Keithley 2304A high speed power supply

[Short Circuit]

This afternoon, I finally made some time to modify my Keithley 2304A power supply.
The power supply is extremely noisy, which made it too annoying for regular use. Also, it has screw
terminals on the back for the connections, not very convenient for general lab usage.

So I bought a new Sunon fan and a couple of banana jacks with 'switch' function (jacks with two half
terminals inside, which are shorted when a plug is inserted).

The machine contains a surprisingly large amount of electronics; two 20x29cm boards full of control
and power electronics, and a small user-interface board to control the LCD and keys.

[Short Circuit]

...and closeup of the modifications.

Nice high-accuracy meters in this machine;
no connections shows a current of -0.0005mA (it's a 4-quadrant PSU, so current can be nagative, and in this case, it needs a bit of calibration)
with my Fluke 189 connected, load is increased to +0.0009mA

[alm]

Nice piece of equipment, and some professional looking modifications. Are all those caps in the second picture (top PCB) in parallel?

[Short Circuit]

Yes indeed, 40 mlcc's all parallel, and only 36.6uF... Seems highcap mlcc's were not so common when this PSU
was designed in 1997 (copyright notice on the PCB).

That makes this a 15 year old device, yet the very same model is still available new from Keithley. Not bad.

[muvideo]

Wonderful power supply,  alm also  my eye immediately was attracted by that huge ceramic capacitors bank.
Also it's impressive the meter resolution of 100nA, 14V in 10Mohm is 1.4uA and it shows 0,9uA+0,5uAoffset=1,4uA!

Fabio.

[Mechatrommer]

Keithley 2304A high speed power supply
The power supply is extremely noisy

what it means by high speed? ho high? 100MHz? 1GHz?
why this branded psu with alot of electronics is extremely noisy?

[Jad.z]

Hi Short Circuit

This is a nice unit indeed. I actually have one of this laying around the lab collecting dust.
One day it just quitted at me. Unfortunately, I never had the time to fix it nor the willing to debug a (0.12 m^2) of circuitry 

@Mechatrommer

Here is what the datasheet says:

TRANSIENT RESPONSE TO 1000% LOAD CHANGE:
NORMAL MODE:
Transient Recovery Time:
                                        <50us to within 100mV of previous level.
                                        <100us to within 20mV of previous level.
ENHANCED MODE:
Transient Recovery Time:
                                        <40us to within 100mV of previous level.
                                        <80us to within 20mV of previous level.


BTW, does anyone have the schematic for this one or the service manual please. I searched everywhere and I just can't find it 

[Short Circuit]

I have no idea about the noise characteristics of this power supply, but that's not important either.
This machine is designed to simulate batteries for GSM test & development purposes. Cell phones
tend to switch from very low standby current to 2Amp RF TX pulses. A normal PSU simply collapses
under such load change because the load transient response is not fast enough.

Personally I use the machine for R&D on systems that flip between normal operation with load current
up to few amps and standby current in the mA range. The nice thing abou this PSU is the very accurate
5mA current sense range without limiting the load capacity. With a normal amp-meter you must
arrange to stay in either range because of the voltage drop across the load-sense resistor, which
makes it impossible to jump from 1mA to 1A without goofing around with jumper cables to bypass
the mA meter.

Jad.Z; have you tried asking your local Keithley representative? I found them very helpful when I
asked them for spareparts of this PSU. I bought it from eBay being an ex-rackmount device, so there
were no handles and feet included. It was no problem at all to buy these, costing a mere 25 euro or
thereabout. (Although perhaps it helped being a customer already, as I cannot believe they would
create a customer account for only 25 of spare parts.)

[saturation]

It has great specs; it has linear like stability with digital controls.  The eBay price is near $1000.

[Short Circuit]

That are trader prices. Normal auction price can be much more reasonable; I paid $400 for it.

[Mechatrommer]

It has great specs; it has linear like stability with digital controls.  The eBay price is near $1000.

why this branded psu with alot of electronics is extremely noisy?

now i get it. mixed smps + linear psu tech, thats why more electronics i guess. i've been (currently) eyeing this kind of topology.

[saturation]

Yes, it has the best of both worlds, but I didn't read too much into it, before going this direction with a PSU it helps to confirm:

You can series or parallel it with another capable PSU without issue, to get more volts or amps .. very flexible and preferred, many cheap linear PSU can do this easily but I don't know if they can respond as quickly as this PSU

Its user repairable: the controller circuitry of the Keithley looks like a lot of work to repair if damaged

It has great specs; it has linear like stability with digital controls.  The eBay price is near $1000.

why this branded psu with alot of electronics is extremely noisy?

now i get it. mixed smps + linear psu tech, thats why more electronics i guess. i've been (currently) eyeing this kind of topology.

[SeanB]

Those mains powered fans are easy to fix, they normally only need a few drops of engine oil on the bearing and then they run for years again. After a few years of continuous operation the sleeve bearings will wear out, but the oil normally evaporates before this. A drop or three and away it goes.

[wlanfox]

I just acquire one of this very nice power supply / source (can also be use as a constant source )  ... Does anyone know what that J13 header is for? JTAG, SERIAL, other? try looking for the service manual but so far I have come up empty...

[reagle]

Just grabbed one of these as well. One of the MLCC  in the large parallel group is completely burned. Looking around, nothing else seems unhappy- I presume it just happened to go on its own?

[TheBay]

Wow lovely PSU!,

Get some Retrobrite on it

[eas]

I snagged one of these for ~$100 a month or so ago. It looks like someone stabbed through the clear section of the plastic sheet over the display with an Exacto knife, but is in good shape otherwise.

I have a home server with picoPSU + 12v power brick for quiet operation. The 12v brick died after a bit over a year. I didn't have a suitable replacement, so I ran it off the 2304A for a few days. The thing is freaking LOUD. I'll need to do a fan retrofit if I keep it.

[reagle]

Mine was a cool $75 on ebay, plus a few "extra" parts I bought to "pimp it out".
Ended up adding binding posts on the front, as the terminal block on the back is not optimal for bench use. Then I got fed up with a low contrast and poor viewing angle of the original yellow LCD display and swapped it for a modern drop in replacement. Ended up having to add a potentiometer to adjust contrast voltage as well.
I was really hoping to find an OLED one, but geometry and pinouts did not align. So I used this simple negative LCD (ERM1602SBS-3 from BuyDisplay). The pinout and mechanicals fit nicely, I just had to flip backlight polarity jumpers

[Magnum]

I replaced the fan of my 2304A to lower the noise.
Keithley placed a very big component directly in front of the fan and that makes a horrible high frequency noise due to air turbulence. The original one is 25mm thick, I used now a 20mm thick fan with the same airflow. The noise was reduced by more than 3dB, and all the high frequency noise is gone. It is still not quiet, but that is because of the high airflow. If you want to live with higher temperatures inside the unit you can reduce the noise further, just remember to use a thinner fan.

The one I used: Sunon MB 60202 V1-000U-A99
available from http://www.elpro.org for 3,79 EUR

[mrjoda]

hi guys, i bought 2304A few day ago from ebay.de for 150€ shipped. Inside looks clean, some particle of dust but nothing hardcore. I clean it.

I noticed a little mod board marked as Keithley wired to DG408 and glued with double sided tape. Do you know why ?

Voltage looks good, can be set up, current also. Device switch between CC/CV without issue. Only thing i notice is little chirp noise from inside, no matter if is loaded or not. I try to record it. I will try to find source, maybe is it only bearings inside FAN, maybe is it something more serious.

Is it neccesary to change electrolytic caps ? they are nichicon/nippon chemicon but device is marked 1997 so they are 20 years old.

here is video. I dont know if you will hear it but i can.

https://youtu.be/G78v_7z-M18

here is another vid. Is it definetly from main power supply
https://youtu.be/bdmQ3gakizU



edit :

i try to push components with plastic marker, same pitch. I measured all voltages on testpoints - all correct and stable.

[alm]

I have not heard of any of these units having leaking electrolytics. They seem to have used good quality caps and not overheated them. So I see no pressing need to start shotgunning all electrolytics unless they are showing signs of leakage or increased ESR.

The chirp could be coil whine. Making sure all screws are tight might help a bit.

[mrjoda]

so i finally did some scratch repair and i replaced fan for Sunon MB 60202 V1-000U-A99  (from Magnum post). Is it still noisy but i can live with that noise level. I think about pulling out two leds for ON/OFF status and CV/CC for better readability. There are some testpoints on board marked as OVER-I, maybe it will be possible to use it for led indicator.

I repaired scratches with some wierd clamp but it works perfect. Heatgun + clamp and scratches disappeared.

[macboy]

I've effectively repaired dings like that with just hot air. The ABS plastic has an amazing shape memory. The trick is to use just the right temperature, so that the plastic softens enough to go back to original shape, but doesn't melt (and lose its memory). I've also repaired various LEGO parts (also ABS) with the same technique after my boys bent, chewed or abused them. It's amazing how much deformation can be reversed this way.  Pro tip: experiment with temperature and technique on LEGO bricks before moving on to big-$$$ equipment. 

[bitseeker]

Good to see that heat works OK. I've got a 2700 with a dinged corner, but was reluctant to have a go with the heat gun.

[TheSteve]

Careful application of heat does work wonders. I have before/after pictures of an Agilent 33250A that had pretty crazy damage posted somewhere here.

[mrjoda]

yes, heatgun worked very well. Next step is find the source of annoying chirp sound a remove it...

[technogeeky]

I just picked up one of these (for $125 with free shipping). It was in working order, and it appears to operate properly. The only thing unexpected at all is how long it will stay at a elevated voltage before reducing the voltage to meet the current limiting requirements.

The speed of this supply in enhanced mode is absolutely incredible. I'll post some screenshots soon.

In the meantime, a few questions:

1) In what conditions is this power supply able to be used like the famous Keithley SMUs? It seems like it can't reverse polarity (so that's two quadrants out). So I think the leaves:

• source voltage
• source current^[0]
• sink current
• source resistance^[1]

while measuring:

• voltage (with DMM)
• current (direct measurement of source)
• current (with DMM, via current shunt)

I wonder what kind of things this could be used to characterize? It would seem that most/all diodes, some transistors, some inductors, some sensors, etc... could be characterized. Anyone have some cool/unconventional/non-obvious things to try?

2) What do I need to do to protect this unit?

• The 20V isolation from low means that I just need to not try and float the supply on top of some other supply? Anything else to watch out for here?
• What do I need to do to ensure I don't damage the supply when sinking current? Say, if I'm charging a 12V UPS battery with unknown impedance (but presumably low), will this dump e.g. 100A into the unit, or will it limit its own sinking current?
• Anything else to watch out for?

3) Can anyone offer assistance in calibration? The way I see it, there are 2 economical options:

• I borrow someone else's calibration fixtures (paging TiN?  ) and do the calibration with my 5.5 digit Keithley 197.
• I purchase the calibration parts, send them to someone (paging TiN again?  ) to characterize, and calibrate with my 197.

This isn't urgent since I am not sure it's even out of calibration.

Any other tips or tricks?

Besides being a high speed supply, this is also my lowest noise supply (I don't have any linear supplies at all)! Looking forward to it!
^[0]: I would guess that source current means establishing a set current by varying the voltage, which the 2304 can't do, right?
^[1]: The 2304A doesn't have output impedance programming. Which leads me to the last question:

4).
I expected that the parts for this would be unpopulated, but it appears everything anywhere near the output is fully populated. Does this mean we could somehow enable the output impedance stage? Or would it be possible by using the 2303 firmware?

[technogeeky]

Well, some googling and reading and testing has elucidated some things.

1) The 2304a model I have has an analog board which is marked to be one of two units. However, the other unit is the 2303 which is just a lower power version of the 2304a. Neither unit has output impedance control. So that's out.

2) This unit is, if anything, a voltage source. I don't think a voltage source with current limiting acts like a current source (much less an ideal one), so that's out. As is resistance source (see point 1).

3) TiN is in Taiwan, so both options for calibration (for me) are out.

I still would love some perspective on how to think about and how to achieve protection for this power supply, especially when sinking current.

I still would like to hear about what kind of experiments one could conduct using this power supply characterizing things. For example, I powered a white LED and was able to easily see changes in current caused by heating and cooling modulated by touching it with my fingers!

I am going put banana sockets on the front of the power supply, almost certainly. I was wondering about putting a large touch screen LCD on the front of it, but I don't think it's worth it given that this unit isn't quite as versatile as I thought.

[pigrew]

There are two nagging issues with my 2403A that perhaps someone could comment on?

First, is there a way to trigger current digitization other than with a current pulse? I want to trigger the digitization at will over GPIB.

Secondly, mine has firmware A02. One Keithley forum post mentioned it needed a new digital board to upgrade their firmware. Would it work if I stuck in a ROM flashed with a newer firmware? I've run into some unreliability in GPIB when multiple queries are made at once, and I wonder if they were fixed in later firmware. Is the version history posted anywhere?

[alm]

2) This unit is, if anything, a voltage source. I don't think a voltage source with current limiting acts like a current source (much less an ideal one), so that's out. As is resistance source (see point 1).

The 2304A should work fine as a current source. The current regulation specs are specified in the datasheet. Limitations compared to a SMU will be resolution (only down to 1 mA in a single programming range) and possibly output impedance. An ideal current source will have an infinite output impedance (which means a perfect load regulation).

A common problems with using a voltage source with current limit as current source is that the output capacitor, which lowers the impedance of the voltage source, spoils the load regulation as current source. If you set the current source to 1 mA, and connect a 10 kOhm resistor across it, it will charge the output cap to 10 V. If the output now gets shorted, the output cap discharges through the load. Depending on the specifics, this could easily source multiple amps of current (so a 100 000% load regulation) for a few ms. These high-speed power supplies have little to no capacitors on the output (that is how they achieve the high speed), so I would expect them to perform much better as current source.

I wonder what kind of things this could be used to characterize? It would seem that most/all diodes, some transistors, some inductors, some sensors, etc... could be characterized. Anyone have some cool/unconventional/non-obvious things to try?

The 1.25 mA current programming resolution might be limiting for low power devices.

• The 20V isolation from low means that I just need to not try and float the supply on top of some other supply? Anything else to watch out for here?
• What do I need to do to ensure I don't damage the supply when sinking current? Say, if I'm charging a 12V UPS battery with unknown impedance (but presumably low), will this dump e.g. 100A into the unit, or will it limit its own sinking current?
• Anything else to watch out for?

I would also make sure to tie the output to ground somehow if there are other voltage sources involved (connect one of the output rails to ground, possibly via something like anti-parallel diodes to prevent round loops).

It will limit the current it will sink (see datasheet or manual again), unlike the similar half-rack HP/Agilent units. With batteries, I would be worried about the overvoltage protection that might short the output. That could result in some fireworks. I would use a series fuse when connecting it to large batteries/caps, and a series diode when charging.

I would not worry too much about calibration. From what I remember, it requires some loads and measurement of voltage and current that you input into the power supply. A good 5.5 digit DMM should be sufficient for a decent performance verification (if not exactly within factory specs).

[mrjoda]

hello

Look what i found. There is a RJ64 remote port for external remote panel. So i did some test and i found this :

Inside is one ADM202 High Speed, 5 V, 0.1 µF CMOS RS-232 Driver/Receiver.

RJ45 PINOUT
1-2- +5V DC
3 - REMOTE ENABLE PIN - LOG1 activate the remote acces
4 - TX
5 - RX
8-7-6 - GND

RS232 setup : baudrate 9600, 8Bit , Parity none, handshaking none.

With this setup i am getting default response :

Code: [Select]

<0><0><0><0><0><0>?<0><0><0>  ?<0> start response   model INVALID <0>2304A IEEEADR  6A03   00000 50Hz?<0> start response    rev INVALID  ?<0> start response    key INVALID  <0><0> 0.000 V  NL OFF 0.0014 A
I thought it will be SCPI protocol but is not. It does nothing when SCPI commands are sent or return

 

Code: [Select]

?<0>received INVALID    key press
It looks like there must be some button interaction at remote panel and it does not use SCPI but somethink more simple. What do you think ? Does anybody have this remote panel which can be used for data sniffing ?

I would like to made similar panel but with keyboard and rottary encoder because setup of this unit is not very pleasure.

[alm]

I'm pretty sure the front panel of other Keithley instruments like the Keithley 2000/2001 also uses an asynchronous serial protocol. Try to see if anybody (like TiN) has reversed-engineered those, and then check if the protocol for this instrument is similar. You could also check if the built-in front panel also uses the same protocol, and sniff that.

[bitseeker]

I wonder if the protocol might be like that used by older Keithley meters (e.g., 196) when talking via GPIB. It's more compact than SCPI.

[mrjoda]

i think it will be very simple protocol. I tried sent random hex number via terminal and i got only one reaction - 0xE0 enabled output but nothing else. I can't disable it with any combination and i can't change the output voltage.  I will continue with experiments. My plan is write some simple python code (i never use python before so i want to try it) which will contionusly sending ascii table and combination 0x00  to 0xFF and compare new received data with previous for changes. When changes occures it will save combination and received data.

[technogeeky]

I've installed internal connections to three sets of front panel banana jacks. It looks nice. I'm a little worried about noise pickup but it doesn't seem too bad yet, and I can desolder the wires from the output board to test them again later, if I still think it's a problem.

I still need to figure out what mods you guys did to your fans to make them temperature controlled to some extent. Or whatever you guys did.

Any more details?

[mrjoda]

Hi, i am unable to found a source of chirp noise. It is definitely from primary side of power source. Noise disappear when unit is under heavy load.

[TiN]

It's simple UART with few binary codes for stuff like blinking, cursor etc.

3) TiN is in Taiwan, so both options for calibration (for me) are out.

You give up too easy. Shipping little packet with few resistors and connector is not that expensive.

[petarpera]

Keithley 2304a start up with two errors (514 DC calibration data lost and 515 Calibration dates lost). But after start up, do not respond to any button. Tried to set parameter through GPIB - it is recognised but do not accept commands. I checked voltages on test spots, evertything seems to be ok. I would really appreciate any suggestion.

[garrettm]

Hello gentlemen!

I've been hesitant to grab one of these Keithley 2304A high-speed PSUs because of the small screen and wacky connector on the back. However, a good deal recently popped up on the bay and I finally decided to pull the trigger and ended up snagging it for less than $120 USD with shipping.

The listing photos were pretty terrible (a blurry photo of only the front), and I wasn't sure if my unit would come with the rear screw terminal connector or not so I did some looking and think that the original part is made by Phoenix Contact. After getting my unit in the mail, I was pleased to find out it did come with the rear connector installed.

Thinking that I would like to have a spare connector on hand, in case I misplaced the original, I busted out my calipers and did some more looking. The 8-pin connector uses 0.2 inch / 5.08 mm pitch and can be replaced with Molex part number 395340008 "5.08mm Pitch Eurostyle Horizontal Plug, with Retention Screws, 8 Circuits." These are black connectors instead of green, but they are direct replacements for the Phoenix Contact part and, at least for me, were easier to obtain.

Hopefully this helps anyone who is looking for a replacement connector, as many of these units unfortunately come without one installed.

[rl78family]

Hi. What is the difference between models 2304A and 2304?

[alm]

Hi. What is the difference between models 2304A and 2304?

Very little, I think. I skimmed my 1998 paper catalog and compared the specs for the 2304 to the 2304A, and didn't notice any difference. If it helps, I could scan those three pages from the catalog. There might be differences between the circuits, though. I believe the 2304A came out together with the 2303,so it's possible the revision was so the 2303 and 2304 could share more parts.

[rl78family]

If it helps, I could scan those three pages from the catalog.

Thanks, there is no need for this. Thanks for the answer!