Whilst you're here Robert - do you know if any of these have made their way to the UK yet? I'd love to get my hands on one asap!
Enjoy the benefits of a premium membership!
Upgrade Account
-
TalkBass has been independent since 1998. Add your voice.
Create a free account to reply to discussions, view embedded media, and browse with fewer display ads.
Join free Log in
Want zero display ads or expanded classifieds tools? Compare plans.
New Keeley Bass compressor
- Thread starter Jman2089
- Start date
bongomania
Commercial User
Here's my RTA:
The many lines represent a sampling of the whole range of possible settings; for the larger cluster above, I adjusted their output gain so they'd all be in the same range. Then I selected two for more clear comparison, one at 4:1 (dark blue) and the other at inf:1 (light blue), and set their gain so they'd sit lower. I used a rough eyeball judgment to pick the high level dB points that would be most meaningful. As you can see, there is a 7 to 10 dB drop from low to high frequencies.
For reference, this is with the "chirp" method of RTA, as it's quick and easy. Other methods may show different results.
Again, I am in no way disparaging the pedal on account of this frequency spectrum! We wanted a bass pedal from Keeley, and by gosh we got one. Plenty of bass.
And this type of spectrum is often necessary to achieve perceived equality between lows and highs during periods of compression, as the peaks are obviously reduced, which would make the bigger bass waves seem weak. So by raising the lows up a bit, we get our bass perception "back" even though the peaks are compressed.
The many lines represent a sampling of the whole range of possible settings; for the larger cluster above, I adjusted their output gain so they'd all be in the same range. Then I selected two for more clear comparison, one at 4:1 (dark blue) and the other at inf:1 (light blue), and set their gain so they'd sit lower. I used a rough eyeball judgment to pick the high level dB points that would be most meaningful. As you can see, there is a 7 to 10 dB drop from low to high frequencies.
For reference, this is with the "chirp" method of RTA, as it's quick and easy. Other methods may show different results.
Again, I am in no way disparaging the pedal on account of this frequency spectrum! We wanted a bass pedal from Keeley, and by gosh we got one. Plenty of bass.
And this type of spectrum is often necessary to achieve perceived equality between lows and highs during periods of compression, as the peaks are obviously reduced, which would make the bigger bass waves seem weak. So by raising the lows up a bit, we get our bass perception "back" even though the peaks are compressed.
Cyrus, I have used True RTA on two different PC’s and the output signal amplitude is not constant across the spectrum.
Here is what I measured at various frequencies with the generator set to 0dBu (the sound card was calibrated per True RTA setup and it displays a flat response with output connected to input):
Freq. mVrms
20Hz 816
100Hz 816
500Hz 813
1KHz 806
5KHz 636
10KHz 426
20KHz 112
True RTA internally calculates the dB values based on its calibration error correction. It is a good tool for measuring non-dynamic signals from amplifiers and such. But, since the output has wide variances in amplitude it really can’t show a reliable measurement for compression.
To make bandwidth frequency measurements with a Spectrum Analyzer on a compressor (while it's compressing), you need a white noise source that has a constant output from 20Hz to 20KHz. Even then you may have to set the analyzer to average over several measurements to reduce flutter.
-Frank
Here is what I measured at various frequencies with the generator set to 0dBu (the sound card was calibrated per True RTA setup and it displays a flat response with output connected to input):
Freq. mVrms
20Hz 816
100Hz 816
500Hz 813
1KHz 806
5KHz 636
10KHz 426
20KHz 112
True RTA internally calculates the dB values based on its calibration error correction. It is a good tool for measuring non-dynamic signals from amplifiers and such. But, since the output has wide variances in amplitude it really can’t show a reliable measurement for compression.
To make bandwidth frequency measurements with a Spectrum Analyzer on a compressor (while it's compressing), you need a white noise source that has a constant output from 20Hz to 20KHz. Even then you may have to set the analyzer to average over several measurements to reduce flutter.
