Thanks, Mark. The use of isolated DC-DC converters to implement boosted rails is a great idea!
Did he implement them at the main power supply and run an extra pair of boosted rail wires to the IPS-VAS, or did he implement them somewhere on the amplifier board. I would assume the former (especially for a stereo amplifier). The latter arrangement might require a little bit more thought regarding those converters as a noise/EMI source.
Cheers,
Bob
Did he implement them at the main power supply and run an extra pair of boosted rail wires to the IPS-VAS, or did he implement them somewhere on the amplifier board. I would assume the former (especially for a stereo amplifier). The latter arrangement might require a little bit more thought regarding those converters as a noise/EMI source.
Cheers,
Bob
The PCB layout is included in Winfield Hill's thread, linked above.
If you're feeling especially nervous ("conservative"), a 12V DC-DC module provides plenty of headroom, permitting capacitance multiplier circuits which smooth out & filter the boosted rails. Properly implemented, these can achieve high bandwidth and quite excellent PSRR at high frequency. John Curl used them in several of his power amplifiers for Parasound. Encircled in red below.
_
If you're feeling especially nervous ("conservative"), a 12V DC-DC module provides plenty of headroom, permitting capacitance multiplier circuits which smooth out & filter the boosted rails. Properly implemented, these can achieve high bandwidth and quite excellent PSRR at high frequency. John Curl used them in several of his power amplifiers for Parasound. Encircled in red below.
_
Attachments
Bob, you might enjoy taking another look at Winfield Hill's 10 MHz, 1000 V/usec, 100W_8R power amplifier. If his name rings a bell, that's because he's one of the two authors of a famous book called "The Art Of Electronics" ... and he's the one who did the fantastic work on ultra low noise BJTs that you can actually buy today (hint: better than the obsolete former champion, 2SB737), see table 8.1 of the 3rd edition.
His amplifier is here on diyAudio:
You may find his amp's boosted supply rails to be thought provoking. He uses small, PCB mounted, isolated DC-to-DC converter modules which generate XX VDC output, and connects them in series with the amplifier's main rails. They're isolated, you can connect their output however you wish. Voila, boosted rails for the IPS and VAS. Without buying an extra transformer, without sourcing a weirdo custom transformer having numerous secondaries, without winding your own extra coils on a vendor-bought toroid (Cordell 1st ed, p.348)
Since the IPS + VAS current requirements are modest, you have the option to use DC-to-DC modules with 12 VDC output, and still remain comfortably within the output current and output power capabilities of low cost modules. Twelve volts allows the freedom to HEAVILY filter the SMPS output, incurring plenty of IR voltage drop across the filters if you wish, and still get a very smooth rail that's boosted more than enough to do the job.
It's a pretty cute idea, and the cost of the modules is surprisingly modest. Check out the Murata ones on DigiKey. Winfield happened to choose DC-DC modules from Micro Power Direct, but that company doesn't distribute through Digikey or Mouser or Element14, so we do not speak of them here at diyAudio
Some purists may complain about using SMPSs in an otherwise fully analog design. The output can be conveniently filtered, however not so easy with the radiated RF and the PCB layout to minimize injection of garbage in the ground, in particular for a wideband design.
SMPSs always have a lead when it comes to convenience, space/volume, and power losses, but still I'd rather use extra PCB mount transformers to generate the voltage boost, like this L01-6302 Amgis, LLC | Transformers | DigiKey They come at the same price as a good quality SMPS module.
Of course, if one decides to go for SMPSs to generate the power rails, then also using SMPSs for the booster is a no brainer decision.
VAS RE
I got curious about the claims made by Cherry regarding the VAS emitter resistor in his JAES paper (Feedback, Sensitivity, and Stability of Audio Power Amplifiers). He did an analysis of the sensitivity of the forward-path gain to various parameters, denoted S[parameter] (not to be confused with S parameters something that appears in RF/microwave work). He concluded that S[gm_VAS] ~0, where gm_VAS is the transconductance of the VAS. However S[beta_VAS] was significant. As Re in the VAS effects transconductance but not beta this would explain why its impact on forward-path gain is negligible.
