Thank you again for pointing out some of these basic practical reasons to use these types of devices and about bandwidth. At the same time isn't there a basic upper limit to the bandwidth necessary and the slew rate for a practical power amplifier? Doesn't this go with your statement about the -3db point of the input filter, is there any practical improvement in sound by pushing this past a certain point?
Hi Mr. Cordell
Do you mine posting the schematic. I would like to try that when I have time.
I absolutely agree with this. Using fast transistor has a lot of advantages. Not to mention you have lower input distribution capacitance described in your book.
In my original 3EF Diamond, there is no particular sign of oscillation until I changed out the KSC/KSA pre drivers to slower transistors with higher Ccb, then was oscillation big time.
Hi Alan,
I apologize for not getting back to you sooner. I designed the parasitic oscillation sniffer a number of years ago and never finished it or tested it. Not sure where the schematics are now anyway. I would generally not want to post something like that that I had never even finished or tested it. The concept is quite straightforward, however. Fairly steep high-pass filters with lots of attenuation at 20kHz and below. For odd-order filters, put the real zeros passive at the front end to get rid of high-amplitude audio test or program signals first, the do the rest of the HPF with wideband amplifiers in between. I think I used a shottky diode for the detector.
Cheers,
Bob
Hi Mr. Cordell
I designed a detector with 4 order Butterworth highpass driving a CFA AD8009 to drive the detector, it's only in simulation stage. I am planning to use the detector output to trigger a LM555 with say 2sec output pulse to drive a LED to detect any burst above 100KHz. The highpass fc is set to 200KHz. Tell me what you think.
Attached is the .asc file and the opamp AD8009.txt I used. I am planning to use AD8011 400MHz CFA amp, that's the only one that comes with DIP package. I use AD8009 because I used it in my other projects.
I use inverting amp so the 150ohm resistor current limit at the input to protect the opamp. The AD8011 use even a higher value resistor for better protection. A lot of models from Analog devices cannot be used directly by LTSpice, I have to modify it to make it work, I am just too lazy.
Thanks
Last edited:
Hi Alan,
This circuit looks like a good start. Here are a couple of objectives to keep in mind. First, we do not want the circuit to trigger at full amplifier output up to at least 20kHz. This full output should be defined to be at least 40V rms, corresponding to 200w/8 ohms.
Secondly, we want the circuit to trigger on a parasitic oscillation as small as perhaps 100mV rms. Bear in mind that VHF parasitic oscillations evident at the output of an amplifier may be quite small in amplitude, especially if they originate as a local parasitic oscillation in a stage earlier than the output stage. This means that the circuit gain at 20kHz must be at least 52dB down from that at high frequencies. Your filter achieves this, with a difference of about 80dB.
Third, you want the rectifier circuit to be fast enough to capture a parasitic burst of only a few cycles, even if that burst is at 100MHz, where 1 cycle is only 10ns. At the same time, the filter capacitor output time constant must be long enough to store the burst long enough and with little enough ripple for the 555 or comparator to see it.
It looks like your circuit has a 3-dB bandwidth of a little over 500MHz up to the detector diode. This is probably more than necessary, and may make the circuit unnecessarily vulnerable to some RFI sources. I’d keep it to between 100MHz and 200MHz.
Your circuit does not have enough gain to see a 100mV rms parasitic oscillation, with the net gain of the circuit being only about –3dB at HF in front of the diode. Consider upping the op amp feedback resistor to 1k to get some more gain. This will reduce the bandwidth to about 150MHz. This will give you about 10dB, bringing your net HF gain up to about 7dB. A 100mV rms oscillation with a peak level of 141mV will then put about 300mV peak at the diode. Not really enough, given the forward drop of even the Shottky diode. AC-couple the output of the op amp to the diode circuit and forward bias the diode by a replica diode drop of the same type of diode. Accept a couple tens of mV positive offset from the diode at no signal. It will be thresholded out by the subsequent thresholding circuit anyway.
Look carefully at the output filter of the rectifier circuit in light of what is mentioned above with regard to capturing bursts of just a few cycles and holding them long enough for the comparator, or whatever, circuit to see them.
You might also consider putting back-back silicon diodes shunting to ground at the output of the filter so as not to seriously over-drive the op amp in the case of a high-amplitude burst. Not sure about this one.
You might also want to consider implementing some kind of a sensitivity control.
One of these days I might get back to the Parasitic Oscillation Sniffer. I think it is a worthwhile tool to evaluate amplifiers. I wonder how many commercial high-end amplifiers would show parasitic oscillation bursts on some program material.
Cheers,
Bob
Take a look at the Linear Technology LTC5507 " 100 KHz to 1 GHz RF Power Detector" as a replacement for the Schottky diode. See http://cds.linear.com/docs/en/datasheet/5507f.pdf It includes the "replica diode" (don't recall hearing that term before!) as well as the forward-biasing current circuitry and some temperature compensation, all in a 6-pin SOT-23 package.
No, I don't have personal experience with this exact IC, but recently I DID use a very close cousin - the LTC5505 - as the detector in a UHF field-strength meter. The biggest problem was a requirement for capacitor-coupling the input signal.
Dale
It is more complicated as I look deep into the design. The input voltage is very high, +/-40V to +/-80V!!! It will blow any IC RF detector!!!
I was going to change to active filter using R and C and opamp, BUT the input is way too high. Even if I clamp the input, the low audio frequency gets clipped and generates high frequency harmonics and cause false trigger to the detector.
The only way to prevent large audio signal to pass through is to have the passive 4 pole filter like in the original design. But even using the original design with the 4 pole LC Butterworth filter, the first inductor and capacitor is going to see a lot of voltage and current. The voltage of the pass band can be very high also. The inductor can get big. Also, we are working with 100MHz+, the parasitic of the inductor can get in the way because their resonance frequency is low.
It might not be as easy as it seems.
I was going to change to active filter using R and C and opamp, BUT the input is way too high. Even if I clamp the input, the low audio frequency gets clipped and generates high frequency harmonics and cause false trigger to the detector.
The only way to prevent large audio signal to pass through is to have the passive 4 pole filter like in the original design. But even using the original design with the 4 pole LC Butterworth filter, the first inductor and capacitor is going to see a lot of voltage and current. The voltage of the pass band can be very high also. The inductor can get big. Also, we are working with 100MHz+, the parasitic of the inductor can get in the way because their resonance frequency is low.
It might not be as easy as it seems.
Last edited:
Attached is the .asc file. I put a voltage divider at the input of the passive filter to limit the voltage to +/-15V to keep the current down. I add an extra opamp to get the gain back up.
ignore the active filter design, as I said before, the high input voltage is not going to work for this design.
ignore the active filter design, as I said before, the high input voltage is not going to work for this design.
Hi Dale,
This is a very good idea. Thanks for bringing it to our attention.
Cheers,
Bob
Alan you don't want an attenuator at the input. You already don't have enough gain. The passive HPF gets rid of the high audio amplitude. The passive HPF can handle high amplitudes. The kind of parasitic oscillations we are looking for are HF bursts, usually of only modest amplitudes. An amplifier that has a very high amplitude parasitic oscillation at a lower frequency, as from a global loop instability, will likely be able to be seen without need of the Sniffer. Nevertheless, the op amp should have some protection against very high amplitude oscillations above 200kHz. That should be taken care of by the clamp diodes I suggested. Proper simulation will enable a reasonable assessment of protection.
The key is that you have to think these things through, ponder for awhile, let the thoughts marinate, and do explorative simulations. Patience and good critical thinking are key.
Cheers,
Bob
Hi Mr. Cordell
Yes, definitely need time to stew on it for a while. It is not as obvious until I get deeper into it. So far, I feel the passive filter right at the input is a MUST as no RF opamp or other device can take the high amplitude of the power amplifiers. On top, if the signal is clipped at the front end, you generate higher harmonics and can become RF that false trigger the detector. That is, the detector treat the higher harmonics from clipping of the audio signal as oscillation.
I was think about the attenuation. The reason is the first capacitor and inductor. If I don't attenuate the signal, it's going to be high current driving into the first two component. You look at the spec of the inductors, it's hard to get a RF inductor that can handle over 500mA, also, the first cap has to handle high current. The second inductor and cap is not as critical, the signal is attenuated by the first pair already.
The saving grace is we don't worry about noise and distortion, we just want to rectify the RF to trigger the 555. That's the reason I attenuate the signal right at the input. Use the 4th order filter to get rid of the large audio signal first, then amplify the signal up for detection. I did not have enough amplifier to get higher gain as you can only get so much gain per amplifier. I can easily add one more opamp to jack the gain up.
After the passive filter, using clipping diode for protection is not going to affect the result. Because any signal passes through the filter is oscillation already, any higher harmonics generated by clipping is not a false signal anymore. I did not use diode, instead of using a current limiting resistor because the current is still significant even with the input attenuation. Using a big diode adds to the capacitance. But I can go either way. But without attenuation at the front, the signal can get too high and current through the diode can get to 1A.
My contracting work is heating up, I cannot spend a lot of time on this lately.
Thanks
Alan
Last edited:
Can you drive full power into this? It would be nice if it can survive. Then the output will be 40dB lower for the audio signal. Then just put another high pass filter to further attenuate the audio signal to drive the opamp circuit.
You use 1:100 divider before driving the Mini Circuit?
Well, if I put a 1:100 divider, I can use the 4th order active filter or even the RF detector suggested by Dchisholm for a lot less money. The difficult part is the large amount of voltage that burn any SS device and stress the first two components of even the passive LC filter. Mr. Cordell think we need to detect RF oscillation burst as low as 100mV amplitude. That's 1mV after the 1:100 divider. You still have insertion loss through the Mini Circuit.
I think I can safely use a 20:1 divider to drive the active filter using AD8011 without even using clamp diode, just current limit by a series resistor.
I was hoping you can drive the Mini Circuit directly from the output of the amp.
Last edited:
Since the minickts filter wants to be driven from 50 ohms source anyway, feed it through a 47ohm resistor. That will provide the specified filter shape and limit bad currents into the device. The voltage at the device will be about half the amp's voltage, applied power from a 100W audio amp would be about 4 watts into the 50 ohm filter (20^2 / (2*50)).
Last edited:
The Mini Circuit is spec 1W absolutely max. If you end up have to put a divider, then it's right back to what I had with my circuit that is a whole lot cheaper. The Mini Circuit is only 40dB attenuation, we need a lot more than that. My 4 components 4th order high pass is about -70 to -75dB at 20KHz which is the highest audio frequency.
I just pulled out a design with Butterworth filter. If I design with Chebychev filter, I can get even more attenuation of audio frequency. OR I can easily use 5th or 6th order filters with only 5 or 6 LC components.
I just pulled out a design with Butterworth filter. If I design with Chebychev filter, I can get even more attenuation of audio frequency. OR I can easily use 5th or 6th order filters with only 5 or 6 LC components.
Last edited:
