Slew rate . . . .

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

strangeandbouncy

Well-known member
Joined
Aug 8, 2004
Messages
2,112
Location
West Sussex. UK
Dear All,

Now I know that there are many measurable factors that determine how something "sounds", but I must confessed i am somewhat confused. I am currently "hot-rodding" my old La4a. i have replaced the 3 op-amps in the input (ic1a-c) with a transformer - unbelievable difference! No longer muddy and unclear. Now my attention is turning to the output amp. I figure that a 2520 might fit the bill, especially as the way i like to use it is at unity gain, with no gain make-up. I think I can easily substitute ic1d, and maybe use the existing 600:600 output tranny to balance unit.

Where i am confused is as follows. 4136 op-amps are not regarded as being very good for audio, and the main reason is cited as the poor slew-rate, being 1.7V/us. I spotted that 2520 is rated as merely 3V/us. Surely they should sound "bad" too? Not in my experience! I have plenty in use daily. None of them sound "Muddy". Are they perhaps past some mythical threshold of slew-rate after which they sound groovy? Sontec is rated as 200V/us . . . and I must accept that they are seriously hifi compared to API. But API stuff is SO cool, so perhaps it is a combination of factors that count?

Any input on the "nature" of sound with relation to slew-rate/distortion/bandwidth/settling time/etc would be most revealing. Perhaps there are other factors at play that we don't know how to measure yet? - Sunspots?


Thanks,


ANdyP
 
Walt Jung looked at this, back in the day we wondered what was wrong with 741.

You need to slew in a microsecond. For 1V hi-fi signals, 1V/uS is not horrible. For 10V hot-pro signals, you may need 10V/uS. Jung tested 10V signals for many +/-15V-powered circuits, but if you can prove your signal never goes to 10V then this could be overkill.

Just as important: you must slew symetrically and cleanly. Good triangle wave. Many inputs are not balanced for slew and grow saw-tooth waves. Slew-enhanced inputs actually slew slowly until they are in great distress, and then a booster kicks-in and posts the big numbers. That may even be useful for some sampling chores, but is horrible for audio.

Run a sine through your amp, peak level similar to your normal audio level, and scope it as you run from 1KHz to 100KHz. How awful does it get?

IIRC, the 4136 is also modest gain-bandwidth. 3MHz? Maybe not that good? How wideband is the 2520? It does not take much gain in a 1MHz 741 to have nearly zero feedback in the top octave. The internal gain stage is not so linear that you can get away with this. And the class AB output stage may be closer to B than to A: that is one thing the audio-specific amps sometimes do better than the general-purpose chips.

Slew happens on whatever signal the amp gets, not just the audio. If you have tape-bias or digi-crap coming in, that slews bad and slashes the audio. A low-pass filter is sometimes the wisest path. The proper low-pass with enough slew, and the amp never slews even with full-voltage hypersonic garbage.

It was a problem back in the 1970s and may be a worse problem today. Treble power used to be limited by smoke from the disc cutter. You can not put big 6KHz on wax. Today we can put full power 10KHz on CD, and some records do. Smack them cymbals and peak 'em up!

There's also common mode rejection, which has different meaning in audio than on a chip-test bench. CMRR is supposed to be infinite, never is, and the residual leakage can be very non-linear. In instrumentation, as long as it is small, it might be trimmed away or ignored. In audio our ears pick up on this side-path around the amp.

> Perhaps there are other factors at play that we don't know how to measure yet?

We have not begun to measure audio properly, and have forgotten things the Dead Men knew.
 
Dear PRR,


thank you for the most informative reply! If anyone is interested, I have put the 2520 data sheets on my iDisk, and they can be seen here.

http://homepage.mac.com/strangeandbouncy


i see that GAin bandwidth of 2520 is 50mHz with operating bandwidth of 40kHz at full power, so I guess there is some serious milage in this. I guess it's a bit like a small compact car and a muscle car. They can both do 80 miles an hour, but one does so with ease, and the other will fall to pieces, or something . . . .


Kind regards,


ANDyP
 
The gain bandwidth of the 2520 is 50 mHz.
Which is of course a typo in the datasheet--it is 50 MHz.

Sontec is rated as 200 V/us.
Where do you have that number from? It sound pretty high for that time.

I think part of the answer to your question includes the (probably) fact that "good sound" does not necessarily corresponds to "as transparent as possible". Specs tend to correlate with the later.

Samuel
 
Samuel,


I got the figure from Sontec Mep250 info. Actually, it says "Greater than 200V/us" . . . .


see it here:


http://homepage.mac.com/strangeandbouncy


All the best,



ANdyP
 
Dear PRR,

THanks for the link! That's what got me thinking about this! JLMAudio does a 4136 to TLO74 adapter too. I decided to use a transformer for the input since I already have several spare. I settled for a Sowter 4652 600:600 since it was easier to mount than the Gardner's 10k:10k which is octal mount. I prefered the Gardners, since it had more extended top end. This must be a function of the circuit, since Sowter 4652 is flat out to 120kHz! CJ mentioned that with an La2a, substituting a 10K:10k requires a reduction of the 68k loading resistor to 11k to maintain the bandwidth, and that 600 was really too low. I will replace the Sowter with a GArdners or some other 10k:10k when i've worked out how to fit it! Still a VAST improvement over 4136. SO much more detail, and l.ack of muddiness. I've always loved my La4a for it's action. Now It sounds better too, with high hopes of another quantum leap when I attack the output stage too. I was thinking of just replacing the buffer amp ic1-d with a 2520 or similar, missing out the rest of it, since I generally use my LA4a at unity gain to fascilitate recalls, and easily duplicate recording chain set-ups. I could even use the existing output transformer, which surely rocks!


Kind regards,


ANdyP[/u]
 
Yes, that's what the datasheet at Andy's site says too. Interesting. I guess I'll have to build one of these and listen too it.

/Anders
 
Sorry, when I read the first post more closely, I can see that you allready know that. :oops:

But when we talk about API and slew rate, I remember when I read the papers, that they also made a 2525 op-amp under the 1980`s with 20V/uS, that can handle supply voltage up to +/- 35 volts, and have open loop gain around 75 dB. (compare with 2520`s 120 dB)

So, now have I at least add something new. :wink:

--Bo
 
Andy-

Your question about slew-rate:

Today Walt Jung posted the Audio magazine version of his late-1970s SID/TIM (slew-rate) articles:

http://waltjung.org/Classic_Articles.html

(Note that the manuscript was rejected by a professional society.....)

There's no breakthrough concept here, just level-headed consideration of what a specific (and then-overlooked) amp limit really does to sound.

There is also an Audio Amateur article, covering the same thoughts but with more geeky details. If the Audio article leaves you hungry, maybe the AA version can be posted, but it is a LOT of scanning.
 
[quote author="strangeandbouncy"]... I've always loved my La4a for it's action. Now It sounds better too, with high hopes of another quantum leap when I attack the output stage too. [/quote]
for the KSOC and KDOC
( Single and Dual Opto Comp )
I chose to keep the 4136 in the side chain but upgraded the input and output signal paths
I figured the slew problems in the side chain were all part of the LA4 sound.
 
[quote author="strangeandbouncy"] ... not sure what KSOC and KDOC are, ...[/quote] ??
:?

[quote author="Kev"] ... for the KSOC and KDOC
( Single and Dual Opto Comp ) ...[/quote]
KevSingleOptoComp

KevDualOptoComp
image

KDMP
KevDualMicPre


these and more all used to Google when they were on the GroupDIY site
but alas
:green:
 
I did a similar mod to my pair of LA*4's and was equally excited about the difference.
I used Jensen input transformers and replaced the remaining 4136 stages with better op-amps. This required some clever soldering of dual op-amp packages into the 4136 footprint, but as others here have pointed out there are adapters available.
The biggest improvement in performance not related to sonics was delivered by replacing the opto with a faster releasing unit. My biggest complaint about the LA*4 was that it was too slow on its release. Once it engaged it almost never let go in time to be musical. By careful selection of a Vactrol or similar, you can get the compression ballistic to come fairly close to that of an FET type compressor.
I also recommend removing the peak led and the zener that feeds it. I use the led in series with the one inside the opto so it gives me a front panel peek into the compressors doings. Granted, it rarely lights visibly, and when it does the compressor is really digging in, but at least it doesn't compromise the sonics the way the overload indicator did.
There are further refinements to the signal rectifier circuit (LED drive) I'd like to do but haven't found the time. I think there is some subtle distortion created here somehow and would like to track it down. It may be switching noise in the power supply or it may be that the faster opto is responsible.
As for the output transformer, I have both original OT's in mine and have done this mod to a friend's unit that came to me with a bad output transformer. I used a Jensen in his and it sounds a lot more open and clean, but there's that special voodoo you get from less-than-Gucci iron that the originals have, and I'm not changing those in mine.
 
dabo,


thanks for posting. But I have to disagree with you vv release times! I find the interactive release of the LA2/3/4's to be their most endearing feature! The initial attack and release is fast, and the latter part of the release is slow, kind of like the auto fuction on the SSL . . . VERY musical if you ask me! I am certain that there are many others that agree. If I wanted a release like a fet compressor, then I guess I'd use use a fet compressor. Each to his/her own, after all ,if we all did it in the same way, music would all sound the same!

As far as indicator LED, I am using a transformer in place of ic1a-c, so this no longer applies to me.

Vive la Differance!

ANdyP
 
> replaced the 3 op-amps in the input (ic1a-c) with a transformer - unbelievable difference! No longer muddy and unclear. Now my attention is turning to the output amp.

May be less improvement there.

Look at the input and output levels of a limiter. Take a make-believe condition: limiter is unity-gain below threshold, you want to whack peaks down 30dB (a lot, but some folks here do that to tracks). If the output level is 1V, the input level touches 30V!

We don't usually have 30V sources, we don't work unity-gain, we put gain in the limiter. Even so, the input to a hard-worked limiter can easily be higher than the fixed output. We KNOW the output won't go (much) past the limited level; the input level is unlimited and often hot-peaked (which is why we turned to a limiter in the first place).

Someone mentioned this as a feature of the LA2: no active parts before the opto-resistor. Nothing to clip of slew, just a little iron-tone at high input.

The LA2 also has a remarkably lame output stage for Pro gear: isn't very important, because the output level is automatically limited.

> 2520 is rated as merely 3V/us. Surely they should sound "bad" too? Not in my experience!

Working at +4dBu with reasonable headroom, 3V/uS is just a bit below Jung's criteria. While we are pushing the criteria, the huge 50MHz GBW is holding the amp linear as long as possible, unlike 4136's 3MHz GBW.

> unbelievable difference! No longer muddy and unclear

I don't know what levels you were putting in. I do know that if you fed more than about 12V peak, you'd have obvious clipping. But you hear a difference without the 4136, so something was straining. Jung, in 1977, said that anything less than 0.5V/uS per volt of peak output audibly degraded the sound. Assuming you were running 10V peaks from the IC1a-c amplifier, this says you need 5V/uS. The 4136 is 1V/uS to 2V/uS, which is far below Jung's criteria. Especially working with hot-treble tracks, you could well be exceeding the amp's slew limit. Jung actually says you should not approach it, because sound is degraded well before it hard-slews. In this range where you are near slew but not yet making triangles, open-loop gain will hold distortion down, but the 4136 does not have a lot of that at the top of the audio band.

> perhaps past some mythical threshold of slew-rate after which they sound groovy?

It's like clipping. If the core amp is good, with ample excess gain, it is nearly faultless up to clipping, then it gets very UN-groovy. Slew-rate is a second limit that hurts the highs. It is not as familiar as good old clipping, and may be lost in other faults. It also differs from clipping in that many amps can get to 95% of clip-level without any rise of distortion, while most slew-limited amplifiers are rising distortion at 25% of the gross-slew condition. So there is a large range where details are degraded without actually being flattened (made into ramps).

So if a 1.6V/uS amp is slightly degraded, another amp with twice the slew and several times the feedback may be much less degraded, AT that level for that signal. At a lower level, the 1.6V/uS amp may be cruising smooth.

If you re-wired for TL08x or 5532, the slew-rate is just at Jung's criteria and you get a bit more GBW. If not for the 4136's odd pinout, this would be the obvious swap.
 
Back
Top