[hagtech]

You still need the 3rd opamp stage in the BUGLE, as it provides the last 15dB or so of gain.  The CF won't have any voltage gain.  Also, the lower tube in the Aikido is a triode?  If so, then it's plate impedance is probably not a lot higher than just using a power resistor for a pulldown (which would also provide more stability).  A pentode would be far better in this spot.  The alternative would be using the triode in a grounded grid configuration (like the TRUMPET). 

jh

[cengclimbing]

Thanks Jim.  This sounds like a project that is way out of my ability level.  Maybe someone else wants to take it on?  It would be interesting to hear the results.

Cheers,
Chris

[poty]

The last stage is definitely provides some gain, but the gain can be obtainable from the previous two stages just by altering the resistors.
Second consideration is not the cathode follower, but tube gain stage as the last stage. There are a lot of examples of that: SRPP, mu-follower, Allen Wrights gain stages... The way opens more possibilities like transformation the Bugle + tubed output to full-fledged preamp with integrated phono stage.
IMHO it is better to made this more advanced (like new trumpet, based on JFET input stages) than to continue to use opamps, but who knows...

[hagtech]

One of the beauties of the BUGLE / BUGLE2 design is that of gain management.  I do NOT ask too much from a single stage, but rather spread out the gain duties as much as possible.  This does several things.  It prevents overloading of the next stage (critical when using passive EQ!).  It maximizes the bandwidth of each stage.  It minimizes distortion.  It improves stability. 

There are LOT of these little things going on in the BUGLE2 design that are NOT obvious.  Those extra 316/332 ohm resistors that seem to make no sense at all are actually some of the most important in the design. 

Go ahead and make the changes, but don't be surprised if some of that magic goes missing...

jh

[poty]

One of the beauties of the BUGLE / BUGLE2 design is that of gain management.  I do NOT ask too much from a single stage, but rather spread out the gain duties as much as possible.  This does several things.  It prevents overloading of the next stage (critical when using passive EQ!).  It maximizes the bandwidth of each stage.  It minimizes distortion.  It improves stability. 

Jim, as you may already noted I differ from you in recognition of some "magic" things and put my power solely into technical abilities of the schematic and listening practices. Then I don't agree with you about almost everything you wrote above.
First of all, you designed the usage of the (both) Bugles for a 20dB more gain and ironically saying - the gain is spread across the first two stages. Then - use the stages in this field is completely approved by you, the only question is about overloading, because MM carts have more than 10 times more output than MC.
Overloading: Passive EQ does not have any relations to overloading. Initially the overall output of most of MM catridges is 5mV (amplitude is around 7.2mV). Lets use twice as much of the amplitude - 15mV. Output of the first opamp (before the first filter) is 165mV (MM version) and 512mV (MC version). Well in the range of the ouput voltages even for 9V power supply (max 9V-2.5V=6.5V). After the filter there is: almost 0dB in low frequencies to -20dB in high frequencies. Then the most prominent part of the spectrum (due to RIAA it is high-freq) is -20dB lower: 16.5mV (MM), 51.2mV (MC). At the output of the second opamp there is: 181,5mV (MM) and 1.75V (MC). There is no overloading either. 
Bandwidth: 20dB has around 600kHz (-3dB), 30dB - around 350kHz, but it seems -40dB at 20kHz of reverse RIAA says more about the bandwidth...
Distortion: According to the datasheets there is no significant changes in THD+Noise in the output voltage region first of all. Second consideration - the voltage amplitudes exist in the 3-rd stage in the original design. More input voltage is better for signal/noise. The only possible problem - the input opamp gain, but IMHO 3 stages has more THD+noise than two.

There are LOT of these little things going on in the BUGLE2 design that are NOT obvious.  Those extra 316/332 ohm resistors that seem to make no sense at all are actually some of the most important in the design.

It is one of the "magic" thing in which I do not believe. The resistors is for "mythical 3.18ms corner in the RIAA response" which is someones opinion on the cutting head limitation. Not very trustworthy opinion sort to say. Lengthy reasoning and no argumentation. I tried several times to incorporate the pole in a design and never hear any difference. So the matter of importance of the resistors is: believing or not.

[hagtech]

Well, first of all I meant "magic" of the sound, not implying that there was any magic going on in the circuitry...

