Isn't this?
An externally hosted image should be here but it was not working when we last tested it.
john k... said:
Sorry,, I have to fix this. Too many typos.
"When connected in series Rg = Ro + Rg of the second driver..."
should have been, "When connected in series Rg = Ro + Re of the second driver..."
(where Re = voice coil resistance).
Also,
"So Ib = 2Vback/(2Rg + Ro) ."
should have been: "So Ib = 2Vback/(2Re + Ro) .
Re: John Swenson transconductance amp
I generally think that JS is one of the smarter guys around, so I frequently save schematics and such that he posts. I believe he has significatly revisited the PS for the amp, but to the best of my knowledge the Signal section is unchanged.
Regards,
John
mikey_audiogeek said:
I generally think that JS is one of the smarter guys around, so I frequently save schematics and such that he posts. I believe he has significatly revisited the PS for the amp, but to the best of my knowledge the Signal section is unchanged.
Regards,
John
Attachments
No. At the same volume setting the parallel connection will be 6dB louder than a singel driver. With a series connection the SPL will be at the same level as a single driver. This is because the voltage across each driver with be reduced to 1/2 the total. But recognize that with the parallel connection the total power into both driver will be twice that of the single driver. With the series connection the total power will be 1/2 that into a single driver.
Again, these differences are the differences between voltage sensitivity and efficiency. The parallel connection yields a +6dB increase in voltage sensitivity while the series connection yield the same voltage sensitivity as a single driver. But both configurations result is a 3dB increase in efficiency.
Again, these differences are the differences between voltage sensitivity and efficiency. The parallel connection yields a +6dB increase in voltage sensitivity while the series connection yield the same voltage sensitivity as a single driver. But both configurations result is a 3dB increase in efficiency.
I tend to think of it like this:
1. Double the number of drivers = +3db.
2. Connect two drivers in parallel = +3db.
3. Connect two drivers in series = -3db.
EX. 1
2 * 8 ohm drivers connected in series =
#1 & #3 above, or (stated),
The 2 drivers present an acoustic gain of +3db that is then canceled out by the -3db of wiring in series, resulting in a net gain/loss of 0db.
EX. 2
2 * 8 ohm drivers connected in parallel =
#1 & #2 above, or (stated)
The 2 drivers present an acoustic gain of +3db that is then added to with the +3db of wiring in parallel, resulting in a net gain of +6db.
1. Double the number of drivers = +3db.
2. Connect two drivers in parallel = +3db.
3. Connect two drivers in series = -3db.
EX. 1
2 * 8 ohm drivers connected in series =
#1 & #3 above, or (stated),
The 2 drivers present an acoustic gain of +3db that is then canceled out by the -3db of wiring in series, resulting in a net gain/loss of 0db.
EX. 2
2 * 8 ohm drivers connected in parallel =
#1 & #2 above, or (stated)
The 2 drivers present an acoustic gain of +3db that is then added to with the +3db of wiring in parallel, resulting in a net gain of +6db.
LND150 substitutes
Nah not really that complicated. The LND150 just creates the bias voltage for the MosFETs - you can use batteries here instead (9V). You can throw away the tube as well - in this circuit it's just an "intelligent resistor" with no part to play in the amplification process. If you throw away the tube you just gotta heatsink the top MosFET and keep it below max dissipation, easy to do. In fact, it would end up just like the circuit on this page:
Gary Pimm battery bias CCS
A conventional transformer actually works "better" than the parallel feed version shown - better PSRR. I use parafeed, but with the OPT coupled across the choke to get the same PSRR as series feed. I trim the output impedance with a resistor across the choke, in my case 10K 2W.
It's a GREAT wee amp and I totally agree with John Swenson's assessment - it comfortably beat everything else I've built and is easily scalable for higher outputs. I've also built John's DAC output stage (based on the same approach) with similar results.
In fact, a push-pull version (think cascode Moskido) using low impedance Plitron output transformers (340R:5R) would be almost unbeatable, IMHO.
Regards,
Mike
Nah not really that complicated. The LND150 just creates the bias voltage for the MosFETs - you can use batteries here instead (9V). You can throw away the tube as well - in this circuit it's just an "intelligent resistor" with no part to play in the amplification process. If you throw away the tube you just gotta heatsink the top MosFET and keep it below max dissipation, easy to do. In fact, it would end up just like the circuit on this page:
Gary Pimm battery bias CCS
A conventional transformer actually works "better" than the parallel feed version shown - better PSRR. I use parafeed, but with the OPT coupled across the choke to get the same PSRR as series feed. I trim the output impedance with a resistor across the choke, in my case 10K 2W.
It's a GREAT wee amp and I totally agree with John Swenson's assessment - it comfortably beat everything else I've built and is easily scalable for higher outputs. I've also built John's DAC output stage (based on the same approach) with similar results.
In fact, a push-pull version (think cascode Moskido) using low impedance Plitron output transformers (340R:5R) would be almost unbeatable, IMHO.
Regards,
Mike
Thanks, JohnK and ScottG, for clearing up that circular question about series-parallel connections. I only plan to series-connect identical drivers - series-connecting dissimilar drivers is asking for trouble. As mentioned a couple of days ago, each "series" arm of the series-parallel connection has its own lowpass filter, thus giving the ability to shape the response of the OB system.
I'm glad Magnetar has actually built a speaker embodying some of the ideas I've batting around since March - in the simplest terms, a large-area, high-efficiency dipole, using a minimum of "boost" equalization, and staying away for high-priced, low-efficiency audiophile drivers.
What happens all too often in audio are "category errors", or typecasting the sound of a given approach based on social-cultural reasons. Taking this out of the abstract, most audiophiles associated the sound of high-efficiency speakers with Altec, JBL, Klipsch, Edgarhorn, or Avante-Garde, simply because that's what they've heard in the past. Similarly, audiophiles have associated the sound of OB with Linkwitz or similar audiophile-driver based systems.
But neither are in fact true. High-efficiency systems don't need to be restricted to horns, or any other technology. OB sytems do not need to be restricted to heavily equalized low-efficiency drivers. When I designed the Ariel 15 years ago, there was a preconception that transmission-line speakers "had" to be inefficient and large. Not so.
In looking at OB's, I'm less interested in the polar pattern shape than the reduction, or complete elimination, of cabinet coloration. These colorations fall in the frequency range where cone drivers are intrinsically flat (piston band) and operating with reasonably low colorations, leaving cabinet standing-wave modes as the dominant coloration (falling in the 200 to 800 Hz range). This is a frequency range where damping materials are only moderately effective, and interior shapes of the cabinet again only have modest effect.
If you don't think cabinets have a lot of coloration, try singing or speaking into a cabinet with the driver removed and hear the colorations for yourself. Those same colorations are always present with a driver installed, no matter how good the driver is. That includes a Feastrex or Fertin, by the way. If you want further confirmation, listen to an electrostat with a box in the rear - horrible sound.
Thus, my interest in OB as the most direct and effective method of reducing the bass/midbass/lower midrange colorations - but also taking into account the serious limitations of an OB speaker, which is a very substantial reduction in headroom (and increase in IM distortion) compared to a conventional cabinet. These are not wished away to claiming the headroom decrease and IM distortion increase are "inaudible", when a quick audition and simple math says otherwise.
Increasing cone area (Sd) to compensate for dipole (or quasicardioid) rolloff is the only way to offset the efficiency loss without increasing distortion. This isn't a complicated concept, but it is undermined by the powerful desire to make the speaker compact and WAF-acceptable. Sorry, but this is Hoffman's Iron Law with an additional term needed to offset the baffle losses.
If the speaker is going to be small:
1) There isn't going to be much (undistorted) bass, or
2) There's going to be a LOT of EQ and amplifier power going into the drivers, or
3) You're going to have to accept cabinet coloration, or
4) A combination of the above
I'm glad Magnetar has actually built a speaker embodying some of the ideas I've batting around since March - in the simplest terms, a large-area, high-efficiency dipole, using a minimum of "boost" equalization, and staying away for high-priced, low-efficiency audiophile drivers.
What happens all too often in audio are "category errors", or typecasting the sound of a given approach based on social-cultural reasons. Taking this out of the abstract, most audiophiles associated the sound of high-efficiency speakers with Altec, JBL, Klipsch, Edgarhorn, or Avante-Garde, simply because that's what they've heard in the past. Similarly, audiophiles have associated the sound of OB with Linkwitz or similar audiophile-driver based systems.
But neither are in fact true. High-efficiency systems don't need to be restricted to horns, or any other technology. OB sytems do not need to be restricted to heavily equalized low-efficiency drivers. When I designed the Ariel 15 years ago, there was a preconception that transmission-line speakers "had" to be inefficient and large. Not so.
In looking at OB's, I'm less interested in the polar pattern shape than the reduction, or complete elimination, of cabinet coloration. These colorations fall in the frequency range where cone drivers are intrinsically flat (piston band) and operating with reasonably low colorations, leaving cabinet standing-wave modes as the dominant coloration (falling in the 200 to 800 Hz range). This is a frequency range where damping materials are only moderately effective, and interior shapes of the cabinet again only have modest effect.
If you don't think cabinets have a lot of coloration, try singing or speaking into a cabinet with the driver removed and hear the colorations for yourself. Those same colorations are always present with a driver installed, no matter how good the driver is. That includes a Feastrex or Fertin, by the way. If you want further confirmation, listen to an electrostat with a box in the rear - horrible sound.
Thus, my interest in OB as the most direct and effective method of reducing the bass/midbass/lower midrange colorations - but also taking into account the serious limitations of an OB speaker, which is a very substantial reduction in headroom (and increase in IM distortion) compared to a conventional cabinet. These are not wished away to claiming the headroom decrease and IM distortion increase are "inaudible", when a quick audition and simple math says otherwise.
Increasing cone area (Sd) to compensate for dipole (or quasicardioid) rolloff is the only way to offset the efficiency loss without increasing distortion. This isn't a complicated concept, but it is undermined by the powerful desire to make the speaker compact and WAF-acceptable. Sorry, but this is Hoffman's Iron Law with an additional term needed to offset the baffle losses.
If the speaker is going to be small:
1) There isn't going to be much (undistorted) bass, or
2) There's going to be a LOT of EQ and amplifier power going into the drivers, or
3) You're going to have to accept cabinet coloration, or
4) A combination of the above
hello Lynn
i can recommend the Radian 950pb compression driver. As you know, Romy uses the Vitavox S2 driver, which has a aluminium 3" diaphragm. This Radian uses a 4" alu diaphram, and are much more modern drivers. So i guess, the performance could be similar, the Radian might sound even better. John Hasquin likes them bouth. Since i compared in the last couple of days directly with the BMS 4592nd, the BMS sounds thin and liveless, it doesnt come even close. The shear realism , speed, color discrimination of the instruments, and fullbodied natural sound is addicting. I think with your RAAL tweeter, that would be a great match. Just using a auto trafo to pad down the Radian, will do it.
I use it without crossover, so in my 49cm horn it does match the bass, which i cross at 300hz. With a view watts, it does not hurt the driver. The ideal crossover recomended by radian however is 800hz
Angelo
i can recommend the Radian 950pb compression driver. As you know, Romy uses the Vitavox S2 driver, which has a aluminium 3" diaphragm. This Radian uses a 4" alu diaphram, and are much more modern drivers. So i guess, the performance could be similar, the Radian might sound even better. John Hasquin likes them bouth. Since i compared in the last couple of days directly with the BMS 4592nd, the BMS sounds thin and liveless, it doesnt come even close. The shear realism , speed, color discrimination of the instruments, and fullbodied natural sound is addicting. I think with your RAAL tweeter, that would be a great match. Just using a auto trafo to pad down the Radian, will do it.
I use it without crossover, so in my 49cm horn it does match the bass, which i cross at 300hz. With a view watts, it does not hurt the driver. The ideal crossover recomended by radian however is 800hz
Angelo
Lynn Olson said:
Yes, but it depends on what "identical" means. Even small difference between "identical" drivers will have an effect on the intermodulation due to the back EMF.
Also, the arguments regarding efficiency only apply to identical drivers. For example, with two drivers in parallel which have the same Fs, but one has Qts = 1 and the other Qts = .3, At Fs the increase in SPL would only be 2.3dB over the Qts = 1 driver.
I did some simulations of this for an opened baffle that was representative of Lynn's configuration. There are two very different issues here. 1) Equalization of the dipole roll off and 2) whether the woofer system is capable of sufficient volume velocity to produce the required maximum SPL with acceptable distortion. Assuming that the second issue is satisfied then dipole eq, either passive or active should result in the same equalized response. For example consider the system Lynn proposed a while back, two high Q woofers and two low Q woofers on an open baffle. The high Q woofers are low passed at 150 Hz or so and the high Q woofer at 300 Hz or so. For the example I considered the low Q woofer to have Fs = 30 Hz and Qts = 0.33. The high Q woofer also had Fs = 30 Hz with Qts = 1.0. It is assumed that all the woofers have the same midband sensitivity (dB/volt/M).
This first image shows the 2PI, infinite baffle of these woofers and the summed response. Green is the high Q woofer with 150 Hz x-o, red the low Q woofer with 300 Hz x-o and blue the summed resopnse. The summed response is a pretty nice band pass response, but, due to the low Q nature of one pair of woofers only a less than 3db boost at 30 Hz is realized compared to the high Q pair alone. The nice thing about this configuration is that the low Q pair will have lower excursion at low frequency reducing the potention for IM and doppler distortion, at least fromt he low Q pair. However, it is unclear if this would result in a reduction in IM for the summed response. The draw back is that the max low frequency SPL will be limited by the excursion of the high Q woofers.
An externally hosted image should be here but it was not working when we last tested it.
The next figure shows an indication of what could be expected when this woofer system is placed on an open baffle with characteristics approximated to represent the baffle Lynn presented a few pages back. The heavy red line is the 2-Pi infinite baffle result and the green line is what an open baffle result migh look like in 4-Pi space. Note that at 30 Hz the "dipole" response is about 10dB below the 2Pi, infinite baffle response. Changing the crossover points will affect the response in the region of the dipole peak and can affect the slope of the initial roll off, but ultimately the low frequency response will follow the 2pi sumation with a 6dB/octave roll off superimposed at lower frequency.
An externally hosted image should be here but it was not working when we last tested it.
This could be eq'ed either passively or activly. The third figure compared the result of each approach. Regardless of how the Eq is applied, the same power is required to be delivered to the woofers to produce a given SPL.
An externally hosted image should be here but it was not working when we last tested it.
If the choice is to use passive eq then we are really tying the system sensitivity to the sensitivity of the woofer system at it low frequency cut off and dissapatine power in the crossover system. If the 2Pi sensitivity of the woofer system were in the 100 to 106 dB/volt range this would mean we would have a system target of about 90 to 96 dB. This would seem to be defeating the intent of the design. If active eq is applied then ampliifier power is optimally used and a minimum amount of power is dissapated in the crossovers. The system would target a nominal midband sensitivity similar to that of the 2Pi woofer sensitivity.
I guess to put it in a nut shell we can look at a hypothetical system where all woofers are flat to DC. In this case adding a second pair raises the woofer sensitivity by 6dB and staggering the crossover points controls the transition between the single pair and dual pair result as shown below. Here the lower violet line is a single pair of woofers with 300 Hz LP, the green line is a dual pair, both crossover at 300 Hz, the violet line showing a transition between the two has one pair with a 300 Hz LP and the second pair with a 50 Hz LP and the black line represent the 150Hz 300 Hz LP configuration. It is apparent that some type of Eq will be needed even if all the woofers are identical. To maximize the use of amplifier power this needs to be done actively.
I'll be away for a while so I won't be able to respond until next week unless I happen to be near a computer.
An externally hosted image should be here but it was not working when we last tested it.
"A conventional transformer actually works "better" than the parallel feed version shown - better PSRR."
As I understand it, there are three primary reasons why parafeed is usually better than conventional feed:
The absence of the DC across the transformer allows for better core material, such as permalloy, which allows for better coupling, thus better sound in areas such as dynamics.
The parafeed dc blocking capacitors are lower in value which usually means that they are better in quality which results in less impact on the AC signal compared to capacitors usually used in the power supply (in conventional feed, the AC signal traverses through the power supply).
The paralles feed topology allows for the transformer to be wired as a autoformer which allows for a more transparent sound.
As far as PSRR is concerned, just design the power supply more robustly.
Retsel
As I understand it, there are three primary reasons why parafeed is usually better than conventional feed:
The absence of the DC across the transformer allows for better core material, such as permalloy, which allows for better coupling, thus better sound in areas such as dynamics.
The parafeed dc blocking capacitors are lower in value which usually means that they are better in quality which results in less impact on the AC signal compared to capacitors usually used in the power supply (in conventional feed, the AC signal traverses through the power supply).
The paralles feed topology allows for the transformer to be wired as a autoformer which allows for a more transparent sound.
As far as PSRR is concerned, just design the power supply more robustly.
Retsel
Hi Retsel, firstly I'll apologise to Lynn for getting off topic, any further discussion could be shifted to the Magnequest forum.
Yes I'm aware of the advantages of parafeed which is why I have Magnequest TFA2004NiJr and nickel EXO-050 transformers at my disposal.
When building the transconductance amp, the first transformers I used were TFA204 (Series feed) which sounded OK and quiet.
Then I went to BCP-15/TFA2004NiJr parafeed as per John Swenson, which sounded great but more PSU noise. Final iteration is parafeed but with the transformer coupled across the choke, still parafeed but collecting the signal in the same way as the series feed.
Remember that the textbook cascode connection "has almost zero PSRR", according to most references. This statement is correct, but missing the other part of the sentence, "with respect to ground".
Guess what - this means that it has almost 100% PSRR with respect to B+! A series feed transformer collects the signal between the cascode connection and B+, resulting in almost 100% PSRR.
Remember that the influence of the final PSU cap on sound is related to the PSRR. In a single-ended tube amp, the impedance of the tube is usually much lower than the impedance of the load, which is why parafeed works better referenced to ground. With a cascode topology, the opposite applies.
What's more: I've tried it and it sounds better!
Best regards,
Mike
Yes I'm aware of the advantages of parafeed which is why I have Magnequest TFA2004NiJr and nickel EXO-050 transformers at my disposal.
When building the transconductance amp, the first transformers I used were TFA204 (Series feed) which sounded OK and quiet.
Then I went to BCP-15/TFA2004NiJr parafeed as per John Swenson, which sounded great but more PSU noise. Final iteration is parafeed but with the transformer coupled across the choke, still parafeed but collecting the signal in the same way as the series feed.
Remember that the textbook cascode connection "has almost zero PSRR", according to most references. This statement is correct, but missing the other part of the sentence, "with respect to ground".
Guess what - this means that it has almost 100% PSRR with respect to B+! A series feed transformer collects the signal between the cascode connection and B+, resulting in almost 100% PSRR.
Remember that the influence of the final PSU cap on sound is related to the PSRR. In a single-ended tube amp, the impedance of the tube is usually much lower than the impedance of the load, which is why parafeed works better referenced to ground. With a cascode topology, the opposite applies.
What's more: I've tried it and it sounds better!
Best regards,
Mike
Hi Dobias,
My take in the series-parallel -
Two identical drivers, series connected in phase, on the same baffle, and driven by a voltage amplifier source (lo-Z SS or tube with NFB) - the sound level remains the same, 0dB,
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced by the drivers due to reduced current flow is halved (-3dB), compared to one.
Two identical drivers, series connected in phase, on the same baffle, and driven by a current amplifier source (hi-Z tube or SS without NFB) the sound level increases by 6dB,
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced by the drivers due to increased voltage development is also doubled (+3dB), compared to one.
Two identical drivers, parallel connected in phase, on the same baffle, and driven by a voltage amplifier source (lo-Z SS or tube with NFB) - the sound level increases by 6dB,
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced by the drivers is doubled due to increased current flow (+3dB), compared to one.
Two identical drivers, parallel connected in phase, on the same baffle, and driven by a current amplifier source (hi-Z tube or SS without NFB) the sound level remains the same, (0dB),
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced is halved due to decreased voltage development (-3dB), compared to one.
So in response to your Post#2738 question Dobias, I would respond
YES, if you use a voltage amplifier - normal global-NFB SS or tube, and
NO, if you use a current amplifier - specialist non-global-NDB SS or tube.
Anything in between, or where passive crossovers are used, might be less easy to define.
Cheers ......... Graham.
My take in the series-parallel -
Two identical drivers, series connected in phase, on the same baffle, and driven by a voltage amplifier source (lo-Z SS or tube with NFB) - the sound level remains the same, 0dB,
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced by the drivers due to reduced current flow is halved (-3dB), compared to one.
Two identical drivers, series connected in phase, on the same baffle, and driven by a current amplifier source (hi-Z tube or SS without NFB) the sound level increases by 6dB,
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced by the drivers due to increased voltage development is also doubled (+3dB), compared to one.
Two identical drivers, parallel connected in phase, on the same baffle, and driven by a voltage amplifier source (lo-Z SS or tube with NFB) - the sound level increases by 6dB,
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced by the drivers is doubled due to increased current flow (+3dB), compared to one.
Two identical drivers, parallel connected in phase, on the same baffle, and driven by a current amplifier source (hi-Z tube or SS without NFB) the sound level remains the same, (0dB),
- because the total transducer cone area x force factor is doubled (+3dB), whilst the total energy transduced is halved due to decreased voltage development (-3dB), compared to one.
So in response to your Post#2738 question Dobias, I would respond
YES, if you use a voltage amplifier - normal global-NFB SS or tube, and
NO, if you use a current amplifier - specialist non-global-NDB SS or tube.
Anything in between, or where passive crossovers are used, might be less easy to define.
Cheers ......... Graham.
Lucky me. I found a computer.
I would caution about making arguments about SPL increases or decreases based on cone area or other dirver properties for multiple driver systems. Driver properties really have nothing to do with it other than how they affect the efficiency of each individual driver, frequency response and power division (impedance)between the drivers. The SPL radiated by any two equal strength, correlated sources is the sum of the amplitudes times the cosine of 1/2 the phase difference between them. That is always true. If they are in phase this reduces to 1 + 1 = 2 = 6dB. Two identical drivers will always produce 6dB greater SPL than a single driver provided the power into each driver remains constant. That is the caveat. When connected in series or parallel, and depending on whether they are driven by a voltage or current source amplifier, the power supplied to each driver either remains the same (parallel connection for a voltage source, series for a current source) or is reduced by 6dB (series voltage source, parallel current source). Cone area, BL, etc don't enter the picture.
If the drivers aren't identical then we have to consider the frequency response of each driver along with the impedance to determine how the power is divided between each driver and what the result amplitude response of each individual driver is vs frequency before we can perform the sum.
Regardless of driver type or connection, the change in efficiency of a multiple driver system relative to a single driver is the changed in radiated SPL (in dB) minus the change in input power (in dB).
I would caution about making arguments about SPL increases or decreases based on cone area or other dirver properties for multiple driver systems. Driver properties really have nothing to do with it other than how they affect the efficiency of each individual driver, frequency response and power division (impedance)between the drivers. The SPL radiated by any two equal strength, correlated sources is the sum of the amplitudes times the cosine of 1/2 the phase difference between them. That is always true. If they are in phase this reduces to 1 + 1 = 2 = 6dB. Two identical drivers will always produce 6dB greater SPL than a single driver provided the power into each driver remains constant. That is the caveat. When connected in series or parallel, and depending on whether they are driven by a voltage or current source amplifier, the power supplied to each driver either remains the same (parallel connection for a voltage source, series for a current source) or is reduced by 6dB (series voltage source, parallel current source). Cone area, BL, etc don't enter the picture.
If the drivers aren't identical then we have to consider the frequency response of each driver along with the impedance to determine how the power is divided between each driver and what the result amplitude response of each individual driver is vs frequency before we can perform the sum.
Regardless of driver type or connection, the change in efficiency of a multiple driver system relative to a single driver is the changed in radiated SPL (in dB) minus the change in input power (in dB).