-Frank
Last edited:
I received the Keeley today. I tried it out and it is a very good compressor, but unfortunately I did not prefer the overall tone of this unit compared to the Aguilar TLC. So for me the TLC remains my favorite compressor!
bongomania
Commercial User
I know you've been trying to tell me about this for a while now, but there are a couple reasons my head has been stubborn about accepting it. I need to buy a proper noise generator, and I need to find time to do the multiple scope readings, take screen shots, transfer them to my PC, and figure out how to superimpose them into one frame. It takes between 10 and 20 shots to cover the available settings of a typical comp. Kind of a big hassle, TBH--though I know, like flossing, it is what I _should_ be doing.
Update on TrueRTA output: My DMM must have a filter for the AC measurements, which is very odd because the DMM has the ability to measure frequency. I looked at the output of my USB sound module at 50Hz, 500Hz, 5KHz and 10KHz with an O-Scope and the TrueRTA generator output amplitudes were nearly the same for each of those frequencies. So, if you have a decent sound card or external sound module then the output should be somewhat accurate.
I want to clarify why an FFT plot that uses a sweeping generator (chirp) on a compressor (that is in a dynamic state of compressing) yields meaningless results.
All compressor attack and release specifications are in units of time and are never specified in units of frequency. FFT plots are used for measuring signal amplitudes and/or signal phase relationships vs frequency, not time. What you are seeing with those plots is dynamic amplitude measurements due to the attack response of the compressor. It’s essentially an envelope response with regards to units of time, not frequency.
I measured the ‘chirp’ sweep from TrueRTA with a 1/24 oct resolution from 10Hz to 20KHz to be approx 250mS in duration. Since it takes 250mS for the generator to sweep 10Hz to 20KHz, you could look at the horizontal axis of that plot from 10Hz to 20KHz as a linear ‘time’ snapshot of 250mS. Instead of the frequency units, substitute 0mS for 10Hz and 250mS for 20KHz. Then visualize linear graduations of the horizontal axis, where 125mS would be exactly at the ½ point (approx at the 450Hz mark). I know it is a bit confusing when there is a logarithmic scale on the plot, but the time units need to be equal divisions across that horizontal axis.
Now if we look at that plot again, the attack of the compression is approx 70mS (where the large dip is located near 80Hz on the plot). Then the detector in the compressor sidechain relaxes from the overshoot from 70mS to 78mS (the bump after the overshoot at approx 110Hz on the plot). Then the attack continues on for several more time constants to the end of the plot. That swell after the dip at 70mS looks like characteristics of some feedback style of compression, but the description on the website says that it is a feed-forward design... so that must be the character of this adaptive timing circuit.
Note: Attack and release times for analog compressors use RC time constants (R x C = T). The RC time constant is called tau and its symbol looks similar to the letter ‘T’. One time constant for charging the capacitor for a RC circuit yields a 63% full charge. To have a 100% full charge it takes a total of 5 time constants or 5T. Thus if the attack is specified at 50mS, it would take a total of 250mS for the compression to reach its static state. (There are other ways to modify these times within the circuit, but I am trying to keep this as simple as possible with typical feed-forward compression using RC timers.)
Looking at that FFT plot again, we see that large dip at 70mS is at approx 63% compression (one time constant). Towards the end of the 250mS ‘chirp’ sweep, we start to see the static value of the compression… 5 time constants. With this compressor having an adaptive attack response, those RC time constants are also varying with the degree of compression applied. It gets really complicated with continuously changing variables.
The best way to view dynamic envelopes is with time based tools, like O-Scopes. A spectrum analyzer is useful for amplitude and phase vs frequency measurements that are not affected by time. To get an accurate freq plot of a compressor while it is compressing, you MUST be at the static value of compression for the FFT measurement to mean anything useful. The easiest way is to use a white noise source with an RTA that averages many measurements over a period of time. A white noise source will hold the compressor at a static level of compression, so those measurements will be more accurate and meaningful. And averaging many measurements over a period of time reduces flutter measurement errors from the white noise.
Another easy way would be to use an external signal generator that you can sweep manually without interruptions in the output signal (constant output while changing frequency, like most analog signal generators). This way you could start the signal generator and allow the compressor to reach its static state of compression and the signal amplitude will hold it there. Then set your RTA to hold peaks while you manually sweep the frequency spectrum. The objective here is to eliminate any effects of the attack response of the compressor.
The white noise generator in TrueRTA sucks or I just haven’t put enough time into making it work properly. When I set it for 0dBu amplitude it has an output near -30dBu and looks like the inverse of pink noise. The pink noise generator in RTA looks a lot more like a white noise generator, but its output is also about -30dBu from the generator amplitude setting. So I suggest using a more reliable white noise source. BTW, I have seen free white noise files available for download that could possibly be used for a signal source with TrueRTA as the measuring tool. There are also several free white noise generators available. You would need to verify that the white noise source/generator has constant amplitude across the spectrum to be used as a good source for measurements. If you have access to two PC’s then you could set the one without TrueRTA as a white noise signal source and verify a flat response at the correct amplitude with the RTA. It would only be a little extra work with the tools available to you, if you want an accurate frequency response of a compressor that’s compressing. If you want to look at the frequency response of a compressor that is set to no compression, then TrueRTA’s sweep mode works pretty well.
Hey Cyrus, you have battered me enough to make my circuits better... now I have to return the favor!
-Frank
I want to clarify why an FFT plot that uses a sweeping generator (chirp) on a compressor (that is in a dynamic state of compressing) yields meaningless results.
All compressor attack and release specifications are in units of time and are never specified in units of frequency. FFT plots are used for measuring signal amplitudes and/or signal phase relationships vs frequency, not time. What you are seeing with those plots is dynamic amplitude measurements due to the attack response of the compressor. It’s essentially an envelope response with regards to units of time, not frequency.
I measured the ‘chirp’ sweep from TrueRTA with a 1/24 oct resolution from 10Hz to 20KHz to be approx 250mS in duration. Since it takes 250mS for the generator to sweep 10Hz to 20KHz, you could look at the horizontal axis of that plot from 10Hz to 20KHz as a linear ‘time’ snapshot of 250mS. Instead of the frequency units, substitute 0mS for 10Hz and 250mS for 20KHz. Then visualize linear graduations of the horizontal axis, where 125mS would be exactly at the ½ point (approx at the 450Hz mark). I know it is a bit confusing when there is a logarithmic scale on the plot, but the time units need to be equal divisions across that horizontal axis.
Now if we look at that plot again, the attack of the compression is approx 70mS (where the large dip is located near 80Hz on the plot). Then the detector in the compressor sidechain relaxes from the overshoot from 70mS to 78mS (the bump after the overshoot at approx 110Hz on the plot). Then the attack continues on for several more time constants to the end of the plot. That swell after the dip at 70mS looks like characteristics of some feedback style of compression, but the description on the website says that it is a feed-forward design... so that must be the character of this adaptive timing circuit.
Note: Attack and release times for analog compressors use RC time constants (R x C = T). The RC time constant is called tau and its symbol looks similar to the letter ‘T’. One time constant for charging the capacitor for a RC circuit yields a 63% full charge. To have a 100% full charge it takes a total of 5 time constants or 5T. Thus if the attack is specified at 50mS, it would take a total of 250mS for the compression to reach its static state. (There are other ways to modify these times within the circuit, but I am trying to keep this as simple as possible with typical feed-forward compression using RC timers.)
Looking at that FFT plot again, we see that large dip at 70mS is at approx 63% compression (one time constant). Towards the end of the 250mS ‘chirp’ sweep, we start to see the static value of the compression… 5 time constants. With this compressor having an adaptive attack response, those RC time constants are also varying with the degree of compression applied. It gets really complicated with continuously changing variables.
