some 16 gauge red/back speaker wires. put say 6 pairs and twist together, I should get the cross sectional area of almost 8 gauge and still have capacitance of about 1500pF. I should get THD down to about 0.03% or lower at 20KHz.
What do you guys think.
My 10ft 12 gauge monster gave 0.14% THD at 20KHz as the thread starter here. Then I did parallel 3 section of the 3ft 12 gauge Monster cable and pretty much reduced the THD by half. I can't find the specific post on the number. But if we follow this trend, If I parallel 4, I should get about 0.14%/4= 0.035% for 10ft. So it should be comparable to my 10 pairs of 20 gauge red/black wires twisted together.
I measured the capacitance of that cable, it is about 240pF. So 4 pairs in parallel will give me about 1000pF. The 10 pairs of re/black 20 gauge measured 2200pF. This means if I use 4 pairs of 12 gauge monster, I should get the same THD but HALF the capacitance!!!!
I am thinking about ordering some 16 gauge red/back speaker wires. put say 6 pairs and twist together, I should get the cross sectional area of almost 8 gauge and still have capacitance of about 1500pF. I should get THD down to about 0.03% or lower at 20KHz.
What do you guys think.
Dunno myself, solid state amp design is waaay above my pay grade.
I was actually looking at some CAT-5 cables per your suggestion. They are all 24 gauge. No wonder the capacitance of yours is so high. You must be using like 15+ pairs in parallel together.
I am going the other way trying to find a good balance of capacitance and still get the low THD. Let me try the 16 gauge first, 5 or 6 pairs. From results here, I suspect 15ft, 5 or 6 pairs of 16 gauge should give me 0.04%THD at 20KHz and capacitance of about 1500pF. We'll see.
OK, back from a wonderful week in Florida and thinking about this resistor again. I can confirm that the resistor raises THD by a factor of at least 2-3 as the frequency rises to 20 kHz. IMO, just stray inductance wouldn't do that; there has to be a non-linearity involved. Unconfirmed as yet, but I think the problem is that even though the resistor is somehow counter-wound for low inductance, that inductance has a core. All cores are bad if you want low THD. It may be the wire or tape itself, if there's more than one layer, or it may be something else. The end terminals are, as is common, strongly magnetic. What's interesting is so is the entire resistor body. The resistance alloy is obviously magnetic, and it seems to be in the form of tape, with two lays at opposite angles, thus the strong magnetic action. A wire would "probe" differently. I'd need to smash the thing apart to really understand the construction. There are similar resistors on the 'bay but some are made with conventional wire (you can see it through the conformal coating) and others are probably tape. No idea the wire type, Evanohm or other, but they all claim low inductance and good for audio loads. Easily spotted as they're a light to medium green with a matte surface, and you can see the pattern through the coating.
IMHO, the bottom line is these things are probably OK for power measurements, but if you care about THD they should be avoided like the plague.
IMHO, the bottom line is these things are probably OK for power measurements, but if you care about THD they should be avoided like the plague.
Glad you are back, you enjoy the vacation?
It's up to you whether you want to destroy the resistor to look at it. I am keeping mine as I might not find another one. I am busy as I get oscillation with my amp when testing with capacitor load. The tricky thing is it's easy to tame with 0.5uF cap, it's the 3300pF that is hard to tame. It gets into the power transistors particularly I have 9 pairs. I tested one board with only 4 pairs, it's a lot more stable. I am still working on it. So I really don't have time to work on the cables.
The high capacitance speaker cables are a problem with amps. The Nakamichi is definitely behaving funny, a lot of time when first power up, it makes a hum type of sound, I have to turn it off and on to get rid of that. It's definitely the amp. Never happened before until I change to the cable with about 2200pF.
I have been thinking about the capacitance load to test amp, From looking at your analysis and from my understanding of speakers and crossover, The amp never see a pure capacitor. and 10uF load with even 1ohm in series make the whole difference in the world. That 1ohm with stabilize the amp. The issue become much smaller if the capacitance load is on the far end of the cable. Capacitance right at the output of the amp really kills as there is no cable inductance to isolate the capacitance to the output of the amp.
I think it is a lot more important to ensure the amp is stable with say 6000pF ceramic or other good low esr caps right at the output of the amplifier. If an amp can be stable with this load with burst signal, I think it's plenty good. AND don't use a cable with more than 6000pF!!!
Far is I see so far, it's not even the feedback loop of the amp that needs to be tame, it's the output stage where you have slow transistor. The oscillation I see are all in 4MHz arena. Changing compensation does not even change the frequency. So far, increasing the base stop resistor of the output transistor from 2.2 to 10ohm seems to help. I am going to experiment with ferrite beads on the collector of the big output transistor. I have 0.1uF ceramic cap point to point from the collector of the transistors to ground. Maybe I will remove those caps to increase the inductance from the collector to ground and see whether that will stabilize the transistor.
Ha ha, designing the circuit is easy, making it stable is engineering.
It's up to you whether you want to destroy the resistor to look at it. I am keeping mine as I might not find another one. I am busy as I get oscillation with my amp when testing with capacitor load. The tricky thing is it's easy to tame with 0.5uF cap, it's the 3300pF that is hard to tame. It gets into the power transistors particularly I have 9 pairs. I tested one board with only 4 pairs, it's a lot more stable. I am still working on it. So I really don't have time to work on the cables.
The high capacitance speaker cables are a problem with amps. The Nakamichi is definitely behaving funny, a lot of time when first power up, it makes a hum type of sound, I have to turn it off and on to get rid of that. It's definitely the amp. Never happened before until I change to the cable with about 2200pF.
I have been thinking about the capacitance load to test amp, From looking at your analysis and from my understanding of speakers and crossover, The amp never see a pure capacitor. and 10uF load with even 1ohm in series make the whole difference in the world. That 1ohm with stabilize the amp. The issue become much smaller if the capacitance load is on the far end of the cable. Capacitance right at the output of the amp really kills as there is no cable inductance to isolate the capacitance to the output of the amp.
I think it is a lot more important to ensure the amp is stable with say 6000pF ceramic or other good low esr caps right at the output of the amplifier. If an amp can be stable with this load with burst signal, I think it's plenty good. AND don't use a cable with more than 6000pF!!!
Far is I see so far, it's not even the feedback loop of the amp that needs to be tame, it's the output stage where you have slow transistor. The oscillation I see are all in 4MHz arena. Changing compensation does not even change the frequency. So far, increasing the base stop resistor of the output transistor from 2.2 to 10ohm seems to help. I am going to experiment with ferrite beads on the collector of the big output transistor. I have 0.1uF ceramic cap point to point from the collector of the transistors to ground. Maybe I will remove those caps to increase the inductance from the collector to ground and see whether that will stabilize the transistor.
Ha ha, designing the circuit is easy, making it stable is engineering.
Do you have the "standard" series combination of 0.1 uF and 10 ohms from output to ground? With my Tiger amps and their admittedly wonky output stage with gain, I could never get stability without the carefully selected output inductor. It was larger than most recommend, I remember something like 10 or more uH. PMing you something. Yes, vacation was really great. Haven't taken one in a long time.
Yes, I do have 0.1uF film cap in series with 10ohm to ground at the output. I am sure the output inductor will solve all the problem, but it raises distortion. I played with the output stage, increasing the base stop from 2.2 to 10ohm helps a little, but something telling me I did not hit the sore spot. I spent over an hour pulled the transistors out and put a ferrite bead on the collector pin. I expect it should help.......BUT NO!!!! It changed the oscillation a little bit, but it did not improve a bit. I am going back to the driver stage and move backward. It's a painful long process and destroy the board on the way. This is a 3EF particular I do the 3EF with one with opposite transistor like PNP-NPN-NPN 3EF, it can be prone to oscillation. Just take time to go through it. They behave with normal cable, but now I have to test it with 3300pF at the output of the amp.
I think this is going to take some time to tame, I already budget out two boards to destroy going through this.
I feel your pain but I've never been able to do an amp on the 1st run of boards without cuts and jumpers. The thing I don't get is why an output inductor should raise THD. When I did my very low THD custom amp, the output inductor didn't increase the 0.00x numbers if it didn't have a core, accidental or otherwise. It was maybe 0.75" diameter, free wound and held tightly with Ty-Raps. On my Tiger, I made the inductor larger and wound it on a short piece of thin wall plastic pipe. There wasn't room on the board, so I mounted it right at the banana jacks on the rear panel, along with the 0.1/10 parts. Worked fine. No proof, but it's probably important that the coil wires can't move at all- solid magnet wire, varnished or tightly tied. With the Tiger, wire connection points turned out to be very critical to stability. No ground point is entirely equal to any other ground point. Local filter cap returns can't be common with any signal returns- that's a mistake many commercial designs suffer from. FWIW, I've never had any luck with ferrite beads or base stoppers on amps. If they help, something else is probably wrong. Keep hammering at it- keeps you off the streets and out of the bars!
