Some stupid LA2A questions

[NewYorkDave]

[quote author="jhaible"]So I figure the LA2A designer - certainly not stupid - might have had low output impedance as a design goal more than high output voltage swing?
This is just a guess; maybe low output impedance was important for good low frequency response in connection with the transformer?[/quote]

I would tend to think that low output impedance and high output swing are important in this application. The output transformer is 5:1, right? If we take it as a given that the unit should be able to deliver at least +24dBM (20dB headroom over operating level, which is pretty standard) without clipping, it needs to swing 35V peak-to-peak across the load, and 175V pk-pk across the output transformer primary. That's a pk-pk current of 12mA (or a peak current of 6mA), assuming the transformer losses are negligible. The stock output stage idles at 4mA; and due to the high resistances in series with it, will start limiting before it can swing 175V pk-pk across the output xfmr. Of course, this "running out of steam" at high output levels could very well be part of the signature sound of the LA2A. I've never owned one or had one in my possession, nor have I built a clone, so I can only do a dry analysis from looking at schematics.

As I was Googling around just now, I found that Jensen engineers have had a similar thought, that the output stage of the LA2A could be "improved" by increasing its output capability. Their mod is published here. Notice that they also replace the output transformer with one of their own models having a 4.3:1 ratio. (Naturally, they want to sell transformers!). It's interesting to me that they settled on a 680-ohm plate resistor, just as I did.

Does this make any sense? Maybe the whole WCF thing is aiming for minimum output impedance rather than maximum output level?

If you read the patent, it appears that the intention of the inventor was to deliver maximum power into a difficult load (e.g., the capacitive load presented by a CRT). The patent is here.

I have no doubt that the designer of the LA2A knew what he was doing... But "second-guessing" and looking for ways to improve circuits are what keep this hobby interesting. Otherwise, we're just a bunch of clone builders, which to me is not terribly interesting at all. Sometimes the fun comes from finding that your "improvement" sucks and then the reasons for certain choices made by the original designer become crystal-clear. In those moments, I feel almost as if the designer is looking at me from the next world and grinning just a little bit :wink:

 

[PRR]

> interesting for a headphone amp.

See http://headwize.com , DIY section.

Runeight and others there know a lot about tube headphone amps and WCF optimization.

 

[jhaible]

[quote author="NewYorkDave"]I would tend to think that low output impedance and high output swing are important in this application. The output transformer is 5:1, right? If we take it as a given that the unit should be able to deliver at least +24dBM (20dB headroom over operating level, which is pretty standard) without clipping, it needs to swing 35V peak-to-peak across the load, and 175V pk-pk across the output transformer primary. That's a pk-pk current of 12mA (or a peak current of 6mA), assuming the transformer losses are negligible. The stock output stage idles at 4mA; and due to the high resistances in series with it, will start limiting before it can swing 175V pk-pk across the output xfmr. Of course, this "running out of steam" at high output levels could very well be part of the signature sound of the LA2A. I've never owned one or had one in my possession, nor have I built a clone, so I can only do a dry analysis from looking at schematics.[/quote]


That's my problem, too: I don't have an original to compare. But I rarely have on original when I build something. THat's why I'm asking stupid questions. (;->)

But I can well imagine that a certain way of pleasant clipping that's built-into a limiter can be intentional. All this having ultra high headroom looks good on paper, but in practice it might also mean that when the device with high headroom does not clipp, the next device in the signal chain might clip (asuming its headroom is smaller), and unpleasently so (if that device's designer hasn't taken care about pleasant clipping).
Now as a *Limiter* is dedicated to reducing dynamics, one of its applications being to prevent following devices from clipping, and as the (non-clipping) gain-reduction process cannot be infinitely fast (in order not to produce another kind of distortion), this built-in soft clipping rather than having an ultimate headroom makes sense *especially* on a compressor / limiter.

I tried to explain this on my 1176 clone page
http://www.oldcrows.net/~jhaible/compressor/jh_compress.html
see the section (4) "Distortion". It's all just guesswork, but its the most reasonable explanation I found to explain why a simple FET-based compressor can sound so much better than a Behringer which uses VCAs that look much better on paper.

[quote author="NewYorkDave"]As I was Googling around just now, I found that Jensen engineers have had a similar thought, that the output stage of the LA2A could be "improved" by increasing its output capability. Their mod is published here. Notice that they also replace the output transformer with one of their own models having a 4.3:1 ratio. (Naturally, they want to sell transformers!). It's interesting to me that they settled on a 680-ohm plate resistor, just as I did.[/quote]

This makes perfect sense, and it's a nice validation of your own work.