Yes, MM carts are indeed more of an issue for overloading than MC.  And 5mV is merely a "typical" value at 4.5cm/s velocity.  In practice you can get transient peaks 16dB higher than this.  And note those numbers are at 1kHz!  The values at 20kHz are 20dB higher, which is where you need to do your calculations.  Not only that, but splitting the EQ into two sections means you do NOT get all the treble cut in the first section (it takes the combination of both).  As a result the gain at 20kHz is not greatly reduced heading into the second stage opamp.  You can easily see this by measuring the response on an oscilloscope (especially using square waves via the iRIAA filter).  This again was on purpose, to keep noise to a minimum.  Obviously putting the EQ in a feedback network will result in a lower noise floor, but I do not like the sonic penalties incurred by such a topology.  Good design is about balancing a number of compromises.  As is, the BUGLE does not overload given real world conditions.  Increasing the gain of either the first two stages can compromise that - and lead to adverse affects in the sound.  You'll hear it as a shrillness on transients and/or merely a fatiguing of your ears over time.  I mis-spoke if I had said the "gain is spread across the first two stages".  It's actually across all three. 

And there are several more 300 ohm resistors in the circuit besides the 50kHz corner ones.  I should have specified they were the important ones.  And yes, not everyone can hear the effects of the 3.18us corner.  I can't detect it directly, but more as how it impacts the "air" and ambiance of a recording.  Here's a good test:  Can you hear a Nissan Leaf coming?  I can.  The electronics give off quite a whine.  And it's not even that high up in frequency. 

Bandwidth of the opamps in this circuit is not related to EQ.

jh

[poty]

...MM carts are indeed more of an issue for overloading than MC... 5mV is merely a "typical" value at 4.5cm/s velocity.  In practice you can get transient peaks 16dB higher than this.  And note those numbers are at 1kHz! The values at 20kHz are 20dB higher, which is where you need to do your calculations.  Not only that, but splitting the EQ into two sections means you do NOT get all the treble cut in the first section (it takes the combination of both).  As a result the gain at 20kHz is not greatly reduced heading into the second stage opamp...

I was thinking where to stop the quoting. Decided at this place. You've said about AFC of cutting gear (I mean +20dB at 20kHz), but this is not directly related to the absolute amplitude at this frequencies. The total power of harmonics at this frequencies is usually 10-30dB less than at 1kHz. You can see this dependencies in many professional and Hi-Fi/Hi-End devices: multiband amplifiers, modern acoustic systems and so on. Even in the vinyl era the RIAA +20dB at 20kHz was chosen intentionally - it doesn't greatly affect the width of cutted track which allows to increase the density of tracks per a record surface.
Another consideration: I got the nominal value of the record which already includes the boosted hi-frequencies.
I understand splitting EQ and estimated only the first part of the EQ as -20dB @ 20kHz. Don't you agree with this estimation? Then - count this by yourselves - you'll get the same numbers.
Let's count your example of full power 20kHz as a "musical signal": (5mV(RMS)+16dB)*sqrt(2)=45mV in amplitude. MC-version output of the first opamp would be around 1.5V, after the first part of the filter - 150mV. MC-version output of the second opamp - 5.2V - still under the opamp limit (9V supply) of 6.5V. We are speaking here about 15V supply for which the limits are 12,5V - way more than this never happen amplitude. This signal will definitely fry many ASs twitters and make a listener deaf.

Obviously putting the EQ in a feedback network will result in a lower noise floor, but I do not like the sonic penalties incurred by such a topology.

I didn't say a word about EQ in the feedback network, but if we speak about feedbacks: I do not like any feedback: frequency-dependent or not.
Both Bugle versions have also a mismatched (and also frequency-dependent) source impedances (as described on page 9 of the opamp datasheet) for the second and the third stage. "The effect is increased distortion due to the varying capacitance for unmatched source impedances greater than 2kOhm."
You have not mentioned also my second suggestion - use a tube amplifier circuit instead of cathode follower and save "MM-values" for the first two stages.

I mis-spoke if I had said the "gain is spread across the first two stages".  It's actually across all three.

You mis-readed my explanations:

you designed the usage of the (both) Bugles for a 20dB more gain and ironically saying -  the gain is spread across the first two stages.

And there are several more 300 ohm resistors in the circuit besides the 50kHz corner ones.  I should have specified they were the important ones.

