24 May 2009 H.H. Prabhavishnu Swami bhajan session

Here's a quick mix down I did this morning of Parama Karuna, from the recent (24 May) bhajan session with H.H. Prabhavishnu Swami.

In this mix I worked on getting the response chorus louder in the mix, as per the request of one correspondent. In order to do this the cartal sound (which is strong in the chorus mic) untamed. I've put a multipressor (a multiband compressor) on it to try to bring it down a little. Having the vocals louder is nice, but the cartals become overbearing. With a low-pass filter on the cartals go down in the mix, but the vocals disappear.

For future recordings I think I'll mic up the audience/chorus with a couple of Behringer C2s, strategically placed behind the cartal players. In that case a microphone specifically to close mic the cartals (to avoid that "playing in a cavern effect") might be a good idea. That way their presence in the mix can be controlled later. Highly directional (hypercardioid) mics such as the Shure Beta series seem like a good avenue of investigation, to avoid leakage of the cartals into other tracks.

The 66Hz tone is still being triggered from the bottom end of the mrdanga [see here], but as Janardana was playing "laghu" style it doesn't feature so much. With the emphasis on the top end of the mrdanga it doesn't sound so plastic. It might be worth investigating replacing the bottom end of the mrdanga with a triggered sound, and using the capture of the top end, for future recordings of badly tuned Balarama bottom ends. Great mrdanga performance on this track.

I also left the same "Minimum Voltage, Maximum Illusion" set up with the vocal, routing the uncompressed vocal track through the reverb [see here]. In a proper mix down I'll have to make some adjustments, as you can hear it go over the top in the line beginning "Dekho ore bhai".

I'm not sure how this sounds along with the other mixes. Together they probably aren't going to come out sounding like an album, as I'm experimenting and learning things as I go. I think I'll just mix each one in isolation, then listen to them all together at the end.

Anyway, here it is:

• Parama Karuna (mp3, 16MB, 11 min)

Sorry about the couple of minutes silence at the end..

OK, this one is a bit experimental.

On the second song of the night (and from memory the third) Arjuna didn't play tabla - Janardana played the white Balarama mrdanga in the picture.

The sound of an untuned Balarama mrdanga is truly that of a plastic drum. If you tune it up you can get something half-way decent out of it. However, this was untuned, and to boot, unmiked to begin with. I noticed after the first few moments of the kirtan, which were spent trimming input gain and setting up the live sound, then turned around and adjusted the mics. As a result initially the sound level of the mrdanga is low.

However, this kind of works, as it allows the dynamics to build. Speaking of dynamics, I got a free Faderport - I'm just waiting for it to arrive. Now if I can use that, then I can (relatively) easily do the same thing with a recorded mix that I did on the night, riding the fader and recording it with automation. At least until I get the Zed R16, at which time the whole thing becomes a no-brainer. In the meantime you get a mix where the flute wanders all over the top of the whole kirtan - on the night I raised the volume during the response and lowered it while Maharaja sang.

I boosted the 760Hz on the top end of the mrdanga, and tried something I read about in Owsinski's book "Recording Engineers Handbook" [read online at Google Books].

In the book Owsinski explains that R&B producers put a 32Hz tone through a channel and trigger it from the kick drum. This gives you a real deep sub bass feel. I tried that with the wimpy plastic mrdanga to give it more oomph.

I put an test oscillator on a channel and set it to 66.910Hz. Then I put a noisegate on the channel and set the sidechain to the channel with the mrdanga. The noisegate settings are: Threshold (-17db) | Attack (3ms) | Hold (40ms) | Release (1824.6ms).

What this does is gate the oscillator until the bottom end of the mrdanga is hit hard, at which point the gate opens and the test oscillator's tone comes through. The effect, at least in the small amount of listening I've been able to do, is like a deep djembe playing in the kirtan.

In this mix I also played around with stereo positioning of elements, to make them closer to the actual layout of the room as captured in the stereo room mic. On the night I tried to position people to avoid leakage between mics. Next time I'll also take stereo balance into account.

OK, here's the track. Let me know how it sounds / what you think.

• Nitai Pada Kamala (mp3, 11:27, 16MB)

Oh yeah, at the end, in the last few minutes, Janardana and Rohan go to double time in 6/8 time. This song is in 3/4 (waltz) time. The feel changes going to 6/8, but the timing works. Maharaja was not used to that change but he quickly fell into the groove. The chimpta really helps in a situation like to keep the beat anchored.

Here's a rough mix of H.H. Prabhavishnu Maharaja's first song from Sunday, Ei Baro Koruna Koro.

With this I tried a technique from Stav's Book "Mixing with Your Mind" designed to give "Maximum Illusion with Minimum Voltage", a fundamental principle in audio engineering.

You want to get a big sound, but at the same time you don't want to inflate your output to Dragonforce levels. So you strive to get Maximum Illusion (the biggest sound) with Minimum Voltage (the lowest output).

Prabhavishnu Maharaja's voice lends itself to this particular technique. He has a very powerful voice, and a wide dynamic range. When he cuts loose, it gets pretty powerful, and I had to several times adjust the trim (input gain) control on his channel.

To bring that under control in the mix I used a compressor. I set the compressor threshold to -16dB. This means that when the Swami's voice hits -16db the compressor will kick in. The Attack is set to 6.0ms, which is very fast. This means that 6 milliseconds after the compressor detects that the Swami's voice has hit -16dB it starts compressing the signal. The compression ratio is set to 7.1:1. This means that 6 milliseconds after the Swami's voice hits -16dB for every 7dB of increase in volume in his voice, the sound in the mix goes up 1dB. That's compression. The effect is to put a ceiling on his voice, so that it doesn't go from too quiet to hear to blowing up your speakers.

Now, to preserve the power of his vocal performance I did this:

I put his uncompressed vocal on Track 1. I then routed the output of Track 1 to the input of Track 2. I turned up the reverb send on Track 1. This means that the only thing you hear directly from Track 1 in the final mix is the reverb send. Then I put the compressor on Track 2, and routed its output to the final mix.

The result of doing this is that when the Swami cuts loose his vocal track is compressed - it's squashed down to keep it at an even level. However, the reverb that you hear increases in relation to the uncompressed vocal. The effect of this is to make it sound big. You can really hear this in the final, classic Hare Krishna portion of the track. His voice doesn't get much louder, but it sounds bigger, because the reverb does get louder, and it is bigger.

Maximum Illusion, Minimum Voltage.

I still haven't listened to this a wide variety of systems, so I'm not sure about the relative levels. I might cut some bass from the flute and raise its volume, and boost the sax when it kicks in later in the piece.

Here's the rough mix:

• E Baro Koruna Koro - Rough Mix 1

I said that I focused completely on the audio engineering last night, but that wasn't completely true. I also did a little chimpta playing.

The chimpta is the "musical fire tong" - it's a fire tong that has been adapted to a musical instrument by the addition of small cymbals. In kirtan it plays a single, half, or quarter note beat in the same time as hand claps. It is useful to get people clapping, and it really drives the beat and gets the kirtan to the next level rhythmically.

In Janardana's kirtan you can hear me playing it starting at 6:46. Listen to the original mix (Mix 1), rather than the reduced metal mix, because you can hear more clearly what I am going to describe.

At 6:46 you hear me open the chimptas and start. I realized that the sound was overbearing, so I changed technique to reduce the level. You can hear this at 6:55. The initial technique I used was to grasp the chimptas in the middle. This creates a sound with a pronounced attack and a loud cymbal ring. I switched to holding it at the top. This requires more energy to play, because your hand is further away from the fulcrum, but produces a sound with less attack and more sizzle.

In the reduced metal mix the chimpta practically goes away because of this approach.

I've remixed Janardana's track. I changed the EQ on the room mic to bring the cartals right down.

