Author Topic: Cascode circuit vs. Active load circuit  (Read 3840 times)

0 Members and 1 Guest are viewing this topic.

Offline dzsekiTopic starter

  • Frequent Contributor
  • **
  • Posts: 524
  • Country: hu
Cascode circuit vs. Active load circuit
« on: March 13, 2019, 11:55:31 am »
The other day I was reading this ( www.ti.com/lit/pdf/sboa048 ) application bulletin about CRT amplifiers. in this article they mention the cascode circuit as a thing from the past, and they present a „state-of-the art” active load circuit in the form of Philips CR3425 hybrid. This circuit was not exactly unknown to me as I learned about this earlier in the AN-1013 app note too (www.ti.com/lit/an/snva031/snva031.pdf) Section 5.5. The common thing about the two is that they are rather vague about explaining the active load circuit.

To my understanding in the conventional cascode circuit the common base (Q1) part lessen the effect of the Miller capacitence of the lower common emitter transistor (Q2). The bandwidth of the circuit then ultimately relies on the time constant set by R1 and the corresponding stray capacitance around the collector of Q2. Transition times can be improved by emitter bypassing and charge pump to some extent.

Now, with the active load circuit seemingly there are two common emitter transistors tied against each other. They claim better bandwidth on this circuit but I don’t really see why, because for one, this circuit does nothing to eliminate the Miller effect, so what is the trick?

I was not able to find the datasheet of the CR3425 hybrid, but was able to find the OM976/1's which is comparable. I attach it for reference.
HP 1720A scope with HP 1120A probe, EMG 12563 pulse generator, EMG 1257 function generator, EMG 1172B signal generator, MEV TR-1660C bench multimeter
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 15149
  • Country: de
Re: Cascode circuit vs. Active load circuit
« Reply #1 on: March 13, 2019, 04:04:53 pm »
I think the caskode and active load address 2 different weak points.

The problem addressed with the active load is reducing the power consumption and approximately keeping the standing current constant. With just the resistive load the current through the resistor varies quite a bit and thus a higher power and thus larger transistor has to be used. So the effect on improved BW is due to the possibility to use a smaller transistor.  As an in between step there is bootstrapping of the load. This keeps at least the AC current constant and could with a well behaved signal nearly replace the active load.

Ideally one could combine the active load and cascode.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Cascode circuit vs. Active load circuit
« Reply #2 on: March 13, 2019, 05:02:26 pm »
There is no trick.  There is nothing wrong with the active load which doubles the transconductance (push-pull common emitter output stage) but to make the circuit even faster, both the input transistor and the active load can have cascodes added.

The cascode configuration has a second advantage besides reducing the Miller effect.  It also allows a low voltage high speed input transistor to be used with a higher voltage lower speed cascode transistor gaining the advantages of both.  But these two transistor types would not generally be available in a monolithic IC at the same time which is why hybrid construction with high voltage discrete cascode transistors was common in the fastest video amplifiers.

Oscilloscope z-axis amplifiers used all of the usual active load and cascode configurations found in CRT video amplifiers but among variations, also had another one shown below from the Tektronix 400MHz 7834 and 500MHz 7934 oscilloscopes where the common emitter output stages are replaced with emitter followers and the error amplifier is a 5th transistor.  The emitter follower driving a common base transistor is more commonly known as a differential amplifier although in this case it is made with complementary transistors which is a characteristic shared with the folded cascode.

The older 200MHz 7704A used a configuration like your examples but the common emitter NPN output transistor has no cascode and the PNP active load does as shown below.  This example makes me think that the choice of circuit topology has as much to do with available semiconductors as performance requirements.  A high voltage and fast NPN transistor was available when the 7704A was designed however the cascode configuration was needed for the active load because a suitable high voltage and fast PNP transistor was not.  Just to be sure, I checked the parts and it illustrates this perfectly:

Q41113   151-0221-00   PN4258   6-2&15   PNP 900MHz 12V 3pF Ccb
Q41114   151-0208-00   2N4036   6-3&15   PNP 60MHz 90V 30pF Ccb
Q41115   151-0124-00   2N3501   6-6&12   NPN 150MHz 120V 4.5pF Ccb

Note that the 2N3501 is selected for rb'Cc<50ps (collector base time constant) so the actual Ft is much higher than the datasheet specification.  Add an NPN cascode in there and a faster lower voltage part could be used without selection like with the 7834/7934 example but for whatever reason, Tektronix did it this way.
« Last Edit: March 13, 2019, 05:07:38 pm by David Hess »
 
The following users thanked this post: dzseki

Offline Sylvi

  • Regular Contributor
  • *
  • Posts: 85
  • Country: eg
Re: Cascode circuit vs. Active load circuit
« Reply #3 on: March 13, 2019, 05:20:11 pm »
Hi

As post-2 says, the cascode extends frequency response by shielding the lower Q from the load, keeping the collector voltage for the lower Q nearly constant so it has effectively wider bandwidth. The output impedance of the upper Q is quite high and it needs a light load to provide maximum bandwidth and signal swing.

