Modulus-86 build thread

For the Mod-686, 500 VA is the recommended size for a toroidal power transformer? If smaller, where would you draw the line?

For continuous rail-to-rail sine wave reproduction into an 8 Ω load, you'll need 565 VA per channel says math. At 4 Ω load, this just about doubles to 1109 VA per channel.

Now, as you peel yourself off the ceiling, let me provide some comfort here. For music reproduction, you're looking at:
10 dB crest factor: 265 VA (8 Ω); 510 VA (4 Ω)
14 dB crest factor: 175 VA (8 Ω); 330 VA (4 Ω)

I think 500 VA is a good recommendation. I'd probably draw the line at 400 VA.

Also note that you'll need as much heat sinking as you can afford. When running just below clipping, the amp will dissipate around 200 W into the heat sink with a 4 Ω load, even with a music signal! That's Class AB for you.

Is the gain 26db? Also please tell us how it sounds; similar to your previous composite amps, different, even better?

The lowest supported gain is 26 dB. The gain can easily be increased without sacrificing the amp's performance. Do note that the best overall system performance (best gain structure) is still to have the lowest gain possible in the power amp. The laws of physics haven't changed... :)

I'm not far enough with the prototyping to consider connecting this amp to a speaker. Considering that all parameters measure far below audible, just as my other Modulus amps, I expect the MOD686 to sound like the other Modulus amps ... like nothing. Everybody claims "straight wire with gain" or "transparent". I back such claims up with data. :)

How high is the damping factor?

How long is your speaker cable? The output impedance of the amp is about 20 mΩ. With an 8 Ω load, that's a damping factor of 8/0.02 = 400. The DF will be dominated by the connector resistance and resistance of your speaker cable.

Tom
 
Member
Joined 2014
Paid Member
Sad, strange man that I am I have been on the hunt for music with a very high crest factor. This is not helped by the fact that calculating crest factor on real music in a meaningful manner is tricky and you end up with averages of averages, but the foobar DR plugin gets to the right order of magnitude despite not actually calculating DR! But it does give you RMS level and peak (sadly ignoring crescendos in classical music) so is a pointer for looking further. There are not many recordings with very high crest factor and even fewer you would want to listen to, but I've found a few and am sampling them. I have a Rickie Lee Jones track where the RMS level is -24dB with peaks of 0dB. It's a good test for if you have enough amplifier headroom :).

My heatsink torture test is Rodrigo y Gabriela which is smushed to death...

All of which is making me think that THX reference levels (-20dBFS is 85dBa or 75dBa* per channel at listening position) are a good starting point then adjust gain according to the RMS level of the track.

*The 75dBa spec is probably more applicable for those of us without basement home cinemas and who value our hearing :).
 
Sad, strange man that I am I have been on the hunt for music with a very high crest factor. This is not helped by the fact that calculating crest factor on real music in a meaningful manner is tricky and you end up with averages of averages, but the foobar DR plugin gets to the right order of magnitude despite not actually calculating DR! But it does give you RMS level and peak (sadly ignoring crescendos in classical music) so is a pointer for looking further. There are not many recordings with very high crest factor and even fewer you would want to listen to, but I've found a few and am sampling them. I have a Rickie Lee Jones track where the RMS level is -24dB with peaks of 0dB. It's a good test for if you have enough amplifier headroom :).

foobar2000 1.1.16 / Dynamic Range Meter 1.1.1
log date: 2015-07-20 18:13:13

--------------------------------------------------------------------------------
Analyzed: Rickie Lee Jones / Flying Cowboys
--------------------------------------------------------------------------------

DR Peak RMS Duration Track
--------------------------------------------------------------------------------
DR19 -3.16 dB -26.43 dB 4:50 01-The Horses
DR13 -8.97 dB -25.86 dB 4:49 02-Just My Baby
DR24 -0.49 dB -26.59 dB 6:16 03-Ghetto of My Mind
DR16 -6.39 dB -25.94 dB 4:54 04-Rodeo Girl
DR17 -4.38 dB -25.14 dB 4:58 05-Satellites
DR18 -7.16 dB -32.44 dB 4:18 06-Ghost Train
DR16 -6.03 dB -24.17 dB 5:05 07-Flying Cowboys
DR14 -7.15 dB -26.27 dB 4:20 08-Don't Let the Sun Catch You Crying
DR19 -4.11 dB -25.36 dB 4:55 09-Love Is Gonna Bring Us Back Alive
DR16 -6.24 dB -29.30 dB 5:34 10-Away from the Sky
DR13 -4.31 dB -24.21 dB 5:57 11-Atlas' Marker
--------------------------------------------------------------------------------