If you read the patent, it appears that the intention of the inventor was to deliver maximum power into a difficult load (e.g., the capacitive load presented by a CRT). The patent is here.

I wasn't aware of the patent - thanks for pointing this out!
So my asking stupid questions will bring up more and more information!
Tube circuits are still some kind of unknown land to me. I just know the rough basics, and I'm gladly learning. So when the WCF is optimised to give the maximum output swing, does it still work in class A? or is it going into Class AB operation? (It's probably in the patent - I have to read this first.)

JH.

 

[jhaible]

[quote author="NewYorkDave"][quote author="jhaible"]So I figure the LA2A designer - certainly not stupid - might have had low output impedance as a design goal more than high output voltage swing?
This is just a guess; maybe low output impedance was important for good low frequency response in connection with the transformer?[/quote]

I would tend to think that low output impedance and high output swing are important in this application. The output transformer is 5:1, right? If we take it as a given that the unit should be able to deliver at least +24dBM (20dB headroom over operating level, which is pretty standard) without clipping, it needs to swing 35V peak-to-peak across the load, and 175V pk-pk across the output transformer primary. That's a pk-pk current of 12mA (or a peak current of 6mA), assuming the transformer losses are negligible. The stock output stage idles at 4mA; and due to the high resistances in series with it, will start limiting before it can swing 175V pk-pk across the output xfmr. Of course, this "running out of steam" at high output levels could very well be part of the signature sound of the LA2A. I've never owned one or had one in my possession, nor have I built a clone, so I can only do a dry analysis from looking at schematics.

As I was Googling around just now, I found that Jensen engineers have had a similar thought, that the output stage of the LA2A could be "improved" by increasing its output capability. Their mod is published here. Notice that they also replace the output transformer with one of their own models having a 4.3:1 ratio. (Naturally, they want to sell transformers!). It's interesting to me that they settled on a 680-ohm plate resistor, just as I did.

Does this make any sense? Maybe the whole WCF thing is aiming for minimum output impedance rather than maximum output level?

If you read the patent, it appears that the intention of the inventor was to deliver maximum power into a difficult load (e.g., the capacitive load presented by a CRT). The patent is here.

I have no doubt that the designer of the LA2A knew what he was doing... But "second-guessing" and looking for ways to improve circuits are what keep this hobby interesting. Otherwise, we're just a bunch of clone builders, which to me is not terribly interesting at all. Sometimes the fun comes from finding that your "improvement" sucks and then the reasons for certain choices made by the original designer become crystal-clear. In those moments, I feel almost as if the designer is looking at me from the next world and grinning just a little bit :wink:[/quote]


All right, now I'm convinced I must make a modification, too.
As my LA2 project comes close to be finished, I could make more extensive listening tests, and I found that as it is now, clipping does *not* sound nice, once it happens. So I want as much headroom as possible.

Why does the clipping sound ugly? I guess with the feedback loop the whole thing is not that far from a transistor circuit when overdriven - no soft rounding, at least not for the upper limit.

The first thing I will change: bringing the PSU voltage up as much as possible. I have a 230V secondary power transformer, not 250V as the original. So I'll have to reduce the 3k9 rersistor in the power supply to get a similar voltage for the output stage. Won't increase hum much (I already tested this). And then it's time to run more current thru the 12BH7, and to make the anode resistor smaller. In othe rwords, go along the Jensen mod.

I just wanted to ask - has anybody made calculations how much this will increase the output impedance? (Before I read out the tube parameters from the meagre data sheet and fill in th ebig equations myself ...)

JH.

 

[PRR]

> http://www.tubecad.com/october99/page2.html "... high transconductance that is not matched to a high current draw [...] worsens the sound. I like to use as much current as there is transconductance ..." ...what does this mean? ... I figure that he means "high gm devices like to be used at high current"...

John is usually quite clear, but sometimes his cleverness gets in the way. I know some about tubes and have read much of John's work, and I'm not sure what he really means.

My guess: high transconductance gives large gain but does not give large output. The "super-hot" tubes that have high Gm at low plate current will run out of steam on large signals. In many cases you want a large tube with large plate current, so it does not strain under large signals.

This is very true of the WCF. It has lots of internal feedback so high-Gm tubes appear to give fabulously low output impedance. But it is very wasteful of supply voltage. It can't drive large signals into loads anywhere near as low as the small-signal output impedance suggests.


> low output impedance and high output swing are important in this application. The output transformer is 5:1, right? If we take it as a given that the unit should be able to deliver at least +24dBM (20dB headroom over operating level, which is pretty standard)

I think the ratio is 4:1.

