For a smaller amp (25V rails), why not use an IC with emitter coupled pairs and get rid of the emitter resistors in the design? I'm thinking about trying this out and seeing what the result is for the hell of it.
For example:
http://www.nxp.com/documents/data_sheet/PMP5201V_G_Y.pdf
For example:
http://www.nxp.com/documents/data_sheet/PMP5201V_G_Y.pdf
Many amps use(d) undegenerated LTP's. Now, the trendy doctrine is to kill as much open-loop gain as possible, using degeneration resistors for example, but you don't need to follow every new silly trend....
A word about this trend: it is probably a result of the "Sheckley effect".
Robert Sheckley was a science fiction writer, and in one of his novels, he postulated that something that has huge and obvious inconveniences must also have as huge, but hidden benefits, because of some "universal compensation principle" (I am paraphrasing him, I don't remember his exact words).
Since large loop gains and FB ratios have so obvious and measurable benefits, this means that they must also have some hidden, crippling defects.
By contrast, if you let the main benefits of a topology go down the drain, you necessarily have to have some great reward in return. That's the logic anyway, but in the audio field it is taken for granted by many people
Well, matching the pair is not the key point here - it's much more complex and, let say, "fundamental". Degeneration resistors act a as a local negative feedback for the long tail pair, linearising it, increasing its bandwidth, reducing noise, but also reducing its gain. Now, the question is - do you want to have maximum gain for your LTP and then control it with the global feedback loop (there may be some issues, associated with this approach, including stability issues), or you want to reduce the gain locally to some particular level, and have the global loop less "stressed". Decision also depends on the other stages' gain levels. As Elvee just mentioned, there are different approaches to the balance between the global loop gain and the local ones, so this is something, particular developer decides for particular design. Setting the right combination of gains for the stages within the global feedback loop, in many cases allows better transients handling, influencing transient inter-modulation distortion.
So, again - it's much more complex, than just a matching issue.
Cheers,
Valery
Chips like this have long been used for RF projects, both with external active taol loading and passive.
Same deal with discrete differential amps: some include emitter resistors and some don't. The whole point of having at least some of the external emitter resistance unbypassed is the swamping out of the nonlinear re. It costs some gain, but improves linearity.
Whether it makes a difference in any one design depends. It becomes another case of try and listen, see what/if there's a difference.
Distortion In Power Amplifiers gives a quick presentation of diff pair linearity and degeneration
http://www.ti.com/lit/an/snoa737/snoa737.pdf gives a hard limit for slew/TIM without degeneration in a Miller dominant pole compensated amp
in power amps today the typical case is to bias "hot" - above the noise optimum Ic and use degeneration R as Self describes
it is also possible to use alternative compensation that avoids the Solomon paper limits
particularly with today's 2 Vrms consumer digital source 0 dB full scale standard the input pair noise optimization isn't as important in reasonable home audio power amplification
without degeneration you have to limit input Vdiff to ~100 uV to keep the undegenerated diff pair tanh distortion contribution below -120 dB
this requires high loop gain - really beyond the RCA/Bailey Miller dominant pole compensated "standard" amplifier circuit
http://www.ti.com/lit/an/snoa737/snoa737.pdf gives a hard limit for slew/TIM without degeneration in a Miller dominant pole compensated amp
in power amps today the typical case is to bias "hot" - above the noise optimum Ic and use degeneration R as Self describes
it is also possible to use alternative compensation that avoids the Solomon paper limits
particularly with today's 2 Vrms consumer digital source 0 dB full scale standard the input pair noise optimization isn't as important in reasonable home audio power amplification
without degeneration you have to limit input Vdiff to ~100 uV to keep the undegenerated diff pair tanh distortion contribution below -120 dB
this requires high loop gain - really beyond the RCA/Bailey Miller dominant pole compensated "standard" amplifier circuit
" . . . Now, the trendy doctrine is to kill as much open-loop gain as possible, using degeneration resistors for example, but you don't need to follow every new silly trend . . ."
Degenerating the emitters of an LTP has everything to do with good compensation design and nothing to do with any 'silly new trend'.
If you take a typical undegenerated input pair and plot Ic vs Vdiff you get a very narrow region where the relationship is linear. By degenerating the emitters, the linear region opens up. Additionally, assuming a ULGF of 1~2 MHz, you can design for higher slew rates (MC assumed) - important in power amplifiers where a good design rule is a minimum of 1V/us per volt of peak output
If you take a look at me e-Amp write up on my website I address this issue. Ovation e-Amp: A 180 Watt Class AB VFA Featuring Ultra Low Distortion
Its important to note that in the case of a small signal amplifier, and where noise performance is key, degeneration is normally avoided and because for most applications you do not need huge slew rates. In generic MC VFA opamp designs, you will typically see SR of 10~20 V/us in as a result.
While the most important thing for me personally after taking care of my family is to encourage you all to buy as many NXP products as possible, I can only recommend the linked NXP part(s) for small signal applications when applied to differential amplifiers.
Degenerating the emitters of an LTP has everything to do with good compensation design and nothing to do with any 'silly new trend'.
If you take a typical undegenerated input pair and plot Ic vs Vdiff you get a very narrow region where the relationship is linear. By degenerating the emitters, the linear region opens up. Additionally, assuming a ULGF of 1~2 MHz, you can design for higher slew rates (MC assumed) - important in power amplifiers where a good design rule is a minimum of 1V/us per volt of peak output
If you take a look at me e-Amp write up on my website I address this issue. Ovation e-Amp: A 180 Watt Class AB VFA Featuring Ultra Low Distortion
Its important to note that in the case of a small signal amplifier, and where noise performance is key, degeneration is normally avoided and because for most applications you do not need huge slew rates. In generic MC VFA opamp designs, you will typically see SR of 10~20 V/us in as a result.
While the most important thing for me personally after taking care of my family is to encourage you all to buy as many NXP products as possible, I can only recommend the linked NXP part(s) for small signal applications when applied to differential amplifiers.
Last edited:
It is actually a silly new trend, you only have to look at extreme examples, like zero GNFB designs where degeneration is used at the maximum tolerated by the overall gain
If you adhere to certain topologies and compensation strategies, a (reasonable) degeneration can be necessary, but for some, degeneration is "a good thing", an end in itself, and is used well beyond any technical necessity.
It is true that degeneration opens up the linear region of an LTP, but it is also true that it mechanically reduces its transconductance by approximately the same amount, and thus the loop gain of that amplifier, globally.
Which means that in the end, the error signal the LTP has to process will also be increased by a similar amount.
It may not be a zero-sum game, but even so it will remove part of the loop gain available to correct non-linearities elsewhere in the amplifier.
In an amplifier, gain and non-linearities do not follow the same rules of addition (or convolution). Adding the gains simply means adding the logarithms, but for non-linearities, things are somewhat different, and it is closer to a quadratic mean (depends on a number of parameters).
For example, if you cascade two gain blocks having each 20 dB and 1% THD, the composite will have a gain of 40 dB and a (worst-case) THD of 2%.
If you apply GNFB and reduce this gain to 20dB again, the THD will have dropped to 0.2%.
If you chose instead to degenerate each stage separately to arrive at the same overall gain, the THD of each 10dB stage will be 0.3%, and the overall one 0.6%.
This is of course is grossly simplified, but it shows the trend, and it shows why degeneration should be used sensibly.
It is also obvious that applying large FB factors is not "free" technically speaking, this would be too easy. There is an optimum tradeoff somewhere, and it will very rarely be either massive degeneration, or massive GNFB
Elvee, I don't recall using any terminology that implied degeneration should not be used sensibly.
Read the link to Self's webpage in jcx's post above. That explains succinctly why (sensible) degen is a good thing.
As for low or zero GNFB designs, each to his own. There's a place in the sun for everyone's topolgy and design philosophy. See Cordell for a good example of this latter genre.
Read the link to Self's webpage in jcx's post above. That explains succinctly why (sensible) degen is a good thing.
As for low or zero GNFB designs, each to his own. There's a place in the sun for everyone's topolgy and design philosophy. See Cordell for a good example of this latter genre.
We'll I'm giving this a try....
I'll let you guys know how it goes.
An externally hosted image should be here but it was not working when we last tested it.
I'll let you guys know how it goes.
Many IC op amps (especially older ones) did not include emitter degeneration resistors in order to maximize their open loop gain and dc offset specs. Mismatch in the emitter degen resistors can be a source of dc error in op amps, but is probably less of a concern in audio power amplifiers. What I think would be more interesting would be an audio power amplifier that employed either multi-tanh linearization of the input pair (Barrie Gilbert: IEEE Xplore Abstract - The multi-tanh principle: a tutorial overview ) or a slew boosting technique that modern op amps use to preserve input stage linearity for greater input differential voltage (increasing tail current with input differential voltage)
On the other hand, a major benefit of using a transistor array like the one you show is that the two transistors are thermally coupled and hopefully will drift together as the amplifier self heats. And I do think matching the two has a benefit, the even order harmonics of a differential pair only cancel if the two transistors have identical transfer functions.
On the other hand, a major benefit of using a transistor array like the one you show is that the two transistors are thermally coupled and hopefully will drift together as the amplifier self heats. And I do think matching the two has a benefit, the even order harmonics of a differential pair only cancel if the two transistors have identical transfer functions.
Holes drilled (I know, crappy job) and SMD parts mounted
An externally hosted image should be here but it was not working when we last tested it.
The JE990 uses a matched pair. This description from the designer is interesting:
http://www.technicalaudio.com/pdf/J...ted/Jensen_JE-990_opamp_JAES_reprint_1980.pdf
http://www.technicalaudio.com/pdf/J...ted/Jensen_JE-990_opamp_JAES_reprint_1980.pdf
Hi Guys
Sreten's post-10 proves what Elvee said in post-9 rather than counter it.
Elvee is right.
Just because degeneration has benefits and is ubiquitous in power amp designs, which themselves are pretty much the same tired design, does not negate the benefits that non-degenrated diff amps or gain stages offer. There are world-class designs that use no degeneration on PA or preamp diffs and have performance that an AP can barely measure.
Have fun
Sreten's post-10 proves what Elvee said in post-9 rather than counter it.
Elvee is right.
Just because degeneration has benefits and is ubiquitous in power amp designs, which themselves are pretty much the same tired design, does not negate the benefits that non-degenrated diff amps or gain stages offer. There are world-class designs that use no degeneration on PA or preamp diffs and have performance that an AP can barely measure.
Have fun
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.