The best way to view dynamic envelopes is with time based tools, like O-Scopes. A spectrum analyzer is useful for amplitude and phase vs frequency measurements that are not affected by time. To get an accurate freq plot of a compressor while it is compressing, you MUST be at the static value of compression for the FFT measurement to mean anything useful. The easiest way is to use a white noise source with an RTA that averages many measurements over a period of time. A white noise source will hold the compressor at a static level of compression, so those measurements will be more accurate and meaningful. And averaging many measurements over a period of time reduces flutter measurement errors from the white noise.
Another easy way would be to use an external signal generator that you can sweep manually without interruptions in the output signal (constant output while changing frequency, like most analog signal generators). This way you could start the signal generator and allow the compressor to reach its static state of compression and the signal amplitude will hold it there. Then set your RTA to hold peaks while you manually sweep the frequency spectrum. The objective here is to eliminate any effects of the attack response of the compressor.
The white noise generator in TrueRTA sucks or I just haven’t put enough time into making it work properly. When I set it for 0dBu amplitude it has an output near -30dBu and looks like the inverse of pink noise. The pink noise generator in RTA looks a lot more like a white noise generator, but its output is also about -30dBu from the generator amplitude setting. So I suggest using a more reliable white noise source. BTW, I have seen free white noise files available for download that could possibly be used for a signal source with TrueRTA as the measuring tool. There are also several free white noise generators available. You would need to verify that the white noise source/generator has constant amplitude across the spectrum to be used as a good source for measurements. If you have access to two PC’s then you could set the one without TrueRTA as a white noise signal source and verify a flat response at the correct amplitude with the RTA. It would only be a little extra work with the tools available to you, if you want an accurate frequency response of a compressor that’s compressing. If you want to look at the frequency response of a compressor that is set to no compression, then TrueRTA’s sweep mode works pretty well.
Hey Cyrus, you have battered me enough to make my circuits better... now I have to return the favor!
-Frank
Ha ha. Whoa. I have no idea what any of this stuff means but I can tell you without a doubt that this pedal is great.
Rec'd mine Thursday. Used at rehearsal and will use at a gig tonight. Really nice, clean and beefs up my sound in a very pleasant way. Works very well with my Bogner Harlow. Super fat.
thank you for this answer, you're a great pedal manufacturer.
off-topic: will you re-release the holy fuzz?
Howdy folks!
Boy this is fun! Real good stuff here. Been busy but I owe some well thought out replies around here. I purchased the TrueRTA software and am getting to know it. I hope everyone has a great Thanksgiving! I will be back with data and some more info!
Thanks folks very much for the kind words on the comp... the little sister comp (GC-2) is also doing well from reports... ;-)
rk
Invalid Link Removed
Boy this is fun! Real good stuff here. Been busy but I owe some well thought out replies around here. I purchased the TrueRTA software and am getting to know it. I hope everyone has a great Thanksgiving! I will be back with data and some more info!
Thanks folks very much for the kind words on the comp... the little sister comp (GC-2) is also doing well from reports... ;-)
rk
Invalid Link Removed
Last edited:
So i've been using the pedal for a week now and i'm in love! I suffer with hand problems, but one of the bands I play in I have quite an aggressive tone. This pedal has kept all of that 'vibe' and allowed me to back right off. And as a studio engineer, it really worked in the same way i'd expect the DBX rack comps to work. It was a sound I recognised and found it easy to work with. Nice work Mr Keeley!
And the blue LED is bright enough to be seen from space! I've had to stick some tape over it to stop the danger of blinding
And the blue LED is bright enough to be seen from space! I've had to stick some tape over it to stop the danger of blinding
Yes, way too bright. I put a piece of scotch tape over it and colored the tape with a black Sharpie. It was still too bright so I did it again, now it's much better!
svtb15
Inactive
- Mar 22, 2004
- 13,004
- 4,741
- Disclosures
- I play it all. Whatever works for the gig. Q+
svtb15
Inactive
- Mar 22, 2004
- 13,004
- 4,741
- Disclosures
- I play it all. Whatever works for the gig. Q+
