I ran into this on the web. I'm not an engineer just an audiophile so I'm taking a big chance in posting this, so be kind. But I have never seen this topology used anywhere except here. I'm referring to the complementary long tail pair TR1 and TR2 used in this schematic. Why is this config not used more and can it be used as a constant current output stage?
4QD-TEC: Low distortion Audio amplifier
4QD-TEC: Low distortion Audio amplifier
Complementary long tail pair (tall less pair)
I ran into this on the web. I'm not an engineer just an audiophile so I'm taking a big chance in posting this, so be kind. But I have never seen this topology used anywhere except here. I'm referring to the complementary long tail pair TR1 and TR2 used in this schematic. Why is this config not used more and can it be used as a constant current output stage?
4QD-TEC: Low distortion Audio amplifier
I ran into this on the web. I'm not an engineer just an audiophile so I'm taking a big chance in posting this, so be kind. But I have never seen this topology used anywhere except here. I'm referring to the complementary long tail pair TR1 and TR2 used in this schematic. Why is this config not used more and can it be used as a constant current output stage?
4QD-TEC: Low distortion Audio amplifier
I've used it a few times. It works, but you have to dance around it if you want good overload behavior. I never experienced it to cause bad sound. It doesn't cancel Early voltage effects like the LTP so it can be a source of load-independent distortion. I used it here:
http://www.diyaudio.com/forums/head...cascode-headphone-amp-jlh-output-stage-8.html
Keep in mind I was 16 or so then, so a lot of things have changed. I still want to revisit the circuit but I'm not sure what direction to take yet.
http://www.diyaudio.com/forums/head...cascode-headphone-amp-jlh-output-stage-8.html
Keep in mind I was 16 or so then, so a lot of things have changed. I still want to revisit the circuit but I'm not sure what direction to take yet.
Sorry another beginner question...
Please could you explain what is meant by "inert" decoupling?
Paul
When you parallel a 100n cap with a reservoir lytic, the small cap resonates with the loop inductance through the lytic. This is a parallel resonance, meaning a severe impedance spike. At RF, low PSRR combines with rail impedance to make output phase dependent on rail impedance, so when your ULGF is high relative to PSRR, resonant rails can cause your amp to oscillate. There is also the effect of lead inductance resonating with power transistor capacitances, and this can also disrupt phase at high frequencies.
Good point. I've investigated the problem and it can be a serious concern. See page 350 in my book.
Cheers,
Bob
I've also been investigating, a great deal more crudely, sure, but the soldering iron and much effort was involved.
Most of us are familiar with using 1 of 250v capacitor sized at 1% of the decoupler cap size like this: 470u decouplers and then there is 1 (one) of Cornell Dublier's Mallory SEK 250v 4.7uF capacitor from V+ to V- at the amplifier board.
*common name is rail to rail cap
Benefit: slightly tamer midrange, and slightly cooler amplifier.
HOWEVER, last year I removed that sort of thing and replaced it. . .
More effective alternative:
A diode series with V+ and a diode series with V- and the locale is where the DC cable attaches to the amplifier board. Sizing the diode forward voltage drop depends on the decoupler capacitance, sort of like this:
If decouplers are 100u, then use MBR730 for series elements
If decouplers are 220u, then use MBR1645 for series elements
If decouplers are 330u, then use MUR820 for series elements
If decouplers are 220u||220u (440u), then use 6A05 for series elements
If decouplers are 470u||470u (940u), then use 10A1 for series elements
*If the power supply is also onboard with the amp, so that the DC cable doesn't exist, then as replacement, ferrite beads can be placed over the pin of the series element diodes at whichever side faces away from the amplifier.
*It takes 4 diodes for stereo split rail amplifier and it can have enhanced stereo separation as well.
*common name is virtual dual mono, "VDM"
Anyway, I bought a lovely 4 pack of 6a05's (used for dc power series elements located at the amplifier board) fetched from the local shopping mall, overpaid radioshack for the 6a05's, paralleled up some 220u caps for my decouplers to be low loss 440u per each rail, and had a most excellent experience with the audio quality.
I got a more open sound, tone problems vanished, I was able to set the amplifier gain *slightly* lower, the usable amount of peak dynamic power increased, the stereo separation worked better, and the audio seemed much clearer, which was a nice surprise considering the highly palatable tone.
This bit of simplicity with inexpensive diodes worked ever so much better than excessive cap and cable selection labor.
P.S.
Weirdly, the diodes mentioned can work great on DC power, or very poorly on AC power (doesn't work for rectifier). So, indeed I was using a DC power board which has FairChild Stealth that work great on AC but terrible on DC (doesn't work for the filter). That's just funny.
SO, if you have some diodes that made the worst bridge rectifier ever, you can try re-employing them for my DC filter since it won't be doing any switching.
Most of us are familiar with using 1 of 250v capacitor sized at 1% of the decoupler cap size like this: 470u decouplers and then there is 1 (one) of Cornell Dublier's Mallory SEK 250v 4.7uF capacitor from V+ to V- at the amplifier board.
*common name is rail to rail cap
Benefit: slightly tamer midrange, and slightly cooler amplifier.
HOWEVER, last year I removed that sort of thing and replaced it. . .
More effective alternative:
A diode series with V+ and a diode series with V- and the locale is where the DC cable attaches to the amplifier board. Sizing the diode forward voltage drop depends on the decoupler capacitance, sort of like this:
If decouplers are 100u, then use MBR730 for series elements
If decouplers are 220u, then use MBR1645 for series elements
If decouplers are 330u, then use MUR820 for series elements
If decouplers are 220u||220u (440u), then use 6A05 for series elements
If decouplers are 470u||470u (940u), then use 10A1 for series elements
*If the power supply is also onboard with the amp, so that the DC cable doesn't exist, then as replacement, ferrite beads can be placed over the pin of the series element diodes at whichever side faces away from the amplifier.
*It takes 4 diodes for stereo split rail amplifier and it can have enhanced stereo separation as well.
*common name is virtual dual mono, "VDM"
Anyway, I bought a lovely 4 pack of 6a05's (used for dc power series elements located at the amplifier board) fetched from the local shopping mall, overpaid radioshack for the 6a05's, paralleled up some 220u caps for my decouplers to be low loss 440u per each rail, and had a most excellent experience with the audio quality.
I got a more open sound, tone problems vanished, I was able to set the amplifier gain *slightly* lower, the usable amount of peak dynamic power increased, the stereo separation worked better, and the audio seemed much clearer, which was a nice surprise considering the highly palatable tone.
This bit of simplicity with inexpensive diodes worked ever so much better than excessive cap and cable selection labor.
P.S.
Weirdly, the diodes mentioned can work great on DC power, or very poorly on AC power (doesn't work for rectifier). So, indeed I was using a DC power board which has FairChild Stealth that work great on AC but terrible on DC (doesn't work for the filter). That's just funny.
SO, if you have some diodes that made the worst bridge rectifier ever, you can try re-employing them for my DC filter since it won't be doing any switching.
Last edited:
Yes, the separation is noteworthy but not the star of the show.
Given the series diode's forward voltage drop (prior to the amp board decoupling caps), that is also a cousin of a CRC filter, except that the locale is much different so it doesn't hinder charging your power supply reservoir (aka doesn't hinder bass nor cost a larger transformer).
I believe that it was the filtering that helped the amplifier stability and thereby allowed me to set the amplifier gain lower.
P.S.
Due to bizzare furniture arrangement, I was actually using Monophonic, so in my case, I didn't install the filter for stereo separation.
Last edited:
I wonder why this question of channel separation is so often pointed. Most of the time, with pan pot stereo mix, there is TOO MUCH separation and a little crosstalk is not something to be fear of.
If the sound mixer was crazy enough to put an instrument 100% left or right in the stereo landscape, it is something not natural, and this little crosstalk will bring a more natural stereo image.
And if the stereo position of an instrument was made of a mix of phase and level (mic couples) differences, 40db of separation will not make any difference in the stereo image.
More important, on my point of view is the stability of this image. AS the PSUs tend to decrease their voltages during high levels transients, dual mono configurations will have the rails of the two channels reduced not the same amount. This can produce a variation of the subjective position of the instruments, and a single PSU block for two channels will bring a more stable image position.
More than this, as the reservoirs will be the double of the one requested by a mono block, you will have more reserve for the instruments situated at the sides, means more 'subjective' separation.
Reason why i prefer a single power unit for both channels for all my amps.
If the sound mixer was crazy enough to put an instrument 100% left or right in the stereo landscape, it is something not natural, and this little crosstalk will bring a more natural stereo image.
And if the stereo position of an instrument was made of a mix of phase and level (mic couples) differences, 40db of separation will not make any difference in the stereo image.
More important, on my point of view is the stability of this image. AS the PSUs tend to decrease their voltages during high levels transients, dual mono configurations will have the rails of the two channels reduced not the same amount. This can produce a variation of the subjective position of the instruments, and a single PSU block for two channels will bring a more stable image position.
More than this, as the reservoirs will be the double of the one requested by a mono block, you will have more reserve for the instruments situated at the sides, means more 'subjective' separation.
Reason why i prefer a single power unit for both channels for all my amps.
Last edited:
Well, it seems that you're right. I tested the little filter in stereo and again in monophonic (my main system is mono), and have to report that separation wasn't the point. Perhaps, diode based battery isolator is closer to the point?
The point of post#5628 was power filtering decoupling, for better stability.
Edit: or at least easier stability.
The point of post#5628 was power filtering decoupling, for better stability.
Edit: or at least easier stability.
Last edited:
Yes, Daniel, the evils of resonances resulting from different value/technology paralleled caps are now well known. Paralleling a little film cap with a big lytic one can help as well as destroy stability, depending of the frequency where the resonance occurs and how a particular amp behave at this frequency.
As the solutions of any particular situation can be numerous, (paralleling little value of lytic caps, dumping resonances with little resistances or length of tracks, serial diodes etc.) i just not wanted to enter in this
Correct engineering is to understand the problems and its causes and find a solution in any particular situation. As there is no perfect answer or solution, we just have to find the least bad in *OUR* situation and not consider this solution as a rule
Last edited:
We can do a pretty decent quantitative job on outside loop stability for a dominant pole Miller compensated amp.
