For bipolars the base resistors are usually not needed for stability against local parasitic oscillations, but rather to limit forward base current. If one works really hard at it I suppose one could coax some misbehavior out of the fastest devices, but it would take some really silly-long leads. Now it is important not to confuse parasitics with oscillatory bursts near onset of clipping corresponding to large changes in the system gain-phase response. But clipping ahead of the power amp is the best way to avoid this, indeed. Or, just don't get that close to clipping to begin with.
The situation with MOSFETs is very different, as the very large gain-bandwidth product and the large capacitances readily find resonances in the VHF region with quite-small inductances. I had one guy tell me, having described his secret insights into amplifier design, and some to-him anomalous behavior of output bias and dissipation, that he knew it wasn't oscillation because the impedance of the driving stage was so low
I gathered that he lacked any measurement equipment to probe the circuit at anywhere near the likely frequencies it was screaming, not to mention some lapses in circuit understanding one could drive a truck through.Hi Guys
bcarso: I agree with you completely.
Over time I learned from much greater minds than mine that rating the amp anywhere near the clipping point is folly. Not only does it cause the THD to rise dramatically right at the rated output, but the base-drive requirement goes crazy, too. The latter contributes to the former, no doubt. Why would you want to drive a linear power amp to saturation? It won't be musical - at least as hifi goes - and you likely won't hear it anyway since you'll be deaf.
Besides that, just as with low-level discrete circuits, when the supply voltage is increased on a power amp the THD goes down and can be made made much lower at any given power level than with lower rails.
As I described earlier, I get the best results when the phase curve is the same shape as the closed-loop gain curve. In some cases, the phase number suggests rampant instability but there is none in the sim or in the built circuit. This is what baffles me.
Have fun
bcarso: I agree with you completely.
Over time I learned from much greater minds than mine that rating the amp anywhere near the clipping point is folly. Not only does it cause the THD to rise dramatically right at the rated output, but the base-drive requirement goes crazy, too. The latter contributes to the former, no doubt. Why would you want to drive a linear power amp to saturation? It won't be musical - at least as hifi goes - and you likely won't hear it anyway since you'll be deaf.
Besides that, just as with low-level discrete circuits, when the supply voltage is increased on a power amp the THD goes down and can be made made much lower at any given power level than with lower rails.
As I described earlier, I get the best results when the phase curve is the same shape as the closed-loop gain curve. In some cases, the phase number suggests rampant instability but there is none in the sim or in the built circuit. This is what baffles me.
Have fun
I run totally DC coupled, all discrete Borbely jfet/fet Dac, phono, line amp, and power amp. shunt regulated from large separate power supplies . Speakers are Martin Logan CLX full range ESL and 2 x JL fathom 212 subs and yes Bryston is dull and unexciting even with very low distortion. although i use their SP 3 7.1 processor with Jung super regulator and other PS upgrades in a separate HT system
Last edited:
Hi Guys
bacarso: You can get FLEETS of trucks through the gaps in my knowledge.
ticknpop: I hope you have an electronic crossover in there? Using a passive crossover between the subs and the ESLs will significantly erode the sound. Borbely's stuff is interesting but none if it is hifif. Like most ESLs, the Martin Logan offerings have a critical and narrow sweet-spot for listening and I've never heard good sound through them, so it would not surprise me if even a Bryston amp does not sound great with them. Contrary to that, the Quad ESLs have a fairly wide sweet spot and sound quite good.
To each his own.
Have fun
bacarso: You can get FLEETS of trucks through the gaps in my knowledge.
ticknpop: I hope you have an electronic crossover in there? Using a passive crossover between the subs and the ESLs will significantly erode the sound. Borbely's stuff is interesting but none if it is hifif. Like most ESLs, the Martin Logan offerings have a critical and narrow sweet-spot for listening and I've never heard good sound through them, so it would not surprise me if even a Bryston amp does not sound great with them. Contrary to that, the Quad ESLs have a fairly wide sweet spot and sound quite good.
To each his own.
Have fun
Now I am confused. When you said you don't like roller coaster phase behavior, did you mean in closed loop or open loop gain. You use kind of TMC(you called it output inclusive compensation as Self called it, but Cherry call the one with capacitor from output) and this should give you a roller coaster open loop phase behavior, and in close loop it will be flat. You don't give a values for CC11-1 and CC12-1 and it's not easy to make any guessing without that.
Do you have, by chance, LTspice file for that amp, I would like to simulate it, but not eager to draw it from scratch.
Damir
Hi Guys
Actually, I mentioned way back that it is closed-loop response that I am looking at. Open-loop response may be an indicator of things to come but I don't run my amps open-loop so don't care about that condition.
The single cap from output back to the VAS in the Cherry amp is what Cherry-enthusiasts call the "Cherry cap".
Output-inclusive compensation is what Self calls splitting the Miller cap then adding a resistor from the cap junction to the amp output. There is a frequency transition where the output stage is included or not included in the compensation loop. I've seen this referred to as "transitional miller compensation" prior to Self's discussion.
Leech achieved a similar end using two feedback paths, one high-pass one low-pass that took feedback from the driver output and the final output back to the feedback summing node. Both paths turned over at the same frequency so as one dialled out the other dialled in.
Have fun
Actually, I mentioned way back that it is closed-loop response that I am looking at. Open-loop response may be an indicator of things to come but I don't run my amps open-loop so don't care about that condition.
The single cap from output back to the VAS in the Cherry amp is what Cherry-enthusiasts call the "Cherry cap".
Output-inclusive compensation is what Self calls splitting the Miller cap then adding a resistor from the cap junction to the amp output. There is a frequency transition where the output stage is included or not included in the compensation loop. I've seen this referred to as "transitional miller compensation" prior to Self's discussion.
Leech achieved a similar end using two feedback paths, one high-pass one low-pass that took feedback from the driver output and the final output back to the feedback summing node. Both paths turned over at the same frequency so as one dialled out the other dialled in.
Have fun
But you complained about my amps with a roller coaster phase behavior and that was a Loop Gain plot (feedback). All my amps are with perfectly flat close loop phase, with a phase at 20 kHz with less then -2 degree.
Maybe you've confused a Loop Gain plot with Close Loop Gain plot?
Yes, it's popular here in the us amplifier. I built this about 20 years ago. The main contradiction of this amplifier is that to increase the gain of the differential cascade it is necessary to increase the resistance R15R16, but, after that, will have to increase the resistance R21R23 and R22R24, that will eat up all the winnings to gain.
Since the voltage amplifier is powered by voltage of the output stage, with the use of Q12Q13 2SA1380/2SC3502 if the amplitude of the output voltage U - 7 Volts (the effect of quasi-saturation) will increase distortion. To avoid this, VA should have the voltage to 15 Volts exceeds the voltage of the output stage.
Given you the version with Miller compensation will have a low loop gain at the top of the audible range. Moreover, the transistors of the analog overcurrent protection Q10Q11 work on nano - and micro currents and even introduce distortions in the signal. Diodes D7D8 at the time of switching of the output stage of the shoulders also make a clearly visible distortion. During this defense response, it forms a local feedback loop Q11-Q15-Q17-Q19Q21 R29||R30 / R27, which is in conflict with the main feedback loop, which can lead to outbreaks of generation and forcing to set the node protection elements of the frequency compensation.
Overall, well, if this circuitry will bring us less than 0.005% at 20 kHz at full output.
Last edited:
Hi Guys
Dadod: It would not surprise me if I confused those plots but why would you post an open-loop response anyway other than to show the open-loop gain? Open-loop phase doesn't mean anything, as far as I know.
I don't recall you posting closed-loop phase response. The fact you say the closed-loop phase response of your amps is flat in the pass band confirms my belief that it should be, so thanks for that!
T117: The original leech amp did not have a cascode front end, but acquired one later. There were at least five versions of the amp, with various input filtering added then subtracted. It generally sim's to about 50ppm using the MJL21193/94 outputs, MJE15030/29 drivers, and 2N5401/5551s eleswhere. The gain and phase are smooth. Swapping in higher-Ft devices needs radical change to the compensation.
Have fun
Dadod: It would not surprise me if I confused those plots but why would you post an open-loop response anyway other than to show the open-loop gain? Open-loop phase doesn't mean anything, as far as I know.
I don't recall you posting closed-loop phase response. The fact you say the closed-loop phase response of your amps is flat in the pass band confirms my belief that it should be, so thanks for that!
T117: The original leech amp did not have a cascode front end, but acquired one later. There were at least five versions of the amp, with various input filtering added then subtracted. It generally sim's to about 50ppm using the MJL21193/94 outputs, MJE15030/29 drivers, and 2N5401/5551s eleswhere. The gain and phase are smooth. Swapping in higher-Ft devices needs radical change to the compensation.
Have fun
Last edited:
Unfortunately, transistors any analog overcurrent protection can and do work in the regimes of micro - and nano-currents. Since overcurrent protection, usually, forms a local feedback loop, the nonlinearity of transistors protection introduces distortions in the signal. Pay attention to it.
Here we use good old cent as the basic unit of measurement of the THD. Have you used ppm. Please translate ppm: what's that in percent?
My main reason for posting that schematic was to correct the spelling of the late man's name, which has been unsuccessful so far. Secondarily, because it shows the complementary input configuration.
But yes, consider the year it was put forth, and how expensive transistors were then. I could go on at lugubrious length about things to improve it, but I think the main characteristic of this particular one is the complementarity of the input stage. Was Leach the first to do this?
Marshall Leach did have some interesting circuit designs and was apparently a very good teacher. I liked his common-base preamp that had zero input current with no input coupling cap(s), although the capacitors were there, just concealed as it were elsewhere in the circuit.
As I said, I also liked this amp. Precisely because of what you said.
We are talking about the symmetrical input stage of the Leach, or a separate pre-amp?
Hi Guys
T117: Your statement in post-90 is true. Most protection circuits add THD to otherwise good amplifiers. The protection I use in the schematics is basic current clamping. This only changes THD near clipping since I've sim'd it and tested in place and not in place. I size the output stage to withstand full clamped current at full voltage.
As I stated above, I don't really care too much about the near-clipping performance since i never run the amp there. At all useful power levels - and many nonuseful - THD is unchanged by the protection circuit.
1ppm = 0.000 1% as stated previously many times.
10ppb = 0.000 001%, the lowest number LTspice will show. It is great to see all zeroes.
Have fun
T117: Your statement in post-90 is true. Most protection circuits add THD to otherwise good amplifiers. The protection I use in the schematics is basic current clamping. This only changes THD near clipping since I've sim'd it and tested in place and not in place. I size the output stage to withstand full clamped current at full voltage.
As I stated above, I don't really care too much about the near-clipping performance since i never run the amp there. At all useful power levels - and many nonuseful - THD is unchanged by the protection circuit.
1ppm = 0.000 1% as stated previously many times.
10ppb = 0.000 001%, the lowest number LTspice will show. It is great to see all zeroes.
Have fun
Separate machine altogether. I believe it was touted as a moving-coil cartridge stepup device, and some others did some embellishments. There are those who call such low-input-impedance preamps as using the cartridge as a "current source", which is immensely misleading.As I said, I also liked this amp. Precisely because of what you said.
We are talking about the symmetrical input stage of the Leach, or a separate pre-amp?
I think the circuit might be in a magazine article.
Sure, but I would appreciate a link to it, because I do not know the title of the article.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.