The output inductor is a big factor in increasing THD with the green resistor as explained in post #250 in page 13. Any impedance in series with a load that has distortion increase THD at the load end.
I just gone through putting ferrite beads starting from output transistor, to drivers.....then when I put the beads on the pre-drivers, it changes and seems to be less of the high frequency oscillation with the 3300p. Sadly, something dropped on the board and I heard a pop!!!!...............So I moved onto the second board, instead of putting the beads at all, I know the pre driver is the problem, I put a Zobo network of 180ohm and 6600p at the emitter of the predriver to give it a lower impedance load above 130KHz. It helps a lot. Not quite there, but it does help. The compensations are a lot more relaxed. The load impedance for the pre-driver is too high to be stable.
The inverted 3EF is giving a lot of problems, but I really like it as the Vbe temperature drift of the PNP and NPN really cancels out and all I have to do is compensate the big output transistors. Also, I gain 1.2V at the output closer to the rail. It is important for me as I use lower rail voltage, every volt count.
Still working on it, I don't think I got to the bottom, but definitely a step forward.
EDIT:
So far, I have 150ohm and 3300p Zobal for the pre-driver, 50ohm and 3300p on the driver. It really helps. I am still going to play with it more, But it really looks a lot better. With 0.022uF, still looks good.
Those EF need to have a high frequency load ( Zobal ) to keep it tame. I am going to try ferrite bead on the pre-driver tomorrow as it takes a lot of work to pull it out and put it in. Those transistor are discontinued, I only have about 20 ea, I cannot just cut the lead and replace with new one. Have to carefully desolder them without ruining the pcb, put on the beads and solder back on.
I just gone through putting ferrite beads starting from output transistor, to drivers.....then when I put the beads on the pre-drivers, it changes and seems to be less of the high frequency oscillation with the 3300p. Sadly, something dropped on the board and I heard a pop!!!!...............So I moved onto the second board, instead of putting the beads at all, I know the pre driver is the problem, I put a Zobo network of 180ohm and 6600p at the emitter of the predriver to give it a lower impedance load above 130KHz. It helps a lot. Not quite there, but it does help. The compensations are a lot more relaxed. The load impedance for the pre-driver is too high to be stable.
The inverted 3EF is giving a lot of problems, but I really like it as the Vbe temperature drift of the PNP and NPN really cancels out and all I have to do is compensate the big output transistors. Also, I gain 1.2V at the output closer to the rail. It is important for me as I use lower rail voltage, every volt count.
Still working on it, I don't think I got to the bottom, but definitely a step forward.
EDIT:
So far, I have 150ohm and 3300p Zobal for the pre-driver, 50ohm and 3300p on the driver. It really helps. I am still going to play with it more, But it really looks a lot better. With 0.022uF, still looks good.
Those EF need to have a high frequency load ( Zobal ) to keep it tame. I am going to try ferrite bead on the pre-driver tomorrow as it takes a lot of work to pull it out and put it in. Those transistor are discontinued, I only have about 20 ea, I cannot just cut the lead and replace with new one. Have to carefully desolder them without ruining the pcb, put on the beads and solder back on.
Last edited:
Learned a bit more about the infamous green "non-inductive" 4 ohm 100 watt load resistor. The coating was showing a trace of wire at some points, so I scraped enough off to see how the thing is built. AFAICT, it's a winding of 1.5 mm wide resistance tape in the usual spiral manner, coated with a thin layer of the green stuff, then another winding on top, with the opposite twist. The two windings are in parallel and should cancel out the inductance. Sounds good on the surface, but they somehow act as a metallic core for each other, increasing the THD at high frequencies substantially.
You are the expert, I have no idea. I am just happy I got a resistor like this for testing cables. If you use this resistor and test THD with or without the output inductor, you will see a huge difference in THD. That 10 turn inductor doubled the THD of my amp, I have no choice but to get rid of it. I cannot let all my work to be drowned out by one single inductor.
I am going somewhere with taming the amp. It's funny I already have a Zobal on the pcb of 10ohm and 0.1uF film cap. But it really really help when I put a 50ohm and 0.022uF right on the output connector of the amp. Even putting it inside the amp helps. The only difference is this is after the SS relay and the two 10" 10 gauge wire. I think it's the ground return. The return of the load is to the ground plane of the filter cap where the one point ground is, the OPS board ground connects to the speaker connector ( like both the Nakamichi and Acurus). The speaker return current does not go to the OPS PCB. I think the Zobal has to be across the current path. I don't know for sure, this is just my guess.