The other instabilities are different loops but are subject to the same criteria of loop gain and phase, and so should be understandable and controllable in the same way. But few people seem to have done much on this, I want to understand it so it is under control, not just accept a rule of thumb.
I have started to work on this. Not sure yet which are the crucial inductances or other parasitics, hence my question. Any ideas?
Yes, I recommended that site to someone else when they asked a similar question earlier. Need to study it more. I like Dennis Feucht's work too, any other links?
Best wishes
David
Sorry to be a little slow to reply. I have a back up of mail to deal with due to the collapse of my internet provider.
Dan's solution does help with the sound, but is not a good solution to the local resonance, because the inductance through the diode loop is usually too large, and the resistance of the diode changes depending on load. It does however break some resonances that may occur with the power umbilical. It shouldn't be confused with decoupling resonance, which is a liability for amplifier stability and may not respond to the diode trick.
So if you want to have rails that promote stability of the amplifier, you need to have a small-value lytic (Pana VR 100V 47uF is a good start) right across any film decoupling, and the film decoupling should as large as possible with the lowest ESL possible. It's hard to beat the WIMA MKS2 3.3u in this position. Of course for "locally inert" rails you don't actually need film bypass, so it's perfectly possible to save the trouble of damping film caps for a later stage and omit them until then.
If properly damped rails reduce RFI ingress, then it would make sense if the reduction in IMD could cause an improvement in stereo image, which may seem like a reduction in crosstalk. Who knows?
Dan's solution does make an audible improvement in my experience, and while I don't like the bass harmonics it seems to generate (unless you use very large local lytics), it can really improve the enjoyment factor of an amp. As for locally damped film decoupling, this is something I do in order to reduce sibilance when I try to use film bypass.
PS. My comments on the sound are dubious as expected, but everything else is easily measured and verified to my knowledge.
So if you want to have rails that promote stability of the amplifier, you need to have a small-value lytic (Pana VR 100V 47uF is a good start) right across any film decoupling, and the film decoupling should as large as possible with the lowest ESL possible. It's hard to beat the WIMA MKS2 3.3u in this position. Of course for "locally inert" rails you don't actually need film bypass, so it's perfectly possible to save the trouble of damping film caps for a later stage and omit them until then.
If properly damped rails reduce RFI ingress, then it would make sense if the reduction in IMD could cause an improvement in stereo image, which may seem like a reduction in crosstalk. Who knows?
Dan's solution does make an audible improvement in my experience, and while I don't like the bass harmonics it seems to generate (unless you use very large local lytics), it can really improve the enjoyment factor of an amp. As for locally damped film decoupling, this is something I do in order to reduce sibilance when I try to use film bypass.
PS. My comments on the sound are dubious as expected, but everything else is easily measured and verified to my knowledge.
Last edited:
The high inductance of the large ripple reducing/storage caps used in power supplies leads to a slow current transient response... esp noticable with low or no GNFB designs.
I use the largest value film capacitors (not 0.1mfd) for their faster energy release. Mounted close to OPS. Several microfarads for a power amp is not too small....And, usually a series R of about .2-.3 Ohms will critically damp them.
THx-RNMarsh
I use the largest value film capacitors (not 0.1mfd) for their faster energy release. Mounted close to OPS. Several microfarads for a power amp is not too small....And, usually a series R of about .2-.3 Ohms will critically damp them.
THx-RNMarsh
This is no different than using a Pana VR 100V 47u lytic instead of the film cap. The ESL is even much less, and much more stored energy is available, at both a higher and lower frequency. So if you have found benefits to doing this, I wouldn't say they have anything to do with actual energy release ability.
most of us are "releasing energy" for audio reproduction thru ~ 1-2 uH output isolation series L, 10s of ft of cable into dynamic loudspeakers with order of mH Le - ps electrolytic Cap ESL isn't going to register
heavy 100 mA/device output stage bias even means the Class B switchover has low edge rate
heavy 100 mA/device output stage bias even means the Class B switchover has low edge rate
Last edited:
A lot of conjecture... you need some real world tests, me thinks - from a paper I wrote around 1990:
Impulse response -- same charge voltage and capacitance --- each cap type discharged thru its own esr. One on top has been critically damped with a small series R. The bottom - bipolar- has no added series R as its own esr is higher and over-damps the impulse response.
[Larger C values give similar results - longer smear time constant - but the Amps would exceed the current probes limits.]
View attachment cap impulse.pdf
THx-RNMarsh
Test done on a LeCroy transient capture scope with built in FFT
Impulse response -- same charge voltage and capacitance --- each cap type discharged thru its own esr. One on top has been critically damped with a small series R. The bottom - bipolar- has no added series R as its own esr is higher and over-damps the impulse response.
[Larger C values give similar results - longer smear time constant - but the Amps would exceed the current probes limits.]
View attachment cap impulse.pdf
THx-RNMarsh
Test done on a LeCroy transient capture scope with built in FFT
Last edited: