crazy high rail voltage = better????

[clintrubber]

There's an old AES article that runs BJTs at high voltages but the aim there was to mimic hollow state euphonics.

I recall Rafa (who was/is around here) was a BJT@HT-experimenter, but I don't know if he came to firm conclusions.


Bye,

  Peter

 

[Winston OBoogie]

There's an old AES article that runs BJTs at high voltages but the aim there was to mimic hollow state euphonics.

Not talking about euphonics, that's for the 12AX7 crowd.  I'm talking about having a solid state stage that has products that are low to start with and then are brought a bit lower by small amounts of feedback.  I don't think BJT's are the way to go on this btw.
 

Anyway, measure a good valve sometime, it isn't euphonic unless pushed really hard.  Unfortunately, people are programmed into believing that the 12AX7/12AU7 sound is the sound of tubes.  A good valve will have harmonics that are 12 - 20dB lower than a 12A**.

I recall Rafa (who was/is around here) was a BJT@HT-experimenter, but I don't know if he came to firm conclusions.

He's around, we emailed a few times in the last week.

Cheers.

 

[PRR]

> If your voltage rail is large compared to your signal voltage

Emphasis "compared to".

I have a 600V amp, and it has plenty of nonlinearity, 'cuz it swings 550V.

Current over-kill also helps.

On hollow triodes, THD is almost proportional to signal/supply voltage ratio. On BJTs and many FETs, gross excess voltage isn't much help. BJT current linearity is awful, tho.

The product (multiply) of voltage and current is Power. Which can also be spelled C-O-S-T. Bigger devices, bigger heatsinks, bigger PTs, bigger boxes/racks, bigger A/C, and if crazy enough, a dent on yor electric bill.

Unless your needs are very modest, or your money is unlimited, hi-V design carried too far is usually not the best use of resources.

 

[JohnRoberts]

If your voltage rail is large compared to your signal voltage, and you use appropriate devices, your linearity is better.  You need less feedback.  Use all the tricks you can to linearize each stage by itself and use less feedback... 
I realize I'm probably pissing into the wind here.

I'm afraid old age has made me more practical.

I suspect pushing the absolute performance limits of DOAs would be improved more by clever circuit topology and good devices than some brute force high voltage PS, or "less is more" philosophy (out out damn evil negative feedback).

In the old APi consoles there was some merit in having X dB more headroom on the bus when mixing down lots of channels, but a full high voltage DOA path is IMO more of a marketing thing these days.  Modern electronics with much lower noise floors than back in the old days can get you similar dynamic range without the heavy lifting,

Not pissing in the wind... By all means if you have an approach that you think will improve the SOTA, pursue it. Build it and measure it. There are bench tests that can confirm or deny the result, while it is getting progressively more difficult to measure the new high performance silicon. The linearity is so good you have to run them at a ton of gain** just to see anything on the best modern bench measurement systems.

I remain in awe of the new high performance chips. Some seriously good stuff, available off the shelf. And I'm jealous because they weren't available 30 years ago when I had some good applications for them, but that's life.

JR


** I sometimes wonder if they game the designs knowing that high gain is the only way people can measure them?  Of course who's going to know if there's some spurious unity gain instability down around -140 dB.....  I can be a little skeptical  ;D

 

[Winston OBoogie]

Current over-kill also helps.

Yes. 

On BJTs and many FETs, gross excess voltage isn't much help. BJT current linearity is awful, tho.

As I said, you need the right devices. 
I figure out my typical signal voltage first and then size the rail accordingly.
I generally don't see any improvement beyond a certain point and it's then that I use other means to fudge it.

Unless your needs are very modest, or your money is unlimited, hi-V design carried too far is usually not the best use of resources.

80V single-sided or +/-40 Bipolar have done it for me.

 

[Winston OBoogie]

I suspect pushing the absolute performance limits of DOAs would be improved more by clever circuit topology and good devices than some brute force high voltage PS,

I've tried lots of clever designs, most of them not mine.  I do use good devices.  Also, I don't generally use a 'one size fits all' op-amp.  If I need to convert impedance only, why do I need a device with a diff. pair, a V.A. and a current amplifier? + all the other doo-dats that come with the territory. 
I'll add that I don't just, willy-nilly, throw money and voltage at something and expect it to be better.

(out out damn evil negative feedback).

It isn't evil.  But I think more can be had by using less of it.

For each stage, and for the amp as a whole, I try to use principles brought to some folks' attention by Matti Otala many years ago. 

By all means if you have an approach that you think will improve the SOTA, pursue it. Build it and measure it.