I did a LTSpice simulation of a PNP (2N5401 from the LTSpice library) with the base driven by a current source, a resistor from emitter to ground, a current source of 10 mA at the collector and a 10 k resistor from collector to -25 VDC.
The current source at the base was swept to find the value that gave -25 V at the collector node. This means the collector current was 10 mA and no current was flowing through the 10 k resistor. The base current source was then set to that value (69.8 uA).
I did .TF analysis (small signal, DC transfer function) for two values of Re: 0.01 and 68 ohms). This is basically the change of voltage at the collector node divided by the change in the current into the base. The value of the transfer function for the two cases was within 1%. The output impedance (resistance actually) was within 1% as well. The input impedance for Re of 0.01 was 390 and 5728 for Re of 68. The trend of the latter makes sense but the actual values less so.
An AC sweep gave f_3dB of 70 kHz for both, voltage at collector, and gains that were within less than 0.1 dB.
From the Cherry paper: 'For practical purposes S[gm_VAS] is zero, and gm_VAS does not occur in the other sensitivities. It might have been expected that the inclusion of an unbypassed emitter resistor in the VAS would reduce stage gain, hence reduce overall loop gain, and thereby increase distortion associated with nonlinearity in the OPS. In fact, such a resistor has no first-order effect on distortion. We shall see later that such a resistor provides a powerful method for eliminating VHF oscillation. (The author cannot give a simple but convincing physical explanation for this result. A nonsimple explanation has its basis in the fact that the second stage is fed from a current source.)' NOTE: I changed the nomenclature to be consistent current and familiar literature, i.e. Bob's book
I hope that this sparks some interest. There might be another lever to pull when stabilizing an amp particularly when using schemes beyond Miller compensation (I don't have the acronyms memorized).
Jeff
I got curious about the claims made by Cherry regarding the VAS emitter resistor in his JAES paper (Feedback, Sensitivity, and Stability of Audio Power Amplifiers). He did an analysis of the sensitivity of the forward-path gain to various parameters, denoted S[parameter] (not to be confused with S parameters something that appears in RF/microwave work). He concluded that S[gm_VAS] ~0, where gm_VAS is the transconductance of the VAS. However S[beta_VAS] was significant. As Re in the VAS effects transconductance but not beta this would explain why its impact on forward-path gain is negligible.
I did a LTSpice simulation of a PNP (2N5401 from the LTSpice library) with the base driven by a current source, a resistor from emitter to ground, a current source of 10 mA at the collector and a 10 k resistor from collector to -25 VDC.
The current source at the base was swept to find the value that gave -25 V at the collector node. This means the collector current was 10 mA and no current was flowing through the 10 k resistor. The base current source was then set to that value (69.8 uA).
I did .TF analysis (small signal, DC transfer function) for two values of Re: 0.01 and 68 ohms). This is basically the change of voltage at the collector node divided by the change in the current into the base. The value of the transfer function for the two cases was within 1%. The output impedance (resistance actually) was within 1% as well. The input impedance for Re of 0.01 was 390 and 5728 for Re of 68. The trend of the latter makes sense but the actual values less so.
An AC sweep gave f_3dB of 70 kHz for both, voltage at collector, and gains that were within less than 0.1 dB.
From the Cherry paper: 'For practical purposes S[gm_VAS] is zero, and gm_VAS does not occur in the other sensitivities. It might have been expected that the inclusion of an unbypassed emitter resistor in the VAS would reduce stage gain, hence reduce overall loop gain, and thereby increase distortion associated with nonlinearity in the OPS. In fact, such a resistor has no first-order effect on distortion. We shall see later that such a resistor provides a powerful method for eliminating VHF oscillation. (The author cannot give a simple but convincing physical explanation for this result. A nonsimple explanation has its basis in the fact that the second stage is fed from a current source.)' NOTE: I changed the nomenclature to be consistent current and familiar literature, i.e. Bob's book
I hope that this sparks some interest. There might be another lever to pull when stabilizing an amp particularly when using schemes beyond Miller compensation (I don't have the acronyms memorized).
