Comparators and LED Meters...THAT's annnoying!!!

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Swedish Chef

Well-known member
Joined
Jun 7, 2004
Messages
351
Location
London
I've been playing around with this http://www.thatcorp.com/datashts/dn112.pdf
but try as I might I can't get the LEDs to work, dagnabit! :mad:
I'm using discreet LEDs instead of an array but that shouldn't matter should it?
I've checked that the appropriate voltages appear at the divider junctions, and that the comparators go high and low when they should. the LEDs are all in the correct way, all solder joints are solid etc. etc.
The only thing which strikes me is that when U1A is low i.e. Ground potential??!!??!!! it is straddling Ground and Neg rail with no current limit resistor in sight! this may be why that one keeps blowing!!!!
I can't work out what the current source is supposed to be spitting out, so could some kind Sir jump in and let me know the maths on that? :oops:

Cheers in advance!

chef
 
Yep it's on the V- rail...
Perhaps I'll scap it and try something else!
Perhaps an email to THAT Corp first though... :green:

chef

p.s. just managed to measure the current source at 0.6mA
Does that sound about correct?
 
Mr Chef...

Assuming +/- 15V rails, R4+R26 is nearly 30k, meaning 1mA down them.
1mA through R26 (1k6) will drop 1.6V, so there will be about 1V across R21, which at 100R, will establish Q1 as a 10mA current source (a reasonable choice).

So your measured 0.6mA is a bit off. Try checking the voltages I mentioned across the various resistors there.

It might come down to Q1, as if that is gone, then you won't see any leds (although curent sources are usually fairly robust, being to a degree self-protecting).

Hope that helps.
Alan
 
0.6mA seems low if that's supposed to drive your LED's. They might glow very, very faintly. It looks like it should be about 8 mA or so. What's your V+ and V- rails - it needs at least +8 and -8 volts to work - assume 2 volts (roughly) per LED, but if you aren't on +/- 15 volt rails the resistors in the current source will need to be tweaked. Can you get the LED's to glow by removing U5 (and maybe U1?) If nothing glows, that's because either U5C is holding the output low, preventing any of the LED's from glowing, or the current source isn't working.
 
awesome! the 1st 6 are working fine now! Much obliged gentlemen :green: :guinness: :guinness: :sam: :guinness: :guinness:
the top LED is still blowing though! Also without that LED in, when it gets to the point where that LED should come on they all go out. I have a sneaker that this is to be expected but am still scratching my head!

chef

p.s. it was dodgy +V to the Current Source...
 
Ok THAT Corp need a smack! That TOP LED should be connected to ground, NOT -V. I thought that could not be right... :green: :thumb:

chef
 
Ok THAT Corp need a smack! That TOP LED should be connected to ground, NOT -V.

Hmmm... I don't see that myself - in fact, it shouldn't make too much difference, but shouldn't be responsible for killing the LED...

Alan
 
Chef, I made a quad version of this and got it to work with a potential divider that gave me minus a few volts on the last LED, the neg rail was making the last few light up constantly otherwise.
Built it with the 'simple effective soft-knee compressor". Very good, very cheap.
 
> when U1A is low i.e. Ground potential??!!??!!!

Power connections to LM339 are not shown. And while the 339 can be used single-supply, in this circuit it MUST be wired to V+ and V-, so the "low" condition should be about V-, not ground.

> LED should be connected to ground, NOT -V.

It has to be connected to the same place as the negative supply of the LM339. Ground is possible, but then it injects (small) spikes into Ground when the LEDs switch. Returning both 339 and LEDs to V- works the same (slightly higher heat in the current source), but craps-up both supply rails instead of your sacred/holy ground bus.

> I can't work out what the current source is supposed to be spitting out

Neither can I. Supply voltage is not stated. The R4 R26 divider is steep, leaving small voltage from Q1 Base to V+. Q1's Vbe is not known precisely (varies with maker, lot, and temperature). ASSuming +/-15V, then there is about 15V * (1.6K/(1.6K+27K)) = 0.84V across Q1 BE and R21. ASSuming Q1 Vbe will be around 0.7V, then we have 0.84V-0.7V= 0.14V across R21= 100 ohms, and 14 milliAmps. However at +/-12V supply we get nearly no voltage across R21, and no LED current. Small changes in temperature cause not-small changes in current. Change in Q1 die area (substitute an "equivalent" part) can make a significant difference in current.