No, you meant the resistors I answered you about, because the only more around 300Ohm resistor in the original Bugle is at the output. Resistors at the output are nor novel not very important ones. Very low output impedance of the opamp nulls all line instabilities. Certainly the precautions are a good things, but not important ones. The (input to opamp) resistors in the Bugle 2 also (I think) are the part of the decoupling circuits, but (surely lowering frequency dependence from the preceding filters) adds to impedance mismatching.

And yes, not everyone can hear the effects of the 3.18us corner.  I can't detect it directly, but more as how it impacts the "air" and ambiance of a recording.

I wonder - how people with such "ears" differentiate things? For example, in the Bugle there are rather prominent frequency-to-phase dependence in the hearing area, but Hi-End persons hear only some (always positive!) changing way over the hearing ability of any person and don't hear the obvious.

Here's a good test:  Can you hear a Nissan Leaf coming?  I can.

Haven't ever come across the car, so can't say surely.

Bandwidth of the opamps in this circuit is not related to EQ.

Yes, but artificially lowering the signal to 40dB@20kHz says more than 3dB@ >300kHz.

[hagtech]

Actually, the 300 ohm resistors on the outputs are the most important.  It's just not obvious what they do.  You have to take the big picture into account - and by that I mean interconnect cables and the load of the linestage / volume control / amplifier being driven.

At audio frequencies the interconnects present essentially a capacitive load to the output stage.  Without the 300 ohm decoupling, this capacitance is sitting right on the output of the opamp.  Adding such a capacitance leads to instability, ultrasonic ringing, and worst of all, slew rate limiting which breaks the feedback loop (and hence no relation between input and output until it catches up).  Decoupling the load from the feedback network prevents such issues.  The opamp can do its thing without having to deal with some unknown load. 

These things are usually ignored by most designers, as they assume the capacitance to be negligible.  Well, audio is about things that are practically unmeasurable (I still can't define how to measure good soundstaging).  And it is the little things that add up.  That's what I mean when I say all the little things in the BUGLE add up. 

Similarly, the input 300 ohm resistors do two things.  They provide a little bit of ESD protection by working with the input diodes on the opamp to limit current.  They also form a parasitic RC low pass filter with the input capacitance of the opamp.  This helps to cut RF energy above 10MHz, where radio signals can still be demodulated by the junctions in the input JFET pair.  Again, none of this is obvious to the lay person and many people who copy my circuits leave them out not understanding their intended function. 

I'll give you another example.  We were talking about the difference between 20dB and 30dB gain per stage.  The differences are easy to neglect, no?  Well, let's do a thought experiment.  What if I had five 20dB stages cascaded.  That would give me 100dB of reasonably wide bandwidth and low distortion.  Now what if we tried to do the same thing with one stage?   100dB with a single opamp presents a number of issues.  The open loop gain is not high enough to result in much negative feedback leading to high distortion.  The feedback resistor will be ridiculously high, and bandwidth will necessarily be low.  Much lower than with the five stages.  Ok, that's a pretty extreme case, but I use it because it makes things more obvious.  The same issues exist in the 20dB / 30dB example, just to a lesser degree.  My point is that they do exist and you should not always ignore them. 

jh

[hagtech]

FYI - I did do a sound check on my ears.   Made it to 12kHz easily, but then it dropped off super quick and 13kHz was gone.  My 11 year old daughter made it to 18kHz with the same equipment.

And having done this calibration, I'd guess the Nissan Leaf to be in the 8kHz to 10kHz range...

jh

[poty]

Sorry for not answering quickly - on vacation now. Trying to estimate the example you've given.
I did the (frequency) check 2-3 years ago. In my case the result greatly depended on my conditions. Sometimes I recognized 16-17kHz, sometimes even 12kHz was in doubt. I don't think it is much important, because at the frequencies a person can mainly only detect "something audible" and use it (maybe) for recognizing the direction to the source, nothing more.
According to the interconnects... I think the Bugle is not positioned as Hi-End gear and users will use "matched class" interconnects. While the resistor at the output definitely solves (at least eases) the problem of stability for the opamp I can't agree with your estimation of slew rate. Instead of "active circuit problems" we'll have "passive filter problems". The 330 Ohm resistor connected to - say - 1500pF cable (remember matched class) gives us a filter with 320kHz pole. I agree that in this case you may decide for the less difficult problem, but in some other cases - it may be preferable to use some other output stage layout, supplying successfully to higher class interconnects without additional parts.