Bootstrapping the load on the cascode helps maintain the cascode bandwidth and requires that the upper collector is buffered from the bootstrap point and load.

The "active" load is just a current source, which is a standard feature in amplifier voltage gain stages. As the name implies, it keeps the current through the circuit constant and is the same effect as bootstrapping but without some of the turn-on problems a bootstrap might introduce. A benefit of the current source is that the circuit has full performance right from a very low supply voltage up to the full design value.

In a CRT, there is a need for fast-rising voltage steps and to handle high-frequencies, so anything that reduces the need for high-value compensation caps is good: cascoding, current-sources, bootstrapping.

Note that all current sources are not made equally. The 2-BJT type with local feedback has wider bandwidth than the simple buffered voltage references. The speed can be improved if you add a small base-stop to the current-sense BJT, then add a small cap from its base to the collector of the pass element of the current source. 15pF is typical. The current source can also be cascoded.

I think you would look at it as adding the current source is the best first step to improve the circuit, then add the cascode as the next step UNLESS the circuit already has an EF buffered output with a bootstrapped base resistor, then add the cascode first.
 
The following users thanked this post: dzseki

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Cascode circuit vs. Active load circuit
« Reply #4 on: March 13, 2019, 05:51:29 pm »
I think you would look at it as adding the current source is the best first step to improve the circuit, then add the cascode as the next step UNLESS the circuit already has an EF buffered output with a bootstrapped base resistor, then add the cascode first.

And if a transistor with sufficient speed and voltage is not available, the cascode is step zero.
 

Offline dzsekiTopic starter

  • Frequent Contributor
  • **
  • Posts: 524
  • Country: hu
Re: Cascode circuit vs. Active load circuit
« Reply #5 on: March 13, 2019, 08:07:14 pm »
Thank you for the valuable information.
So far my go to circuit was the simple cascode circuit with complementary emitter follower buffer like pictured in the opening post. I was playing with variations of this circuit previously, and I hit a performance barrier. With this 70-80V supply range a 400 Ohm load resistor is the practical minimum that can be handled by transistors (dissipation and current). I have been using mostly BFQ232/252 and MRF548/9 transistors for the common base part, they all have ft in excess of 1GHz.
Text books would suggest that the common base circuit should have bandwidth up to the ft of the transistor, then the common emitter transitor easily can be a 5GHz part, still on the output I never could break the 3ns rise/fall time barrier. In fact with slower transistors (3-500 MHz) the same performance was possible. Obviously the bandwidth is limited on the first hand by the time constant of the 400 Ohm resistor and the stray capacitances around the collector of the common base transistor which couldn't be much less than 5pF. So I would be curious to build the active load circuit. But as Kleinstein also suggested I was under the impression that the sole intention of the active load circuit was to lower the disspation. The attached Philips hybrid datasheet proved otherwise, alsoshort form tables suggesting that hybrids with sub 2ns rise times also exsited (OM977, CR4424) -although no datasheets or pictures to be found from these parts...
HP 1720A scope with HP 1120A probe, EMG 12563 pulse generator, EMG 1257 function generator, EMG 1172B signal generator, MEV TR-1660C bench multimeter
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Cascode circuit vs. Active load circuit
« Reply #6 on: March 14, 2019, 12:08:02 am »
The resistive load can have the advantage of using all NPN transistors and the fastest oscilloscope vertical CRT amplifiers worked this way.  Some stacked the output cascode to use faster but lower voltage transistors with a higher total voltage but I do not remember seeing this done above 100MHz/3.5ns.

I suspect the output capacitance of the cascodes is limiting performance.  There are special low output capacitance transistors which were used for this very application but they are difficult to find now.  The alternative is raising the quiescent current to charge the capacitance faster.

T-coil loads will double output stage Ft.

Ft doubling (unrelated to t-coils) could be used in the input stage but it means 4 times as many input transistors.

A distributed design could cancel out some of the output capacitance but means lots of stages.  This may not be feasible at this frequency without special construction.

 

Offline dzsekiTopic starter

  • Frequent Contributor
  • **
  • Posts: 524
  • Country: hu
Re: Cascode circuit vs. Active load circuit
« Reply #7 on: March 14, 2019, 08:20:52 am »
The resistive load can have the advantage of using all NPN transistors and the fastest oscilloscope vertical CRT amplifiers worked this way.  Some stacked the output cascode to use faster but lower voltage transistors with a higher total voltage but I do not remember seeing this done above 100MHz/3.5ns.


The vertical output of my HP 1720A scope also use sliding casode circuit with two common base transistors in series with self biasing, but this help little with the stray capacitances, there only 50v is used as supply, and the work resistor is 165 Ohms, so that is a much better point to start with.

For normal CRT drivers a single +80-90V supply is enough to achieve 40-50Vpp output, my application is though a little different, needing about double the output voltage, but at the same speed (or even faster :P). This is more easily done with „bridged” output. One could use two such positive drivers in antiphase and a clamped DC level shifting on the output for the negative side. The downside of this application is that the cascode circuit has an inherent assymetry in its performance. A better solution is to build a fully symmetric cascode (made from NPN transistors for the positive side and PNP transistors for the negative side).

I suspect the output capacitance of the cascodes is limiting performance.  There are special low output capacitance transistors which were used for this very application but they are difficult to find now.  The alternative is raising the quiescent current to charge the capacitance faster.

The above mentioned MRF548/549 transistors are such and I have plenty of them to play with. They are cased in TO117 ceramic package with isolated mounting screw, and they have two base legs symmetrical to collector-emitter pins for better shielding. The BFQ232/252 transistors while comparable on paper are cased in simple TO126 package with collelctor connected to the metal mounting surface, which is really no help of keeping the collector capacitance low, so the performance of these transistors in real life is inferior.

HP 1720A scope with HP 1120A probe, EMG 12563 pulse generator, EMG 1257 function generator, EMG 1172B signal generator, MEV TR-1660C bench multimeter
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Cascode circuit vs. Active load circuit
« Reply #8 on: March 15, 2019, 02:26:00 am »
The resistive load can have the advantage of using all NPN transistors and the fastest oscilloscope vertical CRT amplifiers worked this way.  Some stacked the output cascode to use faster but lower voltage transistors with a higher total voltage but I do not remember seeing this done above 100MHz/3.5ns.

The vertical output of my HP 1720A scope also use sliding casode circuit with two common base transistors in series with self biasing, but this help little with the stray capacitances, there only 50v is used as supply, and the work resistor is 165 Ohms, so that is a much better point to start with.

I forget which Tektronix vertical CRT amplifier did this but I think it was just more economical at the time and not for performance.

Quote
For normal CRT drivers a single +80-90V supply is enough to achieve 40-50Vpp output, my application is though a little different, needing about double the output voltage, but at the same speed (or even faster :P).

I suggested it because it is a way to use lower voltage but faster transistors at a higher total voltage.  In some cases it will be the only option.

Quote
I suspect the output capacitance of the cascodes is limiting performance.  There are special low output capacitance transistors which were used for this very application but they are difficult to find now.  The alternative is raising the quiescent current to charge the capacitance faster.

The above mentioned MRF548/549 transistors are such and I have plenty of them to play with. They are cased in TO117 ceramic package with isolated mounting screw, and they have two base legs symmetrical to collector-emitter pins for better shielding. The BFQ232/252 transistors while comparable on paper are cased in simple TO126 package with collelctor connected to the metal mounting surface, which is really no help of keeping the collector capacitance low, so the performance of these transistors in real life is inferior.

I have seen this issue come up twice.  Once was in the Tektronix 547 oscilloscope which uses a couple of these very low output capacitance transistors for performance.  Trying to find a modern replacement for these transistors was not a trivial exercise like I though it would be.  The other is in the VAS stage of audio power amplifiers where designers used the transistors designed for high voltage CRT cascodes.  However these transistors are now rare with CRTs having become obsolete.

Example parts include the still manufactured KSA1381/KSC3503 which are 300 volts, 150 MHz, and specifically have low output capacitance and low reverse transfer capacitance.

There are two other things Tektronix did to help with this type of circuit:

1. They had special versions of these transistors made with the base tied to the case or if an RF package was used, a dual base instead of dual emitter package.  Broadband push-pull common base cable TV distribution amplifiers also used this RF package.

2. If a suitable low capacitance part with the required current and power rating was not available, they sometimes used two in parallel.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf