What phase inverter style has the best output to linearity ratio?

[Macguyver]

I am wondering if there is a phase inverter out there that will stay more linear than the standard long tail pair at high outputs. The kind of music I play, and everyone I do amp stuff for is metal, which the most common requirment being a tight, huge sounding low end. I found that with any version of the long tail pair, I am sacrificing either linearity or power when I try to modify the design. My goal is to have a powamp that will stay transparent at louder volumes, and it doesn't matter to me how many tubes it takes to make it happen, is there any other design that is more up my alley?


:compress: :help: :fishing: :smoke:

[Ray Ivers]

M,

It seems like I'm recommending this one a lot lately, but IMO it's hard to beat the Marshall 1967 (Major) setup, for both simplicity and performance:

http://www.schematicheaven.com/marshallamps/major_1967u_lead_200w.pdf

This uses a 12AX7 concertina phase inverter followed by a 12AU7 differential amp. A 12AU7 diff-amp will have a very large voltage swing - much more than is required for full output - so at the normal reduced swing the linearity is improved.

The problem with maintaining good driver-stage performance (which I define as "forcing the output tube grids to do exactly as the driver tells them to do") at high volumes is that the driver output swing on positive signal peaks is up around the 0-volt area, where the grid circuit impedance drops almost instantly to a few thousand ohms; IME you need both direct-coupling and a very low driver Zout to successfully fight this battle. But if your power amp, OT, and PS are all beefy, the Major setup should work just fine; in fact, the whole Major PI/PA/PS circuit (minus the NFB) would work extremely well for bass-heavy metal IMO.

Ray

[Carl / Zwengel Amps]

From your description of they style of music you're working with, linearity is the LAST thing you need to worry about. The good ole schmidt differential ala Fender/Marshall is the defacto go-to driver for this particular tone. Pretty much all the metal amps use this for a reason and as the saying goes...if it aint broke dont fix it! The devil is definetly in the details with this thing. The trick to tight huge low end is not in the the type of design but how you schedule the gain and how you curve the freqiency response of the circuit, particularly in the preamp. Small changes can make huge differences in the character of the amp. And as always, having really good iron in the OPT.

Now, if you're looking to get all your gain and distortion completely from the preamp with a very clean power section a good way to do this is with some sort of differential driver feeding a CF(AT7, AU7, or BH7) coupled to the power tube grids so you can get into AB2. And of course complete overkill on the power output...100-120 watts minimum! The down side here is that it'd be a scratch design so there's really no "canned" circuit you can just stuff in and have it work right.

-Carl

[Shea]

Now, if you're looking to get all your gain and distortion completely from the preamp with a very clean power section a good way to do this is with some sort of differential driver feeding a CF(AT7, AU7, or BH7) coupled to the power tube grids so you can get into AB2. And of course complete overkill on the power output...100-120 watts minimum! The down side here is that it'd be a scratch design so there's really no "canned" circuit you can just stuff in and have it work right.

-Carl

I was thinking an interstage transformer might be good, if he's going to have a completely clean power amp. Maybe that would work nicely with an ultralinear power section and no negative feedback loop.

Shea

[Bruce / Mission Amps]

I bought an interstage transformer to try myself but have not had time and energy to get with it...:(
love to read what you find.

[DirtyGeorge]

A while back there was a thread about replacing the tail and cathode resistors with a choke. The Vk would then be only the bias voltage above ground instead of raised by the tail resistor voltage drop.

Wouldn't this approach give more voltage swing, all other things being equal? So maybe this PI would drive the PA harder wiht less PI distortion.

DG

[Ray Ivers]

DG,

Sure, chokes can definitely increase your signal swing; using one in the cathode circuit will give you mainly more negative voltage swing (which might come in really handy in a low-mu triode output stage using a high B+ voltage or something similar), and using a CT plate choke (or two separate ones) in the plate circuit would greatly increase your positive swing from the quiescent point.

FWIW, using a constant-current source - either the conventional transistor stage or several constant-current diodes in series - as a diff-amp Rk will give you the extra swing of a cathode choke along with rock-solid DC conditions, no phase shift, and micro-buck expenditure; pretty cool IMO.

Ray

[Shea]

If you have a negative voltage source that can supply enough current for the driver, you can feed a negative voltage through a cathode resistor that has a resistance in the 100's of k's, and reference the grids of the pi to ground. Kind of a faux constant current source. I think it's probably easier and cheaper than using either a choke tail or a true constant current source. I tried it once, and it works well.

Shea

[Guitarist]

Shea, are you referring to a HiWatt style circuit then? Perhaps one could sub in IRF820 instead of the 1/2 tube. On this route is anyone interested, please, in erasing the feedback/presence circuit of the Hiwatt output so something along those lines may be easier to experiment with in keeping with this topic. I personally have little use for for 6KHz.

I was just reading how the Children Of Bodom guitarist uses an Ampeg SVT for his heavy detuned to C tones. "Size matters." Any ideas to improve that Major circuit? Does that 12au7 shove alot of current at the KT88s vs a CF arrangement?

[oc disorder]

Just thought you guys may be interested in an Aussie site...
Similar p/i to the Maj 200 in an extinct "Holden " + some info

http://www.ozvalveamps.elands.com/holden.htm

thats all

[Shea]

Shea, are you referring to a HiWatt style circuit then?

No, it's something that Ken Gilbert talked me through, or maybe it was Ray Ivers or RG Keen. Imagine a long tail pair. Instead of having a chain of two or three resistors running from the cathodes to ground, you use a single resistor in the 100's of k's - connect one end to the cathodes and the other to a negative voltage source. When I did this, I think I used a 330k cathode resistor and a negative voltage source that was around -430 volts (I just rectified it off the HT winding and filtered it the same way as the B+ supply). The grids are referenced to ground instead of to a positive voltage. Just run a 500k or 1meg resistor from the main input to ground, and the other grid can be connected directly to ground unless you're using negative feedback. If you are using negative feedback, then you just make a simple voltage divider from the appropriate OT tap to ground, and connect the grid to the junction of the two resistors. With this type of phase inverter, the plate load resistors should be of equal value.

Shea

[Ray Ivers]

Shea,

Was your circuit (excluding NFB if any) something like this?

http://music-electronics-forum.com/attachment.php?attachmentid=87&stc=1&d=1152629032

With the kind of voltages you were using, I don't think you can really get any more linear than this, for these kinds of output signal swings; balance should be outstanding as well. A current source in the tail will get you similar results with less negative voltage, and a McIntosh-style bootstrap will get you more total swing, but neither would be appreciably more linear IMO.

Ray

[Shea]

That's it, Ray. The two resistors marked Rl would be referenced to the bias voltage for the output tubes instead of ground.

Shea

[Amp Kat]

There are some Interstage trannys at Mouser but they aren't to beefy and are actually real close to mic-input trannys used in Telephone circuits. They are also quite noisy unless moved in the exact right spot as they will pick up almost anything and amplify it. It's one of those moving wires around things but the frequency response of the Mousers aren't to good so maybe the old ones would be better if you can find one or maybe you have an old one ? I'm also with you guys on the LTP but then again Carl had some good input for the type of application your using it for.

[R.G.]

The suggestions already given are good. I missed this thread early on, or I'd have replied sooner.

The standard diffamp/Schmitt PI is going to have enough linearity and swing if it's done correctly.

With no current source in the cathode circuit, input does not get fully transmitted through the cathodes to the second stage. That's why a negative supply and big resistance is good for this. But rather than making a negative supply, use a current source, as ... Ray?... suggested. I'd use a high voltage MOSFET to ensure longevity. Something like an 800-1000V rated MOSFET. The smallest current rating will be overkill for a PI tube stage. The high impedance of the tail current source will let you use all of the voltage swing available for the tubes.

Triodes don't really "saturate". It's hard to get Vpk less than about 50V. But that leaves you a lot of volts to play with. Most power tubes are driven from cutoff to Vgk=0 with less than 70V, so you could in theory need up to 140V of swing on the plates of the PI. That gets hard by the time you go through the whole voltage budget unless you do have more voltage at the cathodes or a solid state current source.

Two things can help here. One is a current mirror load for the two PI stages. A simple design with a couple of high voltage p-channel MOSFETs or a complicated, difficult, and touchy bipolar design would get you a plate load that would let you use almost the entire B+ for output swing. However, you have to watch that, as 12A?7's are really only 300V rated devices. You could have them arcing over internally. You might have to go to higher voltage rated tubes for the PI if you allow more of the B+ to be available for the PI.

An alternate if you're not listening for and liking triode distortion in the PI is to replace the PI with solid state entirely. You can put in a JFET diffamp which uses some high voltage N-channel MOSFETs to cascode them. The JFETs do the gain, the MOSFETs do the voltage swing. That can get you almost the entire B+ for output voltage swing, and no arcing on the PI tube. This thing is going to be linear far beyond the range of the output tubes and transformer, and probably beyond the range of the tube driving it. The gain will be high, so you'll have to pad the signal driving it down a lot to stay within the few volts' range of the JFET gates. Because the SS PI is linear through a bigger range than the tubes, any possible SS distortion hides inside the nonlinearities of the tubes, and is no audible as such. If you hear and like the distortion of wide range tube swing in the PI, it will be missing because the SS stuff will not be distorting.

As already mentioned, direct drive on the output tube grids will also do a great job of providing more output power range, although it may not be perfectly clean.

[Ray Ivers]

R.G.,

There's nothing I can really add to your last post, except maybe a couple of diagrams as visual aids (JFET's to be subbed for the lower tubes):

http://music-electronics-forum.com/attachment.php?attachmentid=103&stc=1&d=1153238704http://music-electronics-forum.com/attachment.php?attachmentid=104&stc=1&d=1153239218

Last month's issue of AudioXpress magazine has a tube PA design using current-mirror cathode bias - I'll try and scan/post as soon as I can.

Ray

PS: FWIW, your posts championing the enhancement of tube-amp performance using high-performance semiconductor technology never fail to make my day (not a stroke, just the truth!).

[StevieP]

The typical PI cathode resistor arrangement can be replaced by a CLD.
MESA Eng used this technique in the SOB - perhaps others.
The CLD is easy and cheap to implement - essentially a JFET with gate tied to source.
Steve

[Bruce / Mission Amps]

There are some Interstage trannys at Mouser but they aren't to beefy and are actually real close to mic-input trannys used in Telephone circuits. They are also quite noisy unless moved in the exact right spot as they will pick up almost anything and amplify it. It's one of those moving wires around things but the frequency response of the Mousers aren't to good so maybe the old ones would be better if you can find one or maybe you have an old one ? I'm also with you guys on the LTP but then again Carl had some good input for the type of application your using it for.


Andy at MOJO gave me a couple prototype trannys from Heyboer to try ... but the PT-124B at AES for $25 should be good enough Kerry.

Bruce

[Ray Ivers]

Steve,

Thanks for that info - I never looked at the SOB schematic before. Wow, 93V across a 50V part (and that's at idle!) - they really are brave S.O.B.'s. :D

Ray

[R.G.]

Brave is one word for it. Lucky is another.

I've always disliked CLDs for no particularly good reason. Not available enough in well-specified enough ways, I guess.

With so much voltage to play with, it's easy to put a low voltage current source, like a JFET or a bipolar device, and then use an 800-1000V MOSFET cascode to stand off the voltage. The voltage may get down to under 20V, but it's hard to say it won't go to B+ on some odd combination of transients and start up or shut down. The low voltage JFET and/or bipolar lets me adjust current closely by changing a resistor.

[paulh]

I get them at Antique Electronics. I have had real good results using transformers as a phase splitter. You can't get much more linear than that.

[aletheian-alex]

My current gigging amp is running a LTPI with a CCS in the tail. The CCS is just a simple BJT cascode with a trimmer to set the current, which is tied to a negative DC supply (about -30v IIRC). I tied in a small "error correcting" network, which is just a fre high impedance resistors from the plates with the junction of them tied into the CCS. Seemed to clean things up a bit.