[hagtech]

Regarding slew rate - it helps to think in extremes, then the problems become apparent.  For example, let's place a 1uF film capacitor on the output of that opamp.  I can pretty much guarantee the slew rate will go to hell.  Once the input signal demands a fast transient, the output can only go so fast, and slews at the current limit of the output stage, resulting in a ramp waveform.  During this time, any other musical content contained in the input signal does NOT show up at the output.  Only the ramp.  We have gone open loop.  Eventually, the output catches up to where it should be and the loop closes and acts like an amplifier again.  Well, unless of course there is some integration component in the feedback which will cause something called "windup" (but that's for another day).

Change the load capacitance to 1pF and the problem pretty much goes away.  The BUGLE resides somewhere in the middle.  We DON"T know exactly what the loading capacitance is going to be.  I play it safe and add inherent protection, such that it becomes very difficult to cause problems.  Most other designs do not have this protection.

Oh, one other thing.  The stability of that opamp depends on how it was designed, it's output drive capability, and the gain it is operated at.  Low gains (such as +1) are the most unstable of conditions.  That is why I shifted a bunch of the BUGLEs distributed gain to the output stage.  Not only does it improve stability, it also ends up increasing overload capability (as discussed previously).

Lots and lots of little things adding up to make one big difference.   

jh

[Jaxn]

Thank you for this most educational discussion, Gentlemen. I love reading the posts of those who know way more than I do...........I always manage to learn a little something.

[poty]

Well, let's do a thought experiment.  What if I had five 20dB stages cascaded.  That would give me 100dB of reasonably wide bandwidth and low distortion.  Now what if we tried to do the same thing with one stage?   100dB with a single opamp presents a number of issues.  The open loop gain is not high enough to result in much negative feedback leading to high distortion.  The feedback resistor will be ridiculously high, and bandwidth will necessarily be low.  Much lower than with the five stages.

Well, better late than never. 100dB is a very high gain. Of course it can't be achieved by only one stage, but let's be more practic. Let's take OPA2134 as the base opamp powered by +/-15V and compare two-stage device with 50dB each to five-stages device with 20dB each. I used the official datasheet of Burr-Brown OPA134, OPA2134, OPA4134 and counted s/n, IMD and bandwidth for both versions:
Version........S/N..........IMD............BW(-3dB)........dGain@20kHz..........S/N@20kHz
2 stage........37dB........0.03%........23kHz.............-2.34dB...................17.7dB
5 stage........36dB........0.78%........490kHz...........0dB.........................16.5dB
As you can see even at this extreme case the results are comparable and bearable. The OPA2134 is not very good at GBWP and slew rate which adds to difference in the bandwidth.

Regarding slew rate - it helps to think in extremes, then the problems become apparent.  For example, let's place a 1uF film capacitor on the output of that opamp.  I can pretty much guarantee the slew rate will go to hell.  Once the input signal demands a fast transient, the output can only go so fast, and slews at the current limit of the output stage, resulting in a ramp waveform.  During this time, any other musical content contained in the input signal does NOT show up at the output.  Only the ramp.  We have gone open loop.  Eventually, the output catches up to where it should be and the loop closes and acts like an amplifier again.

It is difficult to count this params, so I used a simulation program to measure the things. There are the results:
1. Directly connected 1uF capacitor to the output of +/-15V powered OPA134, connected as 20dB gain, 0.25V input signal:
Difference in Gain @20kHz: +0.4dB
BW (-3dB) on the capacitor: 147kHz
90grad phase at the output of the opamp @ 94kHz
Slew rate on the capacitor: 42mV/us
Result: there is a possibility to oscillate at the frequencies more than 94kHz, but the bandwidth is OK. Slew rate is small.
2. 1uF capacitor conected through 332 Ohm to the output of +/-15V powered OPA134, connected as 20dB gain, 0.25V input signal:
Difference in Gain @20kHz: -32.4dB
BW (-3dB) on the capacitor: 477Hz
90grad phase at the output of the opamp @ 3.2MHz
Slew rate on the capacitor: 6.5mV/us
Result: there is a lot of stability, but the bandwidth is completely unusable. Slew rate is even smaller than in the first example.