Jeff
Yes it is, and note who brought it to you first
This article originally appeared in Linear Audio Volume 2, September 2011. Linear Audio, a book-size printed tech audio resource, is published half-yearly by Jan Didden.
Jan
Ovidiu, I know what you mean, and not wanting to split hairs, but an SMPS is a fully analog thing too. Working with analog quantities voltage, current, time.
Jan
I think the difficulties are because stability issues in EC are not obvious. There are second order effects like feed forward thru the feedback network that are usually not considered, but in EC they start to matter. Presumably you used TRANsient simulation to ensure it was stable, did you use Tian probe(s) to check deeper, and if so where placed? For EC we really need the whole Middlebrook GFT theory.
Best wishes David
Differential and common mode balun adapters for the Tian probes are required. Using them should be self explanatory for the Spice afficionados.
Below, I have no idea if these would work for LTSpice or changes are required.
Below, I have no idea if these would work for LTSpice or changes are required.
Code:
-------------
.SUBCKT BALUN_DM_PROBE s1a s2a da s1b s2b db
*
* s1a = difference-mode floating signal 1 for input/output "a"
* s2a = difference-mode floating signal 2 for input/output "a"
* da = difference-mode ground-referenced signal for input/output "a"
* s1b = difference-mode floating signal 1 for input/output "b"
* s2b = difference-mode floating signal 2 for input/output "b"
* db = difference-mode ground-referenced signal for input/output "b"
*
* Connect da and db together externally with a loop gain probe
*
X1 db 0 s1b NCM TRANSFORMER
X2 db 0 NCM s2b TRANSFORMER
X3 da 0 s1a NCM TRANSFORMER
X4 da 0 NCM s2a TRANSFORMER
* R1, R2 and R3 prevent floating nodes
R1 da 0 1E9
R2 db 0 1E9
R3 NCM 0 1E9
*
.ENDS BALUN_DM_PROBE
-------------------
.SUBCKT BALUN_CM_PROBE s1a s2a ca s1b s2b cb
*
* s1a = diff mode signal 1 for input/output "a"
* s2a = diff mode signal 2 for input/output "a"
* ca = common mode signal for input/output "a"
* s1b = diff mode signal 1 for input/output "b"
* s2b = diff mode signal 2 for input/output "b"
* cb = common mode signal for input/output "b"
*
* Connect ca and cb together externally with a loop gain probe
*
X1 NDM 0 s1b cb transformer
X2 NDM 0 cb s2b transformer
X3 NDM 0 s1a ca transformer
X4 NDM 0 ca s2a transformer
* R1, R2 and R3 prevent floating nodes
R1 ca 0 1E9
R2 cb 0 1E9
R3 NDM 0 1E9
*
.ENDS BALUN_CM_PROBE
----------------
Ovidiu, I know what you mean, and not wanting to split hairs, but an SMPS is a fully analog thing too. Working with analog quantities voltage, current, time.
Jan
i agree wholeheartedly......
Working with analog quantities voltage, current, time and square waves with duty cycle...
Hi Ovidiu
I posted a simultaneous differential and common mode balun in DIYaudio a few years back, while you were away from the forum so presumably you missed it. It's a little piece that definitely works in LTspice, your use of "TRANSFORMER" would be problematic in LTspice as written, I think. But that was not my point, in Bob's circuit there's feed-forward that the Tian probe doesn't handle well, sometimes the phase plots just can't be true. I still don't fully understand these Tian anomalies, hence my interest in LKA's results. I had to do the whole Middlebrook GFT analysis, perhaps there are simpler ways.