This is, in my opinion, a bad design. It sure will work as intended, but if anything changes it will work different. And there is no need for such low voltage on R21. For supplies like +/-12V or above, Q1 Collector will never rise much above Ground (assuming we return the LM339 and the LEDs to V-, which is probably best). We can use most of the positive supply as bias.

Re-design: Try R26= 2K2, R4= 4K7. That puts about 1/3rd of rail voltage across Q1 BE and R21, or 3V to 5V for rails from 9V to 15V. With that much voltage, we don't care if Q1 Vbe is 0.6V or 0.8V, that's small change. R21 will get 2.1V to 4.2V worst-case, rails from 9V to 15V and Vbe from 0.6V to 0.8V. Assume we like about 12mA or 15mA current in the LEDs with 12V or 15V rails, a bit dimmer on 9V rails, R21= 270 ohms will work good.

For better light regulation with supply variation: replace R26 with a Red LED, R4 about 4K7, and R21 about 68 ohms. But this is probably not worth the effort unless you happen to need a pilot LED anyway.

Actually, there is darn little reason you HAVE to use the current source. For bench-testing with a solid +/-15V supply, replace it with a 2K2 1/2-Watt resistor. LED brightness will vary with supply rails and with number of LEDs lit: when all are lit each one will be nearly half as bright as with one lit. And you get more trash thrown into the supply rails, which is why for a very good finished product you would want to use the current source. But while still struggling with misunderstood supply connections and possibly blown parts, a simple resistor will work quite well to get you on the tracks again.
 
Mr PRR Sir,

ASSuming +/-15V, then there is about 15V * (1.6K/(1.6K+27K)) = 0.84V

With +/- 15V, there is 30V across those two Rs, giving double your V figure across the 1k6 - which makes the voltage across the 100R a bit more predictable - not perfectly so, but close enough for Jazz (and analog work...)

Rgrds
Alan
 
I thought long and hard before posting this (but hey - what do I know?)

When I were a lad, I was taught that the aim was to keep the signal level such that the distortion from clipping was completely avoided. That was the reason for metering - the purpose of metering (and lots of discussion was had on why that metering was not really up to the job to the degree we would like).

Sure, there was 'soft' tape distortion/compression, but that was a fact of life (that you could, or in fact, would have to, work with). Mainly, we were trying to mitigate these effects.

Certainly, we never selected equipment on the basis of how it sounded when we abused it! (That must be a 'modern' invention)....

Have I been in a time-warp since the '70s (I sure have...)

Alan
 
> there is 30V across those two Rs

You are right. I read wrong. (WHY does THAT Corp have to use that grid-paper background???)

Seems inefficient to me to pull power out of the far rail when ground will do, and there is no crap involved to contaminate ground. But hey, only 15mW.

Ha, and it makes the arithmetic dead-easy. 27K plus a little extra, divided by (+/-15V)= 30V, is 1 milliAmp, near-nuff. So the 1.6K resistor drops 1.6 Volts. If you assume the transistor BE drops 0.6V, you have 1V across R21; if Vbe is higher then R21 gets slightly less voltage. 1V across 100 ohms gives 10mA, or maybe a very-little less if Vbe is a little more.

> we never selected equipment on the basis of how it sounded when we abused it!

I don't see where that point arises in this thread. The only abuse has been some LEDs.

I was a boy when low-low distortion at nominal levels became fairly common. But this clearly overlaps the period when electric guitarists were using overdrive as a tool on the artistic palette. As a small boy, few top-10 recordings featured heavy distortion on guitars, but there were a lot of non-pop players doing it. Just off the top of my head, "Telstar" was reverb-crazy but also overdriven. Later came the Stones' "Satisfaction", and the deriviative opening riff in "Inagadavida". Many early rock recordings were done low-budget and clearly pushing their boards far past the 0.5% IMD point, and that is part of The Sound. A few were beat to pulp: "Louie Louie" sounds like a bootleg from a wild party; Spector used layers of overdubbing because he liked what generation loss did to the base tracks. And never mind what was on records: in real performance, most bands had to push amps way into overdrive to cut through loud clubs, and vocal systems were usually putting out square-waves.

On the Fender guitar, the use of distortion is clear. Devoid of body resonances, with low inharmonicity, with frets to make string vibrato hard, the naked sound from a guitar is very plain and can be expressionless. Soft tube amps add a nice haze of low-order distortion, and a rise of distortion with level that is not unlike the change in un-assisted vocal timbre from normal to loud. (While some vocalists have large range of power, many have a limited range of fundamental power but increase the overtone power when speaking or singing "loud". Same is true of most woodwinds and brass.)

Whether from artistic intent or necessity, both musicians and listeners came to accept amplifier distortion as a musical tool. As precedent, think what Steinway would have thought of ragtime, honky-tonk, or the antics of Jerry Lee Lewis; or what Saxe would have thought of the way sax players "abused" his instrument in Swing and especially Jazz. W.E., Macintosh, Scott, Langevin and others worked to make amps clean and transparent at extremes; on the other hand musicians adopted the faults of amplifers as tools.

What has really changed is that the recording technician does less verbatim capturing of the band sound (that part is now easier), and is now expected to shape and flavor the work. This comes from people like Buddy Holley's recordist (see thread last week), the Beatles and George Martin in their Sgt Pepper days, the early Led Zeppelin. And adding just the right amount of distortion is now such an expected "spice" in many musical styles that a "perfect verbatim" recording won't do. Yes, the artists may have selected distortion in their guitar amps, but are at the mercy of a sound technician for vocal sounds.

And a problem I was just reminded of, on an all-percussion concert: smacked drums do not record well. Tape saturation is very helpful in getting a high average level and loud-sounding peaks without over-cutting the LP. Even then, a really fat drum sound needs help: that piece-of-crap Shure PA limiter has a following because Led Zeppelin used it to mash their drums onto tape and then LP. Drums can be manipulated (distorted) a lot without sounding distorted, because of their transient nature and the wide variety of sounds that can plausably come off a modern drum-set. I got my percussion concert mastered at an acceptably high average level (without tape) by pushing peaks to 0dB or +2dB FS. (At +2dB FS, on percussion, way less than 1% of samples are clipped.) But to make some of those passages hold-up against more steady state notes, or to get the obscenely high average level modern buyers and DJs think is "good", I woulda hadda put a clipper or limiter on the tracks.
 
sorry i am confusing things by not stating things like what rails I'm running! Sorry :oops: !
At the moment I am indeed on +-15V...
I am going to read all of the suggestions, try to digest and then apply. I was running the comparator on +V and Gnd as that's what was mentioned in the App note. Stupid of me for not thinking really! +-15V on that and away we go again... :shock:

chef
 
> we never selected equipment on the basis of how it sounded when we abused it!

I don't see where that point arises in this thread. The only abuse has been some LEDs

Ha-ha. Now that was *my* mistake. I read another thread, wandered about for a while, then came back here to the wrong thread to reply....

... that was the thread at http://www.groupdiy.com/index.php?topic=1440 with the long piece about how different preamps sounded when driven heavily into clipping.

...Ah well...

Alan
 
[quote author="SonsOfThunder"][quote author="PRR"]pushing peaks to 0dB or +2dB FS.[/quote]howwjewdoodat?[/quote]
I read it as meaning that the absolute signal peaks signal reaching the convertor are up to 2dB higher than the maximum undistorted signal level. Since the impulse peak is a tiny fraction of the signal as a whole, the distortion is easily ignored.

..Of couorse, I could be reading it wrong...

Also, -and importantly- meters which read the absolute bit state ignore that the possible sampling & reconstruction filter levels can actually exceed their DC equivalents...

Just more complexity! :green:

Keith
 
> howwjewdoodat?

My live-concert tracks often peak around -10 to -20dB below digital zero.

I decide what average and peak level I want. For "full" sounds like choir, average in the -25 to -15 range and peaks of -5dB to -1dB work OK, though are not as "hot" as a commercially mastered and polished recording.

On occasion, I'll take a track that has max peak at, say -10dB below digital zero, and apply 12dB of digital gain. The peaks come out 2dB above digital zero. They overflow the 16-bit number. The sound editor is smart enough to know to "clip", not wrap-around (what the CPU wants to do when a number overflows its space). Zoomed-in on a waveform editor, the peaks are flat. But to the ear, a millisecond or less of clipping is not a big deal on most sounds, often just inaudible, especially if it only happens once or twice a track. I have semi-commercial CDs with far more clipping, and on brass crescendos you hardly notice (though I think it would be nicer a few dB lower). And pushing the max peak 2dB above clipping also raises the average 2dB. This is important when a track is loud, must play loud, but the live performers got over-loud on one note.
 

Latest posts

Back
Top