"The devil is the detail"
Thank you for the replies.
This is invaluable information. Considering going surface mount with my current creation and soldering surface mount films always fills me with dread whereas ceramics appear to be quite tough.
NP0/COG seem fine for compensation caps. If using class II caps for decoupling it seems wise to go for X7R, over rate the voltage and use several times the required capacitance. The low Q is appealing though.
Thank you for the replies.
This is invaluable information. Considering going surface mount with my current creation and soldering surface mount films always fills me with dread whereas ceramics appear to be quite tough.
NP0/COG seem fine for compensation caps. If using class II caps for decoupling it seems wise to go for X7R, over rate the voltage and use several times the required capacitance. The low Q is appealing though.
As long as the BJT models are reasonably accurate, they should correspond well to real life. If you have two simulations with different models that are still accurate, then the simulations will get different results but both should still be realistic.
A 192KHz samplerate with +-2V max output is normal for modern PC audio. That is a maximum of 4V in 5.2uS which is 770mV/uS. If your amp has a gain of 40 it will need a slew rate of at least 30V/uS. But this is only the minimum requirement.
Most stability compensation methods load down the internals of an amp at RF so that the output can't change before feedback is in-phase. For this reason high-speed signals can and will load VAS or IPS out of their linear region. This in turn will reduce linearity for all other signals. A 25mV high-frequency triangle wave can cause highly compensated amplifiers to go out of spec; this kind of signal is common for computer sound where there is lots of interference from SMPS supplies.
Considering this it is not surprising that amps with high internal RF load capacity are well-liked. Possible solutions may include:
1: Input filter.
A: loads the preceeding circuits so may be counterproductive.
B: Multi-pole bootstrapped input filters are possible without problem A, but who ever does this? Furthermore, for better or worse, many are apprehensive about any complex filter in the signal path.
2: High innate slew-rate.
3: High RF internal load capacity (running VAS hot, etc.)
3: Instant recovery from high-speed signals.
High-Q bypass caps under 470nF almost certainly will resonate badly with the onboard lytics. Since most amps have low PSRR at RF, the rail resonance actually modifies the stability margin, for better or worse (usually worse). People using very fast opamps have found that 10u lytics soldered as close to the power pins as possible are the best solution to prevent oscillation because their loss dampens any resonance on the rails. I think this would be good practice for fast power amps as well. Perhaps bypass resonance could be used strategically to improve stability, but it would also constantly be resonating with RFI, like an antenna. I believe JC when he says 100nF styrenes work, but I think there is a lot more to the story.
PS. Styrene caps are often of terrible quality and you can pull the leads right out. I guess stick them back in with some dielectric grease? Of course the extra contact loss in this case may be helping instead of hindering.
best cap film source -
There is only one remaining highest quality mfr of polystyrene specifically for capacitors. It is in China. If you want them, you need to spec them as source to your cap film buyer/mfr. Or, you can save the trouble and get the ps types from REL-CAP in southern california. They will make you the best quality cap with the best film to what ever tolerance you like. Dont expect the leads to fall off or drop out.... but they aint cheap, either. 100% tested. But for .1mfd, the cost is not terrible. Cant buy just 1-2 though from them. They can refer you to a distributor, however. They also make the best audio caps IMHO... I share a patent with them (MultiCap) which takes a custom built cap machine to make them. The owner (Bas Lim) has a PHd in material science... he and I worked out the details of how to make a great cap and those and his standard line are superior.
THx-RNMarsh
There is only one remaining highest quality mfr of polystyrene specifically for capacitors. It is in China. If you want them, you need to spec them as source to your cap film buyer/mfr. Or, you can save the trouble and get the ps types from REL-CAP in southern california. They will make you the best quality cap with the best film to what ever tolerance you like. Dont expect the leads to fall off or drop out.... but they aint cheap, either. 100% tested. But for .1mfd, the cost is not terrible. Cant buy just 1-2 though from them. They can refer you to a distributor, however. They also make the best audio caps IMHO... I share a patent with them (MultiCap) which takes a custom built cap machine to make them. The owner (Bas Lim) has a PHd in material science... he and I worked out the details of how to make a great cap and those and his standard line are superior.
THx-RNMarsh
Last edited:
Agreed completely.
A designer who needs to use such large base stoppers to avoid HF instability has done a poor job somewhere in the output stage in regard to layout, decoupling, or something else. There are many ways to screw up in this way.
Although not a savior for some HF instability problems, I like to use distributed Zobel networks (the series R-C to ground from the output line) when multiple output pairs are used, one close to each output pair, with the PNP and NPN devices interleaved. Smaller values of capacitor and larger value of resistor can then be used in the Zobel, and the Zobel's effect persists to much higher frequencies. With today's faster output transistors, behavior of the whole output stage is important to higher frequencies.
In the off chance that a designer is using large gate stopper resistors in a misguided attempt to maintain stability without the use of an output inductor when feeding capacitive loads - this is just foolish.