This is mix 3. If you compare it to mix 1 [available here] you'll find that it doesn't have the same "live" feel to it. However, if you bring the volume down on each mix until you can only hear one element you'll find that with mix 1 that element is the cartals, and with mix 3 it's the main vocal.

Cartals are easily the loudest thing in any kirtan. In this particular recording there is only one pair of cartals, but they still dominate. Cartals produce high frequency sound, which requires less energy to propagate. With low frequency sound you are moving more air, so you need more energy to get equivalent volume. With high frequencies you're moving less air, so a small thing will produce a big sound.

On Saturday night a guest found our place by the sound of the cartals from the street. Other sounds die off quickly as their energy is expended - cartals carry. Listen as you approach a street harinam. All you can hear from most distances is cartals. It's only when you are really close that you can hear the singing.

Further to the points about dynamics yesterday, check out this video by Jared Falk over at freedrumlessons.com - Dynamic Drumming: Drumming with Groove. In it he explains that beginners will drum without dynamics for at least a year or two before developing feel. In the meantime they are unable to modulate the dynamics of their playing. Players of other instruments, such as cartals, are in a similar situation.

Unfortunately, many cartal players in ISKCON never develop the ability to play dynamically. They play at one constant volume - "too loud". This is because there are so many players who never develop the ability to play dynamically, and everyone is copying everyone else. Too loud cartals do not sound good, they sound familiar. Once you get used to them they evoke memories of previous kirtans and experiences that you've had. When I started listening to the recent 24 hour kirtan webcast from Birmingham at first the excessive cartals detracted from my listening experience. Then after some time I found myself enjoying the immersion in a familiar sonic environment. It wasn't the sound I was enjoying, but the familiarity.

Listen to descriptions by people who are unfamiliar with cartals and kirtan. This will give you more of a perspective devoid of the enculturation and habituation that takes place as loud cartals become an integral part of your life experience. People often describe it as "ringing bells". That's what it sounds like. People ringing bells.

Players who can play dynamically are rare, and are regarded as "awesome". As Owsinski mentions in his book "How to Make Your Band Sound Great", the audience can't put their finger on why, but they know it rocks the kirtan and it makes the kirtan group sound tight.

To reduce the cartals in this recording I've had to reduce the frequency range that contains the chorus. The cartals washed out a lot of this anyway, so there wasn't much to begin with.

Aindra's "Vraja Vilasa" albums were hailed at the time of their release as "his best recordings ever" by at least one 30-year veteran of kirtan that I know. One of the things that Aindra did with that album was reduce the 'metal on metal" of the cartals. He did this by compressing the heck out of the ambient mics, which also reduced the mrdangas. The effect for me is an over-compressed sound that lacks dynamics and is fatiguing to listen to for long periods. However, you do experience a (mostly) live kirtan liberated from excessive metal on metal.

So I've done something with this track to reduce the cartals. It's a compromise, because it destroys other elements of the live kirtan. However, it's all in service of the vision: "Better than being there". Ideally, however, we'd have good control of the dynamics in the kirtan by people who are aware of their playing. One way to get this is to record yourself and critically analyze your performance afterwards. I've found that to be invaluable in developing my own musicianship, whether it's timing, tone, pitch, or dynamics.

Here's the remixed track:

• Janardana's track Mix 3 (mp3, 13:58, 18MB)

OK, you've waited long enough.

Here's a rough mix of the first kirtan, which was a sound check / crowd warm up number. Janardana is leading this one.

There was some feedback through the harmonium mic early on in the piece, which was picked up also in the room mic and the SM57 on the top end of the tabla. I used a channel EQ on each of these tracks to remove it. It's from around 92Hz down, peaking between 86 and 92Hz. I used the analyzer to find it, then heavily EQ'd this part of the spectrum. I then used automation to turn the EQ on as the feedback hits, then off as it passes. I stopped it pretty quickly when it actually happened. I must have been raising the harmonium, either through its input gain, or output to PA (using the input pan); noticed it happen; and backed off immediately.

I got the idea for this particular "fix-it-in-the-mix" technique from a Youtube video by Bob Katz, author of Mastering Audio: The Art and Science. The video is "Bass Frequency Surgery".

Here's a screenshot of the EQ settings:

And of the automation (click for larger view):

See my previous post for a list of the mics used. There are two mrdangas as well as the tabla in this - they are coming through the room mic only.

And here is the track:

• Janardana Kewin - Tribute to Birmingham 24hr Kirtan (mp3, 18MB, 13:38)

OK, I'm off to hang upside down in my closet for a few minutes before I go to work...

This evening I took a break from watching the 24 hour kirtan in Birmingham to record a couple of hours of kirtan at the temple. His Holiness Prabhavishnu Swami is a great kirtan singer.

I tried my idea of using the post-record pan control as a volume control for a live mix (see Using the BR-1600CD to do a live mix while recording). It worked.

I had my attention divided between two things:

1. Monitoring the gain control for each channel to make sure that the signal coming in was hot, but not clipping.

2. Monitoring the input pan controls to give an expressive live mix, without getting feedback.

Making a live kirtan mix rock

Yesterday Dominic lent me two audio engineering books by Bobby Owsinski. I looked him up and found that he's written quite a few good ones. There is an excerpt on his website from his book "Making Your Band Sound Great". The excerpt is from Chapter 8: The Keys to Greatness.

Dynamics

If you only learn one thing from this book it's that playing with dynamics is the greatest key to making your band sound great. It's an improvement that both you (the band) and the audience will notice immediately, and will automatically separate you from about 90% of other bands on the planet. ... Most bands are oblivious to dynamics and play at one volume throughout the entire song, song after song, set after set.

Owsinski makes an interesting observation on how mastering dynamics is perceived by an audience:
"suddenly they'll start telling you how tight you sound"

I learned an interesting principle from Stav's book "Mixing With Your Mind". When you introduce an element in the mix you can bring it up in volume to make the listener aware of it, then fade it back. It doesn't need to stay up once it's made itself known.

Based on these two things, and what I saw Dominic do at Dave Stringer's gig here, I manipulated the live mix to give it more dynamics.

At the beginning of the kirtan I had the main vocal on 50%, and the harmonium between 25% and 40%.

As the kirtan progressed I would raise the vocal, ending on 100%. I also brought up the tablas and backing vocal.

Whenever Maharaja would sing I would drop the bottom end of the tabla, and Sridhar's mic (flute or sax). Then on the response I would bring the tabla and sax back up.

Over the course of the evening I also pushed the overall volume up. I had to contend with some feedback at different points, but was able to isolate and stop it quickly. That's why I think supercardioid mics are the way to go for live kirtan, which is usually performed in small spaces with lots of potential for leakage and feedback.

One thing I learned from the Christmas Kirtans was to use a new song on the BR-1600CD for each song. Initializing them takes time. I did it while the translation was being read afterwards. I figured out how to save the compressor settings as a user patch, which saved me time, and allowed me to refine them over successive songs. I put an 8:1 compressor on the vocal with a 30ms attack and 120ms release, which I then changed to 250ms on a later track. I also put some compression on the tabla bottom end. The other tracks were set to limiting at -6db.

I'll refine my compressor technique and create some custom setups. The BR-1600CD ships with some standard layouts, like "rock band" or "horn ensemble" with compressors and EQs configured on each track for a different instrument. I'll create my own user set up with the tracks configured for standard kirtan instruments, like "input 3 - mrdanga top end, input 4 - mrdanga bottom end". I can probably copy the compressor and EQ settings from the existing input presets, or from Logic Pro's channel strips.

I have a spiral bound notebook where I keep notes about each recording I do, with the date, the mics used, the channels, the channel settings and anything else that comes up. This way I can systematically improve my game.

I'll have some of the tunes to post in the next few days.

(Photos by Krishnapada, who organised and hosted the evening, via facebook.)