Expensive diode limiter?

[AMZ-FX]

I'm thinking the clipper is like this:

You only want the extreme peaks to be clipped and as little as possible. Most likely the clipped signal will be some low frequency impulse and it should not be very noticeable.  If you clip a high vocal note, it certainly will be heard! The clipper is an overload protector so the A/D will not splatter - it's not an effect.

Germaniums are not very useful in this type circuit...  they do not have a real  knee at which clipping begins but instead begin to turn on gradually over a fairly wide range.  They are so inconsistent you would have to buy a large number and match them for use.

The Headwize reference to my work with clippers is not really accurate concerning the sound that I found when analyzing different diodes.

regards, Jack

 

[NewYorkDave]

Yes, that's the circuit I was describing earlier in the thread. I agree that this is what's probably inside those $200 XLR barrels :wink:

 

[bcarso]

Again, if you are gong to duplicate this I would strongly recommend at least using a single zener in the middle of a diode bridge, so you get nearly perfect symmetry without needing to do any zener matching.

 

[CJ]

And they do make Schtky bridges by the way. I know, becuase I had to hunt them down, which was a major PITA.

Or, much less hassle to make a bridge from some 58xx diodes.

 

[AMZ-FX]

If you put the zener in the middle of the bridge then you have to match the four diodes if you want perfect symmetry. Just get a pair of 1% zeners and that will be close enough.

Besides, perfect symmetry is not necessary... audio waveforms are not symmetrical most of the time so a 1% mismatch will not be audible.

regards, Jack

 

[bcarso]

A pair of 1% zeners will most likely be a worse match (and harder to come by) than the mismatch from four out-of-the-same-batch 4148's.

For example: 7.5V 1% zeners might have a 75mV mismatch. If we are particularly unlucky they could be in opposite directions for a 150mV mismatch. There is also no spec on Vf, although that's likely to be a pretty tight number.

It's unlikely that 1N4148's in a reel are going to have 75mV mismatches in Vf. I'll measure some and get back with the statistics.

 

[bcarso]

So I tested 16 1N4148's out of a drawer. I don't remember the manufacturer or when acquired, but they are presumably from the same purchase.

The forward voltages are clearly distributed bimodally, but if treated as a whole set the mean value of Vf is 614.9mV @ 1mA, with a standard deviation of 13.6mV.

If I sort out the two different populations six are mean value 598.1mV, s.d. 1.64mV, and ten are mean value 625.0mV and s.d. 2.48mV.

 

[clintrubber]

How would matching of those four diodes be if a bridgecell was used ?
Come to think of it, I have no idea how these are actually made. All on one chip?

 

[bcarso]

[quote author="clintrubber"]How would matching of those four diodes be if a bridgecell was used ?
Come to think of it, I have no idea how these are actually made. All on one chip?[/quote]

I think the product is just a couple of discrete zeners.

I don't know if integrated bridges are done with monolithic devices or simply four chips in a package. Of course monolithic would be better for out-of-the-box matching, although it would have the substrate C floating there to contend with.

I think RCA made an ancient diode array chip once but I haven't seen it used for many years.

There are some high frequency diode ring arrays I believe for double-balanced mixers, but that is not the right topology for this app.

Anyway, as pointed out by Jack (AMZ-FX) a small asymmetry is not probably a big deal, although I hate to have a built-in second-harmonic + d.c. process imposed that I don't know about and can't eliminate.

Again, this is an emergency stopgap clipper, not a fine compressor, just to keep you from running into digital clipping, so one wouldn't want to fuss too much.

 

[clintrubber]

Hey, I got a notification but there wasn't any news ....  new message deleted again ?

Whatever, let's reply anyway 

I think RCA made an ancient diode array chip once but I haven't seen it used for many years.

Might be the CA3019 ? (just guessing) I have a few, RCA gave it a nice box as well. A balanced diode limiter (like the 33609-topology) might be a very application.

Bye,

  Peter

 

[gyraf]

Why not make them variable-zeners, like the ones you see for setting the bias current in class-AB amplifiers?

Jakob E.

 

[clintrubber]

Why not make them variable-zeners, like the ones you see for setting the bias current in class-AB amplifiers?

Jakob E.

That'd be a nice one, a.k.a. Vbe-multiplier; a more direct control of clipping-level hard to imagine. 

Potential problem of this 'direct involvement' might be the wires to/from front-panel: the audio-path
gets some additional parasitics from the control-pot, but this could of course be minimized or circumvented alltogether by making it less direct by means of an intermediate step. Mr. Moog would probably have made it voltage controlled.

Regards,

  Peter

 

[craptical]

Hey, I got a notification but there wasn't any news ....  new message deleted again ?

Whatever, let's reply anyway 

Sorry to wake the dead... Intended to quote a post from this thread to the 8ch Prism Barrel limiter & meter etc. thread, but accidentally posted here :

Best,

/Dave

 

[bcarso]

Hey, I got a notification but there wasn't any news ....  new message deleted again ?

Whatever, let's reply anyway 

Sorry to wake the dead... Intended to quote a post from this thread to the 8ch Prism Barrel limiter & meter etc. thread, but accidentally posted here :

Best,

/Dave

That's o.k.  It seems to be one of those days already here in California.

I was awakened out of deep sleep this AM just in time to hear my message machine cut off the message from a soft-speaking nurse at a hospital where my mother is ailing---all I got was her name and that it was the hospital.  By the time I got up and tracked down someone it turned out to be a question about any recent vaccines (flu shots etc.) that my brother who is local to her had already been asked.

But hey I'm up now.

 

[clintrubber]

But hey I'm up now.

It's not nice indeed, but starting to read your words I was fearing it'd end in a quick rush to the hospital,
half awake, one sock, two different shoes, no coffee etc.
This is relatively better, but still, all the best with your mother.

The wake-up call here came from an (early ?  ) postmen delivering some DIY-stuff I don't need, but a decent reason to get going anyway...

 

[chrissugar]

Ha ha, almost 4 years since I started this thread.

I totally forgot about it and to post my results. I experimented with this idea and here are my findings.
I have no idea what is exactly in the Prism limiter but from the info I have it looks like there are only two zenner diodes back to back.

I started to play with this and came up with a solution similar to the one posted by Jack (AMZ-FX)
I found that 100ohm resistors are a good compromise, they produce a non significant attenuation to the signal and reduce distortion by current limiting, by not forcing the source to work into a short when the high peaks appear.
I used matched 8.2Volt zenners back to back.

Now about sound.
Tried them with individual instruments, high dynamic music (classical, acoustic jazz), and low dynamic music (compressed moden music)
-individual instruments with short peaks (for example snare, bass drum) - you can limit up to 2dB without significant sonic degradation, at 3dB it starts to sound bad (or good if you want to use it as distortion effect)
-dynamic music with ocasional peaks - up to 1.5-2dB limit do not produce significant sonic degradation
-compressed music - maximum 1dB, it is almost unusable because peaks are dense and produce a noticeable sound degradation even at low levels of limiting

This is not a real limiter, it is mostly a device that will not let the AD converter to overload. Digital clipping sounds much worse than signal limited with this device, so in some situations it can be a good compromise.
If not necessary, do not use a device like this. Use 24bit recording with quality converter and never go above -6dBfs.
Calibrate the recording chain to not go over -6dB even for the loudest signal. The AD will never clip. You will not loose dynamics and it will not distort.

Here is a quick hand drawn schematic

chrissugar

 

[craptical]

So there was a point in my malposting anyway! ;D

Thanks for sharing, chrissugar!

/Dave

 

[AMZ-FX]

It is interesting to note that the impedance of the zener diodes will vary with the current through the diode. At 20ma the zener impedance will be quite low but as the current decreases down to the knee region, the impedance increases. At low currents such as in an audio signal, the zener impedance can be substantial and makes the diode network more of a gain reducer than hard clipper.

Zzk is the impedance in the knee region of the zener's curve at low current. There are some zeners, like the UDC series, that maintain a low impedance even at small currents and these would clip the audio harder.

The network B in this image exploits the impedance curve and also softens the sound by using actual zener diodes, which supposedly have a softer knee than avalanche diodes (I haven't had the chance to measure this yet). Exact matching isn't necessary and a pack of 10 zeners should give you 4 diodes within 1% or better.

regards, Jack

 

[mikep]

Something that strikes me as ironic: out of all brands of A/D converters, the ones made by prism are very resistant to sounding bad when clipped. especially their dream series 20 and 24-bit, I can't speak for the 8-ch units.  the output flat-tops when overdriven, but if only a dB or 2 is shaved off, the subjective effect can be completely un-noticeable.  Many mastering engineers exploit this property of prism A/D's.  perhaps they offer these barrels as a concession to those who can only afford a cheaper A/D made by someone else, but want the "well known" prism peak tolerance? wouldn't surprise me.

Mike P

 

[Kingston]

in the 8-channel prismsound orpheus unit you can seriously overdrive the AD before you get any clipping, 3-5dB very transparently. But actually the clip protection itself is a separate feature. You can actually hear relays click in the unit when you turn it on from the control panel.

And that clip protection sounds like crap, exactly like a simple zener clipper would.

I find it extremely odd they have the protection as a separate feature (with relays!) when the pure AD conversion itself sounds a whole lot better when clipped. The feature is completely pointless in orpheus.