Number of tracks: 11
Official DR value: DR17

Samplerate: 44100 Hz
Channels: 2
Bits per sample: 16
Bitrate: 715 kbps
Codec: FLAC
 
Member
Joined 2014
Paid Member
In more detail (left and right seperate)
foobar2000 1.3.16 / Dynamic Range Meter 1.1.1
log date: 2017-12-26 17:53:10

--------------------------------------------------------------------------------
Statistics for: 3/11-(03) Ghetto Of My Mind
Number of samples: 16573956
Duration: 6:16
--------------------------------------------------------------------------------

Left Right

Peak Value: -0.62 dB --- -0.40 dB
Avg RMS: -26.47 dB --- -26.75 dB
DR channel: 23.44 dB --- 23.69 dB
--------------------------------------------------------------------------------

Official DR Value: DR24

Samplerate: 44100 Hz
Channels: 2
Bits per sample: 16
Bitrate: 650 kbps
Codec: FLAC
================================================================================

It's an outlier and I don't know if this is actually more realistic (as in less compressed) or just too close miked on the drum kit. But beats Dafos as a headroom test :)
 
in my experience, transcoding LPs to 16/44.1, I found normalizing to -13dB RMS (actually should be "average") resulted in zero clipped samples on most popular Rock / Folk / Blues albums, mostly from the 1970's. I aimed for a maximum level of -1dB Full Scale, so the actual crest factor was 12dB. Rarely would I need to normalize at -14dB RMS, and many could have been normalized to -12dB but I prefer similar average levels when possible.

Using that method, there would of course be portions of songs that had a higher crest factor, but they were down from the peak average levels in the program material anyway, so I didn't measure them.

Actual recording was done at 24/96, using some of the extra headroom to generate clean recordings (zero clipped samples, maximum level -1dB FS), de-click performed, normalized, then downsampled to 16/44.1. Some with dither, some without.

This was done before I really knew what I was doing with digital transcribing, 15~20 years ago, but based on my experience with analog recording and some educated guessing. It seemed to work out OK
 
Last edited:
DF should be output impedance vs:
the sum of:
impedance of speaker wiring (both ways, eg 20 ft for a 10 ft speaker cable)
impedance of any crossover network
impedance of the voice coils of the driver(s)
impedance of any connective elements (bananas, binding posts, internal speaker wiring, etc)

We are most interested in the DF at low frequencies of the largest driver (ie with the greatest back-EMF between the minimum excursion frequency and any crossover frequency), however the "quick and dirty" method is to measure the DC resistance of the speaker wire at the amplifier output point (unconnected at amp end, connected at speaker end).

In the modern Solid State world (and in most cases, in the "Classic" Solid and Vacuum State age) DF was adequate, so it's not something I worry much about. Some suggest any complex DF above 1:20 is adequate.
 
Last edited:
Member
Joined 2014
Paid Member
yeah a lot of 'popular' music does hang around -12dB. Or at least used to. I have a track with -6dB RMS and Tom has worse on his website. It does trip the over sensitive protection on the old amplifier, but can't be bothered to fix that as have something better still waiting installation :)
 
DF should be output impedance vs:
the sum of:
impedance of speaker wiring (both ways, eg 20 ft for a 10 ft speaker cable)
impedance of any crossover network
impedance of the voice coils of the driver(s)
impedance of any connective elements (bananas, binding posts, internal speaker wiring, etc)

We are most interested in the DF at low frequencies of the largest driver (ie with the greatest back-EMF between the minimum excursion frequency and any crossover frequency), however the "quick and dirty" method is to measure the DC resistance of the speaker wire at the amplifier output point (unconnected at amp end, connected at speaker end).

In the modern Solid State world (and in most cases, in the "Classic" Solid and Vacuum State age) DF was adequate, so it's not something I worry much about. Some suggest any complex DF above 1:20 is adequate.

Good discussion. I’ve been reading Toole’s 3rd Edition of “Sound Reproductions” <highly recommended> and he discusses this in detail on Chapter 16. He doesn’t mention a DF ratio though so I presumed 1:10 was adequate. Do you have a source for your 1:20 recommendation?

Best,
Anand.