Ferrite beads can work magic, but don't tell some audiophiles you are using them - they are frowned upon by many, whether justifiably or not. Also, keep in mind that some ferrite beads can saturate at fairly low current, losing some of their properties. Never under-estimate the amount of base current that can flow into the base of a power transistor under some conditions.
Cheers,
Bob
Neat info, Marsh. On that topic, what about polycarbonate capacitors like the old Evox? American Capacitor seems to be the only current source, but I couldn't get through to them:
AMERICAN CAPACITOR CORPORATION - Capacitors proudly made in the United States of America
Also look at the case styles:
PLASTIC CAPACITORS, FILM CAPACITOR CASE TYPES
Yes, I know, much ado over caps, but I'm curious. They can also do feed-thru caps which would clearly be ideal for filtering purposes.
AMERICAN CAPACITOR CORPORATION - Capacitors proudly made in the United States of America
Also look at the case styles:
PLASTIC CAPACITORS, FILM CAPACITOR CASE TYPES
Yes, I know, much ado over caps, but I'm curious. They can also do feed-thru caps which would clearly be ideal for filtering purposes.
But at those currents will the transistor have enough Ft to be unstable? Could we find an Ib current over which oscillation is not a problem and set as the lower limit on bead saturation? Then when the amp is under extreme load
Also, there is the option of choosing a bead with very high resistive loss, so that you can just use it as an inductor and use your own parallel R. I wonder if there are beads like this that have higher saturation points?
There are also inductors like this, but would they withstand the max Ib?
Fest-Induktivitäten, axial bei reichelt elektronik
Polycarb is below polyprop and polysty/teflon. A little like mylar for audio. A good or B grade. Use pp most of the time and ps for your best effort..... but not ps with tube circuits as they can not take the high temps.
As a PSuppply bypass I use high quality ps or pp with a small value (typically .1 Ohm or less) pure series R to critically damp it.
[ Dont depend on esr for damping as it is freq and volt/current and temp sensitive and is usually too high value and overdamped.]
THx-RNMarsh
As a PSuppply bypass I use high quality ps or pp with a small value (typically .1 Ohm or less) pure series R to critically damp it.
[ Dont depend on esr for damping as it is freq and volt/current and temp sensitive and is usually too high value and overdamped.]
THx-RNMarsh
Last edited:
Good points Bob thanks for sharing.
The base input current can be quite high - speaker loads often dip down to 2 or 3 ohms.
Re the ferrite beads, one really has to go to the data sheets to select the right type, but saturation is always an issue I agree. Once many years ago I was designing an SMPSU and had a noise problem which was solved by fitting a bead over the lead of a 'lytic. The bead actually ran hot to the touch!
I've tried ferrite beads on the inputs of preamps and can honestly say I hear no difference - but, GEB types don't like 'em, as of course a few other things.
Re SR, I've seen many calculations about the optimal value, but I think the generally accepted rule of at least 1 V per microsecond per max peak output voltage is a good guide. Of course, quite easy to go a bit above or below this so on a big 250 watter you get a guidance of say 80 V per microsecond minimum. I don't think a figure of double this unreasonable- so 150 to 200 V per microsecond.
Got some links to these mythical beasts? AFAIK, they are rare even in Spice world.
I simulated ASTX's amp , his ppm was lower or equal in real life with 2 pole comp.
- even with the long ribbon cable.
http://www.diyaudio.com/forums/solid-state/235194-2stageef-high-performance-class-ab-power-amp-200w8r-400w4r.html
"The cat" made fun of it for less than 200v/us slew , but it does perform.
OS
- even with the long ribbon cable.
http://www.diyaudio.com/forums/solid-state/235194-2stageef-high-performance-class-ab-power-amp-200w8r-400w4r.html
"The cat" made fun of it for less than 200v/us slew , but it does perform.
OS
How did you calculate this? Slew rate for the sine signal is its maximum slope.
SR = dv/dt = 2*pi*f*Vp
Vp is amplitude and f is frequency.
For 192kHz, theoretical max. frequency is 96kHz. 96kHz with 2V amplitude makes 1.2V/us.
Anyway, it is not very important. There is no 96kHz music signal at full scale. Have you done analysis of real music signals? This is where my question is addressed, I do not speak about non-existent music signals.
I am really able to multiply link level output slope with power amplifier gain, and other trivialities mentioned. My point is the real slew rate of music signals. I have done quite a lot of analysis, and the value is pretty low. I also have no problem to build an amplifier with slew rate numbers high enough, yet not sure it is really necessary.
I use 3.6 ohm base stoppers for MJL1302/3281, is there anything wrong?
Do the base stoppers really affect transfer curve linearity that much? How about VAS output impedance, does anybody care?
keantoken said:
I've also done a lot of sim work on ASTX's excellent amp and variants (which i believe started as a variant of your excellent Badger
)It's taught me a lot about how to move SPICE world closer to 'real world' and we've corrected several important BJT models along the way.