Best wishes David
Last edited:
Hi Bob,
Just wanted to give you the heads up on the latest issue of the Magazine Hi-fi news & Record review. They review the Chord Ultima Power amplifier. There is a side note with some words by John Franks which mentioned that it's based on the error-correction technology conceived by Malcom Hawksford and young Bob Cordell from Bell labs.
That's nice.
Cheers,
Mogens (which is looking forward to the new book)
Just wanted to give you the heads up on the latest issue of the Magazine Hi-fi news & Record review. They review the Chord Ultima Power amplifier. There is a side note with some words by John Franks which mentioned that it's based on the error-correction technology conceived by Malcom Hawksford and young Bob Cordell from Bell labs.
That's nice.
Cheers,
Mogens (which is looking forward to the new book)
VAS RE, continued
In post 9104 I described an experiment involving a single PNP transistor in a common emitter configuration, driven by a current source. I varied the value of RE from negligible to significant (relative to Re=26mV/Ic,mA). I was interested in the gain as defined by the change of voltage at the collector node to the change of current into the base (transresistance gain). I found that there was no significant difference between RE = 0.01 ohm and RE = 68 ohm.
I followed up with another experiment in which I measured (via SPICE simulation) the voltage gain, that is, the change in collector node voltage to change in base node voltage. In this case the gain does change significantly.
The output quantity from a current-source loaded diff pair is a current. This is the input to the VAS. If the VAS is viewed as a transresistance stage then the gain does not depend on RE. If the VAS is view as a voltage stage (or a transconductance) stage then the gain does depend on RE. Which view is more appropriate?
If you write an expression for the transresistance gain (or current gain) of the VAS stage it does not contain re or RE. It does contain beta however. This would seem to support the conclusions of Cherry's paper.
The current mirror loaded diff pair IPS can be reasonably represented as a transconductance stage on the basis of input impedance (high) and output impedance (high). The OPS can be reasonably represented as a voltage gain stage on the basis of input impedance (high) and output impedance (low). The VAS is an odd duck in that the input impedance is somewhere in between low and high and the output impedance is high. If the input impedance is low then it's a current stage. If the input resistance is high then it's a transresistance stage. But low and high is relative to the output impedance of the IPS.
The input to the amplifier is a voltage impressed on the high impedance, the output of the amplifier is a voltage from a low impedance source. The output of the IPS is a current and the input to the OPS is voltage. From this line of reasoning the VAS is a current-in, voltage-out stage or a transresistance (or more generally transimpedance) stage. By that line of reasoning the transconductance of the VAS does not matter.
In post 9104 I described an experiment involving a single PNP transistor in a common emitter configuration, driven by a current source. I varied the value of RE from negligible to significant (relative to Re=26mV/Ic,mA). I was interested in the gain as defined by the change of voltage at the collector node to the change of current into the base (transresistance gain). I found that there was no significant difference between RE = 0.01 ohm and RE = 68 ohm.
I followed up with another experiment in which I measured (via SPICE simulation) the voltage gain, that is, the change in collector node voltage to change in base node voltage. In this case the gain does change significantly.
The output quantity from a current-source loaded diff pair is a current. This is the input to the VAS. If the VAS is viewed as a transresistance stage then the gain does not depend on RE. If the VAS is view as a voltage stage (or a transconductance) stage then the gain does depend on RE. Which view is more appropriate?
If you write an expression for the transresistance gain (or current gain) of the VAS stage it does not contain re or RE. It does contain beta however. This would seem to support the conclusions of Cherry's paper.
The current mirror loaded diff pair IPS can be reasonably represented as a transconductance stage on the basis of input impedance (high) and output impedance (high). The OPS can be reasonably represented as a voltage gain stage on the basis of input impedance (high) and output impedance (low). The VAS is an odd duck in that the input impedance is somewhere in between low and high and the output impedance is high. If the input impedance is low then it's a current stage. If the input resistance is high then it's a transresistance stage. But low and high is relative to the output impedance of the IPS.
The input to the amplifier is a voltage impressed on the high impedance, the output of the amplifier is a voltage from a low impedance source. The output of the IPS is a current and the input to the OPS is voltage. From this line of reasoning the VAS is a current-in, voltage-out stage or a transresistance (or more generally transimpedance) stage. By that line of reasoning the transconductance of the VAS does not matter.
VAS transconductance
Bob,
I'm trying understand the conclusions from Cherry's paper (Feedback, Sensitivity, and Stability of Audio Power Amplifiers) and square that the presentation in your book. In particular on page 128 you discuss the nonlinearity of the VAS due to change in transconductance with change in collector current. But if the conclusion of Cherry's paper is valid then the transconductance of VAS is not of significance.
What's up? Please read my post that proceeds this one for my finding and argument. The best I can surmise is that if you want current gain in the VAS and you want to do it with only one or two transistors (as contrasted with Groner's solution) then you have to use a CE stage and live with some messiness.
And, on a different topic, I preordered the second edition of your book at Amazon. I knew it was on the horizon so I open this thread occasionally to check progress. That's what brought me here recently.
Best,
Jeff
Bob,
I'm trying understand the conclusions from Cherry's paper (Feedback, Sensitivity, and Stability of Audio Power Amplifiers) and square that the presentation in your book. In particular on page 128 you discuss the nonlinearity of the VAS due to change in transconductance with change in collector current. But if the conclusion of Cherry's paper is valid then the transconductance of VAS is not of significance.
What's up? Please read my post that proceeds this one for my finding and argument. The best I can surmise is that if you want current gain in the VAS and you want to do it with only one or two transistors (as contrasted with Groner's solution) then you have to use a CE stage and live with some messiness.
And, on a different topic, I preordered the second edition of your book at Amazon. I knew it was on the horizon so I open this thread occasionally to check progress. That's what brought me here recently.
Best,
Jeff
I think this was to be expected. The Re doen't change the transistor current gain hFE, therefor driving with an input current gives a certain output current that is not depending at all on Re.
If you drive the Vas with the output current of the input stage, then the Vas output current does not depend on Re. If you convert the Vas output current to an output voltage, the current-in-to-voltage-out gain also doesn't change.
Of course Vb does go up or down with Re but since you drive with the (ideally infinitely high output impedance) current source, that Vb with go to whatever is required to support the input current.
Jan
Last edited:
I remember that in a letter to Electonics World + Wireless World (I do not have the exact reference, that may be circa 1995), Douglas Self spoke of a double regime, changing from transconductance at low frequencies to transimpedance at frequencies at which the compensation capacitor becomes effective.
In the sixth edition of his book "Audio Power Amplifier Design", page #183, he wrote :
"It is often stated that adding a resistance in the VAS emitter connection introduces local feedback. This is just not true, because the VAS accepts a current input rather than a voltage input, so the voltage developed across the emitter does not cause negative feedback."
Yes, exactly. When I looked at the output voltage vs. input voltage I was still using a current source to drive the base. The resistance at the base node is due to the resistances in the emitter circuit but then so is transconductance. Those terms cancel.
These are generally good observations. In a Miller-compensated amplifier, I believe it is best to consider the behavior of the VAS with the Miller capacitor considered as a part of it. Then, at higher frequencies and with reasonable component values and conditions, the VAS can be considered a transimpedance stage, at least in the limit. At low frequencies, below where the Miller effect dominates (open loop bandwidth frequency) the VAS can be considered as a voltage gain stage, depending on how you treat and assign the impedance to ground at the input of the VAS (and also if the IPS is loaded with a resistor or a current mirror).
I note that there are some amplifiers out there that do not use Miller compensation, lag compensation being only one example. In those cases, the VAS may often be considered a voltage-in, voltage-out stage.
Semantics can always be a problem
.Cheers,
Bob