But it is a lot more stable now than before.
One thing very interesting if you have time to experiment. You test the amp with speaker cable to a load. It is a lot more prone to oscillation if you put a 3300pF on the amp side of the connector, rather on the load end. Also more importantly, it is a lot harder to tame with a 3300pF on the output connector of the amp IF you remove the cable to the load. That is the amp only have 3300pF across the output connector. With the load, it is a lot more stable. I have no problem with capacitor if there is a load on the amp.
But that's the whole issue, I think people test ( even Marantz specified in the manual) with capacitor on the load side and with a 8ohm load. BUT speaker is not 8ohm at high frequency, it likely to be much higher like 50+ohm or even 100ohm at 100KHz on up. You really don't have a load over 500KHz, if you have high cable capacitance, you'll be in trouble. I experimented, even 50ohm Zobal does miracle. I am going to try 22ohm and 0.022uF. I put 50ohm and 3300pF Zobal on the driver stage and 100ohm 3300pF on the pre-driver stage, this really help to stabilize the amp.
You really cannot test amp with a single cap, I found I have to test with 3300pF, 0.022uF and 0.47uF on the amp side and have to make it stable. If you just use 0.47uF, it does not imply it will be stable with 3300pF at all.
This is really live and learn, no books can cover all these. I hope I can get back to the cable again, I already made the 5 X 16gauge pairs 16ft long cable, I just need to have my setup to run THD with and without twisting.
I am going somewhere with taming the amp. It's funny I already have a Zobal on the pcb of 10ohm and 0.1uF film cap. But it really really help when I put a 50ohm and 0.022uF right on the output connector of the amp. Even putting it inside the amp helps. The only difference is this is after the SS relay and the two 10" 10 gauge wire. I think it's the ground return. The return of the load is to the ground plane of the filter cap where the one point ground is, the OPS board ground connects to the speaker connector ( like both the Nakamichi and Acurus). The speaker return current does not go to the OPS PCB. I think the Zobal has to be across the current path. I don't know for sure, this is just my guess.
But it is a lot more stable now than before.
One thing very interesting if you have time to experiment. You test the amp with speaker cable to a load. It is a lot more prone to oscillation if you put a 3300pF on the amp side of the connector, rather on the load end. Also more importantly, it is a lot harder to tame with a 3300pF on the output connector of the amp IF you remove the cable to the load. That is the amp only have 3300pF across the output connector. With the load, it is a lot more stable. I have no problem with capacitor if there is a load on the amp.
But that's the whole issue, I think people test ( even Marantz specified in the manual) with capacitor on the load side and with a 8ohm load. BUT speaker is not 8ohm at high frequency, it likely to be much higher like 50+ohm or even 100ohm at 100KHz on up. You really don't have a load over 500KHz, if you have high cable capacitance, you'll be in trouble. I experimented, even 50ohm Zobal does miracle. I am going to try 22ohm and 0.022uF. I put 50ohm and 3300pF Zobal on the driver stage and 100ohm 3300pF on the pre-driver stage, this really help to stabilize the amp.
You really cannot test amp with a single cap, I found I have to test with 3300pF, 0.022uF and 0.47uF on the amp side and have to make it stable. If you just use 0.47uF, it does not imply it will be stable with 3300pF at all.
This is really live and learn, no books can cover all these. I hope I can get back to the cable again, I already made the 5 X 16gauge pairs 16ft long cable, I just need to have my setup to run THD with and without twisting.
I use LTSpice to do the bode plot, it is not accurate at all. It is to the point it's useless. I wasted so much time on LTSpice before, neither the bode plot stability and THD is useful at all.
It is not easy or even possible to draw by hand. The reason is the miller effect capacitance change with the open loop gain that is continuous vary with frequency.....that is the equivalent capacitance of the miller effect change with frequency. You can only plot bode plot if the capacitance remains constant. In miller effect, the equivalent capacitance at the output of the LTP keep getting lower when frequency goes up because the Cmiller = openloop gain times the feedback capacitor plus 1. If you have a 22pF cap like me, at 1kHz, open loop gain is 1000, the equivalent capacitance at the output of LTP is 22,000pF. But at 100KHz, gain is about 10 only, the equivalent capacitance is down to 220pF!!! You can't draw by hand.
LTSpice is so inaccurate it's not even funny. All the work to optimize THD from simulation absolutely WRONG!!!! I got so low THD using 22p miller cap in simulation, it real life, there is NO DIFFERENCE between 22, 33pF. The extra lead lag network between the collectors of the LTP raises distortion in simulation, it dose nothing in the real circuit. I literally put trim pots and change capacitors to experiment with it. It is tedious.
Also, LTSpice totally missed the output transistor effect on the stability. All the Zobal I put in does not affect the stability on paper as the frequency effected is so high and resistance really does not affect the EF stage, BUT they do make a difference in real circuit. I think there is so many effect LTSpice is missing. I think this is RF type effect scaling down into MHz arena.
shelly_d
Not An Audiophool
Alan:
I would like to make a suggestion: Look to the distortion in your speakers. By that I mean, you are putting tremendous energy into very small differences in your amp, way below 1% distortion. While it is true that 1% distortion in an amp today is poor performance, I recently saw some dramatic results on my own system. I picked up one of these calibration mics and tested my system using Room Equalization Wizard. In addition to providing a room/speaker/system frequency plot (needs help pretty badly I think) it also gave me THD specs. Also pretty high. My system overall measured out at about 4.7%. Since I am using what I think is a low distortion amp (Mac MA6200) most of that is coming from the speakers.
I view my system as the sum of its various parts and it's what I hear that is important. At 4.7% distortion out of my speakers, that is clearly the component that is contributing the most to that result, although I have read that in comparison to amps and other components, speakers are notoriously high in distortion.
Perhaps you might want to just set yourself a target for your amp and speaker leads and then turn your attention and your time and effort to your speakers, taking a practical overall approach, rather then a perfectionist approach.
Just thought I'd share what I found in my own home for your consideration.
Thanks
Shelly_D
I would like to make a suggestion: Look to the distortion in your speakers. By that I mean, you are putting tremendous energy into very small differences in your amp, way below 1% distortion. While it is true that 1% distortion in an amp today is poor performance, I recently saw some dramatic results on my own system. I picked up one of these calibration mics and tested my system using Room Equalization Wizard. In addition to providing a room/speaker/system frequency plot (needs help pretty badly I think) it also gave me THD specs. Also pretty high. My system overall measured out at about 4.7%. Since I am using what I think is a low distortion amp (Mac MA6200) most of that is coming from the speakers.
I view my system as the sum of its various parts and it's what I hear that is important. At 4.7% distortion out of my speakers, that is clearly the component that is contributing the most to that result, although I have read that in comparison to amps and other components, speakers are notoriously high in distortion.
Perhaps you might want to just set yourself a target for your amp and speaker leads and then turn your attention and your time and effort to your speakers, taking a practical overall approach, rather then a perfectionist approach.
Just thought I'd share what I found in my own home for your consideration.
Thanks
Shelly_D
Agree about Spice, LT or other, though there seem to be people over at diyAudio that are pretty good at it. Truthfully, I'm such a dinosaur I never even considered it when I made the Bode suggestion. I'm Mister Physical Measurement, sworn enemy of Mr. Fancypants Math much of the time. Though I've never done it, I know there are techniques used for opamps to measure it open loop, or somehow derive it. Bob Pease probably had something to say about it, and I'll check my notes here.
On the speaker distortion issue, and this is purely my own unsubstantiated belief, speakers are a physical spring and mass system of fairly low frequencies, albeit somewhat complicated. They have a hard time making "unnatural" and unpleasant sounds. Amplifiers, OTOH, have no trouble at all generating harmonics you wouldn't find in a speaker. Amplifiers I've heard that keep THD below 0.01%, and don't have other flaws, sound about the same to me. Amplifiers above that start to please me less and less. I've heard the "you can't hear it below a few tenths of a percent at best" argument, but I remain a skeptic. The best speaker systems I've had were tri-amplified with active crossovers and were very low distortion.
On the speaker distortion issue, and this is purely my own unsubstantiated belief, speakers are a physical spring and mass system of fairly low frequencies, albeit somewhat complicated. They have a hard time making "unnatural" and unpleasant sounds. Amplifiers, OTOH, have no trouble at all generating harmonics you wouldn't find in a speaker. Amplifiers I've heard that keep THD below 0.01%, and don't have other flaws, sound about the same to me. Amplifiers above that start to please me less and less. I've heard the "you can't hear it below a few tenths of a percent at best" argument, but I remain a skeptic. The best speaker systems I've had were tri-amplified with active crossovers and were very low distortion.
