The caution about bandwidth limiting is where do you do it and how many other upstream filters are there? When you cascade them you can get unpredictable and ugly results. Same for high pass filters to limit LF garbage. I think this was in Putzey's Powerpoint as well. Its supposed to be part of the MQA magic.
Microphones don't have nice Gaussian rolloffs either. It may be possible to download the "convolution" files on the web (Recording Impulse Responses Microphone Impulse Response Project ) for different microphones and test them with different bandlimit filters to see what interesting stuff you get. You have similar with tweeters as well, however those rolloffs are extremly location dependent.
Microphones don't have nice Gaussian rolloffs either. It may be possible to download the "convolution" files on the web (Recording Impulse Responses Microphone Impulse Response Project ) for different microphones and test them with different bandlimit filters to see what interesting stuff you get. You have similar with tweeters as well, however those rolloffs are extremly location dependent.
Part of the bonus of DIYAudio is knowing what happens from 1's and 0's to displacements of air. 🙂
But, yes, an anti-imaging filter on the DAC followed by a cascade of bandlimiters at a (useless, IMO) preamp and (useful!) amplifier can add up for sure.
But, yes, an anti-imaging filter on the DAC followed by a cascade of bandlimiters at a (useless, IMO) preamp and (useful!) amplifier can add up for sure.
I am genuinely surprised that you are getting serious overshoot.
I disable the input filter for the slew tests.
I usually start off by doing the main comp optimization (after sim) without any front end filter. after comp however there is little or no overshoot. Once I have it optimized, then I do the front end BW filter. Typically I set this between 250 and 500 kHz.
I have some graphics showing the compensation design evolution in my sx-amp write-up.
Last edited:
I would like to see more data on using a starter's pistol to generate a true impulse. This study was very well instrumented (G.R.A.S. 1/8" mic,>100kHz) and the plots don't even look close to an impulse. Earthworks says they use a spark at least in a few places which is more like a doublet and needs integration to yield the impulse response. You've seen stroboscopic movies of balloons popping I doubt any two are alike.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4617326/
EDIT - Found one, not even close. http://www.openairlib.net/anechoicdb/content/starter-pistol
Last edited:
Mark, not sure I follow this. Risetime is the time from 5% to 95% (or was it 10% to 90%) of level, but under non-slewrate-limit conditions. It is a small-signal parameter, while slew rate is a large-signal parameter.
I don't think it is possible to directly relate slew rate to rise time as the two parameters are measured or observed under very much different operating conditions.
Jan
Mark J. , Jan is right on this one. Rise-time and slew-rate are not directly related to each other. Slew rate is determined by internal clipping of one or more stages due to not being able to drive the compensation caps with enough peak current.
At low output levels, rise-time is always there, and sometimes, IF you design the amp right, will almost or will completely dominate over any slew rate imposition. Open loop designs like Charles Hansen does, for example, does this regularly. There is just no internal overshoot within a feedback loop to create slew rate limiting, IF you don't have a feedback loop in the first place. The CTC Blowtorch works this way as well. Rise-time is all you get.
However this is not very practical for the vast majority of amps and preamps that we have to design, so we can offset the slew-rate problem by either increasing the inherent slew-rate of the amp design by speeding it up or degenerating the input stage with resistors or jfets, OR we can bandwidth limit the input rise-time (frequency) of the input signal with a filter, either a simple RC or even a more complex 2pole filter. Usually, both techniques are necessary to get an amp that is absolutely 'unslewable'. Once we achieve 50V/us or so, who cares anyway? '-)
At low output levels, rise-time is always there, and sometimes, IF you design the amp right, will almost or will completely dominate over any slew rate imposition. Open loop designs like Charles Hansen does, for example, does this regularly. There is just no internal overshoot within a feedback loop to create slew rate limiting, IF you don't have a feedback loop in the first place. The CTC Blowtorch works this way as well. Rise-time is all you get.
However this is not very practical for the vast majority of amps and preamps that we have to design, so we can offset the slew-rate problem by either increasing the inherent slew-rate of the amp design by speeding it up or degenerating the input stage with resistors or jfets, OR we can bandwidth limit the input rise-time (frequency) of the input signal with a filter, either a simple RC or even a more complex 2pole filter. Usually, both techniques are necessary to get an amp that is absolutely 'unslewable'. Once we achieve 50V/us or so, who cares anyway? '-)
slew rate can be a small signal parameter
its slew rate Limit that's nonlinear, same as I, V vs their limits
"Power Bandwidth" is the frequency at full (linear) output V where slew rate Limiting starts on the sine zero crossing
its slew rate Limit that's nonlinear, same as I, V vs their limits
"Power Bandwidth" is the frequency at full (linear) output V where slew rate Limiting starts on the sine zero crossing
Yes, jcx, slew rate CAN be a small signal parameter, but only with 'slew enhanced' input stages like the 741S, that filled a need 40 years ago. IF the input stage is LINEAR Class A, then slew rate distortion will only happen at the extremes of input drive, not at low levels.
What you mean by slew rate depends on what you are using it for. Is it the rate where all the internal transistors are in saturation (marketing slew rate) or the maximum rate at which there is meaningful information at the output ({\mathrm {SR}}\geq 2\pi fV_{{{\mathrm {pk}}}},) https://en.wikipedia.org/wiki/Slew_rate. I have seen 3% distortion for that limit but you may need lower depending on your application. Or higher for competitive reasons.
Ah.
I might make this a jumper option so the filter can be enabled and disabled easily, without needing a soldering iron and without socketing the cap(s).
As for my sloppy armwaving approximation that risetime ( = 0.35 / bandwidth) is related to slew rate, perhaps the easiest refutation is this: apply the D. Self "Constant gm" method of increasing slew rate without changing bandwidth at all, viz. increase IPS Itail and simultaneously increase IPS degeneration so that IPS gm remains constant. Wallah: slew rate goes up (SR = Itail/Cc) but bandwidth is unchanged (BW = gm/Cc). Thus slew rate and bandwidth are not dependent. Thanks to the bandwidth-risetime relationship, slew rate and risetime are not dependent. The proof she is complete.
I remember "risetime ( = 0.35 / bandwidth)" being quoted in all the TEK manuals to describe 'scope probe/amp characteristics.
That's commonly given for a simple RC time constant, but doesn't work for all cases. More info here: https://en.wikipedia.org/wiki/Rise_time
You guys are mixing different different situations up. There's an old rule of thumb that was meant to apply to undegenerated bi-polar amplifiers ONLY where slew rate and BW are all related by C and the tail current. It appears in lots of tutorials, a basic educational tool.
And thanks Markw4 for reminding me that Jim Roberge was using the tri-wave test even back in 1970. He was my advisor for my last two years, and two years ago I went to his memorial service, a sad closure. John would appreciate that he liked to race Jags.
Hmmm.... I have several pistols to try and capture the waveform. Starter blanks are not even close to live ammo sound. and I have an 1/8 incher mic.
Should be fun experiment....... but not many people are going to use live ammo for their test source. Maybe a fire cracker? Small one? 🙂
THx-RNMarsh
Even more fun trying to get the shock wave (sub audible frequency) before the sound wave hits... Live ammo and explosions are best. I dont know whether a bullet does have an initial shock wave, explosions do.
Found a copy on the 'net and ordered it. Expensive, though; better be worth it! 😡
Jan
A shock wave can get beyond the 194dB limit of normal propagation. Estimates of Krakatoa are 300dB. A spark gap actually does generate a shock wave.
The mis-conception in the article is that pistols, etc. are broadband sources of compact support not necessarily an impulse. When you capture the stimulus and the resulting room response with the same mic you can use deconvolution techniques to extract the room response. They are not useful in capturing the impulse response of the mic. If you look at the spectra in that link this is obvious (at least to me, but I might be wrong).
Last edited:
- Status
- Not open for further replies.