I've spent more time than I probably should have building and measuring.  Time that should have been spent getting paid for 'run o' the mill stuff' 

There are bench tests that can confirm or deny the result,

When I'm happy, my bench tests generally agree with my ears in terms of my preference.  Since I don't have the best ears in town, I also borrow those of a few folks I respect. 


Cheers John,

John

 

[Winston OBoogie]

...I realize I'm probably pissing into the wind here.

No, you are not.

Fred Forssell

Thanks Fred.  Nice to see ya 

 

[Samuel Groner]

If your voltage rail is large compared to your signal voltage, and you use appropriate devices, your linearity is better.

This needs qualification. There are several basic distortion mechanisms in semiconductors and not all of them benefit from higher supply rails. Anything which is associated with voltage dependence of Early effect and junction capacities does; anything which is related to basic transconductance behaviour is only a very week function of supply voltage (sufficiently below clipping, of course). Same applies for collector current dependence of hFE or similar parameters. Distortion from thermal self-heating or thermal feedback even increases with higher supply voltages.

While I do agree that proper use of high supply voltages can lead to excellent results (see e.g. my SGA-HVA-1 opamp) I'm hesitant to sell high voltages as cure-it-all-and-not-obvious-for-the-deaf-only means.

Samuel

 

[abbey road d enfer]

Marketing...

BTW - there are small-signal BJTs that go up to 120v around from firms such as Sanyo.

On a complementary / PP design, you could run them @ +/-120v!!

No. You can run them at +/-60V. Let's take tha case of a basic push-pull with an NPN at the top and a PNP at the bottom. When the output is idle, each transistor is subjected to 60V. But at the top or bottom of a full-swing sinewave, each transistor is subjected to the whole differential voltage, I.E. 120V.

 

[Samuel Groner]

You can run them at +/-60V.

I really hope we do not need to re-discuss the entire topic--see my first post on page one.

Samuel

 

[Winston OBoogie]

If your voltage rail is large compared to your signal voltage, and you use appropriate devices, your linearity is better.

This needs qualification.

Hi Samuel.
My qualification is that I said "and you use the appropriate devices'.

I will also add again that I don't think BJT's are the way to go.  With the caveat, by themselves.

First of all, if we're dealing with an open loop stage, it is nigh on impossible to get a BJT to have low levels of higher order harmonics.  If we apply local FB to linearize it, then our even order products are also converted to high order products.  We are dealing with a device that has non linear beta, non linear voltage gain etc., etc.

My goal is usually to keep the number of singularities (stages) in the signal path to a minimum.  I also aim for high bandwidth (as stated by Otala and others).  And I also want fairly low levels of harmonics that are benign in nature - the low orders.  

Now, just because I am working with fewer stages, does not mean that these stages cannot be compound ones.

If you cascode a good J-fet with a fairly beefy transistor, and pick the appropriate operating points, you will find that the harmonic spread is fairly benign.  Of course, the cascode takes care of the wide bandwidth.  And also the Early effect in the BJTs and Cgd modulation in the J-FET

If your BJT is an appropriate one, the voltage supply rail can be made quite high, which pushes down the already quite benign harmonics.

Should you wish to cancel more even orders, you could extend your compound stage to a complimentary one.  

If you then use a judicious amount of feedback, your distortions will be a bit lower yet but will still be monotonic, which is much better in my opinion.

I'm hesitant to sell high voltages as cure-it-all-and-not-obvious-for-the-deaf-only means.
Samuel

So am I.  Again, as I stated, I don't just willy-nilly throw high voltage at something and expect it to be better.  I will add that I'm not necessarily looking for extra headroom in the classic sense, but headroom in terms of keeping the distortions very benign when operating at normal system levels.


Cheers.

 

[JohnRoberts]

For each stage, and for the amp as a whole, I try to use principles brought to some folks' attention by Matti Otala many years ago. 


Cheers John,

John

Perhaps I didn't give him a fair shake but I was not very impressed with Otala's contributions.. It seemed much ado about slew rate, and some interfacing phenomenon(? I forget what he called it) which competent designers were, or should have been, aware of. I used to joke that I was going to propose "AID" (amplitude induced distortion) to counter the new slew of SID (slew induced distortion) tests.  Of course AID is just clipping, so for any reasonably designed circuit paths it's a go/no go test. The circuit is clipping or not. Likewise for slew limiting the well designed circuit should be slew limiting or not (and Leach published a short but IMO classic AES paper on how to avoid slew limiting).

I invested some time and killed some brain cells trying to quantify a more meaningful slew rate metric for design, something like output rate of change, vs. input differential voltage, or better yet, output rate of change linearity vs. input differential but folks have enough trouble with the basic amplitude domain linearity measures (THD and IM)  and we can parse out a lot of information by properly applying them (as evidenced by Sam's work).

Actually circuits shouldn't be clipping or slew limiting. Power amps often specify output power at X% THD to provide some comparable amount of amplitude clipping. It might be interesting to use a similar approach when specifying slew rate... More meaningful IMO than measuring the slew rate of a signal overdriven to square wave, is slew rate at X% THD.  FWIW I don't even like specifying slew rate. Well designed audio paths IMO should exhibit a rise-time characteristic when driven too fast not clip it's rate of change.

Sorry if I am veering too far off topic...  To bring this back almost on topic, there is a relationship between rail voltages and several mechanisms, but a circuit's performance is influenced by more things than just that. What matters is the final result, not the path taken to get there.

JR 
 

 

[Winston OBoogie]

Perhaps I didn't give him (Otala) a fair shake but I was not very impressed with Otala's contributions.. It seemed much ado about slew rate...

PIM, TIM, JIM, HIM and VIM.  And now, AID 

There have been others that have continued along the same path with these goals though and some of them have designed some mighty fine amplifiers.

...there is a relationship between rail voltages and several mechanisms, but a circuit's performance is influenced by more things than just that. What matters is the final result, not the path taken to get there.

JR 

Well, I can't argue with your last point.  If there is no perceivable difference, then it does not matter what devices or topology we have used.

When someone puts on a CD or record, they probably aren't thinking: "On that vocal, they should have used a mic amp with better slew rate!"

However, I do know people who have been able to detect proportionally high levels of high order harmonics that are below the noise floor...

 

[JohnRoberts]

Well, I can't argue with your last point.  If there is no perceivable difference, then it does not matter what devices or topology we have used.

When someone puts on a CD or record, they probably aren't thinking: "On that vocal, they should have used a mic amp with better slew rate!"

However, I do know people who have been able to detect proportionally high levels of high order harmonics that are below the noise floor...

Sure, just like we can hear signal below the tape hiss..  This is why we have modern codecs with bit quantization way below the analog noise floor. 

It seems some people's hearing goes in the crapper when they participate in double blind ABX tests... I know better than to trust my ears for more than broad strokes. If it sounds bad, it is surely bad, if it sounds good, I still measure it..

JR

 

[Winston OBoogie]

It seems some people's hearing goes in the crapper when they participate in double blind ABX tests... I know better than to trust my ears for more than broad strokes. If it sounds bad, it is surely bad, if it sounds good, I still measure it..

Like I said, my ears are not the best so I borrow some.
A business partner of mine has good ears (recently won an engineering Grammy) and, while he may not know the 1st thing about circuits or care less how it's done, he can tell when things are a little off kilter.  
Of course, I would measure and be reasonably happy before I even wasted his time with listening for nuances.

 

[Winston OBoogie]

In regard to the measuring of non-linearity and THD spec, here's the 1989 Daniel Cheever paper on the correlation between spectra of harmonics and perceived quality of sound.   I'm sure a few have read it years ago but, if not, it's worth a glance in my opinion

It is presented as applies to power amplifiers but I find the same applies for line and pre stages too.
In other words, low levels to zero levels of negative feedback around linear (benign low order partials), wide bandwidth singularities, would seem to be preferred by quite a few listeners over other topologies that might have vastly superior specifications on paper when tested by the usual means and methods.

http://www.next-tube.com/articles/Cheever/cheever.pdf

 

[abbey road d enfer]

I think we drifted away from the original topic, which was basically: How much rail voltage do we need to properly handle +4dBu nominal? with how much crest factor/headroom?

 

[Winston OBoogie]

I think we drifted away from the original topic, which was basically: How much rail voltage do we need to properly handle +4dBu nominal? with how much crest factor/headroom?

Maybe.  Although I posted the article as supplemental reference for my earlier posts in which I had given some reasons for why I think the use of higher rails is sometimes beneficial.

If the mods consider it off topic or my posts irrelevant, they may remove them 

Cheers.

 

[abbey road d enfer]

Not what I meant. In fact I think these two different aspects of the subject deserve two different threads. I still think that, if ADL/SPL/Neve need "90v/120v/600v technology" to handle line-level signals, they should go back to school. Indeed I know they are very competent, so my conclusion is that their designers have been taken hostage by the marketing department. At a more pedestrian level, I have never seen practical advantages in more than +20dBu output capability.

 

[desol]

Indeed I know they are very competent, so my conclusion is that their designers have been taken hostage by the marketing department.

hMM. Not sure how fair it is to speculate here.

From what i've heard Rupert say in the past, i don't think he's the kind of guy that would, how shall we say it, compromise his own integrity with such foolishness. How much time do you think he sets aside to waste?

Money is money tho.

It's clear he believes his approach makes a difference...and if it does, that's really what matters.
I simply appreciate the fact that he's doing what he is...god bless em.