300MHZ 0 bias AMPLIFIER

[ccurtis]

You might consider bootstrapping the video op amp you selected. Bootstrapping to give greater output voltage is described at:

EDN PDF

Linearity is maintained since the bootstraps are included in the feedback path.

Low idle current is maintained as well.

I haven't used this technique, so I can't speak for it personally. I looks workable for the application. Your output power requirement is very small if you are driving electrostatic CRT plates.

 

[C&B]

You might consider bootstrapping the video op amp you selected. Bootstrapping to give greater output voltage is described at:

EDN PDF

Linearity is maintained since the bootstraps are included in the feedback path.

Low idle current is maintained as well.

I haven't used this technique, so I can't speak for it personally. I looks workable for the application. Your output power requirement is very small if you are driving electrostatic CRT plates.

Nice, this is the circuit I was thinking of only didn’t have enough horsepower to bring it to fruition.

I see the concept of the design and realize it will fit into my application with some adjustment; my application is not dual supply.


Thank You!

 

[C&B]

try LTSpice available for free from Linear Technology - Linear Home Page

it draws schematics and simulates the operation of the circuit.

the reason those video amps are dissipating so much heat is that they're operating in class A. a class A amplifier is always in conduction and so dissipates a lot of heat. sounds like you're trying to eliminate the quiescent current, which means using class B operation or class C operation. the problem you will experience is that class B and class C operation is very nonlinear. if the device in question is a color monitor, any nonlinearities will be very evident in the picture. what you are describing sounds like a kind of a push-pull class C amp, which is fine if your drive signal is digital, on or off, but if you're displaying analog video, it won't work. if you're displaying analog video, the quiescent current and the heat are neccesary evils.

Yes, those facts are what drove me to attempt a design that did not consume so much power in the idle state. These video drivers were designed years ago and I thought with today's technology, mainly op amps, it should be possible to design a video amplifier with good linearity and very low or no quiescent current. What I didn't know is how difficult it's to understand how the basic components work, it's like each component is a little of everything. It's not fair that the a diode is a capacitor sometimes!

 

[ccurtis]

Your UHF bandwidth requirement is the hardest to meet. As stated before, it will require careful selection of components and great design/layout skills. It took the best oscilloscope manufactures years to come to make oscilloscopes with high voltage, 500 MHz bandwidth capability, and the cost of those vertical amplifiers is still quit high and they do not have the added requirement of low idle current.

 

[unclejed613]

a class AB design might reduce the quiescent current, but not eliminate it completely. the same is done with audio power amps. class A audio amps run at huge idle currents, like 1 to 5 amps. class AB power amps run idle currents run idle currents of 10 to 50 mA. that's a 99% decrease. if the amp chips (i'm guessing they're the phillips green plastic modules) are class AB already, my guess is that you aren't going to get much of a reduction in idle current. if they're running class A. you need to look at the spec sheet for the module. what's the part number of the module?

 

[C&B]

a class AB design might reduce the quiescent current, but not eliminate it completely. the same is done with audio power amps. class A audio amps run at huge idle currents, like 1 to 5 amps. class AB power amps run idle currents run idle currents of 10 to 50 mA. that's a 99% decrease. if the amp chips (i'm guessing they're the phillips green plastic modules) are class AB already, my guess is that you aren't going to get much of a reduction in idle current. if they're running class A. you need to look at the spec sheet for the module. what's the part number of the module?

The amp hybrid is a VPA30; the data sheet is attached in post #8.

Let me ask a hypothetical question, in the circuit attached on post #5, assuming the op amp slew rate was insignificant compared to the operating frequency of the amplifier, would the output of the circuit create significant distortion?

 

[unclejed613]

ok, i see now.... the VPA30 is already biased class AB, so there's not a lot you can do about that. if you look at the output transistors, you will see a pair of forward biased diodes between the bases. this biases each transistor just barely on, so there's not a lot of quiescent current there to begin with. assuming that the supply voltage is about 100 volts, a 20mA idle current is about 2 watts of dissipation. with signal this will increase quite a bit. the module is a very simple amplifier. i don't think you will get the performance you are looking for using an opamp. power op amps that can handle the high rail voltages have nowhere near the bandwidth (300khz, NOT Mhz) and opamps that have the speed are low voltage devices. you could make a hybrid amp with a high speed op amp and a similar output stage as what's in the module, but you still have about the same amount of heat to contend with. your amp diagram in #8 might work, but regardless of feedback through the op amp, the transition between conduction on the transistors will still cause crossover distortion. part of the problem there is the op amp tries to slew across the conduction potentials and at high frequencies the slew induced distortion and the crossover notch distortion will become very noticeable. high frequenct op amps also have a tendency to oscillate during the swing through the transition zone, because like clipping, it's a circuit condition where there is no useful feedback. in digital logic this would be called an "undefined" state, just it's an analog version of one.

 

[C&B]

ok, Your amp diagram in #8 might work, but regardless of feedback through the op amp, the transition between conduction on the transistors will still cause crossover distortion. part of the problem there is the op amp tries to slew across the conduction potentials and at high frequencies the slew induced distortion and the crossover notch distortion will become very noticeable. high frequenct op amps also have a tendency to oscillate during the swing through the transition zone, because like clipping, it's a circuit condition where there is no useful feedback. in digital logic this would be called an "undefined" state, just it's an analog version of one.

Great response, very clear and on point!

I would like to discuss the distortion caused from TBC “transition between conduction on the transistors”.

If the feedback circuit is designed so the TBC frequency is constant and outside the domain of the amplifier operating frequency; is it possible to filter out the TBC from the output stage eliminating the TBC distortion?

I believe the worse case condition is when the input is at steady state; under this condition the op amp is in continuous oscillation creating output distortion; so let’s study this condition.

If the oscillation is controlled through the feedback circuit, tuned to 400mhz, and a 400mhz filter component is inserted in the output of the transistor pair; in theory, wouldn’t the output be at steady state (no distortion)? The assumption here is the filter will smooth out the 400mhz TBC signal while not affecting the 300mhz signal.

 

[C&B]

Need some help with LTspice. I have a feel for how to use it however my current level of electronics knowledge is slowing me down. I don't know how to create the singal input to properly simulate the circuit. I would like to see the distortion from the transistors, in the attached file, however the output does not show crossover distortion. What am I missing?

A few questions:
1)What methods are available to import third party devices?
2)How to configure the source signal to sweep a frequency range from 100mhz to 300mhz.

Any guidance will be appreciated!

 

[unclejed613]

there is a device in the special functions list called "modulator" you input a ramp waveform (using the PWL source Piece Wise Linear) into the FM input. your input limits are 0-1V. your frequency limits are set by inputting the limits as Mark and Space frequencies, and you set the output level by applying DC bias to the AM input.

importing 3rd party devices can be easy, or it can be difficult. it depends on what SPICE it was written for. many 3rd party models are written for PSpice or some other text-based SPICE. these are usually completely compatible with LTSpice. you usually (in the case of ICs like op amps) make a symbol for them (or edit an existing one and rename it for the device you are adding). the impossible ones are those written for a SPICE that uses binary models (like TI's TINA Spice). the difficult ones are the ones written for text based SPICE programs, but use a different set of internal functions and syntax. you would have to know how to write models well enough to figure out what is being done in the first model, and "translate" it into LTSpice syntax. sometimes these translations only require one or two lines or parameters in the model to be rewritten, sometimes it's more like 90% of the model needing translation or transliteration.


crossover distortion need not be visible to be there, often the distortion only shows up at higher frequencies and under load. part of the reason for this is the collector-emitter capacitance (Cce) discharging through the emitter resistors masks the effect until a load is applied. as the frequency increases, the slew induced distortion gets worse, and the Cce discharge rate can't keep up with the drive waveform. at the same time the frequency response of the feedback is also limiting it's ability to keep up with distortion products in the output. with the sum of all of these factors (and a few others), not only does the crossover notch get larger (actually it's the same duration, but it takes up a higher percentage of the time of each cycle), but the rest of the waveform is beginning to distort as well. another thing that changes during the "dead time" in the crossover region that contributes a lot to crossover notch distortion (especially with a load) is that the output impedance changes nonlinearly. this is an additional effect of having no useful feedback. this effect of output impedance shifts also contributes to oscillation while driving capacitive loads. i've done quite a bit of experimentation with amplifiers and feedback, and the effects that feedback, open loop gain, and physical output resistance has on the output impedance of amplifiers. the results were very interesting, as a matter of fact, i'll post some of my notes in my blog if anybody is interested. as a result of experiments, i came up with a method for setting the proper bias on amplifier output stages, by dynamically measuring the output impedance.


also, the problem you will have by letting the amp oscillate at 400Mhz, is you will worsen a condition that the OP was attempting to alleviate, heat dissipation. as i mentioned above, when you get near the upper limits of an amplifier's bandwidth Cce (and Ccb) have a more difficult time discharging. one of the side effects of this is what's called "cross conduction" or "common mode conduction" where the output transistors never completely turn off while the opposite one is being turned on. at this point both transistors are in conduction, but none of this current is going to the load, it's going from the - rail, directly through the transistors, to the + rail, generating lots of heat (and if you're not careful, letting out the magic blue smoke). by the time this begins happening, the crossover notch has been swamped by the current spikes, and you are now seeing an "opposite" of crossover notch. it causes the same type of harmonic content though. so letting the op amp "idle" at 400Mhz is not desireable. an oscillating video amp also causes smearing of the picture.

 

[C&B]

After studying the different effects you wrote about I have a much better understanding of the design challenges along with a new found respect for electronic engineers. I also realize the thrill of the design challenge and am seriously considering furthering my education in this field.

Regarding the project, I would like to go forward with this however with more realistic design parameters. I admit it is difficult to let go of my original concept however I recognize your expertise and realize the attempt is futile.

You wrote earlier it may be possible to design a similar amp, as the original hybrid, using discrete components. My new design will be just that only with the addition of a feedback loop. The feedback loop is a requirement necessary to eliminate a problem in the current design.

I understand that component selection is critical when designing high speed low noise / distortion video amplifiers and would appreciate any advice you have to offer. At some point, when functional in LTspice and before making a prototype, I will post the design for discussion.

 

[C&B]

Help with LTSpicce please. A simulation of the attached circuit should show some type of output distortion however the output looks clean. It there a option that causes the sim to ignore diode drops?

 

[unclejed613]

where are you sampling your output?

it looks as if this circuit is doing level shifting, in which case there won't be much distortion. there are 3 ways to check for distortion in LTSpice:

1) right click on the waveform and select FFT. this shows your output spectrum

2) insert a .four spice directive into the schematic and read the results in the spice log file

3)subtract your input voltage from the output voltage (remember to use multiplication or division if the circuit has any gain or loss) and plot a result waveform

using any of the above 3 methods will require you consult the help file, but if you get stymied, feel free to ask for help.

i also recommend you join the LTSpice yahoo group
**broken link removed**

 

[C&B]

A question regarding bias voltage. There are two diodes (D1,D2)used to create the bias voltage across output transistors (see attached). The idea is to set the DC current value through the diodes such that the sum of their voltage drops is equal to some ideal value so the output transistors are either in slight (micro amps) cross conduction or just less than that. The problem I see is it requires a significant amount of current through the diodes to achieve the desired voltage drops (to match the drops of the transistors). If I understand the curves for the hybrid we are talking 50ma explaining the incredible amount of heat being generated by the hybrid.

Question; Are there any diode types that have base emitter junction voltage that can be controlled with less current? I.E. Say 1ma emitter current = .4v drop and 5ma emitter current = 1v drop.

 

[unclejed613]

the diodes provide barely enough voltage to turn on the transistors, and act as a voltage reference for the transistor bias. it doesn't take very much current to do this. the diodes are most likely thermally coupled to the transistors to maintain a thermally stable idle current. silicon exhibits a -2.2mV/deg C temperature coefficient. if the transistors heat up without the bias voltage changing, the turn-on voltage of the transistors would drop, causing them to conduct more, increasing the current, increasing the temperature, etc.... until the transistors burn out. so coupling the diodes and the transistors thermally insures that the bias voltage applied to the transistors tracks the turn-on voltage of the transistors, and so the idle current of the transistors doesn't change with temperature. my guess from the schematic is that the idle current is probably around 500uA to 1mA. the diode current and the transistor base current are about equal. with a likely beta of 100 for the transistors, the base current and the diode current would add up to about 20uA (10uA each branch) so the resistors going from the rails to the diodes and bases would be calculated accordingly.

 

[C&B]

Your point on thermally coupling the diodes to the transistors is well taken, thank you. It was not a consideration in my redesign however in now. What is the standard procedure for implementing this type of coupling when using discrete surface mount components?

So why is the hybrid so hot? If the bias current is < 1ma and the output drive is < 3ma peak the max power consumption would be .4 watts. In the working application the hybrid is mounted to a 5” X 4” fin heat sink with a fan mounted on the heat sink. The running temperature on the heat sink averages 54C.

My original thought is the bias current is a function of the input signal level and to design the new circuit such that the bias current was constant. Basically, have three stages in the amplifier. The first stage being inversion and controlling gain;, the second stage being a high side voltage follower driving the bias diodes into a low side constant current source and the last stage being the push pull stage (see attached).

Any comments on the attached design concept are appreciated. I would like to build on this concept assuming a constant bias current does not have a negative effect on the circuit functionality.

 

[C&B]

This spice program is great! I am a bit excited, may have figured out how to see the crossover distortion! Had to post it!

Run the simulation and zoom in on the crossover point and observe the curves are touching, change the value of R1 to 2.2k and notice the results. Does this visual represent the crossover distortion?

 

[unclejed613]

if you look at the VPA15 data sheet, under the heading CURRENT CONSUMPTION, the actual operating current with signal is closer to 100mA. idle current of a class AB amp is just that the current when the amp has no signal and is driving no load. the operating current is the average current with signal and driving a load. the cathode of a CRT has current flowing all the time there's high voltage applied to the CRT anode. that's the beam current, and it comes through the module. you are actually sourcing the beam current through the transistors, so there's a lot more than just the bias currents. the beam current isn't obvious here, because you're treating the cathode as a purely capacitive load, but you're actually feeding a voltage controlled current source (or sink, depending on how you look at it), and at all times except a short blanking period, there's beam current flowing. what you would have to do to simulate the cathode being driven, would be to download a tube model and use it with the triode symbol or the pentode symbol (closer to acting like the real thing, but even the triode would give you an idea what's going on here). granted, you would need a tube model with a -50V grid cutoff voltage, but it would better model the load you're driving.

 

[Dick Cappels]

I have designed amplifiers like this, and if you are planning to use this as a video amplifier, you will spend a very long time to make this work. The best configuration is a push-pull cascode circuit. This was the basis for TRW's (later Motorola's) wideband video cathode driver hybrids.

Once inherited a product that used discreet bipolar transistors, connected as a cascode with a inductively peaked resistive collector load, which achieved about a 2.5 ns rise time, but it required massive heatsinks for the transistors (the transistor's junction to case thermal resistance is pretty high because of the small die) and many watts dissipated in the resistive load. And this only provided about 50 V P-P swing.

Hybrid video amplifiers have been dominant in the CRT monitor industry for the last 15 years or so. I suggest you see if you can find a discarded color monitor and salvage the hybrid amp. And then the fun with layout and decoupling can start.

 

[unclejed613]

the driver stage of the VPA30 and also NatSemi's LM2419 CRT driver chip are almost (but not quite)identical cascodes. you can see that there's a lot of similarity between these two CRT driver chips and the cascode amp shown for reference. the class AB output stage is not just there for current gain, but to isolate the cascode from loading effects. two advantages of the cascode amp are 1)improved linearity due to the nearly constant Vce of the lower transistor, and 2) the capacitances of the two transistors are in series, reducing the total miller capacitance to less than the capacitance of either transistor alone (greatly improving high frequency response).