I ran a bunch of simulation and then built up about a dozen different LTPI circuits and ran real world tests, and as it ended up, that was the one that won out in just about every category. The outputs are nearly identical, the PRSS is better, and it allowed me to crank the current up a bit to set the tube into a sweeter sounding op-point that I had never quite hit with a resistor in the tail. But most importantly, it SOUNDS better... transistor hatred aside.

[Amp Kat]

The CLD may explain why so many MESA's have the switching problems. When the amp heats up it tends to switch channels or crosstalking. I've seen it in several newer amps and thought it was the optoisolators but the newer models don't use the optos it's all relays so maybe they are giving it up ?

[R.G.]

... But most importantly, it SOUNDS better... transistor hatred aside.
It is hard to give up on early hatreds, just like on early loves. Good on you for the open mind.

[aletheian-alex]

It is hard to give up on early hatreds, just like on early loves. Good on you for the open mind.

Thanx R.G. I just go with what works best.

[Arthur B.]

Isodyne (http://www.aikenamps.com/Isodyne.pdf)

From http://www.bonavolta.ch/hobby/en/audio/split.htm#Isodyne:

Pros:
Very linear
Easy to adjust and stable
Low distortion
High gain
Direct coupled
High amplitude output
Low output impedance
Cons:
The tubes must be paired from a twin tube

[Ray Ivers]

Arthur,

Cool! I hadn't seen that one - direct-coupled, too!

Ray

[Zoe_N_Iain]

(Iain) And there was me commenting on 18W the other day that I hadn't seen a guitar amp with an Isodyne! :rolleyes:

Must admit, I like the look of that PI - for some inexplicable reason. And as you say, it's direct coupled.

Zoe's Sound City L-120 is moving up the project queue slowly and we were toying with using an Isodyne in that, just to throw a curve. We want to keep the active tone controls in the preamp, of course. But a change of PI might be fun.

Don't suppose you see an obvious way to fit a variable PI balance control? We're liking the idea of being able to control the balance and, hence, the harmonics generated by the power amp to taste. What's one more knob? :D

[Steve Conner]

I experimented with the see-saw type of PI using a 12AT7. My implementation of it has an adjustable balance control.

http://scopeboy.com/poweramp.html

I never did figure out whether it sounded better or worse than other designs, though. This amp was designed with high-gain metal sounds in mind. I found that it sounded great at medium volumes for recording. But later when I took it out to jam with a drummer who was hitting pretty hard, I had to turn it right up, and I found that it kind of lost that heavy thumping bottom end you need for metal, and went mushy. (It was plugged into a Marshall 1960A IIRC.)

I think, like other posters have mentioned, the fine details aren't important, you just need more power to hold that metal rhythm guitar sound together at gigging volumes. Amps like the Marshall Mode4 and Randall Warhead used several hundred watts of solid-state power.

I think the most awe inspiring metal sound I ever heard was on a recording of Ken Gilbert's Big-Ass Guitar Amp system, which had (iirc) a 400W tube power amp firing through two 4x12"s fitted with EV speakers and stuffed with that acoustic blanket material they put in hi-fi speakers. I almost felt sorry for the poor drummer and bassist trapped in the room with it.

Of course, the best way to get a heavy thumping rhythm guitar sound is to have the bass player double what you're playing. ;-) A lot of good metal riffs seem to have the rhythm guitar, bass and kick drum stomping out the same thing, and you'll never get a guitar played on its own to sound as heavy and tight as this, no matter what you do to the amp.

[Zoe_N_Iain]

(Iain) I've saved that one for future reference. Cheers, bud.

The simple version we designed for a cathodyne/concertina looks like this:

http://www.music-electronics-forum.com/attachment.php?attachmentid=128&stc=1&d=1154780181
varicathodyne1.jpg

Seemed to work Ok in the little test bed we built. Haven't tried it in a complete amp, yet.

[Ray Ivers]

Iain,

The built-in 'Balance adjust' control should work as an AC balance control, but you could also make both plate resistors 180K or something, and fit a 50K pot (probably best to make this a rear-panel, plastic-shaft item) to their ends with its wiper to B+, then compensate for the DC imbalance with the Balance adjust pot (a large cap could be fitted across the 10K bias-circuit resistor to eliminate the AC-imbalance feedback to the input stages).

I've got a Sound City 120 here too - I've always thought it would benefit more from a master-volume installation than any other amp I've seen, as it's got so much gain after the volume controls.

Ray

[Zoe_N_Iain]

(Iain) Thanks, Ray.

Zoe's SC did have a Master Vol in it when she used to gig with it. Stick an overdrive pedal in the front and that amp would scream! Amusingly, the best pedal we ever found to drive it was a no-name POS that she got for £10 brand new from a little back street music shop in Edinburgh. It sounded atrocious through any other amp but absolutely amazing through the SC120.

[Arthur B.]

(Iain) And there was me commenting on 18W the other day that I hadn't seen a guitar amp with an Isodyne! :rolleyes:

Must admit, I like the look of that PI - for some inexplicable reason. And as you say, it's direct coupled.

Zoe's Sound City L-120 is moving up the project queue slowly and we were toying with using an Isodyne in that, just to throw a curve. We want to keep the active tone controls in the preamp, of course. But a change of PI might be fun.

Don't suppose you see an obvious way to fit a variable PI balance control? We're liking the idea of being able to control the balance and, hence, the harmonics generated by the power amp to taste. What's one more knob? :D

The Isodyne already has a balance control. It's between the two CFs.

I experimented with the see-saw type of PI using a 12AT7. My implementation of it has an adjustable balance control.

http://scopeboy.com/poweramp.html

I never did figure out whether it sounded better or worse than other designs, though. This amp was designed with high-gain metal sounds in mind. I found that it sounded great at medium volumes for recording. But later when I took it out to jam with a drummer who was hitting pretty hard, I had to turn it right up, and I found that it kind of lost that heavy thumping bottom end you need for metal, and went mushy. (It was plugged into a Marshall 1960A IIRC.)



I think that's what the cathode followers of the Isodyne were designed to alleviate - low output and poor definition.

[Zoe_N_Iain]

Thanks for that, guys. I think I pretty much understand how to make an Isodyne that can be tuned for balance, now.

I'll draw one up when I get a minute and you can criticise it. :D

[Steve Conner]

Arthur, I don't know about poor output. I checked the amp with a sine wave generator and it puts out a full 50 watts before clipping. (70 with KT88s :eek:) It also seems to clip the same way as other tube guitar amps I've seen: the tops and bottoms of the waveform flatten, and if you keep cranking up from there, the zero crossings grow flat spots from crossover distortion because the grid current pumped the bias too far negative. I concluded from that that the PI can put out more than enough drive for the power tubes.

[Arthur B.]

Arthur, I don't know about poor output. I checked the amp with a sine wave generator and it puts out a full 50 watts before clipping. (70 with KT88s :eek:) It also seems to clip the same way as other tube guitar amps I've seen: the tops and bottoms of the waveform flatten, and if you keep cranking up from there, the zero crossings grow flat spots from crossover distortion because the grid current pumped the bias too far negative. I concluded from that that the PI can put out more than enough drive for the power tubes.

Could it maintain that output with a sgnal of wider bandwidth, like a square or triangle wave?

[Steve Conner]

When I built the amp I only had a sine wave generator. But the dirty channel turned that into a nice square wave, and the power stage seemed to reproduce that OK at levels up to what it could manage with the sine wave. I just bought a new signal generator, so I'll try it with the other waveforms.

When I compare this amp to my other one, a modded Selmer Treble'N'Bass 50, it sounds a lot darker, with less upper mids and treble. I was thinking about this and I remembered that I chose the compensation capacitor in the PI to make it stable with a 12AX7 PI tube and EL34 power tubes. IMO this is the worst case for stability since it has the highest overall loop gain.

I'm now thinking that this is what makes it sound somewhat dark and muddy. Since I use a 12AT7 PI tube now, the loop gain will be less, so the compensation cap (basically 140pF from one PI tube plate to ground) will start rolling off the treble instead of compensating the feedback loop.

