suggestion to optimize impedance matching of linear amplifier output

Thread Starter

shengwuei

Joined Aug 22, 2008
33
Hi Sir,

I have a linear amplifier circuit as below, description of each stage from left to right :
  • linear amplifier : an OPAMP linear amplifier with output resistance around 3 Ohm
  • back-to-back diode : for some reason I have to add this diode to block the noise from amplifier output
  • PCB trace : the amplified signal is routed through a 15 inch trace (with propagation delay about 3ns) and trance impedance is 50 Ohm
  • multiplexer : the signal is switched through a multiplexer(with resistance about 4.5 Ohm) before load
  • load : for simplicity, just disconnect the load for now

1672710434082.png

the circuit is simulated in LTSPICE(https://www.analog.com/en/design-center/design-tools-and-calculators/ltspice-simulator.html), see right figure for the result, waveform of amplifier out is set to a 5 MHz/2 cycle pulse, as you can see there is distortion on peaks of the waveform at load side as marked by the red circles

the simulation result fairly matches real measurement, see the figure below, I had tried to compare simulation result and scope measurement with different loads(50 Ohm, a real transducer...) and believe the simulation can represent real situation of the circuit on my PCB

1672711208930.png

My target is to reduce/avoid the distortion as described above, I had tried to use different diodes in the simulation but the distortion is still obvious in all cases, any suggestion is very appreciated, the simulation file of LTSPICE is also attached(TX_diode_simulation.zip), thanks.
 

Attachments

Thread Starter

shengwuei

Joined Aug 22, 2008
33
The diodes are generating the distortion.
What noise are you trying to suppress with the diodes?
Hi crutschow,

the terminal of load side is also connected to my RX circuit through a T/R switch, without the diode noise from linear amplifier output would effect RX signal SNR, thanks.
 

crutschow

Joined Mar 14, 2008
34,283
the terminal of load side is also connected to my RX circuit through a T/R switch, without the diode noise from linear amplifier output would effect RX signal SNR, thanks.
A diode is a very odd way to eliminate noise.
If you want a linear signal then you need some other way to reduce the noise, such as a filter or a lower noise linear amp.

You are showing us only a part of your "solution", not what your problem is.
What is the configuration of the linear amp?
 
Last edited:

Thread Starter

shengwuei

Joined Aug 22, 2008
33
Hi Crutschow,

The problem is as depicted in the figure below, there is noise coupled onto the trace of linear amplifier output from the traces of adjacent layers, and since the sensitivity of RX circuit is pretty high (down to micro-volt), any noise would be captured by RX circuit and effect SNR.
1672793275634.png

We have a 12-layer PCB, and from the characteristics of noise captured by RX circuit, we can confirm source of noise/interference is from layer 6, while the trace of linear amplifier is on layer 3. Although layer 4 is a pure ground plane and layer 5 is a power plane, the two layers cannot fully block the interference from layer 6 to layer 3.

Use a series back-to-back diode to block the noise of linear amplifier is a common way in my field, below is the block diagram of the linear amplifier, actually there is a built-in back-to-back diode inside the chip to block noise. Though due to the noise on my PCB is coupled "after" linear amplifier chip so the back-to-back diode inside the chip is not able to block the noise, so I have to add my own back-to-back diode. Base on test result we found noise is truly blocked, though transmit waveform is distorted, as described in my first post.
1672794267199.png

Thanks !
 

crutschow

Joined Mar 14, 2008
34,283
Well if you insist on using the diodes, then you will have to live with the non-linearity it causes.

What's the point of having microvolt sensitivity in the RX circuit if the diodes are blocking anything below a half volt??

Sorry, but your circuit just does not make a lot of sense. :confused:
 

Thread Starter

shengwuei

Joined Aug 22, 2008
33
Hi Crutschow,

Sorry I did not make it clear enough. The "load" is actually a transducer, firstly the linear amplifier drive the transducer at maybe +/-20V 5MHz sine pulse, after the pulse is transmitted, then the transducer continuously receives signals at level down to micro-volt for a period of time, during the receiving stage, any noise on linear amplifier output below Vf of the diode can be blocked, thanks.
 

Thread Starter

shengwuei

Joined Aug 22, 2008
33
Hi Crutschow,

Thanks. I indeed noticed that if I choose a diode with lower Vf(ex. schottky diode), waveform distortion is also reduced. Below is the simulation and real measurement result, the schottky diode (SDM40E20LS, purple color) yields the least distortion, though not 100% satisfied with it but guess this is the best that I can achieve, thanks for your help !

1672882041616.png
 

star_earth

Joined Jan 4, 2023
1
Old Dude here. Two things that usually help with this type of distortion: Increase supply voltage and/or current, especially in the output stage, and a bit of negative feedback, either globally or locally.
 

MisterBill2

Joined Jan 23, 2018
18,175
One way to avoid the diode distortion is to bias the diode into full conduction during the transmit time, and then to bias it off during the listen time interval. This is by no means a new concept and there is a large body of information available about diode switching..
 

ericgibbs

Joined Jan 29, 2010
18,766
back-to-back diode : for some reason I have to add this diode to block the noise from amplifier output
Hi Bill,
Note the TS is using the diode pair to block low level noise, if he biasses them ON, this will defeat the object of the diodes.
E
 

MisterBill2

Joined Jan 23, 2018
18,175
Hi Bill,
Note the TS is using the diode pair to block low level noise, if he biasses them ON, this will defeat the object of the diodes.
E
My suggestion was to bias them ON for transmitting and bias them off for listening. I think I made that clear. The TS describes in post #7 using a passive approach that they are trying, but it causes the distortion they are seeing. Deliberate biasing is used in many communication systems and it works well. The transducer may be part of a position transducer for hydraulic cylinders, or some similar. For an application like that I offer a warning that the transmit/receive switching circuit must be perfect or disaster will strike. That caution is based on first hand experience.
 

Thread Starter

shengwuei

Joined Aug 22, 2008
33
Hi MisterBill2,

The diode is indeed turned ON when transmit voltage is larger than +/-Vf, or do you mean I need to add extra circuits to turn ON/OFF the diode ? How ? Could you provide some reference circuit of what you called "deliberate biasing" ? Thanks !
 

MisterBill2

Joined Jan 23, 2018
18,175
No, the diode is not turned on during transmit, it is just being biased into conduction, which means that it is probably far from saturation.
Deliberate biasing is passing a DC current through the diode to force it into the saturated conduction mode where the signal current variations do not affect the voltage drop. DELIBERATE BIASING IN THE OPPOSITE DIRECTION removes all charge carriers and produces a high resistance, low capacitance condition.
If you want example circuits that will require a bit of research, NOTon yootoob or amazon.
 
Top