But its not sub ppm at 20kHz. However all the latest versions are sub ppm at 1kHz.
AFAIK, the only sub ppm 20kHz SPICE world designs are Edmond Stuart's and he has some 'real world' confirmation too.
Small Signal Bandwidth
If the small signal bandwidth is low enough, the amp CANNOT slew limit without also going into voltage limiting. Then the important factor is how quickly and cleanly the amp goes into clipping and recovers from overload.
That's why its VITAL to have a sensible LP filter on the input of ALL amps. The smaller the slew rate, the lower this LP filter MUST be.
______________
Here's an interesting EXPERIMENTAL FACT.
ALL properly conducted Double Blind Listening Tests on bandwidth limitation of electronics show that of those who can reliably tell the difference (ie true golden pinnae), ALL prefer the sound band-limited.
The studies this Millenium (with supa dupa and not so supa digital sources) confirm those of the previous including my own tests. Some notable members of this forum have been involved in these tests which suggest down to even 15kHz may be PREFERRED.
It may be that present amps have insufficient slew to support even 15kHz bandwidth limit ... but I don't think so as the bandwidths & slews from the calculations are quite small. (This is a possible explanation for the anecdotal preference for CFA topologies in listening tests.)
Over to da pedants to explain this EXPERIENTAL FACT .. and ignore the Deaf Golden Pinnae who refuse to conduct Double Blind Tests bla bla
Meanwhile, if you want to incorporate a feature in your amps that will guarantee its audible preference among true golden pinnae, use a brick wall filter at 20kHz ..
Duu.uuh! Red Book CD has just this built in but the 21st century experimental results hold even with these most evil of digital sources.
There is another important factor in this calculation the small signal bandwidth of the amplifier.
If the small signal bandwidth is low enough, the amp CANNOT slew limit without also going into voltage limiting. Then the important factor is how quickly and cleanly the amp goes into clipping and recovers from overload.
That's why its VITAL to have a sensible LP filter on the input of ALL amps. The smaller the slew rate, the lower this LP filter MUST be.
______________
Here's an interesting EXPERIMENTAL FACT.
ALL properly conducted Double Blind Listening Tests on bandwidth limitation of electronics show that of those who can reliably tell the difference (ie true golden pinnae), ALL prefer the sound band-limited.
The studies this Millenium (with supa dupa and not so supa digital sources) confirm those of the previous including my own tests. Some notable members of this forum have been involved in these tests which suggest down to even 15kHz may be PREFERRED.
It may be that present amps have insufficient slew to support even 15kHz bandwidth limit ... but I don't think so as the bandwidths & slews from the calculations are quite small. (This is a possible explanation for the anecdotal preference for CFA topologies in listening tests.)
Over to da pedants to explain this EXPERIENTAL FACT .. and ignore the Deaf Golden Pinnae who refuse to conduct Double Blind Tests bla bla
Meanwhile, if you want to incorporate a feature in your amps that will guarantee its audible preference among true golden pinnae, use a brick wall filter at 20kHz ..
Duu.uuh! Red Book CD has just this built in but the 21st century experimental results hold even with these most evil of digital sources.
Last edited:
There is another important factor in this calculation the small signal bandwidth of the amplifier.
If the small signal bandwidth is low enough, the amp CANNOT slew limit without also going into voltage limiting. Then the important factor is how quickly and cleanly the amp goes into clipping and recovers from overload.
That's why its VITAL to have a sensible LP filter on the input of ALL amps. The smaller the slew rate, the lower this LP filter MUST be.
______________
Here's an interesting EXPERIMENTAL FACT.
ALL properly conducted Double Blind Listening Tests on bandwidth limitation of electronics show that of those who can reliably tell the difference (ie true golden pinnae), ALL prefer the sound band-limited.
The studies this Millenium (with supa dupa and not so supa digital sources) confirm those of the previous including my own tests. Some notable members of this forum have been involved in these tests which suggest down to even 15kHz may be PREFERRED.
It may be that present amps have insufficient slew to support even 15kHz bandwidth limit ... but I don't think so as the bandwidths & slews from the calculations are quite small. (This is a possible explanation for the anecdotal preference for CFA topologies in listening tests.)
Over to da pedants to explain this EXPERIENTAL FACT .. and ignore the Deaf Golden Pinnae who refuse to conduct Double Blind Tests bla bla
Meanwhile, if you want to incorporate a feature in your amps that will guarantee its audible preference among true golden pinnae, use a brick wall filter at 20kHz ..
Duu.uuh! Red Book CD has just this built in but the 21st century experimental results hold even with these most evil of digital sources.
Please make experiments of these conditions:
1. Inputs are limited bandwidth but using 2 amp but different slew rate (ex. one amp has 20kHz power bandwidth and the other have 200kHz power bandwidth).
2. Inputs are not limited bandwidth but using 2 amp but different slew rate (ex. one amp has 20kHz power bandwidth and the other have 200kHz power bandwidth).
Last edited: