Author Topic: AWG Buffer Amp (Read 8926 times)

mawyatt · « **on:** December 13, 2020, 11:32:15 pm »

Been using the Juntek DAP-1698 Dual Channel Power Amp as a AWG buffer, and thought others might be interested in a low cost Dual Channel Low Frequency AWG Buffer.

The design is based around a TI LM1875T power Op-Amp which is direct coupled input and output and has dual 26V power supplies from a XL6012 DC to DC converter and built-in fan which is noisy! Input power is 12VDC and supplied with a 6A AC to 12VDC converter. The LM1875 is a low frequency high power type Op-Amp that has a GBP, power BW and Slew Rate, 5.5MHz, 70KHz and 8V/us. The LM1875 design is a class A/B linear type output with overcurrect protection on both +- output current. It's heavily biased into the Class A for improved distortion, at the expense of static power dissipation with a nominal bias of ~70ma. Static PD is ~3.6 watts. Check out the data sheet on the TI LM1875, looks like a nice IC to base this around.

We just ran into an issue with the DAP-1698 going into a thermal limit cycle on both channels and suspected that the LM1875 was getting too hot because of an improper heatsink cooling. After removing the PCB, the two LM1875 were not mounted securely to the heat sink, causing a high thermal impedance and allowing the IC core temperature to rise and initiate thermal shutdown cycle. So the LM1875 IC did what it was designed to do and protect itself from over temperature

Once back together the DAP-1698 was fine driving a 100 ohm load at 40Vpp square wave or sine wave. Attached are some images and screen captures of the 40Vpp pulse rise and fall, 1, 10 & 100KHz sinewave and a 200KHz showing slew rate limiting. Also included a FFT of the output at 10KHz. Later maybe some tests at higher load currents if I can find some loads.

With the input shorted, the output measures ~20mvpp noise, ~2.5mv SD and ~6.5mv DC offset.

Anyway, hope this helps anyone looking for an inexpensive way to expand their AWG in voltage and current capability at DC & lower frequencies.

Best,

Edit I can confirm that the amp can deliver a 36vpp square wave into 33 ohms, that's ~ 9.8 watts!! Also can deliver ~0.9 amps peak before current limit kicks in.

mawyatt · « **Reply #1 on:** December 13, 2020, 11:34:40 pm »

Another image and the waveforms mentioned above.

Best,

mawyatt · « **Reply #2 on:** December 14, 2020, 05:59:19 pm »

Since there has been discussion of the value of a simple AWG on a DSO, just did a quick Bode Plot of this amplifier using the DSO built-in AWG.

Here you can see the amplifier voltage gain of ~6dB. Note how the gain slightly peaks, indicative of a underdamped amplifier response which shows in the step response with some overshoot. The -3dB bandwidth extends to ~1MHz for a 1vpp input (2vpp output) and ~500KHz for a 3vpp input (6vpp output), 3vpp was the max allowed input for the DUT from the Bode Plot using the DSO AWG.

For a check with an external AWG, the Bode' plot limits the max DUT input to 6vpp (12vpp output), this produces a -3dB bandwidth of ~250KHz, which is respectable considering the spec of 100KHz and amp cost.

Best,

ExaLab · « **Reply #3 on:** December 30, 2020, 05:44:06 pm »

Interesting product, even for its price.

Can you check if the BNC input of the device is properly decoupled from the LM1875?
This is a fundamental trick in order to avoid the damage of the connected instrument (that may be an expensive function generator...) in case of failure of the LM1875.

It would be great if you had the wiring diagram of one of the amplifier stages (it should be really simple).
Just to take a look before purchasing...

mawyatt · « **Reply #4 on:** January 06, 2021, 01:05:14 am »

Don't have a schematic, but looks like they have a 2X attenuator on the input with 10k resistors. This would explain the LM1875 set up as a 4X amplifier with a phase margin of ~45 degrees and a gain margin of ~6dB, so the overall effect is a gain of 2X input to output for the DAP-1698. The LM1875 would be very underdamped with a closed loop gain of just 2X (6dB).

We required a higher voltage than the DAP-1698 can provide for some applications and began development of custom AWG amplifiers capable of suppling 100~120Vpp based upon the LM3886 and TDA7293. So can see more detail on this thread (also one with a 150~180Vpp output at low currents). We'll post information on these devices later.

https://www.eevblog.com/forum/testgear/whats-inside-a-feeltech-fpa-301-better-than-juntek-dpa-2698/msg3307300/#msg3307300

Best,

mawyatt · « **Reply #5 on:** January 16, 2021, 03:36:20 pm »

Something that we've noticed when using this DPA-1698 amp is the output voltage excursions when power is applied and removed. This might damage a circuit under test. The output voltage excursions occur the same regardless if the input is terminated or not, and likely caused by a race condition with the internal +-26V power supplies.

First screen capture is when power is applied, second capture is when power is removed.

Best,

mawyatt · « **Reply #6 on:** January 16, 2021, 10:51:43 pm »

Our Siglent AWG (SDG2042X) has a 20 volt peak to peak output amplitude limit, which creates a maximum peak to peak output with the DPA-1698 of 40 volts (2X). We needed a signal greater than 40 volts peak to peak for a test and haven't another amplifier capable of this level output at the moment, then realized the DPA-1698 should be able to produce greater than 40 volts peak to peak since the internal supply voltages are about +-26V for the LM1875.

We connected the AWG to the DPA-1698 channel 1 input as usual, but connected the channel 1 output to the channel 2 input. Then observed the channel 2 output which should be 2 times 2X the AWG input, or 4X. With the AWG set to 12.5 volts peak to peak we should be able to get ~50 volts peak to peak output. Works well without any nasty signal behavior and produced the expected 50 volt peak to peak clean signal as shown below, even when heavily overdriven as in the second image.

Best,

precaud · « **Reply #7 on:** January 17, 2021, 01:07:17 am »

Accoring to the manual, the LM1875 "is designed to be stable when operated at a closed-loop gain of 10 or greater". Are they doing anything obvious to make it stable at 4X? Or is the spec just conservative?

mawyatt · « **Reply #8 on:** January 17, 2021, 02:19:26 am »

If you look at the LM1875 data sheet Figure 12 Open Loop Gain and Phase vs Frequency this shows a Phase Margin of about 45 degrees at a gain of 12dB (4X). So the response is slightly underdamped and this shows in the step response. I think they mention 10X (20dB) because the phase margin looks to be ~60 degrees and considered the design goal for phase margin in most closed loop amplifier systems.

As you would expect the overshoot/ringing in the step response is not constant and varies as the output voltage swing and load change, indicating a varying phase margin with level and load.

I can't find any specific feedback components to effect the compensation, but haven't traced the circuit out either so there could be a RC network in the feedback. There is a 1 ohm series output resistor to help isolate capacitive loads which are usually the most troublesome. I haven't found any stability issues other than the burst oscillations during the power down glitch waveform shown earlier.

Believe this is a good value amp for AWG buffering. Wish Juntek would consider the LM3886 which has mute capability and supports higher supply voltages. I have just completed a design with the LM3886 that uses a closed loop gain of 5X with compensation. Also have one for the TDA7293 and for very high voltage outputs (~320 VPP) based upon the PA340.

Best,

precaud · « **Reply #9 on:** January 17, 2021, 02:20:27 pm »

Good points, mawyatt. A well-chosen small capacitance across the feedback resistor would help stability, as well. I find this little Juntek amp intriguing and wish I had a need for one

But I have just started building a dual LM1875-based amp for a set of electrostatic headphones and don't need 20dB of gain for them. Distortion should be lower at the lower gain, as well. But I will definitely keep an eye out for the turn-off burst which you showed... the headphones would not be happy to be on the receiving end of that after every use! I'm thinking it might be caused by the XL6012, and not endemic to the LM1875, which has been out there for years and one would think it would have been mentioned before now.

mawyatt · « **Reply #10 on:** January 17, 2021, 03:01:04 pm »

Agree a simple feedback network is in order which is what we have on the designs we've done. Used the classic R and C in series and the combo in parallel with the feedback R. Might consider doing this with the Juntek if stability was an issue.

At first we thought the high frequency content of the waveform at power down was due to the XL6012 but the frequency is ~ 2MHz which indicates it's coming from the LM1875. What's likely happening is the power supply voltages are falling unequally at power down, and the LM1875 internal bias is getting messed up and changing the characteristics allowing this burst oscillation. With a carefully controlled power down this should be avoided. The LM3886 has a mute function which could be employed to mute the output during power up and down, and has a more uniform open loop response without the quirkiness of the LM1875 around 1~4MHz as shown in Figure 12 Open Loop response.

I'm a retired IC designer so interested in the chip details, and if you study the designs you can see that the LM3886 employs local internal feedback to help keep the output well behave during the mute process which is a current controlled gain function as shown in Figure 44 of the LM3886 data sheet.

If you plan to power these amps from a SMPS be sure to pay attention to the PSRR, especially the negative supply PSRR. At the XL6012 cycle rate this is only 20~30 dB rejection. The DPA-1698 uses a dual inductor (4.7uH) power supply filter on both the + and - supplies for this reason, however you have the benefit of the lower audio frequencies so maybe not an issue.

Anyway good luck with your headphone amp design and build as I'm sure you'll find the LM1875 a good device.

Best,

precaud · « **Reply #11 on:** January 19, 2021, 01:28:56 pm »

Yes, the -V PSRR is a red flag waving, for sure, but it's typical of most op amps. For audio, not much to concern oneself with.

A little OT, but I wonder how much of the 1875 distortion is common-mode, and if it could be lessened by the usual tricks; balancing the impedances at the inputs, and/or operating it in inverting mode (it's very easy to swap the speaker leads to restore polarity...). The data sheet only shows it used as a non-inverter.

precaud · « **Reply #12 on:** January 23, 2021, 06:19:40 pm »

I've built the amp, running at 16dB gain into a nominal 16 Ohm load, made a few mods to address specific issues, and I have to say the result is underwhelming. Distortion spectra is particularly ugly. The most significant improvement came from adding a 100pF compensating capacitor across the 6k65 feedback R (-3dB @ 240kHz). This lowered 3rd harmonic distortion by over 12 dB, making the sound much less edgy and brittle (see attached). But it still isn't an impressive amp for high-quality audio by any measure.

BTW mawyatt, I think you were reading the Fig 12 gain-phase plot wrong. Phase margin is 45º at 10X gain, which is probably why they recommend it as minimum...

mawyatt · « **Reply #13 on:** January 23, 2021, 09:59:43 pm »

Quote from: precaud on January 23, 2021, 06:19:40 pm

I've built the amp, running at 16dB gain into a nominal 16 Ohm load, made a few mods to address specific issues, and I have to say the result is underwhelming. Distortion spectra is particularly ugly. The most significant improvement came from adding a 100pF compensating capacitor across the 6k65 feedback R (-3dB @ 240kHz). This lowered 3rd harmonic distortion by over 12 dB, making the sound much less edgy and brittle (see attached). But it still isn't an impressive amp for high-quality audio by any measure.

BTW mawyatt, I think you were reading the Fig 12 gain-phase plot wrong. Phase margin is 45º at 10X gain, which is probably why they recommend it as minimum...

Not having any experience with the LM1875 is why we included the LM3886 and TDA7293 for evaluation.

Don't think we got the gain-phase plot wrong (likely you are not using dB for the Gain axis and why @ 10X shows 45 degrees which is incorrect, 10X is 20dB), included as Fig 12 from the LM1875 data sheet. As you can see the phase is ~45 degrees when the gain crosses 12dB (4X), at 10X which is 20dB gain, the phase is ~65 degrees.

Best,

precaud · « **Reply #14 on:** January 24, 2021, 12:54:23 am »

Ooops, you're right, my bad, I was reading the gain as mag ratio, not dB.

mawyatt · « **Reply #15 on:** January 24, 2021, 01:37:28 am »

I just ran a quick test with the Juntek 1698 (LM1875) at 1KHz with a 10 ohm load. First is directly from AWG (SDG2042X), second is from Juntek 1698 with 10 ohm load (Gain 2X).

Best,

precaud · « **Reply #16 on:** January 24, 2021, 01:57:29 am »

Interesting, thanks for posting it. They're both pretty stinky...

mawyatt · « **Reply #17 on:** January 24, 2021, 02:44:22 pm »

Using an inverting stage should give better HD results, although only slightly I expect. The LM1875 doesn't show CMRR, nor have a spec, the LM3886 shows ~75dB @ 1KHz.

The LM1875 is OK for general purpose use, but not something for very low distortion applications like you've indicated. The LM3886 seems like a better solution overall and shows better distortion characteristics in the data sheet, why we included this in the experimental design.

Best,

precaud · « **Reply #18 on:** January 24, 2021, 02:55:03 pm »

My thoughts as well, I've been reviewing the 3886 data sheet this morning...

ExaLab · « **Reply #19 on:** January 31, 2021, 11:39:22 am »

Quote from: ExaLab on December 30, 2020, 05:44:06 pm

Interesting product, even for its price.

Can you check if the BNC input of the device is properly decoupled from the LM1875?
This is a fundamental trick in order to avoid the damage of the connected instrument (that may be an expensive function generator...) in case of failure of the LM1875.

It would be great if you had the wiring diagram of one of the amplifier stages (it should be really simple).
Just to take a look before purchasing...

Just received! Nice product!
At the moment, only some doubts about the supplied power supply: a fake Liteon without the input earth pin and without an output common mode HF rejection toroid... This is unacceptable given the uses for which the device is intended!!

I have already ordered a certified low noise power supply. The presence of the earth pin is crucial to break down the common mode mains network noise injected on the target device.

mawyatt · « **Reply #20 on:** February 02, 2021, 07:14:00 pm »

The JunTek supplied 12V 5A SMPS is better than most cheap types we have, but certainly not lab grade. You could use a linear ~12V supply if switching noise is an issue. After studying the JunTek some the power input voltage can be greater than 12 volts since it feeds a XL6012 SMPS chip and the input filter cap is rated at 35VDC, it also uses dual LC low pass (2 inductors) filters on the +- 26 volts to reduce the switching noise to the LM1875 amps.

What power supply did you select to replace the JunTek supplied device?

Best,

ExaLab · « **Reply #21 on:** February 03, 2021, 06:05:02 pm »

Safety and noise are the reasons that led me to replace the SMPS.

Many unbranded chinese SMPS do not respect the most basic safety rules and this represents a risk both for the operator (e.g. for me) and for the relative instrumentation.
For less than 25€ I bought a branded chinese device which meets all safety regulations (Mean Well GST60A12-P1J).

More, as mentioned above, the original supplied smps is not equipped with the earth pin and this (for emi reasons) makes it substantially not recommended for lab use (these kind of devices are generally used as strip led drivers...).

If the noise related to the internal XL6012 will be unacceptable for some lab uses, I am thinking of equipping the DPA-1698 with an additional power input so that it can be powered, if required, also with an external dual linear power supply.

mawyatt · « **Reply #22 on:** February 03, 2021, 07:03:21 pm »

Mean Well SMPS seem to have a good reputation, good selection.

I've found that the noise on the DPA-1698 output is higher on one CH1 than CH2 (same for DPA-2698) and doesn't depend on the input SMPS or Linear Supply much.

Bringing in external +-26VDC will power the dual LM1875 but you'll need ~12VDC for the fan. You'll need the the fan since the LM1875 draws considerable power even static.

Best

mawyatt · « **Reply #23 on:** February 04, 2021, 01:08:54 am »

Just got the Juntek DPA-2698. It's based upon a high speed op-amp being buffered by a pair of BUF634 and the outputs tied together via 1 ohm resistors. Power is from a XL6012 creating +-15VDC. The case is the same as the DPA-1698 and also has a fan which is also somewhat loud.

The bandwidth is ~50MHz at 4VPP, 40MHz at 10VPP with 50 ohm load, and drops to ~10MHz at 25VPP (100 ohm load). Rise and fall time is ~12ns at 10VPP with 50 ohm load, and 2-Tone IMD at 1MHz with 2VPP signals is ~-53dB. IMD is better than this but that's as good a source (SDG2042X) as we have.

Noise is rather high and higher on CH1 with ~100mvPP, CH2 27mvPP. The fan seems to have an effect on the noise, as it's lower with the fan unplugged by over 2X.

Here's a couple images disassembled.

Best,

Noy · « **Reply #24 on:** February 04, 2021, 07:39:09 am »

You should add a gapfiller and check the soldering.
I did..

And maybe solder a SMD cap directly on the backside from the fan?


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: AWG Buffer Amp (Read 8926 times)

Share me