In the LA2's world, +8dBm was a more usual reference, with as little as 10dB headroom. Less than 1% of peaks will rise more than 10dB above VU-indicated level. 14dB headroom was plenty for broadcasters. IIRC, the fad for +4dBm came in the recording industry circa 1960s with direct-coupled transistor amps working at lower voltages (+22dBm clip) and a desire for less than 1% clipping, especially in light of the harsh clipping of transistor amps opposed to the soft clipping of push-pull 1620 tubes.

And note that IF a limiter is doing its job, the output level will not contain any significant large peaks, so it can be worked closer to its limits. The lame output stage in the LA2 does not compete with "real" general-purpose line-amps, often 5W-10W stages rated for 1W (+30dBm) nominal level but able to peak 36-40dBm before they went splat.

> the designer of the LA2A knew what he was doing...

To be honest: making a decent limiter cheaper than the Big Boys. The LA2 is much cheaper than the broadcast-quality RCA, GE, and Fairchild monsters of that day. It is just good enough for most uses.


> when the WCF is optimised to give the maximum output swing, does it still work in class A? or is it going into Class AB operation?

For Audio, it must run class A. If the top tube goes into cut-off, there is no drive to the bottom tube. And if the bottom tube is going more than a little bit into cut-off, then the stage is not balanced correctly and not giving best performance for symmetrical signals. (Audio is not allways symmetrical in details, especially music, but over the long run we swing both ways equally.)

The patent also talks about non-Audio loads with asymmetrical signals (sweep-signals), and how to optimize for that. If the wave is very-very asymmetrical, the WCF isn't the best choice; but for a large class of medium-asymmetry signals the WCF can be a good choice.

> Why does the clipping sound ugly?

A plain cathode follower is "perfect" up to a point and then it clips horribly, but one side before the other. You still have peaks on one side. The well-optimized WCF cancels the slight inperfection and can go a little louder before it clips, but when it clips it REALLY clips, both sides, a much harsher sound.

Run more voltage and current on the output stage. For small tubes like 12BH7, you may have to approach the plate dissipation rating to get "impressive headroom". Be sure the top grid sits at very-nearly half the total supply voltage; small deviations here make large differences in maximum output. Don't worry too much about output impedance: the original LA2 has a fairly low output Z (around 100 ohms?) for vintage gear, and no reasonable tweaks will "ruin it". Anyway in a modern 10K-input studio, Z(out) only needs to be well below 1K ohms.

 

[jhaible]

[quote author="PRR"]> http://www.tubecad.com/october99/page2.html "... high transconductance that is not matched to a high current draw [...] worsens the sound. I like to use as much current as there is transconductance ..." ...what does this mean? ... I figure that he means "high gm devices like to be used at high current"...

John is usually quite clear, but sometimes his cleverness gets in the way. I know some about tubes and have read much of John's work, and I'm not sure what he really means.

My guess: high transconductance gives large gain but does not give large output. The "super-hot" tubes that have high Gm at low plate current will run out of steam on large signals. In many cases you want a large tube with large plate current, so it does not strain under large signals.

This is very true of the WCF. It has lots of internal feedback so high-Gm tubes appear to give fabulously low output impedance. But it is very wasteful of supply voltage. It can't drive large signals into loads anywhere near as low as the small-signal output impedance suggests.


> low output impedance and high output swing are important in this application. The output transformer is 5:1, right? If we take it as a given that the unit should be able to deliver at least +24dBM (20dB headroom over operating level, which is pretty standard)

I think the ratio is 4:1.

In the LA2's world, +8dBm was a more usual reference, with as little as 10dB headroom. Less than 1% of peaks will rise more than 10dB above VU-indicated level. 14dB headroom was plenty for broadcasters. IIRC, the fad for +4dBm came in the recording industry circa 1960s with direct-coupled transistor amps working at lower voltages (+22dBm clip) and a desire for less than 1% clipping, especially in light of the harsh clipping of transistor amps opposed to the soft clipping of push-pull 1620 tubes.

And note that IF a limiter is doing its job, the output level will not contain any significant large peaks, so it can be worked closer to its limits. The lame output stage in the LA2 does not compete with "real" general-purpose line-amps, often 5W-10W stages rated for 1W (+30dBm) nominal level but able to peak 36-40dBm before they went splat.

> the designer of the LA2A knew what he was doing...

To be honest: making a decent limiter cheaper than the Big Boys. The LA2 is much cheaper than the broadcast-quality RCA, GE, and Fairchild monsters of that day. It is just good enough for most uses.


> when the WCF is optimised to give the maximum output swing, does it still work in class A? or is it going into Class AB operation?

For Audio, it must run class A. If the top tube goes into cut-off, there is no drive to the bottom tube. And if the bottom tube is going more than a little bit into cut-off, then the stage is not balanced correctly and not giving best performance for symmetrical signals. (Audio is not allways symmetrical in details, especially music, but over the long run we swing both ways equally.)

The patent also talks about non-Audio loads with asymmetrical signals (sweep-signals), and how to optimize for that. If the wave is very-very asymmetrical, the WCF isn't the best choice; but for a large class of medium-asymmetry signals the WCF can be a good choice.

> Why does the clipping sound ugly?

A plain cathode follower is "perfect" up to a point and then it clips horribly, but one side before the other. You still have peaks on one side. The well-optimized WCF cancels the slight inperfection and can go a little louder before it clips, but when it clips it REALLY clips, both sides, a much harsher sound.

Run more voltage and current on the output stage. For small tubes like 12BH7, you may have to approach the plate dissipation rating to get "impressive headroom". Be sure the top grid sits at very-nearly half the total supply voltage; small deviations here make large differences in maximum output. Don't worry too much about output impedance: the original LA2 has a fairly low output Z (around 100 ohms?) for vintage gear, and no reasonable tweaks will "ruin it". Anyway in a modern 10K-input studio, Z(out) only needs to be well below 1K ohms.[/quote]

I did the Jensen mod (15mA thru the tubes, and 680R anode resistor), and I reduced the series resistor in the PSU to 680R. Now I have at least 250V supply voltage under load.

Clipping is very unsymmetrical ... until I connect a 600R load. Now it's almost perfectly symmetrical. Still a quite sharp cutoff. I wonder if I wouldn't have got th esame results with a simple transistor amp running at 30V. (;->)

I have finally read the White patent. I'm totally with you that audio signals should be run in Class A, but reading the patent, which (among other things9 had TV signals in mind, this sounds a lot like leaving class A, doesn't it? He talks a lot about running a low bias current thru the tubes, and providing higher output current for both push and pull operation (I'm paraphrasing), so maybe the main goal of White was providing a more efficient circuit than class A? I'm still quite puzzled about this.

Anyway: I'm happy with the higher output swing of my compressor now (about 100Vpp, or 25Vpp after the 4:1 trnasformer).
I've designed a front panel - it's on the updated web page for download, if anybody is interested in a not-so-original-looking LA2 panel.

http://www.oldcrows.net/~jhaible/opto_compressor/jh_opto_comp.html

JH.

 

[jhaible]

A plain cathode follower is "perfect" up to a point and then it clips horribly, but one side before the other. You still have peaks on one side. The well-optimized WCF cancels the slight inperfection and can go a little louder before it clips, but when it clips it REALLY clips, both sides, a much harsher sound.

Now I have redesigned the whole output stage. I think the worst part of the original design was the negative feedback around the 3 stages of the audio amp, which will make any kind of clipping even so much worse. I went back to a much simpler design and optimised the WCF output to handle the signals that come from the no-feedback voltage gain stage. See
http://www.oldcrows.net/~jhaible/opto_compressor/jh_opto_comp.html
for details.

I've also added the schematics of my meter circuit (no real VU, but better than nothing), and some pictures of the finished compressor.

Thanks to everyone who helped me to get there !!

JH.

 

[PRR]

> I think the worst part of the original design was the negative feedback around the 3 stages of the audio amp, which will make any kind of clipping even so much worse.

But it is a limiter. The output "shouldn't ever" clip. If it is limiting correctly, the output won't exceed a set point. If you need a set-point higher than the fairly weak 1-tube output stage can handle, build a bigger output stage.

If you want "color", I think your revised plan is a perfectly good way to do that. Simple, effective, and very easy to troubleshoot. In theory, the 6K8 plate resistor "should" be smaller, but in this case it makes very slight difference. Something to play with after you have used it a while. You may prefer the larger resistor for asymmetrical "sweet" clipping.

Is that 100 microFarads for the coupling cap between the 12AX7 and the WCF? Likewise for the coupling cap from the top to the bottom of the WCF. Seems a little large. 100 nanoFarads would be more common.

 

[jhaible]

>But it is a limiter. The output "shouldn't ever" clip. If it is limiting >correctly, the output won't exceed a set point.

It did clip. The problem is that strange arrangement of gain reduction
and (makeup) gain pots. If you reduce gain reduction, and don't reduce gain at the same time, your output is prone to overload. I like the 1178-style controls much better, where the sidechain gain / compression threshold is fixed, and you set the threshold indirectly by changing input level and output level. There you can change gain reduction without touching the output level pot, and you're always in the right output level range.


>If you need a set-point higher than the fairly weak 1-tube output stage >can handle, build a bigger output stage.

This will only shift the problem to anything you connect after the compressor. If you have no soft clipping feature, either nothing will clip (with the right setting), or it will clip in an unpleasant way.

I must add that my goal was not the usual (?) vocal processing; I built this for taming electric piano dynamics (to feed the e-piano sound to a chorus device etc.). I want to get the whole range from clean (but over-compressed) to barely compressed & dity. I even think about replacing this extra clean Lundhal transformer with an O.E.P. type for that reason.

>Is that 100 microFarads for the coupling cap between the 12AX7 and the >WCF? Likewise for the coupling cap from the top to the bottom of the >WCF. Seems a little large. 100 nanoFarads would be more common.

No, that's just my ugly handwriting. It's 100n as in the original.

JH.

 

[Bjorn Zetterlund]

Thanks for an enjoyable and informative thread!

Bjorn

 

[thedug]

So what is the ratio for the A-24?

So far on this thread I've seen 5:1 & 4:1.

The spec sheet I found said the windings are 15,000(split- whatever that means) to 600. So that would be 25:1. That sounds crazy.

I got recomended a lundahl L7902 but that can do 4:1 at best.

Thanks,

d./

 

[PRR]

> So what is the ratio for the A-24? So far on this thread I've seen 5:1 & 4:1. The spec sheet I found said the windings are 15,000(split- whatever that means) to 600. So that would be 15:1. That sounds crazy.

How do you get "15"?

15,000/600= 25. That is the impedance ratio. The voltage ratio and the turns ratio is the square root of that: 5.

This is a good value for the original design. And it was a VERY standard part. Used everywhere. I pulled an A-24 out of a classroom.

The original design is a little lame. That may be part of its "sonic signature", but mostly this was a cheap box. 20+ years later, Jensen looked at it and proposed new resistor and bias values for more "oomph". Part of that re-design is a different transformer ratio. It happens to be their standard 10K:600 part, but the design is good and they didn't fudge it just to use their stock-part. And 10K:600 is a stock-part from many other vendors.

The difference is not large. On voltage-gain, less than 2dB, and you have many dB of adjustment in the LA2. The original values won't make huge output dead-clean, and would be a little worse with a 4:1 tranny. But you rarely need huge output unless you are trying to impress with specs.

If you aim for an exact clone, get a vintage A24 and use the original resistors.

If you are not trying to be 101% authentic, use a good modern 4:1 transformer with the Jensen-suggested resistors.

 

[thedug]

I am planning on using a Lundahl LL7902.

Which will be wired 4:1.

So what resisters should I change?

 

[Scenaria]

you dont need to change any resistors...

just decide on the original version or the newer revision... there really is no difference sound wise...

I would prefer the 22k/2w as I think it reduces the amount of wear on the EL panel..

 

[thedug]

So do I need to change the capaciter values too?

You aggreed with leaving off C13, but the other cap values aren't the same.

 

[thedug]

These are the jensen mods, are you saying that aren't necessary for using 4:1?

http://www.jensen-transformers.com/as/as091.pdf

 

[Scenaria]

no you dont need the jensen mod for those lundahls.. isak is around here somewhere.. I know he finished one with lundahls.. maybe we can summon isak

 

[thedug]

Cool, Thanks.

Isak's user is popedh & I im'd him... We'll see.

I have another post on this thread where I am trying to figure out which values to use. I don't want to say too much about the bloo manual, but there are some values from the 79 version and some from the 65 version and that worries me. Please let me know if that mismatch was intentional or not.

http://www.groupdiy.com/index.php?topic=3278&start=195#img_1663

 

[hejsan]

[quote author="Scenaria"]no you dont need the jensen mod for those lundahls.. isak is around here somewhere.. I know he finished one with lundahls.. maybe we can summon isak [/quote]

Hey I was thinking exactly the same thing right now, meaning that I want to use lundahl's for input and output trafo. Are there any more news on this?

I would really want to be able to use the la2a I'm building normally with my other gear.

As I understand, then I will run into problems trying to feed the la2a my signal and vice versa because of the 600ohm impedances, right?

What can I do to make the la2a accept xlr connectors both (or balanced jacks) ways and make me super happy??

Thanks,
Bjarni