But then again, maybe my Toaster amp actually has a flat response, and the tone stacks in the T'n'B have a treble boost even with all the knobs set halfway. Again, something to check out when I get my new generator, I guess :rolleyes:

steve

[Merlinb]

Or just drop in a low ra valve to reduce output impedance to improve driving performance when grid current sets in. 12AY7 is a popular choice to replace the ECC83.

And if you can't get the big negative voltage as discussed above, just use a pentode for a constant currrent sink instead of a resistor tail.

[Guitarist]

I found this circuit which looks like it sounds very clean, simple and no transistors (which are OK but a misunderstood phenomenon). As long as these can drive the tubes to full, clean power it should work as well.

I saw a few builders use this split 12au7 which is very similar to the concertina (is that right) ala Orange but it must be cleaner and better balanced or all these cats wouldn't've used it.

[Arthur B.]

I found this circuit which looks like it sounds very clean, simple and no transistors (which are OK but a misunderstood phenomenon). As long as these can drive the tubes to full, clean power it should work as well.

I saw a few builders use this split 12au7 which is very similar to the concertina (is that right) ala Orange but it must be cleaner and better balanced or all these cats wouldn't've used it.

It's yet another variation on the Williamson amplifier. This time, the feedback is applied to the concertina as well.

[kg]

I think the most awe inspiring metal sound I ever heard was on a recording of Ken Gilbert's Big-Ass Guitar Amp system, which had (iirc) a 400W tube power amp firing through two 4x12"s fitted with EV speakers and stuffed with that acoustic blanket material they put in hi-fi speakers. I almost felt sorry for the poor drummer and bassist trapped in the room with it.

awww, you're making me blush. ;)

fwiw, the driver circuit of the baga was pretty straight-forward: 1x 6n1p diff amp, sourced from symmetric split rail power supplies, plain old resistive cathode and plate loads, cap coupled to 4x el84 cathode followers, directly coupled to 4 banks of 3 kt90s each. the el84 cf's have constant current sinks in each of their cathodes.

in this way you could feed the input stage either a single ended or differential signal.

ken

ps don't feel sorry for the drummer--he's the reason i needed a big ass guitar amp in the first place.

edit: baga preamp schemo: http://ken-gilbert.com/images/pdf/BAGAcfdriverschema.pdf

[Steve Conner]

OMG, it's the man! I'm not kidding about the BAGA, it reminds me of someone gunning a V8 engine with no mufflers. :D

For a while I was all inspired to build a bass version of it, but it turns out the band I play with just now is quiet enough that I can use my old Toaster head for bass. I got a big 2x15" high efficiency cab to help it out. I can get the meter showing about 150mA of plate current before the singer starts complaining the bass is too loud. :rolleyes:

This is kind of relevant to the thread I guess because the Toaster head is the one that uses the see-saw PI I posted earlier. I'm not sure the output is exactly "Clean" but it seems to have plenty. I think KT88s are fairly low gain and need a lot of grid drive, so my PI must be doing all right.

I went back to EL34s though. I tried the KT88s driving an 8 ohm speaker with the 16 ohm tap, and I never really got much more power, because the B+ just sagged drastically. I was expecting 100w but I only got about 70. Using a 16-ohm load they didn't really do anything that 6L6s or EL34s couldn't.

[kg]

Using a 16-ohm load they didn't really do anything that 6L6s or EL34s couldn't.

well, kt88s do tend to drain the wallet a bit quicker, don't they? ;)

[Steve Conner]

Well, I thought so but it turns out the Svetlana 6550Cs and Shuguang KT88s aren't too pricey. A pair of these tubes costs about $10 more than a quad of 5881WXTs and does about the same power.

I found out that 100 watt hi-fi grade output transformers sure do drain the wallet though :eek:

[Smitty]

Bruce,

Is that the chi town copy of the Musicmaster Bass driver transformer? I wonder if that's got enough oomph with a 12AU7 running in parallel to drive the grids of EL34s just enough positive to round off the clipping I'm getting with a convnetional LTP. I'm wondering how far up fc goes during conductance.

I'm also looking at the monster from the 300 PS. Although it would work nicely, it seems like overkill. I really dig the split secondary. I like dual bias.

Perhaps I'm looking for Baby Bear's Driver

Goldilocks